vvEPA
March 2023
EPA Document No.
822P23008
PUBLIC COMMENT DRAFT
APPENDIX: Toxicity Assessment and Proposed
Maximum Contaminant Level Goal for Perfluorooctane
Sulfonic Acid (PFOS) in Drinking Water
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PUBLIC COMMENT DRAFT
APPENDIX: Toxicity Assessment and Proposed Maximum Contaminant
Level Goal for Perfluorooctane Sulfonic Acid (PFOS)
in Drinking Water
Prepared by:
U.S. Environmental Protection Agency
Office of Water (4304T)
Health and Ecological Criteria Division
Washington, DC 20460
EPA Document Number: EPA 822P23008
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Disclaimer
This document is a public comment draft for review purposes only. This information is
distributed solely for the purpose of public comment. It has not been formally disseminated by
the U.S. Environmental Protection Agency. It does not represent and should not be construed to
represent any agency determination or policy. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Contents
Disclaimer ii
Contents iii
Figures vii
Tables xi
Acronyms and Abbreviations xxi
Appendix A. Systematic Review Protocol for Updated PFOS Toxicity Assessment. A-27
A. 1 Overview of Background Information and Systematic Review Protocol A-28
A. 1.1 Summary of Background Information A-28
A. 1.2 Problem Formulation A-29
A. 1.3 Overall Objective and Specific Aims A-31
A. 1.4 Populations, Exposures, Comparators, and Outcomes (PECO) Criteria.... A-32
A. 1.5 Literature Search A-3 5
A. 1.6 Literature Screening Process to Target Dose-Response Studies and PK
Models A-46
A. 1.7 Study Quality Evaluation Overview A-77
A. 1.8 Data Extraction for Epidemiological Studies A-125
A. 1.9 Data Extraction for Animal Toxicological Studies A-134
A. 1.10 Evidence Synthesis and Integration A-139
A. 1.11 Dose-Response Assessment: Selecting Studies and Quantitative Analysis ... A-
144
A. 1.12 Candidate Toxicity Value Derivation and Selection A-147
A.2 Meta-Analysis Table A-150
A.3 Studies Identified After Updated Literature Review A-154
Appendix B. Detailed Toxicokinetics B-l
B.l Absorption B-l
B.l.l Cellular Uptake B-l
B. 1.2 Oral Exposure B-2
B. 1.3 Inhalation Exposure B-2
B. 1.4 Dermal Exposure B-2
B. 1.5 Developmental Exposure B-2
B.l.6 Bioavailability B-2
B.2 Distribution B-3
B.2.1 Protein Binding B-3
B.2.2 Tissue Distribution B-6
B.2.3 Distribution during Reproduction and Development B-16
B.2.4 Volume of Distribution B-40
B.3 Metabolism B-45
B.4 Excretion B-45
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.4.1 Urinary and Fecal Excretion B-45
B.4.2 Physiological and Mechanistic Factors Impacting Excretion B-48
B.4.3 Maternal elimination through lactation and fetal partitioning B-49
B.4.4 Other routes of elimination B-51
B.4.5 Half-life Data B-53
Appendix C. Non-priority Health Systems Evidence Synthesis and Integration C-l
C. 1 Reproductive C-l
C. 1.1 Human Evidence Study Quality Evaluation and Synthesis C-l
C. 1.2 Animal Evidence Study Quality Evaluation and Synthesis C-14
C. 1.3 Mechanistic Evidence C-24
C. 1.4 Evi dence Integrati on C -2 5
C.2 Endocrine C-44
C.2.1 Human Evidence Study Quality Evaluation and Synthesis C-44
C.2.2 Animal Evidence Study Quality Evaluation and Synthesis C-52
C.2.3 Mechanistic Evidence C-71
C.2.4 Evidence Integration C-72
C.3 Metabolic/Systemic C-78
C.3.1 Human Evidence Study Quality Evaluation and Synthesis C-78
C.3.2 Animal Evidence Study Quality Evaluation and Synthesis C-95
C. 3.3 Mechani sti c Evi dence C-104
C. 3.4 Evi dence Integrati on C-105
C.4 Nervous C-l 13
C.4.1 Human Evidence Study Quality Evaluation and Synthesis C-l 13
C.4.2 Animal Evidence Study Quality Evaluation and Synthesis C-121
C.4.3 Mechanistic Evidence C-130
C. 4.4 Evi dence Integrati on C-130
C.5 Renal C-141
C.5.1 Human Evidence Study Quality Evaluation and Synthesis C-141
C.5.2 Animal Evidence Study Quality Evaluation and Synthesis C-146
C.5.3 Mechanistic Evidence C-l49
C. 5.4 Evi dence Integrati on C-150
C.6 Hematological C-156
C.6.1 Human Evidence Study Quality Evaluation and Synthesis C-156
C.6.2 Animal Evidence Study Quality Evaluation and Synthesis C-159
C.6.3 Mechanistic Evidence C-161
C. 6.4 Evi dence Integrati on C-162
C.7 Respiratory C-166
C.7.1 Human Evidence Study Quality Evaluation and Synthesis C-166
C.7.2 Animal Evidence Study Quality Evaluation and Synthesis C-168
C.7.3 Mechanistic Evidence C-172
C. 7.4 Evi dence Integrati on C-172
iv
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
C.8 Musculoskeletal C-176
C.8.1 Human Evidence Study Quality Evaluation and Synthesis C-176
C.8.2 Animal Evidence Study Quality Evaluation and Synthesis C-179
C.8.3 Mechanistic Evidence C-180
C.8.4 Evidence Integration C-181
C.9 Gastrointestinal C-185
C.9.1 Human Evidence Study Quality Evaluation and Synthesis C-185
C.9.2 Animal Evidence Study Quality Evaluation and Synthesis C-187
C.9.3 Mechanistic Evidence C-189
C. 9.4 Evi dence Integrati on C-189
C.10 Dental C-192
C.10.1 Human Evidence Study Quality Evaluation and Synthesis C-192
C.10.2 Animal Evidence Study Quality Evaluation and Synthesis C-194
C.10.3 Mechanistic Evidence C-194
C.10.4 Evidence Integration C-194
C.ll Ocular C-196
C. 11.1 Human Evidence Study Quality Evaluation and Synthesis C-196
C.11.2 Animal Evidence Study Quality Evaluation and Synthesis C-197
C.11.3 Mechanistic Evidence C-198
C.l 1.4 Evidence Integration C-199
C.12 Dermal C-201
C.12.1 Human Evidence Study Quality Evaluation and Synthesis C-201
C.12.2 Animal Evidence Study Quality Evaluation and Synthesis C-202
C.12.3 Mechanistic Evidence C-203
C.12.4 Evidence Integration C-203
Appendix D. Detailed Information from Epidemiology Studies D-l
D. 1 Developmental D-2
D.2 Reproductive D-52
D.2.1 Male D-52
D.2.2 Female D-63
D.3 Hepatic D-73
D.4 Immune D-82
D.5 Cardiovascular D-l 20
D.5.1 Cardiovascular Endpoints D-120
D.5.2 Serum Lipids D-137
D.6 Endocrine D-162
D.7 Metabolic/Systemic D-170
D.8 Nervous D-l89
D.9 Renal D-212
D.10 Hematological D-221
D.ll Respiratory D-224
v
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
D.12 Musculoskeletal D-226
D.13 Gastrointestinal D-230
D.14 Dental D-232
D.15 Ocular D-232
D.16 Dermal D-233
D.17 Cancer D-234
Appendix E. Benchmark Dose Modeling E-l
E.l Epidemiology Studies E-l
E.l.l Modeling Results for Immunotoxicity E-l
E.l.2 Modeling Results for Decreased Birthweight E-18
E. 1.3 Modeling Results for Increased Cholesterol E-25
E. 1.4 Modeling Results for Liver Toxicity E-37
E.2 Toxicology Studies E-47
E.2.1 Butenhoff et al. (2012, 1276144)/Thomford (2002, 5029075) E-47
11.2.2 Lee et al. (2015, 2851075) 11-63
11.2.3 Luebker et al. (2005, 757857) 11-67
11.2.4 NTP (2019, 5400978) 11-70
11.2.5 Zhong et al. (2016, 3748828) 11-74
Appendix F. Pharmacokinetic Modeling F-l
F.l Comparison of Fits to Training Datasets Used in Wambaugh et al. (2013,
2850932) F-l
F.2 Visual Inspection of Test Datasets not Used for Initial Fitting F-4
F.3 Human Model Validation F-6
Appendix G. Relative Source Contribution G-l
G. 1 B ackground G-l
G.2 Literature Review G-2
G. 2.1 Sy stemati c Revi e w G-2
G.2.2 Evidence Mapping G-3
G.3 Summary of Potential PFOS Sources G-3
G.3.1 Dietary Sources G-4
G.3.2 Consumer Product Uses G-9
G. 3.3 Indoor Dust G-10
G.3.4 Ambient Air G-ll
G.3.5 Other Possible Exposure Sources G-l2
G.4 Recommended RSC G-12
vi
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figures
Figure A-l. Overview of Study Quality Evaluation Approach A-77
Figure A-2. Possible Domain Scores for Study Quality Evaluation A-79
Figure A-3. Overall Study Confidence Classifications A-79
Figure C-l. Summary of Study Evaluation for Epidemiology Studies of PFOS and Male
Reproductive Effects C-3
Figure C-2. Summary of Study Evaluation for Epidemiology Studies of PFOS and Female
Reproductive Effects C-8
Figure C-3. Summary of Study Evaluation for Epidemiology Studies of PFOS and Female
Reproductive Effects (Continued) C-9
Figure C-4. Summary of Study Evaluation for Toxicology Studies of PFOS and
Reproductive Effects C-l5
Figure C-5. Gestation Length in Rats Following Exposure to PFOS C-16
Figure C-6. Sperm Parameters in Male Rodents Following Exposure to PFOS C-17
Figure C-l Percent Change in Testosterone Levels Relative to Controls in Male Rodents
and Non-Human Primates Following Exposure to PFOS C-18
Figure C-8. Percent Change in Estradiol Levels Relative to Controls in Male Rodent and
Non-Human Primates Following Exposure to PFOS C-19
Figure C-9. Percent Change in LH and Prolactin Levels Relative to Controls in Male Rats
Following Exposure to PFOS C-20
Figure C-10. Percent Change in Prolactin-Family Hormone Levels Relative to Controls in
Female Mice Following Exposure to PFOS C-21
Figure C-l 1. Percent Change in Estradiol and Testosterone Levels Relative to Controls in
Female Rodents and Non-Human Primates Following Exposure to PFOS C-22
Figure C-12. Summary of Mechanistic Studies of PFOS and Reproductive Effects C-25
Figure C-13. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Endocrine Effects C-47
Figure C-14. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Endocrine Effects (Continued) C-48
Figure C-15. Summary of Study Evaluation for Toxicology Studies of PFOS and Endocrine
Effects C-53
Figure C-16. Percent Change in Adrenal Hormones Relative to Controls in Rodents
Following Exposure to PFOSa'b C-68
Figure C-17. Summary of Mechanistic Studies of PFOS and Endocrine Effects C-72
vii
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Figure C-18. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Metabolic Effects C-81
Figure C-19. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Metabolic Effects (Continued) C-82
Figure C-20. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Metabolic Effects (Continued) C-83
Figure C-21. Summary of Study Evaluation for Toxicology Studies of PFOS and Metabolic
Effects C-96
Figure C-22. Summary of Study Evaluation for Toxicology Studies of PFOS and Systemic
Effectsa C-99
Figure C-23. Summary of Study Evaluation for Toxicology Studies of PFOS and Systemic
Effects (Continued)51 C-100
Figure C-24. Effects on Body Weight in Rodents and Non-Human Primates Following
Exposure to PFOS (logarithmic scale) C-103
Figure C-25. Summary of Mechanistic Studies of PFOS and Metabolic Effects C-104
Figure C-26. Summary of Mechanistic Studies of PFOS and Systemic Effects C-105
Figure C-27. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Neurological Effects C-l 15
Figure C-28. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Neurological Effects (Continued) C-l 16
Figure C-29. Summary of Study Evaluation for Toxicology Studies of PFOS and Nervous
Effects C-122
Figure C-30. Summary of Mechanistic Studies of PFOS and Nervous Effects C-130
Figure C-31. Summary of Study Evaluation for Epidemiology Studies of PFOS and Renal
Effects C-l 43
Figure C-32. Summary of Study Evaluation for Toxicology Studies of PFOS and Renal
Effects C-147
Figure C-33. Summary of Mechanistic Studies of PFOS and Renal Effects C-150
Figure C-34. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Hematological Effects C-158
Figure C-35. Summary of Study Evaluation for Toxicology Studies of PFOS and
Hematological Effects C-l60
Figure C-36. Summary of Mechanistic Studies of PFOS and Hematological Effects C-161
Figure C-37. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Respiratory Effects C-167
viii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure C-38. Summary of Study Evaluation for Toxicology Studies of PFOS and
Respiratory Effects C-169
Figure C-39. Summary of Mechanistic Studies of PFOS and Respiratory Effects C-172
Figure C-40. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Musculoskeletal Effects C-178
Figure C-41. Summary of Study Evaluation for Toxicology Studies of PFOS and
Musculoskeletal Effects C-180
Figure C-42. Summary of Mechanistic Studies of PFOS and Musculoskeletal Effects C-181
Figure C-43. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Gastrointestinal Effects C-187
Figure C-44. Summary of Study Evaluation for Toxicology Studies of PFOS and
Gastrointestinal Effects C-188
Figure C-45. Summary of Mechanistic Studies of PFOS and Gastrointestinal Effects C-189
Figure C-46. Summary of Study Evaluation for Epidemiology Studies of PFOS and Dental
Effects C-193
Figure C-47. Summary of Study Evaluation for Epidemiology Studies of PFOS and Ocular
Effects C-197
Figure C-48. Summary of Study Evaluation for Toxicology Studies of PFOS and Ocular
Effects
.C-198
Figure C-49. Summary of Mechanistic Studies of PFOS and Ocular Effects C-199
Figure C-50. Summary of Study Evaluation for Epidemiology Studies of PFOS and Dermal
Effects C-202
Figure C-51. Summary of Study Evaluation for Toxicology Studies of PFOS and Dermal
Effects C-203
Figure D-l. Overall ALT Levels from Epidemiology Studies Following Exposure to PFOS. D-73
Figure D-2. Overall Levels of Total Cholesterol in Adults from Epidemiology Studies
Following Exposure to PFOS D-l37
Figure E-l. Difference in population tail probabilities resulting from a one standard
deviation shift in the mean from a standard normal distribution, illustrating the
theoretical basis for a baseline BMR of 1 SD E-3
Figure E-2. Difference in population tail probabilities resulting from a '/2 standard deviation
shift in the mean from an estimation of the distribution of log2(tetanus antibody
concentrations at age seven years) E-5
Figure F-l. Experimentally Observed Serum Concentrations {Chang, 2012, 1289832} and
Median Prediction for a Single Oral Dose of 1 or 20 mg/kg PFOS to Female
CD1 Mice F-l
ix
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure F-2. Experimentally Observed Serum Concentrations {Chang, 2012, 1289832} and
Median Prediction for a Single Oral Dose of 1 or 20 mg/kg PFOS to Male CD1
Mice F-l
Figure F-3. Experimentally Observed Serum Concentrations {Chang, 2012, 1289832} and
Median Prediction for a Single IV Dose of 2 mg/kg or a Single Oral Dose of 2
or 15 mg/kg PFOS to Male Sprague-Dawley Rats F-2
Figure F-4. Model Prediction Summary for PFOS Training Data F-2
Figure F-5. PFOS Sensitivity Coefficients of the Adult Model and Developmental Model F-3
Figure F-6. mentally Observed Serum Concentrations {Huang, 2019, 7410147} and
Median Predictions for a Single IV Dose of 2 mg/kg or an Oral Dose of 2 or
20 mg/kg PFOS to Male Sprague-Dawley Rats F-4
Figure F-7. Experimentally Observed Serum Concentrations {Huang, 2019, 7410147} and
Median Predictions for a Single IV Dose of 2 mg/kg or an Oral Dose of 2 or 20
mg/kg PFOS to Female Sprague-Dawley Rats F-4
Figure F-8. Experimentally Observed Serum Concentrations {Kim, 2016, 3749289} and
Median Prediction for a Single IV Dose of 2 mg/kg or an Oral Dose of 2 mg/kg
PFOS to Male Sprague-Dawley Rats F-5
Figure F-9.Experimentally Observed Serum Concentrations {Kim, 2016, 3749289} and
Median Prediction for a Single IV Dose of 2 mg/kg an Oral Dose of 2 mg/kg
PFOS to Female Sprague-Dawley Rats F-5
Figure F-10. Model Prediction Summary for PFOS Test Data F-6
Figure F-l 1. Model Comparison F-7
Figure F-12. Predicted Child Serum Levels Compared to Reported Values F-8
Figure F-13. Comparison of Predicted and Observed Child Serum Levels F-8
Figure F-14. Sensitivity Coefficients F-10
Figure F-15. Predicted Child Serum Levels Compared to Reported Values with Increased
Volume of Distribution in Children as was Implemented in the Minnesota
Department of Health Model F-l 1
Figure F-16. Predicted Child Serum Levels Compared to Reported Values with Constant
Volume of Distribution and Variable Exposure Based on Drinking Water Intake. F-12
Figure G-l. Application of the Exposure Decision Tree {U.S. EPA, 2000, 19428} for
PFOS G-l 3
x
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Tables
Table A-1.Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for a
Systematic Review on the Health Effects from Exposure to PFOA and PFOS A-32
Table A-2. Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for
Absorption, Distribution, Metabolism, and/or Excretion (ADME) Studies A-33
Table A-3. Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for
Mechanistic Studies A-3 5
Table A-4. Search String for April 2019 Database Searches A-36
Table A-5. Search String for September 2020 and February 2022 Database Searches A-38
Table A-6. Key Epidemiological Studies of Priority Health Outcomes Identified from 2016
PFOS Health Effects Support Document A-41
Table A-7. Key Toxicological Animal Toxicological Studies Identified from 2016 PFOS
Health Effects Support Document A-44
Table A-8. Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for a
Systematic Review on the Health Effects from Exposure to PFOA and PFOS A-47
Table A-9. DistillerSR Form for Title/Abstract Screening A-49
Table A-10. SWIFT-Active Form for Title/Abstract Screening A-50
Table A-l 1. Supplemental Tags for Title/Abstract and Full-Text Screening A-50
Table A-12. Mechanistic Study Categories Considered as Supplemental A-51
Table A-13. DistillerSR Form for Full-Text Screening A-53
Table A-14. Health Effect Categories Considered for Epidemiological Studies A-56
Table A-15. litstream Form for ADME Screening and Light Data Extraction A-59
Table A-16. litstream Form for Mechanistic Screening and Light Data Extraction A-69
Table A-17. Study Quality Evaluation Considerations for Participant Selection A-81
Table A-18. Study Quality Evaluation Considerations for Exposure Measurement A-83
Table A-19. Criteria for Evaluating Exposure Measurement in Epidemiology Studies of
PFAS and Health Effects A-86
Table A-20. Study Quality Evaluation Considerations for Outcome Ascertainment A-88
Table A-21. Study Quality Evaluation Considerations for Confounding A-90
Table A-22. Study Quality Evaluation Considerations for Analysis A-93
Table A-23. Study Quality Evaluation Considerations for Selective Reporting A-95
xi
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table A-24. Study Quality Evaluation Considerations for Study Sensitivity A-96
Table A-25. Evaluation Considerations for Overall Study Confidence - Overall
Confidence, Epidemiological Studies A-97
Table A-26. Study Quality Evaluation Considerations for Reporting Quality A-99
Table A-27. Study Quality Evaluation Considerations for Selection and Performance -
Allocation A-102
Table A-28. Study Quality Evaluation Considerations for Selection and Performance -
Observational Bias/Blinding A-104
Table A-29. Study Quality Evaluation Considerations for Confounding/Variable Control... A-108
Table A-30. Study Quality Evaluation Considerations for Selective Reporting and Attrition
- Reporting and Attrition Bias A-l 10
Table A-31. Study Quality Evaluation Considerations for Exposure Methods Sensitivity -
Chemical Administration and Characterization A-l 12
Table A-32. Study Quality Evaluation Considerations for Exposure Methods Sensitivity -
Exposure Timing, Frequency, and Duration A-l 15
Table A-33. Study Quality Evaluation Considerations for Outcome Measures and Results
Display - Endpoint Sensitivity and Specificity A-l 17
Table A-34. Study Quality Evaluation Considerations for Outcome Measures and Results
Display - Results Presentation A-120
Table A-35. Evaluation Considerations for Overall Study Confidence - Overall
Confidence, Animal Toxicological Studies A-122
Table A-36. DistillerSR Form for Data Extraction of Epidemiological Studies A-125
Table A-37. Epidemiological Study Design Definitions A-132
Table A-38. HAWC Form Fields for Data Extraction of Animal Toxicological Studies A-134
Table A-39. Framework for Strength-of-Evidence Judgments for Epidemiological Studiesa A-140
Table A-40. Framework for Strength-of-Evidence Judgments for Animal Toxicological
Studiesa A-141
Table A-41. Evidence Integration Judgments for Characterizing Potential Human Health
Effects in the Evidence Integration51 A-142
Table A-42. Epidemiologic Meta-Analysis Studies Identified from Literature Review A-150
Table A-43. Studies Identified After Updated Literature Review (Published or Identified
After February 2022) A-154
Table B-l. Cellular Accumulation and Retention Relative to Lipophilicity and
Phospholipidicity as Reported by Sanchez-Garcia et al. (2018, 4234856) B-l
xii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-2. PFOS Parameters from Toxicokinetic Studies Informing Bioavailability in
Sprague-Dawley Rats B-3
Table B-3. Concentrations of PFOS in Various Tissues of Male Sprague-Dawley Rats
Exposed to PFOS by Gavage for 28 Days as Reported by Cui et al. (2009,
757868) B-10
Table B-4. Concentrations of PFOS in Various Tissues of Male and Female Sprague-
Dawley Rats Exposed to PFOS by Feed for 28 Days as Reported by Curran et
al. (2008, 757871) B-ll
Table B-5. Distribution of PFOS in Male Wistar Rats Exposed via Drinking Water for 1 or
3 Months as Reported by Iwabuchi et al. (2017, 3859701) B-12
Table B-6. PFOS Levels in the Serum and Liver of Male and Female Sprague-Dawley Rats
Exposed to PFOS in Feed for 2 Years as Reported by Thomford (2002,
5029075) B-12
Table B-7. PFOS concentrations in Human Cord Blood, Maternal Blood, and
Transplacental Transfer Ratios (RCM) B-20
Table B-8. Summary of PFOS Concentrations in Human Maternal Blood, Cord Blood,
Placenta and Amniotic Fluid Studies B-25
Table B-9. Summary of Human PFOS Concentrations in Maternal Serum, Breast Milk, and
Infant Serum B-31
Table B-10. Percent Change in PFOS Ratios in Human Maternal Serum and Breast Milk
and Breast Milk and Infant Serum by Infant Age as Reported by Mondal et al.
(2014, 2850916) B-33
Table B-ll. Percent Change in Human PFOS Serum Concentration by Exclusive, Mixed or
No Breastfeeding Per Month as Reported by Mogensen et al. (2015, 3859839)....B-33
Table B-12. Liver, Serum, Urine, and Feces PFOS Concentrations in Pregnant Sprague-
Dawley Dams and Fetuses {Luebker, 2005, 1276160} B-34
Table B-13. Serum, Liver, and Brain Tissue PFOS Concentrations of Sprague-Dawley
Dams and Offspring as Reported by Chang et al. (2009, 757876) B-36
Table B-14. Serum, Hippocampus, and Cortex PFOS Concentrations of Sprague-Dawley
Rat Pups as Reported by Zeng et al. (2011, 1326732) B-37
Table B-15. Serum and Lung PFOS Concentration of Sprague-Dawley Rat Pups {Chen,
2012, 1276152} B-3 8
Table B-16. Concentration Ratios of 35S-PFOS Maternal Serum to Various Organs of
C57BL/6 Mouse Dams, Fetuses, and Pups {Lai, 2017, 3981375} B-39
Table B-17. Percent Distribution of PFOS in Male and Female KM Mice After 50 mg/kg
Subcutaneous Injection {Liu, 2009, 757877} B-39
Table B-18. Summary of PFOS Volume of Distribution Values Assigned in Human Studies .B-41
xiii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-19. Summary of PFOS Volume of Distribution in Rats B-43
Table B-20. Pharmacokinetic Parameters After Acute PFOS Exposure in Cynomolgus
Monkeysa {Chang, 2017, 3981378} B-45
Table B-21. Enterohepatic Transporters of PFOS B-49
Table B-22. Estimated Percentage of the Sum of PFOS, PFNA, and PFOA in Excreta and
Serum of Male and Female Wistar Ratsa as Reported by Gao et al. (2015,
2851191) B-53
Table B-23. Summary of PFOS Concentration in Blood and Urine in Relation to Half-life
values in Humans B-58
Table B-24. Summary of Human PFOS Half-Life Values B-62
Table B-25. Summary of Animal PFOS Half-Life Values Identified in the Literature
Review B-66
Table C-l. Evidence Profile Table for PFOS Reproductive Effects in Males C-27
Table C-2. Evidence Profile Table for PFOS Reproductive Effects in Females C-37
Table C-3. Summary of Results for Thyroid and Thyroid-Related Hormones in
Toxicological Studies Following Exposure to PFOS C-57
Table C-4. Associations Between PFOS Exposure and Endocrine Organ Weights in
Rodents and Non-human Primates C-69
Table C-5. Evidence Profile Table for PFOS Endocrine Effects C-74
Table C-6. Evidence Profile Table for PFOS Systemic and Metabolic Effects C-107
Table C-l. Associations Between PFOS Exposure and Neurobehavioral Effects in Rodents C-125
Table C-8. Associations Between PFOS Exposure and Neurotransmitters in Rodents C-128
Table C-9. Evidence Profile Table for PFOS Nervous System Effect C-133
Table C-10. Evidence Profile Table for PFOS Renal Effects C-152
Table C-l 1. Evidence Profile Table for PFOS Hematological Effects C-163
Table C-12. Evidence Profile Table for PFOS Respiratory Effects C-174
Table C-13. Evidence Profile Table for PFOS Musculoskeletal Effects C-183
Table C-14. Evidence Profile Table for PFOS Gastrointestinal Effects C-190
Table C-15. Evidence Profile Table for PFOS Dental Effects C-195
Table C-16. Evidence profile table for PFOS Ocular effects C-200
Table C-17. Evidence Profile Table for PFOS Dermal Effects C-205
Table D-l. Associations Between PFOS Exposure and Developmental Effects in Recent
Epidemiological Studies D-3
xiv
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table D-2. Associations Between PFOS Exposure and Male Reproductive Effects in
Recent Epidemiologic Studies D-52
Table D-3. Associations between PFOS Exposure and Female Reproductive Effects in
Female Children and Adolescents D-63
Table D-4. Associations between PFOS Exposure and Female Reproductive Health Effects
in Pregnant Women D-67
Table D-5. Associations between PFOS Exposure and Female Reproductive Health Effects
in Non-Pregnant Adult Women D-71
Table D-6. Associations Between PFOS Exposure and Hepatic Effects in Epidemiologic
Studies D-74
Table D-7. Associations between PFOS Exposure and Vaccine Response in Recent
Epidemiological Studies D-82
Table D-8. Associations between PFOS Exposure and Infectious Disease in Recent
Epidemiological Studies D-95
Table D-9. Associations Between PFOS Exposure and Asthma in Recent Epidemiologic
Studies D-102
Table D-10. Associations Between PFOS Exposure and Allergies in Recent Epidemiologic
Studies D-110
Table D-l 1. Associations Between PFOS Exposure and Eczema in Recent Epidemiologic
Studies D-l 16
Table D-12. Associations Between PFOS Exposure and Autoimmune Health Effects in
Recent Epidemiologic Studies D-l 18
Table D-13. Associations Between PFOS Exposure and Cardiovascular Effects in Recent
Epidemiological Studies D-120
Table D-14. Associations Between PFOS Exposure and Serum Lipid Effects in Recent
Epidemiologic Studies D-138
Table D-15. Associations Between PFOS Exposure and Endocrine Effects in Recent
Epidemiologic Studies D-l62
Table D-16. Associations Between PFOS Exposure and Metabolic Effects in Recent
Epidemiologic Studies D-l70
Table D-17. Associations Between PFOS Exposure and Neurological Effects in Recent
Epidemiologic Studies D-l89
Table D-18. Associations Between PFOS Exposure and Renal Effects in the General
Population D-212
Table D-19. Associations Between PFOS Exposure and Hematological Effects in Recent
Epidemiologic Studies D-221
xv
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table D-20. Associations Between PFOS Exposure and Respiratory Effects in Recent
Epidemiologic Studies D-224
Table D-21. Associations Between PFOS Exposure and Musculoskeletal Effects in Recent
Epidemiologic Studies D-226
Table D-22. Associations Between PFOS Exposure and Gastrointestinal Effects in Recent
Epidemiologic Studies D-230
Table D-23. Associations Between PFOS Exposure and Dental Effects in Recent
Epidemiologic Studies D-232
Table D-24. Associations Between PFOS Exposure and Ocular Effects in Recent
Epidemiologic Studies D-232
Table D-25. Associations Between PFOS Exposure and Dermal Health Effects in Recent
Epidemiologic Studies D-233
Table D-26. Associations Between PFOS Exposure and Cancer in Recent Epidemiologic
Studies D-234
Table E-l. Results specific to the slope from the linear analyses of PFOS measured at age
five years and log2(tetanus antibody concentrations) measured at age
seven years from Table 1 in Budtz-j0rgensen and Grandjean (2018, 5083631)
in a single-PFAS model and in a multi-PFAS model E-2
Table E-2. BMDs and BMDLs for effect of PFOS at age five years on anti-tetanus antibody
concentrations at age seven years {Budtz-j0rgensen, 2018, 5083631} using a
BMR of '/2 SD change in log2(tetanus antibodies concentration) and a BMR of 1
SD change in log2(tetanus antibodies concentration) E-5
Table E-3. Results of the linear analyses of PFOS measured perinatally and tetanus
antibodies measured at age five years from Budtz-j0rgensen and Grandjean
(2018, 7276745) in a single-PFAS model and in a multi-PFAS model E-6
Table E-4. BMDs and BMDLs for effect of PFOS measured perinatally and anti-tetanus
antibody concentrations at age five years {Budtz-j0rgensen, 2018, 5083631} E-8
Table E-5. BMDs and BMDLs for effect of PFOS on anti-tetanus antibody concentrations
{Timmerman, 2021, 9416315} using a BMR of V2 SD change in logio (tetanus
antibodies concentration) and a BMR of 1 SD change in logio (tetanus
antibodies concentration) E-9
Table E-6. BMDLs for effect of PFOS on anti-tetanus antibody concentrations using a
BMR of1 - SD {Timmerman, 2021, 9416315} I>10
Table E-7. Results specific to the slope from the linear analyses of PFOS measured at age
five years and log2(diphtheria antibodies) measured at age seven years from
Table 1 in Budtz-j0rgensen and Grandjean (2018, 5083631) in a single-PFAS
model and in a multi-PFAS model E-10
XVI
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-8. BMDs and BMDLs for effect of PFOS at age five years on anti-diphtheria
antibody concentrations at age seven years {Budtz-j0rgensen, 2018, 5083631}
using a BMR of '/2 SD change in log2(diphtheria antibodies concentration) and a
BMR of 1 SD log2(diphtheria antibodies concentration) E-13
Table E-9. Results of the linear analyses of PFOS measured perinatally and diphtheria
antibodies measured at age five years from Budtz-j0rgensen and Grandjean
(2018, 7276745) in a single-PFAS model and in a multi-PFAS model E-14
Table E-10. BMDs and BMDLs for effect of PFOS measured perinatally and anti-
diphtheria antibody concentrations at age five years {Budtz-j0rgensen, 2018,
5083631} I >15
Table E-l 1. BMDs and BMDLs for effect of PFOS on anti-diphtheria antibody
concentrations {Timmerman, 2021, 9416315} using a BMR of V2 SD change in
logio (tetanus antibodies concentration) and a BMR of 1 SD change in logio
(tetanus antibodies concentration) E-17
Table E-12. BMDLs for effect of PFOS on anti-diphtheria antibody concentrations using a
BMR of1 - SD {Timmerman, 2021, 9416315} E-17
Table E-13. BMDs and BMDLs in ng/mL for effect of PFOS on decreased birth weight, by
using the exact percentage (8.27%) of live births falling below the public health
definition of low birth weight, or alternative study-specific tail E-24
Table E-14. BMDs and BMDLs for effect of PFOS on decreased birth weight by
background exposure, using the exact percentage of the population (8.27%) of
live births falling below the public health definition of low birth weight, or
alternative tail probability E-25
Table E-15. NHANES mean and standard deviation of TC (mg/dL) and mean PFOS
(ng/mL) E-27
Table E-16. BMDs and BMDLs for effect of PFOS on increased cholesterol in Dong et al.
(2019, 5080195) E-28
Table E-17. NHANES mean and standard deviation of ln(TC) (ln(mg/dL)) and mean
ln(PFOS) (ln(ng/ml.)) I>29
Table E-18. BMDs and BMDLs for effect of PFOS on increased cholesterol in Steenland et
al. (2009, 1291109) I>29
Table E-19. Regression Results for Serum Total Cholesterol by Deciles of serum PFOS
from Steenland et al. (2009, 1291109) E-30
Table E-20. Summary of Benchmark Dose Modeling Results for Increase Mean Serum
Total Cholesterol in Steenland et al. (2009, 1291109) E-31
Table E-21. Odds ratios for elevated serum TC by quartiles of serum PFOS from Steenland
et al. (2009, 1291109) 11-32
xvii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-22. Summary of Benchmark Dose Modeling Results for Elevated Total
Cholesterol in Steenland et al. (2009, 1291109) E-33
Table E-23. Adjusted Mean Differences in Serum Total Cholesterol by Quartiles of Serum
PFOS (ng/ml.) from Lin et al. (2019, 1291109) 11-35
Table E-24. Summary of Benchmark Dose Modeling Results for Increase Mean Serum
Total Cholesterol Lin et al. (2019, 5187597) E-36
Table E-25. BMDLs for effect of PFOS on serum total cholesterol using a BMR of 5% E-37
Table E-26. Odds Ratios for Elevated ALT by Decile of PFOS serum concentrations
(ng/mL) from Gallo et al. (2012, 1276142). The NOAEC is bolded E-38
Table E-27. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et
al. (2012, 1276142) Using the Unadjusted Mean PFOS Serum Concentration E-39
Table E-28. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et
al. (2012, 1276142) Using the Adjusted, No Intercept Mean PFOS Serum
Concentration E-40
Table E-29. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et
al. (2012, 1276142) Using the Unadjusted, Median PFOS Serum Concentration..E-41
Table E-30. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et
al. (2012, 1276142) Using the Adjusted, No Intercept Median PFOS Serum
C oncentrati on E-42
Table E-31. NHANES mean and standard deviation of ln(ALT) (In IU/L) and mean PFOS
(In ng/mL) E-43
Table E-32. Prevalence of elevated ALT E-44
Table E-33. BMD and BMDL for effect of PFOS (ng/mL) on increased ALT in Gallo et al.
(2012, 1276142) 11-45
Table E-34. BMD and BMDL for effect of PFOS (ng/mL) on increased ALT in Nian et al.
(2019, 5080307), for 5% and 10% Extra Risk 11-46
Table E-35. BMDLs for effect of PFOS on serum ALT using a BMR of 5% E-46
Table E-36. Dose-Response Modeling Data for Hepatocellular Adenomas in Male Rats
Following Exposure to PFOS {Butenhoff, 2012, 1276144/Thomford, 2002,
5029075} 11-47
Table E-37. Summary of Benchmark Dose Modeling Results for Data for Hepatocellular
Adenomas in Male Rats Following Exposure to PFOS {Butenhoff, 2012,
1276144/Thomford, 2002, 5029075} 11-48
Table E-38. Dose-Response Modeling Data for Incidence of Islet Cell Carcinomas in Male
Rats Following Exposure to PFOS {Butenhoff, 2012, 1276144/Thomford,
2002, 5029075} 11-50
xviii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-39. Summary of Benchmark Dose Modeling Results for Incidence of Islet Cell
Carcinomas in Male Rats Following Exposure to PFOS {Butenhoff, 2012,
1276144/Thorn lord. 2002, 5029075} E-50
Table E-40. Dose-Response Modeling Data for Combined Incidence of Islet Cell
Adenomas and Carcinomas in Male Rats Following Exposure to PFOS
{Butenhoff, 2012, 1276144/Thomford. 2002, 5029075} 11-52
Table E-41. Summary of Benchmark Dose Modeling Results for Combined Incidence of
Islet Cell Adenomas and Carcinomas in Male Rats Following Exposure to
PFOS {Butenhoff, 2012, 1276144/Thomford, 2002, 5029075} E-53
Table E-42. Dose-Response Modeling Data for Hepatocellular Adenomas in Female Rats
Following Exposure to PFOS {Butenhoff, 2012, 1276144/Thomford, 2002,
5029075} 11-55
Table E-43. Summary of Benchmark Dose Modeling Results for Data for Hepatocellular
Adenomas in Female Rats Following Exposure to PFOS {Butenhoff, 2012,
1276144/Thomford, 2002, 5029075} 11-55
Table E-44. Dose-Response Modeling Data for Hepatocellular Adenomas and Carcinomas
in Female Rats Following Exposure to PFOS {Butenhoff, 2012,
1276144/Thomford, 2002, 5029075} 11-57
Table E-45. Summary of Benchmark Dose Modeling Results for Data for Hepatocellular
Adenomas and Carcinomas in Female Rats Following Exposure to PFOS
{Butenhoff, 2012, 1276144/Thomford, 2002, 5029075} 11-58
Table E-46. Dose-Response Modeling Data for Individual Cell Necrosis in the Liver in
Female Sprague-Dawley Crl:CD(SD)IGS BR Rats Following Exposure to
PFOS {Butenhoff, 2012, 1276144/Thomford, 2002, 5029075} E-60
Table E-47. Summary of Benchmark Dose Modeling Results for Individual Cell Necrosis
in the Liver in Female Sprague-Dawley Crl:CD(SD)IGS BR Rats Following
Exposure to PFOS {Butenhoff, 2012, 1276144/Thomford, 2002, 5029075} E-60
Table E-48. Dose-Response Modeling Data for Individual Cell Necrosis in the Liver in
Male Sprague-Dawley Crl:CD(SD)IGS BR Rats Following Exposure to PFOS
{Butenhoff, 2012, 1276144/Thomford, 2002, 5029075} 11-61
Table E-49. Summary of Benchmark Dose Modeling Results for Individual Cell Necrosis
in the Liver in Male Sprague-Dawley Crl:CD(SD)IGS BR Rats Following
Exposure to PFOS {Butenhoff, 2012, 1276144/Thomford, 2002, 5029075} E-62
Table E-50. Dose-Response Modeling Data for Fetal Body Weight in Fi Male and Female
CD-I Mice Following Exposure to PFOS {Lee, 2015, 2851075} E-63
Table E-51. Summary of Benchmark Dose Modeling Results for Fetal Body Weight in Fi
Male and Female CD-I Mice Following Exposure to PFOS (constant variance)
{Lee, 2015, 2851075} 11-65
xix
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-52. Dose-Response Modeling Data for Number of Dead Fetuses in Po Female CD-
1 Mice Following Exposure to PFOS {Lee, 2015, 2851075} E-66
Table E-53. Summary of Benchmark Dose Modeling Results for Number of Dead Fetuses
for Cavg,dam,gest in Po Female CD-I Mice Following Exposure to PFOS
(nonconstant variance) {Lee, 2015, 2851075} E-66
Table E-54. Summary of Benchmark Dose Modeling Results for Number of Dead Fetuses
for Cmax,dam in Po Female CD-I Mice Following Exposure to PFOS
(nonconstant variance) {Lee, 2015, 2851075} E-67
Table E-55. Dose-Response Modeling Data for Pup Body Weight Relative to the Litter
(LD5) in Fi Male and Female Sprague-Dawley Rats Following Exposure to
PFOS {Luebker, 2005, 757857} E-68
Table E-56. Summary of Benchmark Dose Modeling Results for Pup Body Weight
Relative to the Litter (LD5) in Fi Male and Female Sprague-Dawley Rats
Following Exposure to PFOS (nonconstant variance) {Luebker, 2005, 757857} ..E-69
Table E-57. Dose-Response Modeling Data for Extramedullary Hematopoiesis in Male
Sprague-Dawley Rats Following Exposure to PFOS {NTP, 2019, 5400978} E-70
Table E-58. Summary of Benchmark Dose Modeling Results for Extramedullary
Hematopoiesis in Male Sprague-Dawley Rats Following Exposure to PFOS
{NTP, 2019, 5400978} E-71
Table E-59. Dose-Response Modeling Data for Extramedullary Hematopoiesis in the
Spleen in Female Sprague-Dawley Rats Following Exposure to PFOS {NTP,
2019, 5400978} E-72
Table E-60. Summary of Benchmark Dose Modeling Results for Extramedullary
Hematopoiesis in the Spleen in Female Sprague-Dawley Rats Following
Exposure to PFOS {NTP, 2019, 5400978} 11-73
Table E-61. Dose-Response Modeling Data for PFC Response of Splenic Cells in Fi male
C57BL/6 mice Following Exposure to PFOS {Zhong, 2016, 3748828} E-74
Table E-62. Summary of Benchmark Dose Modeling Results for Plaque Forming Cell
Response of Splenic Cells in Fi Male C57BL/6 Mice Following Exposure to
PFOS (constant variance) {Zhong, 2016, 3748828} E-75
Table G-l. Summary of EPA national fish tissue monitoring results for PFOS G-7
xx
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Acronyms and Abbreviations
17-OHP
17-hydroxyprogesterone
BALF
bronchoalveolar lavage
ABC
ATP-binding cassette
fluid
transporter
BBB
blood-brain barrier
aBMD
areal bone mineral
BCERP
Breast Cancer and
density
Environment Research
ACD
anterior chamber depth
Program
ACE
America's Children and
BCRP
breast cancer resistance
the Environment
protein
ACTH
adrenocorticotropic
BD
bolus dose
hormone
BDI
Beck Depression
ADHD
attention deficit
Inventory
hyperactivity disorder
BDI-II
Beck Depression
ADME
absorption, distribution,
Inventory-II
metabolism, and
BMC
bone mineral content
excretion
BMD
benchmark dose
AF:CB
amniotic fluid and cord
BMDL
lower limit of benchmark
blood ratio
dose
AFFF
aqueous film forming
BMDLo.5sd
lower bound on the dose
foam
level corresponding to the
AGD
anogenital distance
95% lower confidence
AIC
Akaike information
limit for a change in the
criterion
mean equal to 0.5
ALSPAC
Avon Longitudinal Study
standard deviation from
of Parents and Children
the control mean
ALT
alanine aminotransferase
BMDLisd
lower bound on the dose
AMH
anti-Mullerian hormone
level corresponding to the
95% lower confidence
ASBT
apical sodium-dependent
limit for a change in the
bile acid transporter
mean equal to one
ASD
autism spectrum disorder
standard deviation from
ASQ
Ages and Stages
the control mean
Questionnaire
BMDLs
lower bound on the dose
ATP III
Adult Treatment Panel III
level corresponding to the
AT SDR
Agency for Toxic
95% lower confidence
Substances and Disease
limit for a 5% response
Registry
level
AUC
area under the curve
BMDLio
lower bound on the dose
AUMC
area under the first
level corresponding to the
moment curve
95% lower confidence
AZI
azithromycin-dihydrate
limit for a 10% change
P
regression coefficients
BMDS
Benchmark Dose
Software
xxi
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
BMI
body mass index
BMR
benchmark response
BRIEF
Behavior Rating
Inventory of Executive
Function
BUN
blood urea nitrogen
BW
body weight
C7,avg
average concentration
over final week of study
CalEPA
California Environmental
Protection Agency
Cavg,pup,gest
area under the curve
normalized per day
during gestation
Cavg,pup,gest,lact
area under the curve
normalized dose per day
during gestation/lactation
Cavg,pup,lact
area under the curve
normalized per day
during lactation
Cavg-pup-diet
average concentration
during the post-weaning
phase
CDI
Comprehensive
Developmental Inventory
C-F
carbon-fluorine
CHCA
a-Cyano-4-
hydroxycinnamic acid
CHECK
Children's Health and
Environmental Chemicals
in Korea
CHEF
Children's Health and the
Environment in the
Faroes
CHO
Chinese hamster ovary
CI
confidence interval
CKD
chronic kidney disease
Cmax
maximum blood
concentration
Cmax,dam
maximum maternal
concentration during
gestation
Cmax,pup,gest
maximum fetal
concentration during
gestation
Cmax,pup,lact
maximum pup
concentration during
lactation
CNS
central nervous system
CRH
corticotropin releasing
hormone
CSF
cancer slope factor
CSM
cholestyramine
CTX
type I collagen
CVD
cardiovascular disease
DFI
deoxyribonucleic acid
fragmentation index
DHEA
dehydroepiandrosterone
DHEAS
dehydroepiandrosterone
sulfate
DNA
deoxyribonucleic acid
DNBC
Danish National Birth
Cohort
DPP
Diabetes Prevention
Program
dU
diurnal urinary
E2
estradiol
EFSA
European Food Safety
Authority
GLP
good laboratory practice
eGFR
estimated glomerular
filtration rate
eNT
equilibrative nucleoside
transporter
EPA
U.S. Environmental
Protection Agency
EYHS
European Youth Heart
Study
Fi
first generation
FDA
U.S. Food and Drug
Administration
FEV1
forced expiratory volume
in one second
FR
folate receptor
xxii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
FSH
follicle stimulating
IQR
hormone
IRIS
FT3
free triiodothyronine
FT4
free thyroxine
IUFD
FTI
free thyroxine index
IV
FVC
forced vital capacity
Kd
GD
gestation day
Kmem/w
GI
gastrointestinal
GM
geometric mean
KpS
Hb
hemoglobin
LC50
HDL
high-density-lipoprotein
HED
human equivalent dose
LD
HEK293
human embryonic kidney
cells
LDL
HERO
Health and
Environmental Research
L-FABP
Online
LH
HESD
health effects support
document
LIFE
HHRA
human health risk
assessment
HOMA-B
Homeostatic Model
Assessment of Beta-Cell
LINC
Function
LLOQ
HOMA-IR
Homeostatic Model
Assessment for Insulin
LOAEL
Resistance
HOME
Health Outcome
LOD
Measures of the
LOQ
Environment
MALDI
HUMIS
Norwegian Human Milk
Study
MCDI
IBD
inflammatory bowel
disease
IC50
median inhibiting
concentration
MCLG
ID
intellectual disability
IMS
imaging mass
spectrometry
MDH
INUENDO
Biopersistent
Organochlorines in Diet
MDI
and Human Fertility
MDL
IQ
intelligence quotient
MDR1
interquartile range
Integrated Risk
Information System
intrauterine fetal death
intravenous
disassociation constant
membrane/water partition
coefficients
tissue-to-plasma partition
coefficients
median lethal
concentration
lactation day
low-density lipoprotein
liver fatty acid binding
protein
luteinizing hormone
Longitudinal
Investigation of Fertility
and the Environment
Study
Linking Maternal
Nutrition to Child Health
lower limit of
quantification
lowest-observed-adverse-
effect level
limit of detection
limit of quantification
matrix-assisted laser
desorption/ionization
MacArthur
Communicative
Development Inventories
for Infants
Maximum Contaminant
Level Goal
Minnesota Department of
Health
Mental Development
Index
minimum detection limit
p-glycoprotein
xxiii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
MeSH
medical subject headings
Mg/kg-day
milligrams per kilogram
per day
MIREC
Maternal-Infant Research
on Environmental
Chemicals
MLR
mixed linear regression
MP AH
N-methyl-PFOSA
mPL-II
mouse placental lactogen
mPLP
prolactin-like protein
MRL
minimum reporting level
mRNA
messenger ribonucleic
acid
MRP
multi-drug resistance-
associated protein
MOA
mode of action
MoBA
Norwegian Mother,
Father, and Child Cohort
Study
MP AH
2-(N-methyl-PFOSA)
acetate
NHANES
National Health and
Examination Survey
NICHD
U.S. National Institute of
Child Health and Human
Development
NJDEP
New Jersey Department
of Environmental
Protection
NO A A
National Oceanic and
Atmospheric
Administration
NOAEL
no-ob served-adverse-
effect level
NOAEC
no observed adverse
effect concentration
NPDWR
national primary drinking
water regulation
NTCP
sodium-taurocholate
cotransporting
polypeptide
NTP
National Toxicology
Program
OATs
organic anion transporters
OATPs
organic anion
transporting polypeptides
OCC
Odense Child Cohort
OCT
organic cation/carnitine
transporter
OECD
Organisation for
Economic Co-operation
and Development
OR
Odds Ratio
ORD
Office of Research and
Development
OVA
ovalbumin
Po
parental generation
PBPK
physiologically-based
pharmacokinetic
Pc
partition coefficient
PC
phosphatidylcholine
PCOS
polycystic ovary
syndrome
PDI
Psychomotor
Development Index
PECO
Populations, Exposures,
Comparator, and
Outcomes
PEF
peak expiratory flow rate
PFAA
perfluorinated alkyl acid
PFAS
per- and polyfluoroalkyl
substances
PFBA
perfluorobutanoic acid
PFBS
perfluorobutane sulfonate
PFC
plaque forming cell
PFCA
perfluorocarboxylates
PFDA
perfluorodecanoic acid
PFHxA
perfluorohexanoic acid
PFHxS
perfluorohexane sulfonate
PFOA
perfluorooctanoic acid
PFOS
perfluorooctane sulfonic
acid
PFNA
perfluorononanoic acid
PFSA
perfluoroalkanesulfonic
acid
xxiv
-------
DRAFT FOR PUBLIC COMMENT
PHQ-9
Patient Health
Questionnaire
Pion
passive anionic
permeability
PFUnDA
perfluoroundecanoic acid
PK
pharmacokinetic
PND
postnatal day
PNW
postnatal week
POD
point-of-departure
PODhed
point-of-departure human
equivalent dose
POI
premature ovarian
insufficiency
POPUP
Persistent Organic
Pollutants in Uppsala
Primiparas
PPARa
proliferator-activated
receptor alpha
Qi
quantile 1
Q2
quantile 2
Q3
quantile 3
Q4
quantile 4
QA
quality assurance
QUICKI
Quantitative Insulin
Sensitivity Check Index
RBC
red blood cell
RCM
ratio of cord blood to
maternal blood
concentrations
RFC
reduce folate carrier
RfD
reference dose
RIS
Research Information
System
ROBINS-I
Risk of Bias in
Nonrandomized Studies
of Interventions
Rpm
ratio of
placental :maternal
concentrations
RSC
relative source
contribution
rT3
reverse triiodothyronine
SAB
Science Advisory Board
MARCH 2023
SE
standard errors
SERT
serotonin transporter
SES
socioeconomic status
SD
standard deviation
SDQ
Strengths and Difficulties
Questionnaire
SDWA
Safe Drinking Water Act
SHBG
sex hormone binding
globulin
SMBCS
Shanghai Minhang Birth
Cohort Study
SWAN
Study of Women's Health
Across the Nation
T3
triiodothyronine
14
thyroxine
TA
thyroid antibody
TC
total cholesterol
TDS
Total Diet Study
TgAB
thyroblobulin antibodies
TiAb
title-abstract
Tmax
maximum plasma
concentration
TPO
anti-thyroid peroxidase
TPoAb
thyroid peroxidase
antibody
TSH
thyroid stimulating
hormone
TT3
total triiodothyronine
TTE
transplacental transfer
efficiencies
UCMR3
third Unregulated
Contaminant Monitoring
Rule
Vi
volume of central
distribution
v2
volume of peripheral
distribution
Vd
volume of distribution
VI
visual impairment
VLDL
very low-density
lipoproteins
XXV
-------
DRAFT FOR PUBLIC COMMENT
VMWM Virtual Morris Water
Maze
WBHGB whole blood hemoglobin
WHO World Health
Organization
ww wet weight
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix A. Systematic Review Protocol for
Updated PFOS Toxicity Assessment
Per- and polyfluoroalkyl substances (PFAS) refers to a large group of fluorinated anthropogenic
chemicals that includes perfluorooctanoic acid (PFOA), perfluorooctane sulfonic acid (PFOS),
and thousands of other chemicals. The universe of environmentally relevant PFAS, including
parent chemicals, metabolites, and degradants, is greater than 12,000 compounds
(https://comptox.epa.eov/dashboard/chemical4ists/PFASMASTER). The Organisation for
Economic Co-operation and Development (OECD) New Comprehensive Global Database of
Per- and Polyfluoroalkyl Substances (PFASs) includes over 4,700 PFAS {OECD, 2018,
5099062}. The number of PFAS used globally in commercial products at the time of the drafting
of this document is approximately 250 substances {Buck, 2021, 9640864}.
PFAS have been manufactured and used in a wide variety of industries around the world,
including in the United States since the 1950s. PFAS have strong, stable, carbon-fluorine (C-F)
bonds, making them resistant to hydrolysis, photolysis, microbial degradation, and metabolism
{Ahrens, 2011, 2657780; Beach, 2006, 1290843; Buck, 2011, 4771046}. There are many
families or classes of PFAS, each containing many individual structural homologues that can
exist as either branched-chain or straight-chain isomers {Buck, 2011, 4771046}. The chemical
structures of PFAS enable them to repel water and oil, remain chemically and thermally stable,
and exhibit surfactant properties; these properties make PFAS useful for commercial and
industrial applications and make some PFAS extremely persistent in the human body and the
environment {Calafat, 2007, 1290899; Calafat, 2019, 5381304}. Due to their widespread use,
physicochemical properties, persistence, and bioaccumulation potential, many different PFAS
co-occur in environmental media (e.g., air, water, ice, sediment) and in tissues and blood of
aquatic and terrestrial organisms, including humans.
To understand and address the complexities associated with PFAS, the U.S. Environmental
Protection Agency (EPA) is developing human health toxicity assessments for individual PFAS,
in addition to other components of the broader PFAS action plan underway at EPA
(https://www.epa.eov/pfas/epas-pfas-action-plan). The updated toxicity assessment that was
developed for PFOS according to the scope and methods outlined in this protocol builds upon
several other assessments, including the Health Effects Support Document for PFOS {U.S.EPA,
2016, 3603365} (hereafter referred to as the PFOS HESD) and Proposed Approaches to the
Derivation of a Draft Maximum Contaminant Level Goal for Perfluorooctane Sulfonic Acid
(PFOS) (CASRN335-67-1) in Drinking Water {U.S. EPA, 2021, 10428576}, which was released
to the public for review by the Science Advisory Board (SAB) in November 2021.
This protocol describes the methods used for conducting the systematic review and dose-
response analyses for the assessment of PFOS (Draft Toxicity Assessment and Proposed
Maximum Contaminant Level Goal (MCLG) for PFOS) and has been updated to address
comments from the SAB. It should be noted that PFOA and PFOS underwent some steps of
systematic review (e.g., literature searches) concurrently.
A-27
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l Overview of Background Information and Systematic
Review Protocol
The methods used to conduct the systematic review for PFOS are consistent with the methods
described in the draft and final EPA ORD Staff Handbook for Developing IRIS Assessments
{U.S. EPA, 2020, 7006986; U.S. EPA, 2022, 10367891} (hereafter referred to as the Integrated
Risk Information System (IRIS) Handbook) and a companion publication {Thayer, 2022,
10259560}. EPA's IRIS Handbook has incorporated feedback from the National Academy of
Sciences (NAS) at workshops held in 2018 and 2019 and was well regarded by the NAS review
panel for reflecting "significant improvements made by EPA to the IRIS assessment process,
including systematic review methods for identifying chemical hazards" {NAS, 2021, 9959764}.
Furthermore, EPA's IRIS program has used the IRIS Handbook to develop toxicological reviews
for numerous chemicals, including some PFAS. Though the IRIS Handbook was finalized
concurrently with this assessment, the alterations in the final IRIS Handbook compared to the
draft version did not conflict with the methods used in this assessment. In fact, many of the NAS
recommendations incorporated into the final IRIS handbook (e.g., updated methods for evidence
synthesis and integration) were similarly incorporated into this assessment protocol {NAS, 2021,
9959764}. However, some of the study evaluation refinements recommended by NAS {2021,
9959764}, including clarifications to the procedure for evaluating studies for sensitivity and
standardizing the procedure for evaluating reporting quality between human and animal studies,
were not included in this assessment protocol, consistent with a 2011 NASEM recommendation
not to delay releasing assessments until systematic review methods are finalized {NRC, 2011,
710724}. The assessment team concluded that implementing these minor changes in study
quality evaluation would not change the assessment conclusions. Therefore, EPA considers the
methods described herein to be consistent with the final IRIS Handbook and cites this version
accordingly.
The Safe Drinking Water Act (SDWA) regulatory process enables EPA to receive comments and
feedback on the systematic review protocol, including the SAB early input and via the public
comment period associated with rule proposal. This protocol has been updated based on SAB
recommendations to improve the clarity and transparency of the methods descriptions. It now
includes information about additional data sources and how they were evaluated and expands the
application of systematic review through dose-response analysis.
A.l.l Summary of Background Information
This section summarizes more detailed sections on these topics from Proposed Maximum
Contaminant Level Goal (MCLG) for PFOS (hereafter referred to as the PFOS MCLG main
document) and is provided for context. Please refer to the PFOS MCLG main document for more
detailed information about chemical identity, physical-chemical properties, and occurrence.
A.l.l.1 Chemical Identity
PFOS is a PFAA that was used as an aqueous dispersion agent and emulsifier in a variety of
water-, oil-, and stain-repellent products (e.g., agricultural chemicals, alkaline cleaners, carpets,
firefighting foam, floor polish, textiles) {NLM, 2022, 10369707}. It can exist in linear- or
branched-chain isomeric form. PFOS is a strong acid that is generally present as the sulfonate
A-28
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
anion at typical environmental pH values. Therefore, this assessment applies to all isomers of
PFOS, as well as nonmetal salts of PFOS that would be expected to dissociate in aqueous
solutions of pH ranging from 4 to 9 (e.g., in the human body). PFOS is stable in environmental
media because it is resistant to environmental degradation processes such as biodegradation,
photolysis, and hydrolysis. In water, no natural degradation has been demonstrated, and
dissipation is by advection, dispersion, and sorption to particulate matter.
A.1.1.2 Occurrence Summary
Key PFOS occurrence information is summarized below. More detail is provided in Chapter 1 of
the PFOS MCLG main document.
A. 1.1.2.1 Biomonitoring
The CDC NHANES has measured blood serum concentrations of several PFAS in the general
U.S. population since 1999. PFOS has been detected in up to 98% of serum samples taken in
biomonitoring studies that are representative of the U.S. general population; however, blood
levels have dropped 60% to 80% between 1999 and 2014, presumably due to restrictions on its
commercial usage in the United States.
A.l.1.2.2 Occurrence in Water
PFOS is one of the dominant PFAS compounds detected in ambient water, along with PFOA
{Ahrens, 2011, 2657780; Benskin, 2012, 1274133; Dinglasan-Panlilio, 2014, 2545254;
Nakayama, 2007, 2901973; Remucal, 2019, 5413103; Zareitalabad, 2013, 5080561}.
Data from the third Unregulated Contaminant Monitoring Rule (UCMR 3), collected from 2013-
2015, are currently the best available national occurrence data for PFOA and PFOS {U.S. EPA,
2017, 9419085; U.S. EPA, 2021, 7487276; U.S. EPA, 2023, 10692764}. Under UCMR 3,
36,972 samples from 4,920 PWSs were analyzed for PFOA and PFOS. The minimum reporting
level (MRL)1 for PFOA was 0.02 |ig/L and the MRL for PFOS was 0.04 |ig/L. A total of 1.37%
of samples had reported detections (> MRL) of at least one of the two compounds.
A. 1.2 Problem Formulation
As described in Chapter 1 of the PFOS MCLG main document, EPA conducted this updated
assessment of PFOS to support development of an MCLG and national primary drinking water
regulation (NPDWR). This problem formulation section will describe the key considerations and
scope of the assessment, which were informed in part by EPA's past human health assessments
of PFOS (2016 HESD and 2021 Proposed Approaches to the Derivation of a Draft Maximum
Contaminant Level Goal for Perfluorooctane Sulfonic Acid (PFOS) (CASRN 335-67-1) in
Drinking Water) as well ongoing EPA assessments of other PFAS (e.g., perfluorobutanoic acid
(PFBA) and draft perfluorohexanoic acid (PFHxA), perfluorohexane sulfonate (PFHxS),
perfluorononanoic acid (PFNA), and perfluorodecanoic acid (PFDA) IRIS assessments).
1 The reporting level is the threshold at or above which a contaminant's presence or concentration is officially quantitated. In the
case of many of EPA's nation-wide drinking water studies, the selected reporting level is known officially as the MRL. The MRL
for each contaminant in each study is set at a level that EPA believes can be achieved with specified confidence by a broad
spectrum of capable laboratories across the nation {U.S. EPA, 2021, 9640861}.
A-29
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The 2016 PFOS HESD identified several adverse health outcomes associated with PFOS
exposure based on results from animal toxicological and epidemiological studies, including:
developmental effects (e.g., low birth weight, accelerated puberty, skeletal variations), cancer
(e.g., testicular, kidney), liver effects (e.g., tissue damage), immune effects (e.g., antibody
production and immunity), thyroid effects (e.g., hypothyroidism), and other effects (e.g.,
cholesterol changes). It concluded that there is "suggestive evidence of carcinogenic potential"
for PFOS. EPA's 2021 Proposed Approaches to the Derivation of a Draft Maximum
Contaminant Level Goal for Perfluorooctane Sulfonic Acid (PFOS) (CASRN 335-67-1) in
Drinking Water) {U.S. EPA, 2021, 10428576} evaluated PFOS in relation to all health
outcomes. The SAB recommended that the scope be narrowed to focus on the five main health
outcomes that have the strongest weight of evidence (immune, developmental, hepatic,
cardiovascular, and cancer), most of which were also identified in the conclusions from the 2016
HESD for PFOS. Therefore, the current assessment provides a comprehensive systematic review
of all health effects literature published through February 2022 for these five health outcomes.
Mechanistic data for these health outcomes were also synthesized. For other health outcomes
beyond the five primary ones, the current assessment summarizes the health effects literature
published prior to 2016 and includes a systematic review of the health effects literature published
from 2016-2020.
The Systematic Review Protocol for the PFBA, PFHxA, PFHxS, PFNA, and PFDA (anionic and
acidforms) IRIS Assessments outlines key science issues related to PFAS in general {U.S. EPA,
2020, 8642427}, many of which are relevant to PFOS. They include: toxicokinetic differences
across species and sexes; human relevance of effects in animals that involve peroxisome
proliferator-activated receptor alpha (PPARa); potential confounding by other PFAS exposures
in epidemiology studies; and toxicological relevance of changes in certain hepatic endpoints in
rodents. Differences in PFOS toxicokinetics across species and sexes were accounted for in the
PFOS-specific animal and human toxicokinetic (see PFOS MCLG main document). The human
relevance of effects in animals that involve PPARa was investigated in the mechanistic syntheses
of the five main health outcomes (see PFOS MCLG main document). Potential confounding by
other PFAS (and other co-occurring contaminants) in epidemiology studies was considered as
part of the confounding domain during study quality evaluations. Specifically, if a study did not
account for potential confounding with other co-occurring PFAS in its statistical analyses, then
the maximum quality rating this domain could receive was adequate. Concerns about potential
confounding by other PFAS were limited when there was evidence that exposure was
predominantly PFOS-based (such as in certain occupational or high-exposure studies) and the
potential for co-exposure was minimal, or the correlations between co-exposures were small. The
toxicological relevance of changes in certain hepatic endpoints in rodents was accounted for by
incorporating the Hall (2012, 2718645) criteria into the animal hepatic synthesis and hazard
conclusions.
An additional key science issue that EPA has encountered for PFAS toxicity assessments is a
general lack of data on human and ecological toxicity. For PFOS, this is less of an issue as there
has been substantial research and publication of both epidemiological and animal toxicological
studies.
A-30
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.3 Overall Objective and Specific Aims
A. 1.3.1 Objective
The primary objective of this draft for public comment is to derive an MCLG for PFOS to
support the NPDWR for PFAS. To derive an MCLG, a cancer classification, toxicity values (i.e.,
a reference dose (RfD) and cancer slope factor (CSF)), and relative source contribution (RSC)
for PFOS are potentially needed. The toxicity values, cancer classification, and RSC derived in
this assessment build upon the work completed in the Proposed Approaches to the Derivation of
a Draft Maximum Contaminant Level Goal for Perfluorooctane Sulfonic Acid (PFOS) (CASRN
1763-23-1) in Drinking Water {U.S. EPA, 2021, 10428576} and in the 2016 PFOS HESD {U.S.
EPA, 2016, 3603365} and Health Advisory {U.S. EPA, 2016, 3982043}.
A.1.3.2 Specific Aims
The specific aims of the PFOS MCLG document, which support the overall objective of deriving
an MCLG for PFOS, are to:
• Provide a description of the literature searches conducted and systematic review methods
used to identify health effects information (epidemiological, animal toxicological studies,
and physiologically-based pharmacokinetic (PBPK) models) published since the 2016
PFOS HESD.
• Describe literature screening methods, including use of the Populations, Exposures,
Comparator, and Outcomes (PECO) criteria and procedures for tracking studies
throughout the literature screening process.
• Identify epidemiological and animal toxicological literature reporting effects of exposure
to PFOS (and its associated salts and isomers) as outlined in the PECO criteria.
• Evaluate and document the available mechanistic information (including toxicokinetic
understanding) associated with PFOS exposure to inform interpretation of findings related
to potential health effects in studies of humans and animals for the five main health
outcomes (developmental, hepatic, immune and cardiovascular effects, and cancer).
• Describe and document study quality evaluations conducted for epidemiological and
animal toxicological studies considered potentially useful for point-of-departure (POD)
derivation.
• Describe and document data from high and medium confidence epidemiological and
animal toxicological studies (as determined by study quality evaluations) that could be
used for POD derivation. For dose-response assessment, only high and medium
confidence studies were used to quantify health effects.
• Synthesize and document the adverse health effects evidence reported across studies,
assessing similar health outcomes using a narrative approach. (The assessment focuses on
synthesizing the available evidence for five main health outcomes—developmental,
hepatic, immune and cardiovascular effects, and cancer—and also provides secondary
syntheses of evidence for dermal, endocrine, gastrointestinal, hematologic, metabolic,
musculoskeletal, nervous, ocular, renal, and respiratory effects; reproductive effects in
males or females; and general systemic toxicity.
• Develop and document strength-of-evidence judgments across studies (or subsets of
studies) separately for epidemiological and animal toxicological lines of evidence for the
five main health outcomes and integrate mechanistic analyses into the judgments.
A-31
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• Develop and document integrated expert judgments across lines of evidence (i.e.,
epidemiological and animal toxicological lines of evidence) as to whether and to what
extent the evidence supports that exposure to PFOS has the potential to be hazardous to
humans. The judgments will be directly informed by the evidence syntheses and based on
structured review of an adapted set of considerations for causality first introduced by
Austin Bradford Hill {Hill, 1965, 71664}.
• Describe and document the dose-response analyses conducted on the studies identified for
POD derivation.
• Derive candidate RfDs and/or CSFs and select the RfD and/or CSF for PFOS and describe
the rationale.
• Determine PFOS's cancer classification using a weight-of-evidence approach.
• Characterize the effects associated with PFOS exposure, including uncertainties and data
gaps.
A.1.4 Populations, Exposures, Comparators, and Outcomes
(PECO) Criteria
This section describes the PECO criteria that were developed and used for this assessment.2 As
described in the IRIS Handbook {U.S. EPA, 2022, 10476098}, the PECO criteria provide the
framework for literature search strategies and are the inclusion/exclusion criteria by which
literature search results will be screened for relevancy to identify epidemiological and animal
toxicological evidence that addresses the aims of the assessment. For the PFOS assessment, the
PECO criteria were used to screen results of the literature searches to identify and prioritize the
dose-response literature and studies containing pharmacokinetic (PK) or PBPK models. For
studies captured in the 2019 and 2020 literature searches, the PECO criteria were used to screen
and categorize ("tag") studies of PFOS absorption, distribution, metabolism, and excretion
(ADME) and studies with mechanistic data for further evaluation using ADME- and
mechanistic-specific PECO criteria. ADME, mechanistic, and other supplemental studies
captured in the 2022 literature search were not tagged or considered further in this assessment.
Table A-l describes the PECO criteria used to screen the results of the literature search (the
literature search is described in Section A. 1.5 of this appendix). ADME- and mechanistic-
specific PECO criteria are outlined in Table A-2 and Table A-3, respectively.
Table A-l.Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for a
Systematic Review on the Health Effects from Exposure to PFOA and PFOS
PECO
Element
Inclusion Criteria
Population Human: Any population and life stage (occupational or general population, including children and
other sensitive populations).
Animal: Nonhuman mammalian animal species (whole organism) of any life stage (including
preconception, inutero, lactation, peripubertal, and adult stages).
In vitro!cell studies or in silico/modeling toxicity studies should be tagged as supplemental.
Exposure Any exposure to PFOA, PFOS, and/or the salts of PFOA/PFOS, including but not limited to:
PFOA {Chemical Abstracts Service (CAS) number 335-67-1).
2 Note: Although this appendix and its accompanying main document pertain to PFOS, the PECO criteria also cover PFOA
because the literature searching and screening were performed concurrently for PFOA and PFOS.
A-3 2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PECO
Element
Inclusion Criteria
Other names: perfluorooctanoate; perfluorooctanoic acid; 2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-
pentadecafluorooctanoic acid; pentadecafluoro-l-octanoic acid; pentadecafluoro-n-octanoic acid;
perfluorocaprylic acid; pentadecafluorooctanoic acid; perfluoroheptanecarboxylic acid; octanoic-
acid, pentadecafluoro-
Relevant Salts of PFOA: ammonium perfluorooctanoate (APFO), sodium perfluorooctanoate,
potassium perfluorooctanoate
PFOS {CAS number 1763-23-1).
Other names: perfluorooctane sulfonate, perfluorooctanesulfonic acid, perfluorooctane sulfonic
acid, perfluorooctane sulphonate, perfluorooctanyl sulfonate, heptadecafluorooctane-l-sulphonic,
heptadecafluoro-l-octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-l-
octanesulfonic acid
Relevant Salts of PFOS: lithium perfluorooctanesulfonate, potassium perfluorooctanesulfonate
(K+PFOS), ammonium perfluorooctanesulfonate, sodium perfluorooctanesulfonate
Human: Any exposure to PFOA or PFOS via oral routes. Other exposure routes, including
inhalation, dermal, or unknown/multiple routes will be tracked during title and abstract screening
and tagged as "potentially relevant supplemental information."
Animal: Any exposure to PFOA or PFOS via oral routes. Other exposure routes, including
inhalation, dermal, injection or unknown/multiple routes, will be tracked during title and abstract
screening and tagged as "potentially relevant supplemental information." Studies involving
exposures to mixtures will be included only if they include exposure to PFOA or PFOS alone.
Studies with less than 28 days of dosing, with the exception of reproductive, developmental,
immune and neurological health outcome studies, should be tagged as supplemental.
Human: A comparison or referent population exposed to lower levels (or no exposure/exposure
below detection limits) of PFOA or PFOS, or exposure to PFOA or PFOS for shorter periods of
time. Case reports and case series will be tracked as "potentially relevant supplemental
information."
Animal: A concurrent control group exposed to vehicle-only treatment or untreated control.
All health outcomes (both cancer and noncancer).
Comparator
Outcome
PBPK Models Studies describing physiologically based pharmacokinetic (PBPK) models will be included.
Epidemiological, animal toxicological, and in vitro studies tagged as containing potentially
relevant ADME data were further screened using ADME-focused PECO criteria (Table A-2).
Key information from each study meeting the ADME-focused PECO criteria was extracted using
ICF's litstream™ software.
Table A-2. Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for
Absorption, Distribution, Metabolism, and/or Excretion (ADME) Studies
PECO
Element
Inclusion Criteria
Population Human: Any population and life stage (occupational or general population, including children and
other sensitive populations): whole organism, tissues, individual cells, or biomolecules.
Animal: Select non-human mammalian animal species: only non-human primates, rats, and mice
(whole organism, tissues, individual cells, or biomolecules) of any life stage (preconception, in
utero, lactation, peripubertal, and adult stages).
Exposure Any exposure to PFOA, PFOS, and/or the salts of PFOA/PFOS, including in vitro, in vivo (by
various routes of exposure), and ex vivo. In silico studies will also be included if the model system
can be linked to a PECO-relevant species.
PFOA (CAS number 335-67-1).
A-33
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PECO
Element
Inclusion Criteria
Comparator
Outcome
Other names: perfluorooctanoate, perfluorooctanoic acid, perfluoroctanoic acid,
2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluorooctanoic acid, pentadecafluoro-l-octanoic acid,
pentadecafluoro-n-octanoic acid, octanoic acid, pentadecafluoro-, perfluorocaprylic acid,
pentadecafluorooctanoic acid, perfluoroheptanecarboxylic acid, octanoic acid,
2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro-, ammonium perfluorooctanoate (APFO), sodium
perfluorooctanoate, potassium perfluoroctanoate
PFOS (CAS number 1763-23-1).
Other names: perfluorooctane sulfonate, perfluorooctanesulfonic acid, perfluorooctane sulfonic
acid, perfluorooctane sulphonate, perfluorooctane sulfonate, perfluorooctanyl sulfonate,
heptadecafluorooctane-1 -sulphonic, heptadecafluoro-1 -octanesulfonic acid,
1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-l-octanesulfonic acid,
heptadecafluorooctanesulfonic acid, lithium perfluorooctanesulfonate, potassium
perfluorooctanesulfonate, ammonium perfluorooctanesulfonate, sodium perfluorooctanesulfonate
Any comparison that informs PFOA or PFOS (1) absorption by the oral, inhalation, or dermal route
of exposure, (2) distribution across biological compartments, (3) metabolism, and/or (4) excretion.
Any examination of PFOA and/or PFOS (1) absorption of dose through gastrointestinal (GI) tract,
lungs, or skin, (2) distribution across biological compartments, (3) metabolism, and/or (4)
excretion. Studies describing PK models for PFOA and/or PFOS will be included.
Information and terms that are typically found in relevant ADME/PK modeling studies include the
following:
Absorption: Bioavailability; absorption rate(s); uptake rates; tissue location of absorption (e.g.,
stomach versus intestine, nasal versus lung); blood:air partition coefficient (PC); irritant/respiratory
depression; overall mass transfer coefficient; gas-phase diffusivity; gas-phase mass transfer
coefficient; liquid- (or tissue-) phase mass transfer coefficient; deposition fraction; retained
fractions; computational fluid (airway) dynamics.
Distribution: Volume of distribution (Vd) and parameters that determine Vd, including blood:
tissue PCs (especially for the target or a surrogate tissue) or lipophilicity; tissue burdens; storage
tissues or tissue components (e.g., serum binding proteins) and the binding coefficients;
transporters (active and passive).
Note: PFOA/PFOS are not metabolized so we are not expecting studies that focus on metabolites.
The terms below are general terms associated with metabolism.
Metabolism: Metabolic/biotransformationpathway(s); enzymes involved; metabolic rate;
maximum rate of transport (Vmax), Michaelis constant (Km);; metabolic induction; metabolic
inhibition, Ki; metabolic saturation/non-linearity; key organs involved in metabolism; key
metabolites (if any)/pathways; metabolites measured; species-, inter-individual-, and/or age-related
differences in enzyme activity or expression ("ontogeny"); site-specific activation (may be
toxicologically significant, but little systemic impact); cofactor (e.g., glutathione) depletion.
Excretion: Route(s)/pathway(s) of excretion for parent and metabolites; urine, fecal, exhalation,
hair, sweat, lactation; elimination rate(s); mechanism(s) of excretion (e.g., passive diffusion, active
transport).
Notes: CAS = Chemical Abstracts Service; PK = pharmacokinetic ADME = absorption, distribution, metabolism, and/or
excretion.
Epidemiological and animal toxicological studies that were tagged as containing potentially
relevant mechanistic data were further screened using mechanistic-focused PECO criteria (Table
A-3). Studies meeting the mechanistic-focused PECO criteria underwent a light extraction of key
study information using ICF's litstream™ software.
A-3 4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table A-3. Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for
Mechanistic Studies
PECO
Element
Evidence
Population Human: Any population and life stage (occupational or general population, including children
and other sensitive populations).
Animal: Select mammals (i.e., non-human primates and rodents (i.e., rats, mice, rabbits, guinea
pigs, other rodent models) and fish (i.e., zebrafish) of any life stage (preconception, in utero,
lactation, peripubertal, and adult stages).
Ex vivo, in vitro, in silico: Cultures of human or animal cells from relevant animal models
(primary, immortalized, transformed), organ slices, organotypic culture, in vitro molecular or
biochemical assay systems. In silico modeling data if it informs PFOA/PFOS MOA.
Exposure Any exposure to PFOA, PFOS, and/or the salts of PFOA/PFOS, including in vitro, in vivo (by
various routes of exposure), and ex vivo. In silico studies will also be included if the model
system can be linked to a PECO-relevant species.
PFOA (CAS number 335-67-1).
Other names: perfluorooctanoate, perfluorooctanoic acid, perfluoroctanoic acid,
2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluorooctanoic acid, pentadecafluoro-l-octanoic acid,
pentadecafluoro-n-octanoic acid, octanoic acid, pentadecafluoro-, perfluorocaprylic acid,
pentadecafluorooctanoic acid, perfluoroheptanecarboxylic acid, octanoic acid,
2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro-, ammonium perfluorooctanoate (APFO), sodium
perfluorooctanoate, potassium perfluoroctanoate
PFOS (CAS number 1763-23-1).
Other names: perfluorooctane sulfonate, perfluorooctanesulfonic acid, perfluorooctane sulfonic
acid, perfluorooctane sulphonate, perfluorooctane sulfonate, perfluorooctanyl sulfonate,
heptadecafluorooctane-1 -sulphonic, heptadecafluoro-1 -octanesulfonic acid,
1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-l-octanesulfonic acid,
heptadecafluorooctanesulfonic acid, lithium perfluorooctanesulfonate, potassium
perfluorooctanesulfonate, ammonium perfluorooctanesulfonate, sodium perfluorooctanesulfonate
Comparator Human: Comparison to group with no exposure or lower exposure.
Animal, ex vivo, in vitro, in silico: Comparison to an appropriate vehicle or no treatment control.
Outcome Any mechanistic data related to the MOA of PFOA/PFOS toxicity. This may include molecular
initiating events with PFOA/PFOS or downstream key events that inform the MOA or adverse
outcome pathway linking PFOA/PFOS exposure to disease.
Notes: MOA = mode of action; CAS = Chemical Abstracts Service.
A.1.5 Literature Search
EPA assembled literature inventories of epidemiological, animal toxicological, mechanistic, and
toxicokinetic studies for this updated toxicity assessment based on three data streams: 1)
literature published from 2014 through 2019 and then updated in the course of this review (i.e.,
through February 3, 2022) identified via literature searches of a variety of publicly available
scientific literature databases, 2) literature identified via other sources (e.g., searches of the gray
literature and studies shared with EPA by the SAB), and 3) literature identified in EPA's 2016
HESDs for PFOA and PFOS, which captured literature through 2013 {U.S. EPA, 2016,
3603279; U.S. EPA, 2016, 3603365}.
A. 1.5.1 Literature Search Strategies
The following sections describe literature search strategies used for databases and for additional
sources. They also describe methods used to incorporate studies from the 2016 PFOS HESD into
the literature inventory. The literature search strategy included searches within core literature
A-3 5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
databases (e.g., PubMed®, Web of Science™) as well as relevant domestic and international
non-periodical "gray" literature, such as technical reports, monographs, and conference and
symposium proceedings prepared by select committees or bodies (e.g., those convened by the
National Academy of Sciences or the World Health Organization (WHO)).
A.1.5.2 Database Searches
The database literature searches for this updated assessment focused only on the chemical name
(PFOS and related salts) with no limitations on lines of evidence (i.e., human, animal, in vitro, in
silico) or health outcomes. These searches comprised all literature related to health effects in
animals and humans resulting from acute, subchronic, and chronic exposure durations, and from
inhalation, oral, dermal, and injection exposure studies. Epidemiological, animal toxicological,
and in vitro studies that provide MOA information were included, and data specifically useful for
addressing risks to children and other susceptible populations (e.g., the elderly, pregnant or
lactating women, genetically susceptible populations) were identified. The searches likewise
included ADME studies and models useful for dose-response assessment, such as dosimetry and
PBPK models. The initial database search covered from January 2013 through April 11, 2019
(the 2019 literature search). That was subsequently updated by a search covering April 2019
through September 3, 2020 (2020 literature search) and another covering September 2020
through February 3, 2022 (2022 literature search). The date field tag used for these searches may
reflect either the date the article was published in print or e-published which may result in small
amounts of literature being captured in a literature search despite being published prior to the
start date. At the recommendation of SAB peer reviewers, the 2022 literature search focused on
the five main health outcomes that have been concluded to have the strongest evidence
(developmental, hepatic, immune, and cardiovascular effects, and cancer). EPA considered
mechanistic and toxicokinetic data identified through the September 2020 literature search, as
well as any more recent studies recommended by the SAB.
The databases listed below were searched for literature containing the search strings identified in
Table A-4 and Table A-5:
• Web of Science™ (Thomson Reuters),
• PubMed® (National Library of Medicine),
• ToxLine (incorporated into PubMed post 2019), and
• TSCATS (Toxic Substances Control Act Test Submissions).
Table A-4. Search String for April 2019 Database Searches
Database Search String Date Run
WoS ((TS="perfluorooctanoic acid" OR TS="perfluorooctane sulfonic acid") AND 4/10/2019
PY=(2013-2019) OR (TS="2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro-
Octanoic acid" OR TS="2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluorooctanoic
acid" OR TS="3,3,4,4,5,5,6,6,6-nonafluoro-2-oxo-Hexanoyl fluoride" OR
TS="3,3,4,4,5,5,6,6,6-nonafluoro-2-oxohexanoyl fluoride" OR TS="Hexanoyl
fluoride, 3,3,4,4,5,5,6,6,6-nonafluoro-2-oxo-" OR TS="Octanoic acid,
2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro-" OR TS="Pentadecafluoro-l-
octanoic acid" OR TS="Pentadecafluoro-n-octanoic acid" OR
TS="Pentadecafluorooctanoic acid" OR TS="Perfluorocaprylic acid" OR
TS="Perfluoroctanoic acid" OR TS="Perfluoroheptanecarboxylic acid" OR
TS="perfluorooctanyl sulfonate" OR TS="Perfluorooctanoic acid" OR
A-36
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Database Search String Date Run
TS="Octanoic acid, pentadecafluoro-" OR TS="Perfluorooctanoate" OR
TS="perfluorooctane sulfonate" OR TS="A 5717" OR TS="EF 201" OR
TS="Eftop EF 201" OR TS="Perfluoro-l-heptanecarboxylic acid" OR
TS="l,l,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-Heptadecafluoro-l-octanesulfonic acid"
OR TS="l-Octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-
heptadecafluoro-" OR TS="l-Perfluorooctanesulfonic acid" OR TS="EF 101"
OR TS="Eftop EF 101" OR TS="Heptadecafluoro-l-octanesulfonic acid" OR
TS="Heptadecafluorooctane-l-sulphonic acid" OR TS="Perfluorooctane
sulfonate" OR TS="perfluorooctane sulfonate" OR TS="Perfluorooctane
sulfonic acid" OR TS="Perfluorooctanesulfonic acid" OR
TS="Perfluorooctylsulfonic acid" OR TS="perfluorooctane sulphonate" OR
TS="perfluorooctane sulfonate" OR TS="l-Octanesulfonic acid,
heptadecafluoro-"OR TS="Heptadecafluorooctanesulfonic acid" OR
TS="Perfluoro-n-octanesulfonic acid" OR TS="Perfluorooctane Sulphonic
Acid" OR TS="Perfluorooctanesulfonate" OR TS="Perfluorooctylsulfonate"
OR ((TS="PFOA" OR TS="PFOS") AND (TS="fluorocarbon*" OR
TS="fluorotelomer*" OR TS="polyfluoro*" OR TS="perfluoro-*" OR
TS="perfluoroa*" OR TS="perfluorob*" OR TS="perfluoroc*" OR
TS="perfluorod*" OR TS="perfluoroe*" OR TS="perfluoroh*" OR
TS="perfluoron*" OR TS="perfluoroo*" OR TS="perfluorop*" OR
TS="perfluoros*" OR TS= "perfluorou*" OR TS="perfluorinated" OR
TS="fluorinated" ORTS="PFAS"))) AND PY=(2013-2019))
PubMed (335-67-1 [rn] OR 1763-23-l[rn] OR 45298-90-6[rn] OR "perfluorooctanoic 4/10/2019
acid"[nm] OR "perfluorooctane sulfonic acid"[nm]) AND
(2013/01/01:3000[pdat] OR 2013/01/01:3000[mhda] OR
2013/01/01:3000[edat] OR 2013/01/01:3000[crdt]) OR
(("2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro-Octanoic acid"[tw] OR
"2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluorooctanoic acid"[tw] OR
"3,3,4,4,5,5,6,6,6-nonafluoro-2-oxo-Hexanoyl fluoride"[tw] OR
"3,3,4,4,5,5,6,6,6-nonafluoro-2-oxohexanoyl fluoride"[tw] OR "Hexanoyl
fluoride, 3,3,4,4,5,5,6,6,6-nonafluoro-2-oxo-"[tw] OR "Octanoic acid,
2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro-"[tw] OR "Pentadecafluoro-1-
octanoic acid"[tw] OR "Pentadecafluoro-n-octanoic acid"[tw] OR
"Pentadecafluorooctanoic acid"[tw] OR "Perfluorocaprylic acid"[tw] OR
"Perfluoroctanoic acid"[tw] OR "Perfluoroheptanecarboxylic acid"[tw] OR
"perfluorooctanyl sulfonate" [tw] OR "Perfluorooctanoic acid"[tw] OR
"Octanoic acid, pentadecafluoro-"[tw] OR "Perfluorooctanoate"[tw] OR
"perfluorooctane sulfonate"[tw] OR "A 5717"[tw] OR "EF 201"[tw] OR "Eftop
EF 201"[tw] OR "Perfluoro-l-heptanecarboxylic acid"[tw] OR
"1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-Heptadecafluoro-l-octanesulfonic acid"[tw]
OR "1-Octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-
"[tw] OR "1-Perfluorooctanesulfonic acid"[tw] OR "EF 101"[tw] OR "Eftop
EF 101"[tw] OR "Heptadecafluoro-l-octanesulfonic acid"[tw] OR
"Heptadecafluorooctane-1-sulphonic acid"[tw] OR "Perfluorooctane
sulfonate "[tw] OR "perfluorooctane sulfonate" [tw] OR "Perfluorooctane
sulfonic acid"[tw] OR "Perfluorooctanesulfonic acid"[tw] OR
"Perfluorooctylsulfonic acid"[tw] OR "perfluorooctane sulphonate" [tw] OR
"perfluorooctane sulfonate" [tw] OR "1-Octanesulfonic acid, heptadecafluoro-
"[tw] OR "Heptadecafluorooctanesulfonic acid"[tw] OR "Perfluoro-n-
octanesulfonic acid"[tw] OR "Perfluorooctane Sulphonic Acid"[tw] OR
"Perfluorooctanesulfonate"[tw] OR "Perfluorooctylsulfonate"[tw] OR
(("PFOA"[tw] OR "PFOS"[tw]) AND (fluorocarbon*[tw] OR
fluorotelomer*[tw] ORpolyfluoro*[tw] ORperfluoro-*[tw] OR
perfluoroa*[tw] ORperfluorob*[tw] ORperfluoroc*[tw] ORperfluorod*[tw]
A-37
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Database Search String Date Run
ORperfluoroe*[tw] ORperfluoroh*[tw] ORperfluoron*[tw] OR
perfluoroo*[tw] ORperfluorop*[tw] ORperfluoros*[tw] ORperfluorou*[tw]
OR perfluorinated[tw] OR fluorinated[tw] ORPFAS[tw]))) AND
(2013/01/01:3000[pdat] OR 2013/01/01:3000[mhda] OR
2013/01/01:3000[edat] OR 2013/01/01:3000[crdt]))
Toxline @AND+@OR+("perfluorooctane sulfonate"+"pfos"+"perfluorooctanesulfonic 4/11/2019
acid"+"perfluorooctane sulfonic acid"+"perfluorooctane
sulphonate"+"perfluorooctane sulfonate"+"perfluorooctanyl
sulfonate"+"Heptadecafluorooctane-1 -sulphonic"+"Heptadecafluoro-1 -
octanesulfonic acid"+"l,l,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-l-
octanesulfonic acid"+"perfluorooctanoate"+"perfluorooctanoic
acid"+"perfluoroctanoic acid"+"pfoa"+"2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-
pentadecafluorooctanoic acid"+"Pentadecafluoro-1 -octanoic
acid"+"Pentadecafluoro-n-octanoic acid"+"Octanoic acid, pentadecafluoro-
"+"Perfluorocaprylic acid"+"Pentadecafluorooctanoic
acid"+"perfluoroheptanecarboxylic acid"+@TERM+@rn+3 35 -67-
l+@TERM+@rn+1763-23-l+@TERM+@rn+45298-90-
6)+@NOT+@org+pubmed+@AND+@RANGE+yr+2013+2019
TSCATS @AND+@OR+@rn+"335-67- 4/11/2019
l"+@AND+@org+TSCATS+@NOT+@org+pubmed
@AND+@OR+@rn+" 1763-23-
l"+@AND+@org+TSCATS+@NOT+@org+pubmed
Table A-5. Search String for September 2020 and February 2022 Database Searches
Database Search String Date Run
PubMed (335-67-1 [rn] OR 1763-23-l[rn] OR 45298-90-6[rn] OR "perfluorooctanoic 9/3/2020, 2/2/2022
Batch IDs: acid"[nm] OR "perfluorooctane sulfonic acid"[nm] OR
39678, 46137 "2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro-Octanoic acid"[tw] OR
"2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluorooctanoic acid"[tw] OR
"3,3,4,4,5,5,6,6,6-nonafluoro-2-oxo-Hexanoyl fluoride"[tw] OR
"3,3,4,4,5,5,6,6,6-nonafluoro-2-oxohexanoyl fluoride"[tw] OR "Hexanoyl
fluoride, 3,3,4,4,5,5,6,6,6-nonafluoro-2-oxo-"[tw] OR "Octanoic acid,
2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro-"[tw] OR "Pentadecafluoro-1-
octanoic acid"[tw] OR "Pentadecafluoro-n-octanoic acid"[tw] OR
"Pentadecafluorooctanoic acid"[tw] OR "Perfluorocaprylic acid"[tw] OR
"Perfluoroctanoic acid"[tw] OR "Perfluoroheptanecarboxylic acid"[tw] OR
"perfluorooctanyl sulfonate" [tw] OR "Perfluorooctanoic acid"[tw] OR
"Octanoic acid, pentadecafluoro-"[tw] OR "Perfluorooctanoate"[tw] OR
"perfluorooctane sulfonate"[tw] OR "A 5717"[tw] OR "EF 201"[tw] OR
"Eftop EF 201"[tw] OR "Perfluoro-l-heptanecarboxylic acid"[tw] OR
"1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-Heptadecafluoro-l-octanesulfonic acid"[tw]
OR "1-Octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-
"[tw] OR "1-Perfluorooctanesulfonic acid"[tw] OR "EF 101"[tw] OR "Eftop
EF 101"[tw] OR "Heptadecafluoro-1-octanesulfonic acid"[tw] OR
"Heptadecafluorooctane-l-sulphonic acid"[tw] OR "Perfluorooctane
sulfonate "[tw] OR "perfluorooctane sulfonate" [tw] OR "Perfluorooctane
sulfonic acid"[tw] OR "Perfluorooctanesulfonic acid"[tw] OR
"Perfluorooctylsulfonic acid"[tw] OR "perfluorooctane sulphonate" [tw] OR
"perfluorooctane sulfonate" [tw] OR "1-Octanesulfonic acid, heptadecafluoro-
"[tw] OR "Heptadecafluorooctanesulfonic acid"[tw] OR "Perfluoro-n-
octanesulfonic acid"[tw] OR "Perfluorooctane Sulphonic Acid"[tw] OR
A-38
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Database Search String Date Run
"Perfluorooctanesulfonate"[tw] OR "Perfluorooctylsulfonate"[tw] OR
(("PFOA"[tw] OR "PFOS"[tw]) AND (fluorocarbon*[tw] OR
fluorotelomer*[tw] ORpolyfluoro*[tw] ORperfluoro-*[tw] OR
perfluoroa*[tw] ORperfluorob*[tw] ORperfluoroc*[tw] ORperfluorod*[tw]
ORperfluoroe*[tw] ORperfluoroh*[tw] ORperfluoron*[tw] OR
perfluoroo*[tw] ORperfluorop*[tw] ORperfluoros*[tw] ORperfluorou*[tw]
OR perfluorinated[tw] OR fluorinated[tw] ORPFAS[tw]))) AND
(2020/09/03:3000[dp])
(TS="perfluorooctanoic acid" OR TS="perfluorooctane sulfonic acid" OR 9/3/2020, 2/3/2022
TS="2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro-Octanoic acid" OR
TS="2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluorooctanoic acid" OR
TS="3,3,4,4,5,5,6,6,6-nonafluoro-2-oxo-Hexanoyl fluoride" OR
TS="3,3,4,4,5,5,6,6,6-nonafluoro-2-oxohexanoyl fluoride" OR TS="Hexanoyl
fluoride, 3,3,4,4,5,5,6,6,6-nonafluoro-2-oxo-" OR TS="Octanoic acid,
2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro-" OR TS="Pentadecafluoro-l-
octanoic acid" OR TS="Pentadecafluoro-n-octanoic acid" OR
TS="Pentadecafluorooctanoic acid" OR TS="Perfluorocaprylic acid" OR
TS="Perfluoroctanoic acid" OR TS="Perfluoroheptanecarboxylic acid" OR
TS="perfluorooctanyl sulfonate" OR TS="Perfluorooctanoic acid" OR
TS="Octanoic acid, pentadecafluoro-" OR TS="Perfluorooctanoate" OR
TS="perfluorooctane sulfonate" OR TS="A 5717" OR TS="EF 201" OR
TS="Eftop EF 201" OR TS="Perfluoro-l-heptanecarboxylic acid" OR
TS="l,l,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-Heptadecafluoro-l-octanesulfonic acid"
OR TS="l-Octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-
heptadecafluoro-" OR TS="l-Perfluorooctanesulfonic acid" OR TS="EF 101"
OR TS="Eftop EF 101" OR TS="Heptadecafluoro-l-octanesulfonic acid" OR
TS="Heptadecafluorooctane-l-sulphonic acid" OR TS="Perfluorooctane
sulfonate" OR TS="perfluorooctane sulfonate" OR TS="Perfluorooctane
sulfonic acid" OR TS="Perfluorooctanesulfonic acid" OR
TS="Perfluorooctylsulfonic acid" OR TS="perfluorooctane sulphonate" OR
TS="perfluorooctane sulfonate" OR TS="l-Octanesulfonic acid,
heptadecafluoro-"OR TS="Heptadecafluorooctanesulfonic acid" OR
TS="Perfluoro-n-octanesulfonic acid" OR TS="Perfluorooctane Sulphonic
Acid" OR TS="Perfluorooctanesulfonate" OR TS="Perfluorooctylsulfonate"
OR ((TS="PFOA" OR TS="PFOS") AND (TS="fluorocarbon*" OR
TS="fluorotelomer*" OR TS="polyfluoro*" OR TS="perfluoro-*" OR
TS="perfluoroa*" OR TS="perfluorob*" OR TS="perfluoroc*" OR
TS="perfluorod*" OR TS="perfluoroe*" OR TS="perfluoroh*" OR
TS="perfluoron*" OR TS="perfluoroo*" OR TS="perfluorop*" OR
TS="perfluoros*" OR TS= "perfluorou*" OR TS="perfluorinated" OR
TS="fluorinated" OR TS="PFAS"))) AND PY=(2020-2022)
TOXLINE TOXLINE taken down, cannot search. -
TSCATS Incorporated into PubMed post 2019. -
The database searches were conducted by EPA and/or contractor information specialists and
librarians on April 11, 2019, September 3, 2020, and February 2 and 3, 2022 and all search
results were stored in the Health and Environmental Research Online (HERO) database
(https://hero.epa.eov/hero/index.cfm/proiect/paee/proiect id/2608). After deduplication (i.e.,
removal of duplicate results) in HERO, the database search results were imported into SWIFT
Review software for filtering/prioritization. SWIFT Review identifies those references most
likely to be applicable to human health risk assessment (https://www.sciome.com/swift-review/;
Web of
Science
Batch IDs:
39681, 46144
A-39
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
see also {Howard, 2016, 4149688}. In brief, SWIFT Review has preset literature search
strategies ("filters") developed and applied by information specialists to identify and prioritize
studies that are most likely to be useful for identifying human health content from those that
likely are not (e.g., studies on analytical methods). The filters function like a typical search
strategy in which studies are tagged as belonging to a certain category if the terms in the filter
literature search strategy appear in title, abstract, keyword, and/or medical subject headings
(MeSH) fields content. The applied SWIFT Review filters focused on the following evidence
types: human (epidemiology), animal models for human health, and in vitro studies. The details
of the search strategies that underlie the filters are available online
(https://hawcprd.epa.eov/media/attachment/SWIFT-Review Search Strategies.pdf).
For all literature searches, the evidence stream filters used were human, animal (all), animal
(human health model), [no tag], epidemiological quantitative analysis, and in vitro (with one
exception—for the 2022 literature search, the in vitro evidence stream filter was not used).
Studies not captured using these filters were not considered further. Studies that were captured
with these SWIFT Review evidence stream filters were exported as a RIS (Research Information
System) file for title and abstract screening using either DistillerSR or SWIFT ActiveScreener
software (described in subsequent sections of this appendix).
A.l.5.3 Additional Sources
The literature search strategies used were designed to be broad; however, like any search
strategy, studies may be missed (e.g., if the chemical of interest is not mentioned in title, abstract,
or keyword content; or if gray literature is not indexed in the databases that were searched).
Thus, additional sources were reviewed to identify studies that could have been missed in the
database searches. Reviews of additional sources included the following:
1. Review of studies cited in assessments published by other U.S. federal agencies, as well
as international, and U.S. state-level agencies (including Agency for Toxic Substances
and Disease Registry (ATSDR) and California Environmental Protection Agency
(CalEPA) assessments that were ongoing at the time of searching).
• Manual review of the reference list from ATSDR's Toxicological Profile for
Perfluoroalkyls {ATSDR, 2021, 9642134} (not date limited).
• Manual review of the reference list from CalEPA's First Public Review Draft of
Proposed Public Health Goals for Perfluorooctanoic Acid and Perfluorooctane
Sulfonic Acid in Drinking Water {CalEPA, 2021, 9416932} (not date limited).
• Manual review of National Toxicology Program (NTP) publications
(https://ntp.niehs.nih.eov/data/index.html). In 2020, the NTP website was searched
for PFOS toxicity study final reports that could provide relevant health effects
information.
• Manual review of PFAS toxicity studies identified by the New Jersey Department of
Environmental Protection (NJDEP).
2. Review of studies identified during mechanistic or toxicokinetic evidence synthesis:
A-40
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• Manual review of the reference lists of studies screened as PECO-relevant after full-
text review were reviewed at the title level for potentially relevance (backward
citation search).
• Manual review of other EPA PFAS assessments or literature searches under
development by IRIS.
3. Review of studies identified by the SAB PFAS Panel peer reviewers in their final report
(published in August 2022).
A. 1.5.4 Incorporation of Data from the 2016 PFOS Health Effects
Support Document
The 2016 HESD for PFOS contained a comprehensive summary of relevant literature based on
searches conducted through 2013, as described in that document and in the related 2016
Drinking Water Health Advisory for PFOS. The HESD underwent a public comment period in
February 2014, and an independent external public panel peer review in August 2014. EPA
incorporated key studies from the 2016 PFOS HESD that addressed one or more of the five main
health outcomes into this updated PFOS assessment, as described below.
Over 140 epidemiological studies were captured in the 2016 PFOS HESD. The 2016 HESD did
not use the epidemiological data quantitatively. For this updated assessment, EPA reviewed the
epidemiological studies that were included in the HESD summary tables and identified those that
were relevant to one or more of the five main health outcomes (i.e., developmental, immune,
hepatic, cardiovascular, and cancer). A total of 51 epidemiological studies were included and are
listed in Table A-6 (studies relevant to more than one health outcome are listed under each
applicable category in the table).
Table A-6. Key Epidemiological Studies of Priority Health Outcomes Identified from 2016
PFOS Health Effects Support Document
HERO ID
Reference
Title
Cancer
4727072
Alexander and Olsen,
2007
Bladder cancer in perfluorooctanesulfonyl fluoride manufacturing workers
1291101
Alexander et al.,
2003
Mortality of employees of a perfluorooctanesulphonyl fluoride
manufacturing facility
2150988
Boncfcld-Jorgcnscn
etal., 2011
Perfluorinated compounds are related to breast cancer risk in Greenlandic
Inuit: a case control study
2851186
Boncfcld-Jorgcnscn
etal., 2014
Breast cancer risk after exposure to perfluorinated compounds in Danish
women: a case-control study nested in the Danish National Birth Cohort
2919344
Eriksen et al., 2009
Perfluorooctanoate and perfluorooctanesulfonate plasma levels and risk of
cancer in the general Danish population
4930271
Grice et al., 2007
Self-reported medical conditions in perfluorooctanesulfonyl fluoride
manufacturing workers
2968084
Hardell et al., 2014
Case-control study on perfluorinated alkyl acids (PFAAs) and the risk of
prostate cancer
Cardiovascular
1291101
Alexander et al.,
2003
Mortality of employees of a perfluorooctanesulphonyl fluoride
manufacturing facility
A-41
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
2919285
Chateau-Degat et al.,
2010
Effects of perfluorooctanesulfonate exposure on plasma lipid levels in the
Inuit population of Nunavik (Northern Quebec)
2919150
Eriksen et al., 2013
Association between plasma PFOA and PFOS levels and total cholesterol in
a middle-aged Danish population
2850962
Fitz-Simon et al.,
2013
Reductions in serum lipids with a 4-year decline in serum perfluorooctanoic
acid and perfluorooctanesulfonic acid
1430763
Frisbee et al., 2010
Perfluorooctanoic acid, perfluorooctanesulfonate, and serum lipids in
children and adolescents: results from the C8 Health Project
3749193
Fu et al., 2014
Associations between serum concentrations of perfluoroalkyl acids and
serum lipid levels in a Chinese population
2850925
Geiger et al., 2014
The association between PFOA, PFOS and serum lipid levels in adolescents
2851286
Geiger et al., 2014
No association between perfluoroalkyl chemicals and hypertension in
children
1290820
Lin et al., 2009
Association among serum perfluoroalkyl chemicals, glucose homeostasis,
and metabolic syndrome in adolescents and adults
1291110
Nelson etal., 2010
Exposure to polyfluoroalkyl chemicals and cholesterol, body weight, and
insulin resistance in the general US population
1290020
Olsen et al., 2003
Epidemiologic assessment of worker serum perfluorooctanesulfonate
(PFOS) and perfluorooctanoate (PFOA) concentrations and medical
surveillance examinations
10228462
Olsen et al., 2001
A longitudinal analysis of serum perfluorooctane sulfonate (PFOS) and
perfluorooctanoate (PFOA) levels in relation to lipid and hepatic clinical
chemistry test results from male employee participants of the 1994/95, 1997
and 2000 fluorochemical medical surveillance program. Final report.
2850928
Starling et al., 2014
Perfluoroalkyl substances and lipid concentrations in plasma during
pregnancy among women in the Norwegian Mother and Child Cohort Study
1290816
Stein et al., 2009
Serum levels of perfluorooctanoic acid and perfluorooctane sulfonate and
pregnancy outcome
2850370
Timmermann et al.,
2014
Adiposity and glycemic control in children exposed to perfluorinated
compounds
Developmental
1429893
Andersen et al., 2010
Prenatal exposures to perfluorinated chemicals and anthropometric
measures in infancy
1290833
Apelberg et al., 2007
Cord serum concentrations of perfluorooctane sulfonate (PFOS) and
perfluorooctanoate (PFOA) in relation to weight and size at birth
1290900
Apelberg et al., 2007
Determinants of fetal exposure to polyfluoroalkyl compounds in Baltimore,
Maryland
1332466
Chen etal., 2012
Perfluorinated compounds in umbilical cord blood and adverse birth
outcomes
2850274
Darrow et al., 2014
PFOA and PFOS serum levels and miscarriage risk
2850966
Darrow et al., 2013
Serum perfluorooctanoic acid and perfluorooctane sulfonate concentrations
in relation to birth outcomes in the Mid-Ohio Valley, 2005-2010
1005775
Fei et al., 2007
Perfluorinated chemicals and fetal growth: A study within the Danish
National Birth Cohort
1290822
Fei et al., 2008
Prenatal exposure to perfluorooctanoate (PFOA) and
perfluorooctanesulfonate (PFOS) and maternally reported developmental
milestones in infancy
2349574
Fei et al., 2008
Fetal growth indicators and perfluorinated chemicals: a study in the Danish
National Birth Cohort
1290814
Hamm et al., 2010
Maternal exposure to perfluorinated acids and fetal growth
1332465
Maisonet et al., 2012
Maternal concentrations of polyfluoroalkyl compounds during pregnancy
and fetal and postnatal growth in British girls
A-42
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
2349575
Monroy et al., 2008
Serum levels of perfluoroalkyl compounds in human maternal and
umbilical cord blood samples
1290816
Stein et al., 2009
Serum levels of perfluorooctanoic acid and perfluorooctane sulfonate and
pregnancy outcome
1291133
Washino et al., 2009
Correlations between prenatal exposure to perfluorinated chemicals and
reduced fetal growth
Hepatic
1291101
Alexander et al.,
2003
Mortality of employees of a perfluorooctanesulphonyl fluoride
manufacturing facility
1276142
Gallo et al., 2012
Serum perfluorooctanoate (PFOA) and perfluorooctane sulfonate (PFOS)
concentrations and liver function biomarkers in a population with elevated
PFOA exposure
4930271
Grice et al., 2007
Self-reported medical conditions in perfluorooctanesulfonyl fluoride
manufacturing workers
1291111
Linetal., 2010
Investigation of the Associations Between Low-Dose Serum Perfluorinated
Chemicals and Liver Enzymes in US Adults
1290020
Olsen et al., 2003
Epidemiologic assessment of worker serum perfluorooctanesulfonate
(PFOS) and perfluorooctanoate (PFOA) concentrations and medical
surveillance examinations
10228462
Olsen et al., 2001
A longitudinal analysis of serum perfluorooctane sulfonate (PFOS) and
perfluorooctanoate (PFOA) levels in relation to lipid and hepatic clinical
chemistry test results from male employee participants of the 1994/95, 1997
and 2000 fluorochemical medical surveillance program. Final report.
Immune
1937230
Dong et al., 2013
Serum polyfluoroalkyl concentrations, asthma outcomes, and
immunological markers in a case-control study of Taiwanese children
1290805
Fei et al., 2010
Prenatal exposure to PFOA and PFOS and risk of hospitalization for
infectious diseases in early childhood
1248827
Grandjean et al.,
2012
Serum vaccine antibody concentrations in children exposed to
perfluorinated compounds
1937228
Granum et al., 2013
Pre-natal exposure to perfluoroalkyl substances may be associated with
altered vaccine antibody levels and immune-related health outcomes in
early childhood
2851240
Humblet et al., 2014
Perfluoroalkyl chemicals and asthma among children 12-19 years of age:
NHANES (1999-2008)
2850913
Looker etal., 2014
Influenza vaccine response in adults exposed to perfluorooctanoate and
perfluorooctanesulfonate
1424977
Wang et al., 2011
The effect of prenatal perfluorinated chemicals exposures on pediatric atopy
Serum Lipids
2919285
Chateau-Degat et al.,
2010
Effects of perfluorooctanesulfonate exposure on plasma lipid levels in the
Inuit population of Nunavik (Northern Quebec)
2919150
Eriksen et al., 2013
Association between plasma PFOA and PFOS levels and total cholesterol in
a middle-aged Danish population
2919156
Fisher etal., 2013
Do perfluoroalkyl substances affect metabolic function and plasma lipids? -
Analysis of the 2007-2009, Canadian Health Measures Survey (CHMS)
Cycle 1
2850962
Fitz-Simon et al.,
2013
Reductions in serum lipids with a 4-year decline in serum perfluorooctanoic
acid and perfluorooctanesulfonic acid
1430763
Frisbee et al., 2010
Perfluorooctanoic acid, perfluorooctanesulfonate, and serum lipids in
children and adolescents: results from the C8 Health Project
3749193
Fu et al., 2014
Associations between serum concentrations of perfluoroalkyl acids and
serum lipid levels in a Chinese population
2850925
Geiger et al., 2014
The association between PFOA, PFOS and serum lipid levels in adolescents
A-43
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
1290820
Lin et al., 2009
Association among serum perfluoroalkyl chemicals, glucose homeostasis,
and metabolic syndrome in adolescents and adults
3981585
Maisonet et al., 2015
Prenatal exposures to perfluoroalkyl acids and serum lipids at ages 7 and 15
in females
1291110
Nelson etal., 2010
Exposure to Polyfluoroalkyl Chemicals and Cholesterol, Body Weight, and
Insulin Resistance in the General US Population
1290020
Olsen et al., 2003
Epidemiologic assessment of worker serum perfluorooctanesulfonate
(PFOS) and perfluorooctanoate (PFOA) concentrations and medical
surveillance examinations
10228462
Olsen et al., 2001
A longitudinal analysis of serum perfluorooctane sulfonate (PFOS) and
perfluorooctanoate (PFOA) levels in relation to lipid and hepatic clinical
chemistry test results from male employee participants of the 1994/95, 1997
and 2000 fluorochemical medical surveillance program. Final report.
2850928
Starling et al., 2014
Perfluoroalkyl substances and lipid concentrations in plasma during
pregnancy among women in the Norwegian Mother and Child Cohort Study
1291109
Steenland et al., 2009
Association of perfluorooctanoic acid and perfluorooctane sulfonate with
serum lipids among adults living near a chemical plant
2850370
Timmermann et al.,
2014
Adiposity and glycemic control in children exposed to perfluorinated
compounds
Notes: NHANES = National Health and Examination Survey.
EPA also reviewed the animal toxicological studies in the HESD summary tables that were
identified as relevant for all health outcomes. A total of 9 animal toxicological studies were
included and listed in Table A-7 (studies relevant to more than one health outcome are listed
under each applicable category in the table).
Table A-7. Key Toxicological Animal Toxicological Studies Identified from 2016 PFOS
Health Effects Support Document
HERO ID
Reference
Title
Cardiovascular
757871
Curran et al., 2008
Altered fatty acid homeostasis and related toxicologic sequelae in rats
exposed to dietary potassium perfluorooctanesulfonate (PFOS)
757857
Luebker et al., 2005
Neonatal mortality from in utero exposure to perfluorooctanesulfonate
(PFOS) in Sprague-Dawley rats: dose-response, and biochemical and
pharamacokinetic parameters
1290852
Seacat et al., 2003
Sub-chronic dietary toxicity of potassium perfluorooctanesulfonate in rats
757853
Seacat et al., 2002
Subchronic toxicity studies on perfluorooctanesulfonate potassium salt in
cynomolgus monkeys
Endocrine
757871
Curran et al., 2008
Altered fatty acid homeostasis and related toxicologic sequelae in rats
exposed to dietary potassium perfluorooctanesulfonate (PFOS)
757854
Lau et al., 2003
Exposure to perfluorooctane sulfonate during pregnancy in rat and mouse. II:
Postnatal evaluation
757857
Luebker et al., 2005
Neonatal mortality from in utero exposure to perfluorooctanesulfonate
(PFOS) in Sprague-Dawley rats: dose-response, and biochemical and
pharamacokinetic parameters
757853
Seacat et al., 2002
Subchronic toxicity studies on perfluorooctanesulfonate potassium salt in
cynomolgus monkeys
Developmental
A-44
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
757873
Butenhoff et al., 2009
Gestational and lactational exposure to potassium perfluorooctanesulfonate
(K+PFOS) in rats: Developmental neurotoxicity
757854
Lau et al., 2003
Exposure to perfluorooctane sulfonate during pregnancy in rat and mouse. II:
Postnatal evaluation
757857
Luebker et al., 2005
Neonatal mortality from in utero exposure to perfluorooctanesulfonate
(PFOS) in Sprague-Dawley rats: dose-response, and biochemical and
pharamacokinetic parameters
1276160
Luebker et al., 2005
Two-generation reproduction and cross-foster studies of
perfluorooctanesulfonate (PFOS) in rats
Hematologic
757871
Curran et al., 2008
Altered fatty acid homeostasis and related toxicologic sequelae in rats
exposed to dietary potassium perfluorooctanesulfonate (PFOS)
1290852
Seacat et al., 2003
Sub-chronic dietary toxicity of potassium perfluorooctanesulfonate in rats
757853
Seacat et al., 2002
Subchronic toxicity studies on perfluorooctanesulfonate potassium salt in
cynomolgus monkeys
Hepatic
757871
Curran et al., 2008
Altered fatty acid homeostasis and related toxicologic sequelae in rats
exposed to dietary potassium perfluorooctanesulfonate (PFOS)
2919266
Kawamoto et al., 2011
Ultrasonic-induced tonic convulsion in rats after subchronic exposure to
perfluorooctane sulfonate (PFOS)
757854
Lau et al., 2003
Exposure to perfluorooctane sulfonate during pregnancy in rat and mouse. II:
Postnatal evaluation
757857
Luebker et al., 2005
Neonatal mortality from in utero exposure to perfluorooctanesulfonate
(PFOS) in Sprague-Dawley rats: dose-response, and biochemical and
pharamacokinetic parameters
1290852
Seacat et al., 2003
Sub-chronic dietary toxicity of potassium perfluorooctanesulfonate in rats
757853
Seacat et al., 2002
Subchronic toxicity studies on perfluorooctanesulfonate potassium salt in
cynomolgus monkeys
Immune
757871
Curran etal., 2008
Altered fatty acid homeostasis and related toxicologic sequelae in rats
exposed to dietary potassium perfluorooctanesulfonate (PFOS)
1290852
Seacat et al., 2003
Sub-chronic dietary toxicity of potassium perfluorooctanesulfonate in rats
757853
Seacat et al., 2002
Subchronic toxicity studies on perfluorooctanesulfonate potassium salt in
cynomolgus monkeys
Metabolic
757871
Curran et al., 2008
Altered fatty acid homeostasis and related toxicologic sequelae in rats
exposed to dietary potassium perfluorooctanesulfonate (PFOS)
757857
Luebker et al., 2005
Neonatal mortality from in utero exposure to perfluorooctanesulfonate
(PFOS) in Sprague-Dawley rats: dose-response, and biochemical and
pharamacokinetic parameters
1290852
Seacat et al., 2003
Sub-chronic dietary toxicity of potassium perfluorooctanesulfonate in rats
Nervous
757873
Butenhoff et al., 2009
Gestational and lactational exposure to potassium perfluorooctanesulfonate
(K+PFOS) in rats: Developmental neurotoxicity
757871
Curran et al., 2008
Altered fatty acid homeostasis and related toxicologic sequelae in rats
exposed to dietary potassium perfluorooctanesulfonate (PFOS)
1276160
Luebker et al., 2005
Two-generation reproduction and cross-foster studies of
perfluorooctanesulfonate (PFOS) in rats
Renal
A-45
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
1276144
Butenhoff et al., 2012
Chronic dietary toxicity and carcinogenicity study with potassium
perfluorooctanesulfonate in Sprague Dawley rats
757871
Curran et al., 2008
Altered fatty acid homeostasis and related toxicologic sequelae in rats
exposed to dietary potassium perfluorooctanesulfonate (PFOS)
1290852
Seacat et al., 2003
Sub-chronic dietary toxicity of potassium perfluorooctanesulfonate in rats
757853
Seacat et al., 2002
Subchronic toxicity studies on perfluorooctanesulfonate potassium salt in
cynomolgus monkeys
Reproductive
757873
Butenhoff et al., 2009
Gestational and lactational exposure to potassium perfluorooctanesulfonate
(K+PFOS) in rats: Developmental neurotoxicity
757854
Lau et al., 2003
Exposure to perfluorooctane sulfonate during pregnancy in rat and mouse. II:
Postnatal evaluation
757857
Luebker et al., 2005
Neonatal mortality from in utero exposure to perfluorooctanesulfonate
(PFOS) in Sprague-Dawley rats: dose-response, and biochemical and
pharamacokinetic parameters
1276160
Luebker et al., 2005
Two-generation reproduction and cross-foster studies of
perfluorooctanesulfonate (PFOS) in rats
757853
Seacat et al., 2002
Subchronic toxicity studies on perfluorooctanesulfonate potassium salt in
cynomolgus monkeys
Systemic
757873
Butenhoff et al., 2009
Gestational and lactational exposure to potassium perfluorooctanesulfonate
(K+PFOS) in rats: Developmental neurotoxicity
757871
Curran et al., 2008
Altered fatty acid homeostasis and related toxicologic sequelae in rats
exposed to dietary potassium perfluorooctanesulfonate (PFOS)
757857
Luebker et al., 2005
Neonatal mortality from in utero exposure to perfluorooctanesulfonate
(PFOS) in Sprague-Dawley rats: dose-response, and biochemical and
pharamacokinetic parameters
1276160
Luebker et al., 2005
Two-generation reproduction and cross-foster studies of
perfluorooctanesulfonate (PFOS) in rats
1290852
Seacat et al., 2003
Sub-chronic dietary toxicity of potassium perfluorooctanesulfonate in rats
757853
Seacat et al., 2002
Subchronic toxicity studies on perfluorooctanesulfonate potassium salt in
cynomolgus monkeys
A.1.6 Literature Screening Process to Target Dose-Response
Studies and PK Models
This section summarizes the methods used to screen the literature search results against the
PECO criteria to identify relevant studies potentially suitable for use in dose-response analyses
and studies featuring PK models. Literature search results were screened at both title/abstract and
full-text levels. These screening steps are described further below.
The PECO criteria used to screen the literature search results are the same as those used to frame
the initial literature search (Table A-l) and are outlined again in Table A-8 below.
A-46
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table A-8. Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for a
Systematic Review on the Health Effects from Exposure to PFOA and PFOS
PECO
Element
Inclusion Criteria
Population Human: Any population and life stage (occupational or general population, including children and
other sensitive populations).
Animal: Nonhuman mammalian animal species (whole organism) of any life stage (including
preconception, inutero, lactation, peripubertal, and adult stages).
In vitro!cell studies or in silico/modeling toxicity studies should be tagged as supplemental.
Exposure Any exposure to PFOA, PFOS, and/or the salts of PFOA/PFOS, including but not limited to:
PFOA {Chemical Abstracts Service (CAS) number 335-67-1).
Other names: perfluorooctanoate; perfluorooctanoic acid; 2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-
pentadecafluorooctanoic acid; pentadecafluoro-l-octanoic acid; pentadecafluoro-n-octanoic acid;
perfluorocaprylic acid; pentadecafluorooctanoic acid; perfluoroheptanecarboxylic acid; octanoic-
acid, pentadecafluoro-
Relevant Salts of PFOA: ammonium perfluorooctanoate (APFO), sodium perfluorooctanoate,
potassium perfluorooctanoate
PFOS {CAS number 1763-23-1).
Other names: perfluorooctane sulfonate, perfluorooctanesulfonic acid, perfluorooctane sulfonic
acid, perfluorooctane sulphonate, perfluorooctanyl sulfonate, heptadecafluorooctane-l-sulphonic,
Heptadecafluoro-l-octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-l-
octanesulfonic acid
Relevant salts of PFOS: lithium perfluorooctanesulfonate, potassium perfluorooctanesulfonate
(K+PFOS), ammonium perfluorooctanesulfonate, sodium perfluorooctanesulfonate
Human: Any exposure to PFOA or PFOS via oral routes. Other exposure routes, including
inhalation, dermal, or unknown/multiple routes will be tracked during title and abstract screening
and tagged as "potentially relevant supplemental information."
Animal: Any exposure to PFOA or PFOS via oral routes. Other exposure routes, including
inhalation, dermal, injection or unknown/multiple routes, will be tracked during title and abstract
screening and tagged as "potentially relevant supplemental information." Studies involving
exposures to mixtures will be included only if they include exposure to PFOA or PFOS alone.
Studies with less than 28 days of dosing, with the exception of reproductive, developmental,
immune and neurological health outcome studies, should be tagged as supplemental.
Comparator Human: A comparison or referent population exposed to lower levels (or no exposure/exposure
below detection limits) of PFOA or PFOS, or exposure to PFOA or PFOS for shorter periods of
time. Case reports and case series will be tracked as "potentially relevant supplemental
information."
Animal: A concurrent control group exposed to vehicle-only treatment or untreated control.
Outcome All health outcomes (both cancer and noncancer).
PBPK Models Studies describing PBPK models will be included.
Notes: PBPK = physiologically-based pharmacokinetic.
Following SWIFT Review filtering (see Section A. 1.5.2), literature search results were imported
into either DistillerSR (Evidence Partners;
https://www.evidencepartners.com/products/distillersr-svstematic-review-software) or SWIFT
ActiveScreener (Sciome; https://www.sciome.com/swift-activescreener/) software and were
screened against the PECO criteria at the title and abstract level to identify PECO-relevant
studies published since development of the 2016 PFOS HESD and which could influence the
derivation of an oral RfD and/or CSF. Studies that did not meet the PECO criteria as determined
by title/abstract screening but did appear to include potentially important supplemental
information were categorized according to the type of supplemental information they provided
(e.g., mechanistic, ADME). Studies that met the PECO criteria were tagged as having relevant
A-47
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
human data, relevant animal data (in a mammalian model), or a PBPK model. Following
completion of title/abstract screening (described further in Sections A. 1.6.3 and A. 1.6.4), the
literature search results were re-screened at the full-text level (described further in Section
A.l.6.5).
The title/abstract and full-text level screenings were performed by independent reviewers using
structured forms in DistillerSR, with a process for conflict resolution. Literature inventories for
PECO-relevant studies and studies tagged as containing potentially relevant supplemental
material during full-text screening were created to facilitate review of studies by topic-specific
experts by identifying evidence types and health effect systems. These procedures are consistent
with those outlined in the IRIS Handbook {U.S. EPA, 2022, 10476098}.
Studies that did not meet the PECO criteria but contained potentially relevant supplemental
information were inventoried during the literature screening process. Potentially relevant
supplemental materials included the following (see Table A-l 1 for full list):
• Mechanistic data (including in vitro/ex vivo/in silico studies),
• Studies in non-mammalian or transgenic mammalian model systems,
• Non-oral routes of administration (for animal toxicological studies),
• ADME and toxicokinetic studies (including the application of existing PBPK models),
• Exposure assessment or characterization studies (no health outcome assessment),
• Mixture studies (animal toxicological studies on mixtures of PFOS and other substances
or epidemiological studies that only report associations based on sum or total PFAS),
• Human case reports (n = 1-3 cases per report),
• Records or other assessments with no original data (e.g., reviews, editorials,
commentaries),
• Conference abstracts, and
• Non-English language studies.
Following title/abstract and full-text level screening, studies tagged as containing potentially
relevant mechanistic, ADME, or toxicokinetic data underwent additional screening and data
extraction steps that were separate from steps followed for PECO-relevant studies. Details on the
screening and data extraction methods for ADME studies are described below.
1.6.1 Screening ADME Studies
Studies identified as containing potentially relevant supplemental ADME data during
title/abstract and/or full-text screening underwent further screening against the ADME-specific
PECO criteria outlined in Table A-2. For studies that met the ADME-specific PECO criteria (see
Table A-2), key study information was extracted using litstream™ software. Methods for this
ADME screening and extraction of some key study information into litstream is described
further in Section A. 1.6.7.
1.6.2 Screening Mechanistic Studies
Studies identified as containing potentially relevant supplemental mechanistic data during full-
text screening underwent further screening against the mechanistic-specific PECO criteria
outlined in Table A-3. Studies that met the mechanistic-specific PECO criteria were extracted
A-48
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
into litstream™. Methods for this mechanistic information screening and extraction of some key
study information into litstream is described further in Section A. 1.6.8.
A. 1.6.3 Title/Abstract Screening Questions - DistillerSR
Studies identified from the 2016 PFOS HESD and recent systematic literature search and review
efforts (searches through 2020) were imported into DistillerSR software for title/abstract
screening. For each study, screeners reviewed the title and abstract and responded to a series of
prompts within structured DistillerSR forms to assess PECO relevance and identify evidence
stream(s). Table A-9 below lists the prompts within the DistillerSR forms used for title/abstract
screening and the response options for each prompt.
Table A-9. DistillerSR Form for Title/Abstract Screening
Question/Prompt
Response Options
1
Does the article meet PECO criteria?
[Select one]
• Yes
• Noa
• Tag as potentially relevant supplemental material
• Unclear
IP
Yes" to Question #1:
2a
What type of evidence?
[Select all that apply]
• Human
• Animal (mammalian models)
• PBPK model
If"Tag as potentially relevant supplemental male rial" to Question #1:
2b
What kind of supplemental
material?b
[Select all that apply]
• Mechanistic0
• Non-mammalian model
• ADME/toxicokinetic
• Acute/short-term duration exposures
• Non-oral route of administration
• Exposure characteristics (no health outcome)
• Susceptible population (no health outcome)
• Environmental fate or occurrence (including food)
• Mixture study
• Case study or case series
• Other assessments or records with no original data (e.g., reviews,
editorials, commentaries)
• Conference abstract
• Bioaccumulation data in fish
Notes: PBPK = physiologically-based pharmacokinetic.
a Erratums and corrections were considered not relevant.
b Refer to list of supplemental tags in Appendix A. 1.6.4.1.
c Refer to list of mechanistic information in Appendix A. 1.6.4.2.
A. 1.6.4 Title/Abstract Screening Questions - SWIFT-Active
Studies identified from the most recent literature search (2020-2022) were imported into
SWIFT-Active Screener software for title/abstract screening. For each study, screeners reviewed
the title and abstract and responded to a set of prompts designed to ascertain PECO relevance
and identify evidence stream(s). Table A-10 below lists the prompts within SWIFT-Active that
were used for title/abstract screening and the response options for each prompt.
A-49
-------
DRAFT FOR PUBLIC COMMENT
Table A-10. SWIFT-Active Form for Title/Abstract Screening
MARCH 2023
Question/Prompt
Response Options
1
Include this reference?
Select "Yes, include the reference" if unsure.
[Select one]
• Yes, include the reference
• No, exclude the reference3
i r
"Yes" to Question #1:
2a
Identify the Type of Evidence
[Select all that apply]
• Human/Epidemiological
• Animal
• Unsure
i r
'No. exclude the reference" to Question #1:
2b
Not Relevant or Supplemental?1"
Select whether the reference is not relevant to
PECO and should be excluded or if the
reference contains supplemental information.
[Select all that apply]
• Exclude/Not Relevant
• Supplemental
Notes:
a Erratums and corrections were considered not relevant.
b Refer to the list of supplemental tags in Appendix A. 1.6.4.1.
A.l.6.4.1 Supplemental Togs
The categories shown in Table A-l 1 were considered supplemental throughout the title/abstract
and full-text screening processes. With the exception of studies tagged as containing mechanistic
or ADME/TK information, which were further considered as described in Section A. 1.6.7 and
Section A. 1.6.8 of this appendix, studies identified as not PECO-relevant but containing
potentially useful supplemental material were not considered for the subsequent steps of the
systematic review process.
Table A-ll. Supplemental Tags for Title/Abstract and Full-Text Screening
Category
Evidence
Mechanistic Studies
Studies reporting measurements related to a health outcome that inform the biological
or chemical events associated with phenotypic effects, in both mammalian and non-
mammalian model systems, including in vitro, in vivo (by various routes of
exposure), ex vivo, and in silico studies. When possible, mechanistic studies will be
sub-tagged as pertinent to cancer, non-cancer, or unclear/unknown.
PK or PBPK Models
Studies reporting the application of existing PK or PBPK models.
Non-Mammalian Model
Studies in non-mammalian model systems, e.g., fish, birds, C. elegans
Systems
ADME and Toxicokinetic
Studies designed to capture information regarding absorption, distribution,
metabolism, and excretion, including toxicokinetic studies. Such information may be
helpful in updating or revising the parameters used in existing PBPK models.
Acute/Short-Term Duration
Animal studies of less than 28 days (unless the study is a developmental/reproductive
Exposures
study)
A-50
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Category
Evidence
Only One Exposure Group
Animal studies with only one exposure group, e.g., control and 1 mg/kg/day
PFOA/S.
Non-Oral Routes of
Exposure
Studies not addressing routes of exposure that fall outside the PECO scope, include
inhalation and dermal exposure routes
Exposure Characteristics
(No Health Outcome)
Susceptible Populations
(No Health Outcome)
Exposure characteristic studies include data that are unrelated to toxicological
endpoints, but which provide information on exposure sources or measurement
properties of the environmental agent (e.g., demonstrate a biomarker of exposure).
Studies that identify potentially susceptible subgroups; for example, studies that focus
on a specific demographic, life stage, or genotype.
Environmental Fate or
Occurrence (Including Food)
Studies that focus on describing where the chemical will end up after it is used and
released into the environment.
Mixture Studies
Case Studies or Case Series
Records With No Original
Data
Mixture studies that are not considered PECO-relevant because they do not contain
an exposure or treatment group assessing only the chemical of interest.
Case reports and case series will be tracked as potentially relevant supplemental
information.
Records that do not contain original data, such as other agency assessments,
informative scientific literature reviews, editorials, or commentaries.
Other Assessments or Records Secondary studies (e.g., reviews, editorials, commentaries, assessments) that do not
With No Original Data (e.g., provide any primary research/results.
Reviews, Editorials,
Commentaries)
Conference Abstracts Records that do not contain sufficient documentation to support study evaluation and
data extraction.
Bioaccumulation in Fish Retained records relevant to other EPA projects mentioned in the PFAS Action Plan.
Non-English Reports
Studies not reported in English.
Notes: PK = pharmacokinetic; PBPK = physiologically-based pharmacokinetic; ADME = absorption, distribution, metabolism,
and/or excretion; C. elegans = Caenorhabditis elegans.
A. 1.6.4.2 Mechanistic Study Categories and Keywords
The following categories were considered mechanistic throughout the title/abstract and full-text
screening (Table A-l 1). Studies tagged as containing potentially relevant supplemental
mechanistic information were further considered as described in Section A. 1.6.8 of this
appendix.
Table A-12. Mechanistic Study Categories Considered as Supplemental
Category
Examples of Keywords
genotoxicity, micronuclei, DNA strand break, sister chromatid exchange, aneuploidy,
genomic instability, gene amplification, epigenomics, DNA methylation, DNA
methyltransferase, histone, DNA repair, base excision repair, nucleotide excision
repair, DNA mismatch repair
Chromosome or DNA
structure, function, repair, or
integrity
Gene expression and
transcription
individual genes, pathway-related genes, transcriptomics, epigenetics, transcription
factors, microRNAs, noncoding RNAs
Protein synthesis, folding,
function, transport,
localization, or degradation
proteomics, translation, ribosomes, chaperones, heat shock proteins, ubiquitin,
proteasome, ER stress, UPR, PERK
anabolic or catabolic pathways for lipids, carbohydrates, amino acids, nucleotides;
energy metabolism; biochemical pathways; metabolomics; lipidomics; enzyme or
coenzyme activity or function.
Metabolism
A-51
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Category
Examples of Keywords
ligand interactions with membrane, cytoplasmic and nuclear receptors (e.g., AHR,
ER, AR, CAR, RAR, neurotransmitter receptors, insulin receptor, G-protein coupled
receptors), tyrosine kinases, phosphatase, phospholipases, GTPase, second
messengers (calcium, diacylglycerol, ceramide, NO), signaling pathways (NF-RB.
MAPK/ERK, AKT, mTOR, IP3/DAG, cAMP-dependent, Wnt, (3-catenin, TGF(3,
etc.)
Cell signaling or signal
transduction pathway
Cell or organelle structure,
motility, integrity
membrane integrity, cell scaffolding, cytoskeleton, actin, microtubules, ER, Golgi,
mitochondria, lysosome, endosome, phagosome, nucleus, chemotaxis, atrophy,
hypertrophy
Extracellular matrix or
molecules
ECM proteins (collagens, elastins, fibronectins and laminins), proteoglycans, matrix
metalloproteinases (MMPs)
cell cycle (Gl, S, G2, M), cyclins, CDKs, p53, p27, Rb, E2F stem cell, progenitor,
apoptosis, Annexin V, TUNEL, necrosis, blebbing, pyknosis, Bax, Bcl-2,
hyperplasia, dysplasia
Cell growth, differentiation,
proliferation, or viability
Activation of intrinsic cell
defense molecules or
systems
cytokines, chemokines, caspases, MHC/HLA molecules, pattern recognition receptors
(PRRs), NLR, proteasomes, autophagy
reactive oxygen species (ROS), oxidative stress, hydroxyl radical, hydrogen peroxide,
reactive nitrogen species, superoxide anion, peroxyl radicals, antioxidant response,
catalase, superoxide dismutase, EROD, glutathione (GSH), GSH peroxidase,
glutathione-S-transferase, 8-OHdG
Oxidative stress
GnRH, CRF, ADH/vasopressin, FSH, LH, ACTH, GH, TH, TSH, PTH, Cortisol,
epinephrine/norepinephrine, melatonin, oxytocin, estrogen, testosterone, adiponectin,
leptin, insulin, glucagon
Hormone function
Biomarkers of cerebral
function
Apoptotic neurodegeneration protein markers, cerebral glucose metabolism, brain
glucose levels
Please provide specific details regarding reason for supplemental tag in the notes
section.
Other (provide details)
Notes: DNA = deoxyribonucleic acid; microRNA = micro ribonucleic acid; RNA = ribonucleic acid; ER = estrogen receptor;
UPR = unfolded protein response; PERK = protein kinase R-like endoplasmic reticulum kinase; AHR = aryl hydrocarbon
receptor; CAR = constitutive androstane receptor; RAR = retinoic acid receptor; GTPase = guanosine triphosphate; NO = nitric
oxide; NF-RB = nuclear factor kappa B; mTOR = rapamycin; DAG = diacylglycerol; TGFp = transforming growth factor beta;
ECM = extracellular matrix;; CDK = cyclin-dependent kinase; Bcl-2 = B-cell lymphoma 2; TUNET = terminal deoxynucleotidyl
transferase-mediated dUTP nick end labeling; MHC/NHLA = major histocompatibility complex/human leukocyte antigen;
NLR = nucleotide-binding oligomerization domain-like receptors; EROD = ethoxyresorufin-O-dealkylase; 8-OHdG = 8-
hydroxy-2' -deoxyguanosine; GnRH = gonadotropin-releasing hormone; CRF = corticotropin-releasing factor;
ADH = Antidiuretic hormone; FSH = follicle stimulating hormone; LH = luteinizing hormone; ACTH = adrenocorticotropic
hormone; GH = growth hormone; TH = thyroid hormone; PTH = parathyroid hormone.
A-52
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.1.6.5 Full-Text Screening Questions
All studies identified as PECO-relevant from title/abstract screening advanced to full-text screening—which was performed in
DistillerSR. Screeners reviewed each full study report and any supplemental study materials to respond to prompts pertaining to
PECO relevance, evidence stream, and health outcome(s), and whether PFOS and/or PFOA was evaluated (some screening efforts for
PFOA and PFOS were performed concurrently). Table A-13 below lists the prompts and response options that were used for full-text
screening.
Table A-13. DistillerSR Form for Full-Text Screening
Question/Prompt
Response Options
1 Source of study if not identified from database
search.
[Select one]
• Source other than HERO database search
2 Does the article meet PECO criteria?
[Select one]
• Yes
• No
• Tag as potentially relevant supplemental material
• Unclear
If "Yes" to Question #1:
3a If meets PECO, what type of evidence?
[Select all that apply]
• Human
• Animal (mammalian models)
• PBPK model
4a If meets PECO, which health outcome(s) apply?"
[Select all that apply]
• General toxicity, including body weight, mortality, and survival
• Cancer
• Cardiovascular, including serum lipids
• Endocrine (hormone)
• Gastrointestinal
• Genotoxicity
• Growth (early life) and developmental
• Hematological, including non-immune/hepatic/renal clinical chemistry measures
• Hepatic, including liver measures and serum markers (e.g., ALT, AST)
• Immune/inflammation
• Musculoskeletal
• Nervous system, including behavior and sensory function
• Nutrition and metabolic
• Ocular
A-53
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt Response Options
• PBPK or PK model
• Renal, including urinary measures (e.g., protein)
• Reproductive
• Respiratory
• Skin and connective tissue effects
• Dermal
• Unsure
• Other
If meets PECO and endocrine outcome, which endocrine tags apply?
[Select all that apply]
• Adrenal
• Sex hormones (e.g., androgen, estrogen, progesterone)
• Neuroendocrine
• Pituitary
• Steroidogenesis
• Thyroid
If "Unsure" or "Other" is selected for health outcome, write reasoning in the respective
free text-box.
[Free-text]
If"Tag as potentially relevant supplemental material" to Question #1:
3b
If supplemental, what type of information?b c
• Mechanistic
[Select all that apply]
• Non-mammalian model
• ADME/toxicokinetic
• Acute/short-term duration exposures'1
• Non-oral route of administration
• Exposure characteristics (no health outcome)
• Susceptible population (no health outcome)
• Environmental fate or occurrence (including food)
• Mixture study
• Case study or case series
• Other assessments or records with no original data (e.g., reviews, editorials, commentaries)
• Conference abstract
• Bioaccumulation data in fish
4b
If "Acute," which health outcome(s) apply?
[Select all that apply]
A-54
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
• General toxicity, including body weight, mortality, and survival
• Cancer
• Cardiovascular, including serum lipids
• Endocrine (hormone)
• Gastrointestinal
• Genotoxicity
• Growth (early life) and developmental
• Hematological, including non-immune/hepatic/renal clinical chemistry measures
• Hepatic, including liver measures and serum markers (e.g., ALT, AST)
• Immune/inflammation
• Musculoskeletal
• Nervous system, including behavior and sensory function
• Nutrition and metabolic
• Ocular
• PBPK or PK model
• Renal, including urinary measures (e.g., protein)
• Reproductive
• Respiratory
• Skin and connective tissue effects
• Dermal
• Unsure
If "Yes." "Tag as potentially relevant supplemental material."
" or "Unclear" to Question # 1:
5 Which PFAS did the study report?
• PFOA
[Select all that apply]
• PFOS
• Other PFAS
Notes: PBPK = physiologically-based pharmacokinetic; ALT = alanine transaminase; AST = aspartate aminotransferase; PK = pharmacokinetic; ADME = absorption, distribution,
metabolism, and/or excretion.
a Refer to list of health outcomes and examples in Appendix A. 1.6.5.1.
b Refer to list of supplemental tags in Appendix A. 1.6.4.1.
c Refer to list of mechanistic information in Appendix A. 1.6.4.2.
dRefer to definition of acute/short-term duration exposures in Appendix Error! Reference source not found..
A-55
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.6.5.1 Health Effect Categories and Example Outcomes for Epidemiological
Studies
The following health effects categories were considered throughout the full-text screening and
subsequent steps of the systematic review process for epidemiological studies (Table A-14.
Health Effect Categories Considered for Epidemiological StudiesTable A-14).
Table A-14. Health Effect Categories Considered for Epidemiological Studies
Health Effect Category
Example Health Outcomes
Notes
Cancer
Tumors
Precancerous lesions (e.g., dysplasia)
-
Cardiovascular
Serum lipids (e.g., cholesterol, LDL,
HDL, triglycerides)
Blood pressure
Hypertension
Atherosclerosis
Coronary heart disease
Other cardiovascular disease
Dermal
Skin sensitivity
-
Developmental
Birth size (birth weight; birth length; smallMarkers of development specific to other
for gestational age) systems are organ/system-specific (e.g.,
Preterm birth tests of sensory maturation are under
Sex ratio
Postnatal growth
Nervous System)
Pubertal development is under
Reproductive.
Endocrine
Thyroid hormones (e.g., T3, T4, TSH) Reproductive hormones (e.g., estrogen,
Thyroid weight and histopathology progesterone, testosterone) are under
Hormonal measures in any tissue or blood Reproductive.
(non-reproductive)
Gastrointestinal
Symptoms of the stomach and intestines
(e.g., diarrhea, nausea, vomiting,
abdominal pain and cramps)
Hematologic
Blood count
Red blood cells
White blood cell counts and globulin are
under Immune.
Blood Hematocrit or hemoglobin
Serum lipids are under Cardiovascular.
Corpuscular volume
Blood Platelets or reticulocytes
Blood biochemical measures (e.g.,
Serum liver markers are under Hepatic.
sodium, calcium, phosphorus)
Hepatic
Liver enzymes (e.g., ALT; AST; ALP)
Liver disease
Serum lipids are under Cardiovascular.
Biochemical markers, such as albumin,
Liver-specific serum biochemistry (e.g.,
albumin)
are under Hepatic. Liver tissue cytokines
are under Immune.
Globulin is under Immune.
Serum glucose is under Metabolic.
Immune
Asthma
Allergy
Atopic dermatitis/eczema
Vaccine response
Red blood cells are under Hematological.
Non-immune measures of pulmonary
function are under Respiratory.
Interleukin 6 (IL-6) is considered a
IgE
Mechanistic outcome.
Autoimmune or infectious disease
Hypersensitivity
General immune assays (e.g., white blood
cell counts)
A-56
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Health Effect Category
Example Health Outcomes
Notes
Immune responses in the respiratory
system
Stress-related factors in blood (e.g.,
glucocorticoids or other adrenal markers)
Metabolic
Obesity
BMI
Adiposity
Diabetes (including gestational diabetes)
Insulin resistance
Blood glucose
Waist circumference, ponderal index,
BMI SDS, BMI z-scores, are all included
here.
Gestational weight gain, adult weight
change also included here.
Musculoskeletal/Connective
Tissue
Bone health
Osteoporosis
Bone density
Nervous
Cognition
Behavior
Autism
Attention (ADHD)
Depression
Communication
Motor
Ocular
Vison changes
Eye irritation
-
Reproductive, female
Reproductive hormones
Breastfeeding
Fecundity
PCOS
Spontaneous abortion
Menopause
Endometriosis
Pubertal development
Menstrual cycle characteristics
Anogenital distance (females)
If data indicate altered birth parameters
are likely attributable to female fertility,
these data may be discussed under Female
Reproductive.
Reproductive, male
Reproductive hormones
Semen parameters
Sperm DNA damage
Pubertal development
Anogenital distance (males)
Respiratory
Non-immune measures of pulmonary
(lung) function (e.g., FEV1, FVC, lung
capacity)
Asthma, wheeze, lower/upper respiratory
trat infections are Immune.
Renal
GFR
Uric acid
Creatinine
Renal function
Urinary measures (e.g., protein; volume;
pH; specific gravity)
Other
Select this category if the outcome does
not fit in any of the above categories.
-
Notes: LDL = low-density lipoprotein; HDL = high-density lipoprotein; T3 = triiodothyronine; T4 = thyroxine; TSH = thyroid
stimulating hormone; ALT = alanine transaminase; AST = aspartate aminotransferase; ALP = alkaline phosphatase;
IgE = immunoglobulin E; BMI = body mass index; ADHD = attention deficit hyperactivity disorder; PCOS = polycystic ovary
syndrome; DNA = deoxyribonucleic acid; FEV1 = forced expiratory volume in one second; FVC = forced vital capacity;
GFR = glomerular filtration rate.
A-57
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1.6.6 Animal Toxicological Study Design Definitions
The following definitions were used throughout full-text screening and data extraction for animal
toxicological studies:
• Acute/short-term: Exposure duration between 1-28 days.
• Sub-chronic: Exposure duration between 28-90 days.
• Chronic: Exposure duration greater than 90 days.
• Developmental: Exposure occurs during gestation and dams are sacrificed prior to birth.
These studies are typically focused on the pups and evaluate viability, developmental
milestones, and other growth and developmental effects in pups.
• Reproductive: Exposure begins prior to mating and may continue through birth and, in
some cases, through a second generation. These studies will typically evaluate
reproductive outcomes in the dams (e.g., copulation and fertility indices, numbers of
corpora lutea and implantation sites, pre- and post-implantation loss).
A-58
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.6.7 ADME Screening and Light Data Extraction
All studies identified as containing ADME data during title/abstract or full-text screening were imported into litstream and underwent
additional screening. Studies that met certain criteria (e.g., PECO relevant and evaluated multiple timepoints, tissues, and/or dose
levels) and underwent light data extraction. For each study, at least two reviewers (one primary screener/extractor and one quality
assurance (QA) reviewer) reviewed the full study and any supplemental study materials to respond to prompts pertaining to key study
elements (e.g., tested species or population, tissues evaluated, dose levels tested, ADME endpoints measured). Table A-15 below
describes the prompts and response options that were used for ADME screening of epidemiological or animal toxicological studies.
Table A-15. litstream Form for ADME Screening and Light Data Extraction
Question/Prompt
Response Options
Suggested Considerations
1 General Questions
1.1
Does the article meet PECO
criteria?
[Select one]
• Yes
• No
• Use ADME-specific PECO statement (See PFOS
Main Document) and "Draft EPA IRIS Handbook:
Principles and Procedures for Integrated Risk
Information System (IRIS) Toxicological Reviews" to
inform the answer.
• Examples of exclusions may include abstract-only,
foreign language, secondary data sources, exposure
studies, physical-chemical properties, and species that
aren't relevant.
• If "No" is selected, do not move forward with the light
extraction. Finish filling out Section 1 - General
Questions (if applicable) and add a note in Section 5 -
Notes under "Notes from Initial Extractor to QA/QC
team" briefly explaining why the study does not meet
PECO.
1.2
What PFAS did the study report?
[Select all that apply]
• PFOA
• PFOS
-
1.3
Does this study contain multiple
time points, multiple tissues,
and/or multiple doses?
[Select one]
• Yes
• No
• If "No" is selected, do not move forward with the light
extraction. Finish filling out Section 1 - General
Questions (if applicable) and add a note in Section 5 -
Notes under "Notes from Initial Extractor to QA/QC
team" briefly explaining why the study meets PECO
but does not contain multiple time points, multiple
tissues, and/or multiple doses.
A-59
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
1.4
Does this study contain supporting
epidemiological information?
[Select one]
• Yes
• No
• Supporting epidemiological information includes
studies that compare PFAS levels in women of
different parity status or weeks of breast feeding as
well as studies that compare PFAS levels across
multiple age groups or multiple time points even if it
is not the same individuals who are being followed
over time (e.g., a cross-sectional study that enrolls
people of various ages and compares PFOS/PFOA
levels in a specific tissue in children vs. older adults).
1.5
Indicate if there is supplemental
data for this study.
[Select all that apply; Free-text]
• MO A/Mechanistic
• Exposure Study
• Use the free text field below to provide a brief
description of the type of MO A/mechanistic (refer to
Appendix A. 1.6.4.2 for examples) and/or exposure
information that is available.
• Examples of exposure information include studies of
PFAS levels in environmental media not directly
linked to human exposure (e.g., soil, sediment,
microbes, water [except drinking water], birds, or fish
[except those typically consumed by humans]).
1.6
Identify the species, system, or
model.
[Select all that apply]
• Human
• Non-human primate
• Rat
• Mouse
• Mammalian cells (in vitro studies)
• PBPK/TK models (or in silico studies)
• If a study only contains PBPK/TK models, do not
move forward with the light extraction. Finish filling
out Section 1 - General Questions (if applicable) and
add a note in Section 5 - Notes under "Notes from
Initial Extractor to QA/QC team" briefly describing
the model.
2
Human Studies Sub-Form
If the study docs not contain a human study, skip this section and move on to Section 3
- Animal Studies Sub-Form.
2.1
Population Name
[Free-Text]
• Name a population (e.g., Females - pregnant, PFOS)
• Separate populations should be made for each
chemical, population sex, life stage where ADME data
was collected, exposure route, etc. combination.
2.2
Select whether the study looks at
absorption, distribution,
metabolism, and/or excretion.
[Select all that apply]
• Absorption
• Distribution
• Metabolism
• Excretion
• Note: PFOA and PFOS are not metabolized so
"metabolism" is an unlikely selection.
2.3
List the specific ADME endpoints
addressed.
-
• List all the ADME endpoints analyzed for this
population.
A-60
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
[Free-text]
2.4
Exposure Category
Use the free text field if additional
information is needed (e.g., it is a
unique exposure, occupational
setting, etc.).
[Select one; Free-text]
• General environmental
• Poisoning
• Occupational
• Developmental
2.5
Identify the Exposure Route
[Select one; Free-text]
• Inhalation
• Oral
• Dermal
• Lactational transfer
• In utero/placental transfer
• Other (e.g., intraperitoneal, intramuscular, intranasal)
• If "other" option is selected, use the free text field to
describe exposure route.
• If the study population is exposed through more than
one route (e.g., oral and dermal), select one route from
the list and use the free text field to describe the other
exposure routes listed in the paper.
• If the study population is offspring that were exposed
"in utero/placental" AND by "lactational transfer",
select "in utero/placental" and use the free text field to
note that lactational transfer also occurred.
• If exposure route is unknown, select "other" option
and write in "Unknown" in the free text field.
• If the route is unspecified or multiple routes were
suspected based on the exposure vehicle, select
"other" and write in suspected exposure route in the
free text field.
2.6
What is the exposure vehicle?
[Select one]
• Drinking water
• Diet
• Breast milk
• In utero/placental transfer
• Occupational
• Unknown
• Other
• If "other" option is selected, use the free text field to
describe exposure vehicle.
• If the study population is offspring that were exposed
"in utero/placental" AND by "breast milk", select "in
utero/placental" and use the free text field to note that
lactational transfer also occurred via breast milk.
• If "occupational" option is selected, use the free text
field to describe exposure vehicle.
2.7
What is the sex of the population?
[Select one]
• Male
• Female
• Unspecified
• If results are given separately for each sex, separate
sub-forms should be used for each population.
2.8
Number of Subjects
-
• Example: Total number of subjects = 428
A-61
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt Response Options Suggested Considerations
Use the free text field to add
additional details on number of
subjects if they are broken up by
groups or quartiles.
[Free-text]
2.9 What is the life stage when the
ADME data was collected?
Use the free text field to provide
additional life stage notes.
[Select one; Free-text]
• Prenatal: conception to birth
• Infancy: 0-12 months
• Childhood: 13 months to 11 years
• Adolescence: 12 to 20 years
• Adult: 21 to 65 years
• Elderly: > 65 years
• If there is more than one life stage when ADME data
was collected, add an additional population in another
form.
2.10 Exposure Levels
Use the free text field to enter the
numeric exposure levels (if
known/estimated in an environmental
medium such as air, water, dust,
food, breast milk, etc.).
[Free-text]
• Do not report levels in serum or urine for this
question.
2.11 Exposure Units
Use the free text field to report the
exposure units as presented in the
paper.
[Free-text]
• Examples: mg/kg-d; mg/m3; ppm
• Use "Not Reported" if appropriate
2.12 Exposure Duration
Use the free text field to enter the
details of the exposure duration if
known.
[Free-text]
• Use abbreviations (h, d, wk, mon, y).
o Examples: 28 d; 13 wk; 2 y
• Use "Not Reported" if appropriate.
2.13 Time Points Analyzed
Use the free text field to enter the
time points data were analyzed.
[Free-text]
• Use abbreviations (h, d, wk, mon, y).
o Examples: 28 d; 13 wk; 2 y
• Use "Not Reported" if appropriate.
2.14 Measured Tissues
Use the free text field to enter the
tissues measured in the study (e.g.,
plasma, breast milk, cord blood).
[Free-text]
A-62
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
3 Animal Studies
IF the study docs not contain ;in animal study, skip (his section and move on to Section 4 - Mammalian Cells///? vitro.
3.1 Population Name
[Free-text]
• Name a population (e.g., Females dams, PFOS).
• Separate populations should be made for each
chemical, species, population sex, life stage where
ADME data was collected, exposure route, etc.
combination.
3 .2 Select whether the study looks at
• Absorption
• PFOA and PFOS are not metabolized, so
absorption, distribution,
• Distribution
"metabolism" is an unlikely selection.
metabolism, and/or excretion.
• Metabolism
[Select all that apply]
• Excretion
3.3 List the specific AD ME Endpoints
addressed.
Use the free text field below to list all
the ADME endpoints analyzed for
this population.
[Free-text]
3.4 Identify the Exposure Route
[Select one]
> Inhalation (nose only)
' Inhalation (whole head exposure)
' Inhalation (whole body exposure)
' Oral (diet)
' Oral (drinking water)
' Oral (gavage)
> Dermal
' Lactational transfer
' In utero/placental transfer
' Other (e.g., intraperitoneal, intramuscular,
intravenous, intranasal)
• If "other" option is selected, use the free text field
below to describe exposure route
• If the study population is offspring that were exposed
"in utero/placental" AND by "lactational transfer",
select "in utero/placental" and use the free text field to
note that lactational transfer also occurred
• If there is more than one exposure route identified, add
an additional population in another form.
3 .5 What is the exposure vehicle?
[Select one]
Diet
Water
Breast milk
In utero/placental transfer
Corn oil
Filtered air
Olive oil
Ethanol
• If "other" option is selected, use the free text field
below to describe exposure vehicle
• If the study population is offspring that were exposed
"in utero/placental" AND by "breast milk", select "in
utero/placental" and use the free text field to note that
lactational transfer also occurred via breast milk.
A-63
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
• DMSO
• Mineral oil
• Corn oil:acetone
• Other
3 .6 What is the strain?
Use the free text field to list the strain
(e.g., Sprague Dawley).
[Free-text]
• If there is more than one species studied, add an
additional population in another form.
3 .7 What is the sex?
[Select one]
• Male
• Female
• Male and Female
• If results are given separately for each sex, add an
additional population in another form.
3.8 What is the life stage when the
animal was dosed?
[Select all that apply]
• Prenatal
• Weaning
• Adolescent
• Adult
• Elderly
• Prenatal
o Non-human primates: conception to birth
o Rodents: GDO to birth
• Weaning
o Non-human primates: 1-130 days (0.35 years)
o Rodents: PND 1-21
• Adolescent
o Non-human primates: 130-1,825 days (0.35-
5 years)
o Rodents: 21-50 days (3-7 weeks)
• Adult
o Non-human primates: 5-35 years
o Rodents: > 50 days (> 7 weeks)
• Elderly
o Non-human primates: > 35 years
3 .9 What is the reported average age
of the animals when dosing began?
[Free-text]
• Use "Not Reported" if appropriate.
3.10 What is the average initial body
weight of the animals when dosing
began?
[Free-text]
• Use "Not Reported" if appropriate.
3 .11 What is the life stage when the
ADME data was collected?
[Select all that apply; Free-text]
• Prenatal
• Weaning
• Adolescent
• Prenatal
o Non-human primates: conception to birth
o Rodents: GD 0 to birth
A-64
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
• Adult
• Weaning
• Elderly
o Non-human primates: 1-130 days (0.35 years)
o Rodents: PND 1-21
• Adolescent
o Non-human primates: 130-1,825 days (0.35-
5 years)
o Rodents: 21-50 days (3-7 weeks)
• Adult
o Non-human primates: 5-35 years
o Rodents: >50 days (> 7 weeks)
• Elderly
o Non-human primates: > 35 years; use the free text
field to provide additional life stage notes.
• If there is more than one life stage when ADME data
were collected, add an additional population in another
form.
3 .12 What is the number of animals per -
• Example: Control = 10, low dose = 20, high
dosing group?
dose = 20; All groups = 20
Use the free text field to report the
• Use "Not Reported" if appropriate.
number of animals per dosing group.
[Free-text]
3 .13 Dose Levels
• Example: 0, 450, 900
Use the free text field to enter the
numeric dose levels.
[Free-text]
3 .14 Dose Units
• Examples: mg/kg-d; mg/m3; ppm
Use the free text field to report the
• Use "Not Reported" if appropriate.
dosage units as presented in the
paper.
[Free-text]
3 .15 Dose Duration -
• Use abbreviations (h, d, wk, mon, y).
Use the free text field to enter the
• For reproductive and developmental studies, where
details of the dose duration if known.
possible instead include abbreviated age descriptions
[Free-text]
such as "GD1-10" or "GD2-PND10"
o Examples: 14 d, 13 w (6 h/d x 5 d/wk); GD 2-
PND 10
• Use "Not Reported" if appropriate.
A-65
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
3 .16 Time Points Analyzed
-
• Use abbreviations (h, d, wk, mon, y)
Use the free text field to enter the
o Examples: 14 or 28 d; 13 wk; 2 y
time points data were analyzed.
• Use "Not Reported" if appropriate.
[Free-text]
3.17 Measured Tissues
-
-
Use the free text field to enter the
tissues measured in the study (e.g.,
plasma, liver, adipose).
[Free-text]
4
Mammalian Cells///? vitro
If the study docs not contain an in vitro component, skip this section and move on to Section 5
- Notes.
4.1
Population Name
-
• Name a population (e.g., Primary Human Hepatic,
[Free-text]
PFOA; A549, PFOS)
• Separate populations should be made for each
chemical, population sex, life stage where ADME data
was collected, exposure route, etc. combination. Use
the "Clone" button to copy forms/information for
easier extraction if the study populations are similar.
4.2
Select whether the study looks at
• Absorption
• PFOA and PFOS are not metabolized so "metabolism"
absorption, distribution,
• Distribution
is an unlikely selection.
metabolism, and/or excretion.
• Metabolism
[Select all that apply]
• Excretion
4.3
List the specific ADME Endpoints
-
-
addressed.
Use the free text field below to list all
the ADME endpoints analyzed for
this population.
[Free-text]
4.4
Does the study present data on
• Yes
• If "Yes" option is selected, use the free text field to list
protein binding?
• No
the binding proteins.
[Select one; Free-text]
4.5
Does the study present data on
• Yes
• If "Yes" option is selected, use the free text field to list
active transport?
• No
the transporters.
[Select one; Free-text]
4.6
Cell Line Name or Tissue Source
-
• Examples: A549; liver tissue from adult Sprague
Dawley female rats
A-66
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
Use the free text field to list the cell
line name or tissue source the cells
were derived from.
[Free-text]
• If there is more than one cell line name or tissue
source studied, add an additional population in another
form.
4.7 In vitro System
[Select one; Free-text]
• Mammalian cells
• Cell-free system
• In silico system
• Other
• If "other" option is selected, use the free text field
below to describe the in vitro system.
• If there is more than one in vitro source studied, add
an additional population in another form.
4.8 Select all study design elements
that apply.
[Select all that apply]
• Multiple time points
• Multiple cell/tissue types
• Multiple dose levels
4.9 Exposure Design - -
Use the free text field to describe the
exposure design, be as succinct as
possible.
[Free-text]
4.10 What is the exposure vehicle?
Use the free text field to describe the
exposure vehicle, be as succinct as
possible
[Free-text]
4.11 Dose Levels
Use the free text field to enter the
numeric dose levels.
[Free-text]
• Example: 0, 450, 900
4.12 Dose Units
Use the free text field to report the
dosage units as presented in the
paper.
[Free-text]
• Examples: ppm; mg/mL
• Use "Not Reported" if appropriate.
4.13 Dose Duration
Use the free text field to enter the
details of the exposure duration.
[Free-text]
• Use abbreviations (h, d, wk, mon, y)
o Examples: 28 d; 13 wk; 2 y
• Use "Not Reported" if appropriate.
4.14 Time Points Analyzed
Use the free text field to enter the
time points data were analyzed.
• Use abbreviations (h, d, wk, mon, y)
o Examples: 28 d; 13 wk; 2 y
A-67
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
[Free-text]
• Use "Not Reported" if appropriate.
5 • Notes
5.1 General Study Notes
[Free-text]
Use the free text field to add any
general study notes not captured
above that may be of interest to the
QC reviewer or PBPK modelers
• Please indicate whether the study contains information
on PFOA/PFOS that is broken up by linear/branched
isomers. Use the following phrase: "Contains
linear/branched isomer information"
5.2 Notes from Initial Extractor to - -
QA/QC Team
Use the free text field to add any
general study notes not captured
above that may be of interest to the
QC reviewer.
[Free-text]
5.3 Notes from QA/QC Team
Use the free text field to add any
general study notes not captured
above that may be of interest to the
PBPK modelers.
[Free-text]
Notes: ADME = absorption, distribution, metabolism, and/or excretion; QA/QC = quality assurance/quality control; MOA = mode of action; PBPK = physiologically-based
pharmacokinetic; TK = toxicokinetic; GD = gestational day; PND = postnatal day; ppm = parts per million.
A-68
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.6.8 Mechanistic Screening and Light Data Extraction
All studies identified as mechanistic in title/abstract or full-text screening were imported into litstream and underwent additional
screening. Studies that were confirmed to be PECO relevant underwent light data extraction. For each study, at least two reviewers
(one primary screener/extractor and one QA reviewer) reviewed the full study and any study materials to respond to prompts
pertaining to key study elements (e.g., tested species or population, mechanistic endpoint(s) evaluated, lifestage(s) at which
evaluations were performed). Table A-16 below describes the prompts and response options that were used for studies with
mechanistic evidence.
Table A-16. litstream Form for Mechanistic Screening and Light Data Extraction
Question
Options
Suggested Considerations
1 General Questions
1.1
Does the article meet PECO
criteria?
[Select one]
• Yes
• No
1.2
What PFAS did the study report?
[Select all that apply]
• PFOA
• PFOS
-
1.3
Publication Type
[Select one]
• Primary research
• Review article
-
1.4
Indicate if there is hazard ID or
supplemental data for this study.
[Select all that apply; Free-text]
• Animal tox
• Epi
• ADME
• Use free text field to provide an explanation.
2
Human Studies Sub-Form
If the study docs not contain a human study, skip this section and move on to Section 3
- Animal Studies Sub-Form.
2.1
Population/Study Group Name
[Free-text]
-
-
2.2
Exposure Category
[Select one; Free-text]
• General environmental
• Poisoning
• Occupational
• Developmental
• Controlled experimental
• Free text field if additional information is needed.
2.3
Identify the Exposure Route
[Select all that apply]
• Inhalation
• Oral
• Dermal
• Lactational transfer
• Free text field to elaborate on "other" and "unknown"
options.
A-69
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question
Options
Suggested Considerations
• In utero/placental transfer
• Other (e.g., intraperitoneal, intramuscular, intranasal)
• Unknown
2.4
What is the exposure vehicle?
• Drinking water
• Free text field to elaborate on "other" and "unknown"
[Select one]
• Diet
options.
• Breast milk
• In utero/placental transfer
• Occupational
• Unknown
• Other
2.5
What is the life stage when the
• Prenatal
• Free text for life stage notes.
mechanistic data was collected?
• Infancy
[Select one; Free-text]
• Childhood
• Adolescence
• Adult
• Elderly
2.6
What is the corresponding health
• Cancer
• Free field for "other" option, includes endpoints that
outcome system?
• Cardiovascular
do not fit neatly into any one health outcome system.
[Select one]
• Dental
• Dermal
• Developmental
• Endocrine
• Gastrointestinal
• Hematologic
• Hepatic
• Immune
• Lymphatic
• Metabolic
• Musculoskeletal/connective tissue
• Nervous
• Ocular
• Renal
• Reproductive
• Respiratory
• Systemic/whole body
• Other
A-70
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question
Options
Suggested Considerations
2.7
Mechanistic Category
[Select all that apply; Free-text]
• Epigenetics
• Chromosome/DNA structure, function, repair or
integrity
• Gene expression and transcription
• Protein expression, synthesis, folding, function,
transport, localization, or degradation
• Metabolomics
• Cell or organelle structure, motility, or integrity
• Structure, Morphology, or Morphometry
• Other
• Free text field for "other" option.
2.8
Mechanistic Pathway
[Select all that apply; Free-text]
• Angiogenic, antiangiogenic, vascular tissue
remodeling
• Atherogenesis and clot formation
• Big data, non-targeted analysis
• Cell growth, differentiation, proliferation, or viability
• Cell signaling or signal transduction
• Extracellular matrix or molecules; Fatty acid
synthesis, metabolism, storage, transport, binding, (3-
oxidation
• Hormone function
• Inflammation and Immune Response
• Oxidative stress
• Renal dysfunction
• Vasoconstriction/vasodilation
• Xenobiotic metabolism
• Other
• Free text field for "other" option.
2.9
Mechanistic Endpoints
[Free-text]
-
• Free text field to list mechanistic endpoints.
3
Animal Studies Sub-Form
• If llic study docs not contain an animal study, skip this section and move on to Section 4 - In vitro Sub-Form.
3.1
Population/Study Group Name
[Free-text]
-
-
3.2
What is the species?
[Select one; Free-text]
• Non-human primate
• Zebrafish
• Rat
• Mouse
• Free text field to list species for "other rodent model"
option.
A-71
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question
Options
Suggested Considerations
• Rabbit
• Guinea pig
• Other rodent model
3.3
What is the strain?
[Free-text]
-
-
3.4
Identify the Exposure Route
[Select one]
• Inhalation (nose only)
• Inhalation (whole head exposure)
• Inhalation (whole body exposure)
• Oral (diet)
• Oral (drinking water)
• Oral (gavage)
• Dermal
• Lactational transfer
• In utero/placental transfer
• Other (e.g., intraperitoneal, intramuscular,
intravenous, intranasal)
• Free text field for "other" option.
3.5
What is the exposure vehicle?
[Select one]
• Diet
• Water
• Breast milk
• In utero/placental transfer
• Corn oil
• Filtered air
• Olive oil
• Ethanol
• DMSO
• Mineral oil
• Corn oil: acetone
• Other
• Free text field for other "other" option.
3.6
What is the life stage when the
animal was dosed?
[Select one; Free-text]
• Prenatal
• Weaning
• Adolescent
• Adult
• Elderly
• Free text field for life stage notes.
3.7
What is the life stage when the
• Prenatal
• Free text field for life stage notes.
mechanistic data was collected?
• Weaning
[Select one; Free-text]
A-72
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question Options Suggested Considerations
• Adolescent
• Adult
• Elderly
3 .8 What is the corresponding health
• Cancer
• Free text field for
'other'
option, includes endpoints
outcome system?
• Cardiovascular
that do not fit neatly into any one health outcome
[Select all that apply; Free-text]
• Dental
• Dermal
• Developmental
• Endocrine
• Gastrointestinal
• Hematologic
• Hepatic
• Immune
• Lymphatic
• Metabolic
• Musculoskeletal/connective tissue
• Nervous
• Ocular
• Renal
• Reproductive
• Respiratory
• Systemic/whole body
• Other
system.
3.9 Mechanistic Category
• Epigenetics chromosome/DNA structure, function,
• Free text field for
'other'
option.
[Select all that apply; Free-text]
repair, or integrity
• Gene expression and transcription
• Protein expression, synthesis, folding, function,
transport, localization, or degradation
• Metabolomics
• Cell or organelle structure, motility, or integrity
• Structure, Morphology, or Morphometry
• Other
3.10 Mechanistic Pathway
• Angiogenic, antiangiogenic, vascular tissue
• Free text field for
'other'
option.
[Select all that apply; Free-text]
remodeling
• Atherogenesis and clot formation
• Big data, non-targeted analysis
A-73
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question Options Suggested Considerations
• Cell growth, differentiation, proliferation, or viability
• Cell signaling or signal transduction
• Extracellular matrix or molecules
• Fatty acid synthesis, metabolism, storage, transport,
binding, (3-oxidation
• Hormone function
• Inflammation and Immune Response
• Oxidative stress
• Renal dysfunction
• Vasoconstriction/vasodilation
• Xenobiotic metabolism
• Other
3.11 Mechanistic Endpoints
-
• Free text field to list mechanistic endpoints
[Free-text]
4
In vitro Sub-Form
If the study docs not contain an in vitro component, skip this section and move on to Section 5
- Notes.
4.1
Population/Study Group Name
[Free-text]
-
-
4.2
Does the study present data on
protein binding?
[Select one; Free-text]
• Yes
• No
• Free text field if "Yes" to list binding proteins.
4.3
Does the study present data on
active transport?
[Select one; Free-text]
• Yes
• No
• Free text field if "Yes" to list transporters.
4.4
In vitro System
[Select one; Free-text]
• Mammalian cells
• Cell-free system
• In silico system
• Other
• Free text field for "other" option.
4.5
If a cellular model is used, is it a
cell line or primary cells?
[Select one]
• Cell line
• Primary cell
4.6
Cell Or Tissue Source for In
vitro/Ex Vivo Studies
[Select one; Free-text]
• Human
• Zebrafish
• Non-human primate
• Rat
• Mouse
• Free text field to list "other rodent model" option.
A-74
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question
Options
Suggested Considerations
• Rabbit
• Guinea pig
• Other rodent model
4.7
What is the corresponding health
outcome system?
[Select all that apply; Free-text]
• Cancer
• Cardiovascular
• Dental
• Dermal
• Developmental
• Endocrine
• Gastrointestinal
• Hematologic
• Hepatic
• Immune
• Lymphatic
• Metabolic
• Musculoskeletal/connective tissue
• Nervous
• Ocular
• Renal
• Reproductive
• Respiratory
• Systemic/whole body
• Other
• Free text field for "other" option, includes endpoints
that do not fit neatly into any one health outcome
system.
4.8
Mechanistic Category
[Select all that apply; Free-text]
• Epigenetics chromosome/DNA structure, function,
repair, or integrity
• Gene expression and transcription
• Protein expression, synthesis, folding, function,
transport, localization, or degradation
• Metabolomics
• Cell or organelle structure, motility, or integrity
• Structure, morphology, or morphometry
• Other
• Free text field for "other" option.
4.9
Mechanistic Pathway
[Select all that apply; Free-text]
• Angiogenic, antiangiogenic, vascular tissue
remodeling
• Atherogenesis and clot formation
• Big data, non-targeted analysis
• Free text field for "other" option.
A-75
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question
Options
Suggested Considerations
' Cell growth, differentiation, proliferation, or viability
' Cell signaling or signal transduction
1 Extracellular matrix or molecules
' Fatty acid synthesis, metabolism, storage, transport,
binding, (3-oxidation
> Hormone function
> Inflammation and immune response
' Oxidative stress
' Renal dysfunction
' Vasoconstriction/vasodilation
' Xenobiotic metabolism
> Other
4.10 Mechanistic Endpoints
[Free-text]
5 • Notes
5.1 General Study Notes
Use the free text field to add any
general study notes not captured
above that may be of interest to the
QC reviewer or PBPK modelers.
[Free-text]
5.2 Notes from Initial Extractor to
QA/QC Team
Use the free text field to add any
general study notes not captured
above that may be of interest to the
QC reviewer.
[Free-text]
5.3 Notes from QA/QC Team
Use the free text field to add any
general study notes not captured
above that may be of interest to the
PBPK modelers.
[Free-text]
Notes: ADME = absorption, distribution, metabolism, and/or excretion; DNA = deoxyribonucleic acid; DMSO = dimethyl sulfoxide, PBPK = physiologically-based
pharmacokinetic; QA/QC = quality assurance/quality control.
• Please indicate whether the study contains information
on PFOA/PFOS that is broken up by linear/branched
isomers. Use the following phrase: "Contains
linear/branched isomer information"
A-76
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.1.7 Study Quality Evaluation Overview
After literature search results were screened and inventoried, epidemiological and animal
toxicological studies that met PECO criteria underwent study quality evaluation to assess each
study's validity and utility. As outlined in the IRIS Handbook {U.S. EPA, 2022, 10476098}, the
key concerns during the review of epidemiological and animal toxicological studies are potential
bias (factors that affect the magnitude or direction of an effect in either direction) and
insensitivity (factors that limit the ability of a study to detect a true effect; low sensitivity is a
bias toward the null when an effect exists). Study quality evaluations produce overall judgments
about confidence in the reliability of study results. The general approach for study quality
evaluation is outlined in Figure A-1, which has been adapted from Figure 4-1 in the IRIS
Handbook {U.S. EPA, 2022, 10476098} (previously Figure 6-1 in the draft IRIS Handbook
{U.S. EPA, 2020, 7006986}). Study quality evaluations were performed using the structured
platform for study evaluation housed within EPA's Health Assessment Workplace Collaborative
(FIAWC).
(b)
(a)
Develop assessment-
specific considerations
Pilot testing (and possible
refinement)
Independent evaluation
by two reviewers
Conflict resolution
Finalization of domain
judgements and overall
ratines
Study Quality Evaluation Domains
Animal Toxicological Studies
Epidemiological Studies
Reporting quality ¦ Allocation ¦ Observational bias'blinding
Confounding/variable control ¦ Selective reporting and attrition
Chemical administration and characterization
Exposure timing, frequency, & duration
Endpoint sensitivity and specificity ¦ Results presentation
Exposure measurement ¦ Outcome ascertainment
Participant selection ¦ Potential confounding
Analysis ¦ Sensitivity ¦ Selective reporting
Domain Scores
^ Good (Appropriate study conduct relating to the domain; minor deficiencies not expected to influence results)
Adequate (Some limitations relating to the domain, but not likely to be severe or to have a notable impact on results)
Deficient (Identified biases or deficiencies interpreted as likely to have had a notable impact on die result or prevent
reliable interpretation of study findings)
^ Critically Deficient (Serious flaws that make observed effects uninterpre table)
Overall Confidence Ratings
High (No notable deficiencies or concerns identified: potential for bias unlikely or minimal: sensitive methodology)
Medium (Possible deficiencies or concerns noted, but resulting bias or lack of sensitivity is unlikely to be of a notable degree)
Low (Deficiencies or concerns were noted, and the potential for substantive bias or inadequate sensitivity could have a
significant impact on the study resuks or their interpretation)
Unin form a rive (Serious flaws make study results unusable for hazard identification or dose-response)
Figure A-1. Overview of Study Quality Evaluation Approach
(a) An overview of the study quality evaluation process; (b) Evaluation domains and ratings definitions (i.e., domain scores and
overall confidence ratings, performed on an outcome-specific basis as applicable).
The overall aims of study quality evaluation are the same for both epidemiological and animal
toxicological studies, but some aspects of the approaches are different. Therefore, study quality
evaluation procedures for epidemiological and animal toxicological studies are described
separately in the following sections. In brief, at least two primary reviewers independently
judged the reliability of the study results according to multiple study quality evaluation domains
presented in the IRIS Handbook. Domain-specific core and prompting questions are provided to
guide the reviewer in assessing different aspects of study design and conduct related to reporting,
risk of bias, and study sensitivity. For each domain, each reviewer assigned a rating of good,
A-77
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
adequate, deficient (or "not reported," which carried the same functional interpretation as
deficient), or critically deficient (see Figure A-l and Figure A-2). A QA reviewer (in accordance
with protocols outlined in the IRIS Handbook) engaged in conflict resolution with the two
independent reviewers as needed and made a final determination (reflected as study confidence
ratings; see Figure A-l and Figure A-3) regarding each health outcome or outcome grouping of
interest; thus, different "judgments" were possible for different health outcomes within the same
study. The overall confidence rating should, to the extent possible, reflect interpretations of the
potential influence on the results (including the direction and/or magnitude of influence) across
all domains. The rationale supporting the overall confidence rating is documented clearly and
consistently and includes a brief description of any important study strengths and/or limitations
and their potential impact on the overall confidence.
The specific study limitations identified during study quality evaluation were carried forward to
inform the synthesis of findings within each body of evidence for a given health effect (i.e.,
study confidence determinations were not used to inform "judgments" in isolation).
Studies containing mechanistic or ADME data did not undergo study quality evaluation as study
quality domains for these types of studies are not currently available in HAWC.
A-78
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Good
Intended to represent a judgment that there was appropriate study conduct relating to the
domain (as defined by consideration of the criteria listed below), and any minor deficiencies
that were noted would not be expected to influence interpretation of the study findings.
Adequate
Indicates a judgment that there were study design limitations relating to the domain (as
defined by consideration of the criteria listed below), but that those limitations are not likely
to be severe and are expected to have minimal impact on interpretation of the study findings.
Deficient
Denotes identified biases or limitations that are interpreted as likely to have had a substantial
impact on the results or that prevent reliable interpretation of the study findings.
Note: Not reported indicates that the information necessary to evaluate the domain was not
available in the study. Generally, this term carries the same functional interpretation as
Deficient for the purposes of the study confidence classification.
Critically Deficient
Reflects a judgment that the study design limitations relating to the domain introduced a flaw
so serious that the study should not be used without exceptional justification (e.g., it is the
only study of its kind and may highlight possible research gaps). This judgment should only
be used if there is an interpretation that the limitations) would be the primary driver of any
observed effect(s), or if it makes the study findings uninterpretable.
Figure A-2. Possible Domain Scores for Study Quality Evaluation
High Confidence
No notable concerns were identified (e.g., most or all domains rated Good).
Medium Confidence
Some concerns are identified but expected to have minimal impact on the interpretation of
the results (e.g., most domains rated Adequate or Good; may include studies with
Deficient ratings if concerns are not expected to strongly impact the magnitude or
direction of the results). Any important concerns should be carried forward to evidence
synthesis.
Low Confidence
Identified concerns are expected to significantly impact the study results or their
interpretation (e.g., generally. Deficient ratings for one or more domains). The concerns
leading to this confidence judgment must be carried forward to evidence synthesis.
Uninformative
Serious flaw(s) make the study results unusable for informing hazard identification (e.g.,
generally. Critically Deficient rating in any domain; many Deficient ratings).
Uninformative studies are not considered further in the synthesis and integration of
evidence.
Figure A-3. Overall Study Confidence Classifications
A. 1.7.1 Study Quality Evoluotion for Epidemiological Studies
Study quality evaluation domains for assessing risk of bias and sensitivity in epidemiology
studies of health effects are: exposure measurement, outcome ascertainment, participant
selection, potential confounding, analysis, study sensitivity, and selective reporting. As noted in
the IRIS Handbook, this framework is adapted from the Risk Of Bias in Nonrandomized Studies
of Interventions (ROBINS-I) tool (https://methods.cochrane.org/methods-cochrane/robins-i-
tool), modified by IRIS for use with the types of studies more typically encountered in EPA's
work. As outlined in Section A. 1.7 of this appendix, study quality evaluations are performed for
a set of established domains, and core and prompting questions are provided for each domain to
guide the reviewer. Each domain is assigned a score of Good, Adequate, Deficient, Not
A-79
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reported or Critically Deficient, and rationales to support the scores are developed. Once all
domains are evaluated, a confidence rating of High, Medium, or Low confidence or
Uninformative is assigned.
The tables presented in the following sections describe the epidemiological study quality
evaluation domains and the prompting questions and considerations for assessing study quality in
relation to each domain.
A-80
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.7.1.1 Participant Selection
The aim of study quality evaluation for this domain is to ascertain whether the reported information indicates that selection in or out of
the study (or analysis sample) and participation was not likely to be biased (i.e., the exposure-outcome distribution of the participants
is likely representative of the exposure-outcome distribution in the overall population of eligible persons) (Table A-17).
Table A-17. Study Quality Evaluation Considerations for Participant Selection
Core Question: Is there evidence that selection into or out of the study (or analysis sample) was jointly related to exposure and to outcome?
Prompting Questions
Follow-Up Questions
Suggested Considerations
For longitudinal cohort:
Did participants volunteer for the cohort based on
knowledge of exposure and/or preclinical disease
symptoms? Was entry into the cohort or
continuation in the cohort related to exposure and
outcome?
For occupational cohort:
Did entry into the cohort begin with the start of
the exposure?
Was follow-up or outcome assessment
incomplete, and if so, was follow-up related to
both exposure and outcome status?
Could exposure produce symptoms that would
result in a change in work assignment/work status
("healthy worker survivor effect")?
For case-control study:
Were controls representative of population and
time periods from which cases were drawn?
Were differences in
participant enrollment and
follow-up evaluated to
assess the potential for bias?
If there is a concern about
the potential for bias, what
is the predicted direction or
distortion of the bias on the
effect estimate (if there is
enough information)?
Were appropriate analyses
performed to address
changing exposures over
time in relation to
symptoms?
Is there a comparison of
participants and
• Minimal concern for selection bias based on description of
recruitment process (e.g., selection of comparison
population, population-based random sample selection,
recruitment from sampling frame including current and
previous employees) such that study participants were
unlikely to differ from a larger cohort based on
recruitment or enrollment methods (or data provided to
confirm a lack of difference)
• Exclusion and inclusion criteria specified and would not
be likely to induce bias.
• Participation rate is reported at all steps of study (e.g.,
initial enrollment, follow-up, selection into analysis
sample). If rate is not high, there is appropriate rationale
for why it is unlikely to be related to exposure (e.g.,
comparison between participants and nonparticipants or
other available information indicates differential selection
is not likely).
• Comparison groups are similar with respect to factors
expected to influence exposure-outcome relationship
(confounders. effect measure modifiers).
A-81
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Is there evidence that selection into or out of the study (or analysis sample) was jointly related to exposure and to outcome?
Are hospital controls selected from a group
whose reason for admission is independent of
exposure?
Could recruitment strategies, eligibility criteria,
or participation rates result in differential
participation relating to both disease and
exposure?
For population based-survey:
Was recruitment based on advertisement to
people with knowledge of exposure, outcome,
and hypothesis?
nonparticipants to address
whether differential
selection is likely?
Adequate
Deficient
Critically
Deficient
» Enough of a description of the recruitment process (i.e.,
recruitment strategy, participant selection or case
ascertainment) to be comfortable that there is no serious
risk of bias.
»Inclusion and exclusion criteria specified and would not
induce bias.
» Participation rate is incompletely reported for some steps
of the study, but available information indicates
participation is unlikely to be related to exposure.
» Comparison groups are largely similar with respect to
factors expected to influence exposure-outcome
relationship (confounders, effect measure modifiers) or
these are mostly accounted for in the study analysis.
> Little information on recruitment process, selection
strategy, sampling framework and/or participation OR
aspects of these processes raises the likelihood of bias
(e.g., healthy worker effect, survivor bias).
Example: Enrollment of "cases" from a specific clinic
setting (e.g., diagnosed autism), which could be biased by
referral practices and services availability, without
consideration of similar selection forces affecting
recruitment of controls.
' Aspects of the processes for recruitment, selection
strategy, sampling framework, or participation result in
concern that the likelihood of selection bias is high
(e.g., convenience sample with no information about
recruitment and selection, cases and controls are recruited
from different sources with different likelihood of
exposure, recruitment materials stated outcome of interest
and potential participants are aware of or are concerned
about specific exposures).
> Convenience sample, and recruitment and selection not
described.
> Case report, case series, or other study designs lacking a
comparison group (these should be excluded if they do not
meet assessment PECO criteria).
A-82
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l.7.1.2 Exposure Measurement
This domain may need to be evaluated multiple times for a single study if more than one measurement of exposure is assessed.
Therefore, different sets of criteria may be applied for different exposure assessments in the same study. Table A-18 outlines criteria
that apply across exposure assessments (first row), and specific additional criteria for specific types of exposure assessments (e.g.,
biomarkers, occupational) in subsequent rows.
Table A-18. Study Quality Evaluation Considerations for Exposure Measurement
Core Question: Does the exposure measure reliably distinguish between levels of exposure in a time window considered most relevant for a causal
effect with respect to the development of the outcome?
Prompting Questions
Follow-Up Questions
Suggested Considerations
Does the exposure measure capture the variability
in exposure among the participants, considering
intensity, frequency, and duration of exposure?
Is the degree of exposure
misclassification likely to
vary by exposure level?
Does the exposure measure reflect a relevant time If the correlation between
window? If not, can the relationship between
measures in this time and the relevant time
window be estimated reliably?
Was the exposure measurement likely to be
affected by a knowledge of the outcome?
Was the exposure measurement likely to be
affected by the presence of the outcome (i.e.,
reverse causality)?
Good
exposure measurements is
of concern, is there an
adequate statistical
approach to ameliorate
variability in
measurements?
If there is a concern about
the potential for bias, what
is the predicted direction or
distortion of the bias on the
effect estimate (if there is
enough information)?
Adequate
Deficient
Critically
Deficient
> Valid exposure assessment methods used, which represent
the etiologically relevant time period for reported effects
(e.g., exposure during a critical developmental window or
exposure preceding the evaluation of the outcome).
1 Exposure misclassification is expected to be minimal.
> Valid exposure assessment methods used, which represent
the etiologically relevant time period of interest.
> Exposure misclassification may exist but is not expected
to greatly impact the effect estimate.
> Specific knowledge about the exposure and outcome raise
concerns about reverse causality, but there is uncertainty
whether it is influencing the effect estimate.
> Exposed groups are expected to contain a notable
proportion of unexposed or minimally exposed
individuals, the method did not capture important
temporal or spatial variation, or there is other evidence of
exposure misclassification that would be expected to
notably change the effect estimate.
» Exposure measurement does not characterize the
etiologically relevant time period of exposure or is not
valid.
» There is evidence that reverse causality is very likely to
account for the observed association.
» Exposure measurement was not independent of outcome
status.
A-83
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Does the exposure measure reliably distinguish between levels of exposure in a time window considered most relevant for a causal
effect with respect to the development of the outcome?
Additional prompting questions for biomarkers Additional suggested considerations for biomarkers of exposure (should be
of exposure: evaluated in addition to the general considerations above):
Is a standard assay used? What are the intra- and
inter-assay coefficients of variation? Is the assay
likely to be affected by contamination? Are values
less than the limit of detection dealt with
adequately?
Good
• Use of appropriate analytic method such as [specific gold
standard exposure assessment method for the exposure of
interest].
Adequate
• Use of appropriate (but not gold standard) analytic
What exposure time period is reflected by the
method.
biomarker? If the half-life is short, what is the
correlation between serial measurements of
exposure?
Deficient
• Did not identify analytical methods used to measure
exposure.
• Failure to report LOD, percentage less than LOD, and
methods used to account for values below the LOD.
• Failure to report QA/QC measures and results.
Critically
Deficient
• Use of inappropriate analytical method or use of an
appropriate method with measurement issues that are
likely to impact the interpretation of results.
Additional prompting questions for case-control Additional suggested considerations for occupational exposures (should be
studies of occupational exposures: evaluated in addition to the general considerations above):
Is exposure based on a comprehensive job history
describing tasks, setting, time period, and use of
specific materials?
Good
• Describes the use of personal protective equipment.
• Confirmed contrast in exposure between groups using
biomarker measurements.
• Expert assessment method based on a detailed lifetime
occupational history and using a high-qualify, validated
job exposure matrix (JEM) or a JEM that incorporates
industry, time period, population/country, tasks, and
material used.
Adequate
• Describes the use of personal protective equipment.
• Confirmed contrast in exposure between groups using
biomarker measurements.
Deficient
• Expert assessment method based on incomplete
occupational history information (lacking job titles,
employers, industries, start and finish years, number of
hours worked per day, number of days worked per week.
A-84
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Does the exposure measure reliably distinguish between levels of exposure in a time window considered most relevant for a causal
effect with respect to the development of the outcome?
tasks performed, or materials used) - may be Critically
Deficient, depending on severity of this limitation.
Critically
Deficient
• JEM with data indicating it cannot differentiate between
exposure levels over time, area, or between individuals.
A-85
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l.7.1.3 PFAS-Specific Exposure Measurement Study Quality Evaluation
Criteria
Standard analytical methods of individual PFAS in serum or whole blood using quantitative
techniques, such as liquid chromatography triple quadrupole mass spectrometry, are considered
well-established methods (Table A-19).
Table A-19. Criteria for Evaluating Exposure Measurement in Epidemiology Studies of
PFAS and Health Effects
Rating
Criteria
Good
Adequate
Deficient
Critically
Deficient
• Evidence that exposure was consistently assessed using well-established analytical methods
that directly measure exposure (e.g., measurement of PFAS in blood, serum, orplasma).
OR
• Exposure was assessed using less established methods (e.g., measurement of PFAS in breast
milk) or methods that indirectly measure exposure (e.g., drinking water concentrations and
residential location/history, questionnaire or occupational exposure assessment by a certified
industrial hygienist) that are supported by well-established methods (i.e., inter-methods
validation: one method vs. another) in the target population of interest.
And all the following:
• Exposure was assessed in a relevant time-window (i.e., temporality is established, and
sufficient latency occurred prior to disease onset) for development of the outcome based on
current biological understanding.
• There is evidence that sufficient exposure data measurements are above the limit of
quantification for the assay.
• The laboratory analysis included data on standard quality control measures with demonstrated
precision and accuracy.
• Exposure was assessed using less established methods or indirect measures that are validated
but not in the target population of interest.
OR
• Evidence that exposure was consistently assessed using methods described in Good, but there
were some concerns about quality control measures or other potential for non-differential
misclassification.
And all the following:
• Exposure was assessed in a relevant time-window for development of the outcome.
• There is evidence that sufficient exposure data measurements are above the limit of
quantification for the assay.
• The laboratory analysis included some data on standard quality control measures with
demonstrated precision and accuracy.
Any of the following:
• Some concern, but no direct evidence, that the exposure was assessed using methods that have
not been validated or empirically shown to be consistent with methods that directly measure
exposure.
• Exposure was assessed in a relevant time window(s) for development of the outcome, but
there could be some concern about the potential for bias due to reverse causality3 between
exposure and outcome, yet no direct evidence that it is present; or has somehow been
mitigated by the design, etc.
| Any of the following:
» Exposure was assessed in a time window that is unknown or not relevant for development of
the outcome. This could be due to clear evidence of bias from reverse causality between
exposure and outcome, or other concerns such as the lack of temporal ordering of exposure
and disease onset, insufficient latency, or having exposure measurements that are not reliable
measures of exposure during the etiologic window(s).
A-86
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Rating Criteria
¦ • Direct evidence that bias was likely because the exposure was assessed using methods with
poor validity.
• Evidence of differential exposure inisclassification (e.g., differential recall of self-reported
exposure).
• There is evidence that an insufficient number of the exposure data measurements were above
the limit of quantification for the assay.
Notes:
a Reverse causality refers to a situation where an observed association between exposure and outcome is not due to causality from
exposure to outcome, but rather due to the outcome of interest causing a change in the measured exposure.
A-87
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l.7.1.4 Outcome Ascertainment
This domain may need to be evaluated multiple times for a single study if more than one PECO-relevant outcome is reported.
Therefore, different sets of criteria may be applied for different outcomes in the same study. Table A-20 presents criteria that apply
across outcomes.
Table A-20. Study Quality Evaluation Considerations for Outcome Ascertainment
Core Question: Does the outcome measure reliably distinguish the presence or absence (or degree of severity) of the outcome?
Prompting Questions
Follow-Up Questions
Suggested Considerations
Is outcome ascertainment likely to be affected by Is there a concern that any Good
knowledge of, or presence of exposure
(e.g., consider access to health care, if based on
self-reported history of diagnosis)?
For case-control studies:
Is the comparison group without the outcome
(e.g., controls in a case-control study) based on
objective criteria with little or no likelihood of
inclusion of people with the disease?
For mortality measures:
How well does cause of death data reflect
occurrence of the disease in an individual? How
well do mortality data reflect incidence of the
disease?
For diagnosis of disease measures:
Is the diagnosis based on standard clinical
criteria? If it is based on self-report of the
diagnosis, what is the validity of this measure?
For laboratory-based measures (e.g., hormone
levels):
Is a standard assay used? Does the assay have an
acceptable level of inter-assay variability? Is the
sensitivity of the assay appropriate for the
outcome misclassification
is nondifferential,
differential, or both?
What is the predicted
direction or distortion of
the bias on the effect
estimate (if there is enough
information)?
• High certainty in the outcome definition (i.e., specificity
and sensitivity), minimal concerns with respect to
misclassification.
• Assessment instrument was validated in a population
comparable to the one from which the study group was
selected.
A-88
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Does the outcome measure reliably distinguish the presence or absence (or degree of severity) of the outcome?
outcome measure in this study population? Were
QA/QC measures and results reported?
Adequate • Moderate confidence that outcome definition was specific
and sensitive, some uncertainty with respect to
misclassification but not expected to greatly change the
effect estimate.
• Assessment instrument was validated but not necessarily
in a population comparable to the study group.
Deficient • Outcome definition was not specific or sensitive.
• Uncertainty regarding validity of assessment instrument.
Critically
• Invalid/insensitive marker of outcome.
Deficient
• Outcome ascertainment is very likely to be affected by
knowledge of, or presence of, exposure.
Note: Lack of blinding should not be automatically
construed to be Critically Deficient.
Notes: QA/QC = quality assurance/quality control.
A-89
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l.7.1.5 Potential Confounding
The aim of evaluating this domain is to ascertain whether confounding of the relationship between the exposure and health outcome of
interest is likely to exist, and if so, what the direction and magnitude of the effect of the confounder might be and whether it was
considered in the design and/or analysis of the study (Table A-21).
Table A-21. Study Quality Evaluation Considerations for Confounding
Core Question: Is confounding of the effect of the exposure likely?
Prompting Questions
Follow-Up Questions
Suggested Considerations
Is confounding adequately addressed by
considerations in:
• Participant selection (matching or restriction)?
• Accurate information on potential confounders
and statistical adjustment procedures?
• Lack of association between confounder and
outcome, or confounder and exposure in the
study?
• Information from other sources?
Is the assessment of confounders based on a
thoughtful review of published literature,
potential relationships (e.g., as can be gained
through directed acyclic graphing), and
minimizing potential overcontrol (e.g., inclusion
of a variable on the pathway between exposure
and outcome)?
If there is a concern about Good
the potential for bias, what is I
the predicted direction or
distortion of the bias on the
effect estimate (if there is
enough information)?
• Conveys strategy for identifying key confounders. This
may include: a priori biological considerations, published
literature, causal diagrams, or statistical analyses; with
recognition that not all "risk factors" are confounders.
• Inclusion of potential confounders in statistical models
not based solely on statistical significance criteria
(e.g., p < 0.05 from stepwise regression).
• Does not include variables in the models that are likely to
be influential colliders or intermediates on the causal
pathway.
• Key confounders are evaluated appropriately and
considered to be unlikely sources of substantial
confounding. This often will include:
o Presenting the distribution of potential confounders by
levels of the exposure of interest and/or the outcomes
of interest (with amount of missing data noted);
o Consideration that potential confounders were rare
among the study population, or were expected to be
poorly correlated with exposure of interest;
o Consideration of the most relevant functional forms of
potential confounders;
o Examination of the potential impact of measurement
error or missing data on confounder adjustment;
o Presenting a progression of model results with
adjustments for different potential confounders, if
warranted.
A-90
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Is confounding of the effect of the exposure likely?
Adequate • Similar to Good but may not have considered all potential
confounders (though all key confounders were
considered), or less detail may be available on the
evaluation of confounders (e.g., sub-bullets in Good). It is
possible that residual confounding could explain part of
the observed effect, but concern is minimal.
Deficient • All key confounders were not considered by design or in
the statistical analysis.
• Assessed an outcome based on report of medical
diagnosis that would have required access to a health
professional (e.g., autism, ADHD, depression) and failed
to consider some marker of socioeconomic status (e.g.,
maternal education, household income, marital status,
crowding, poverty, job status) as a potential confounder.
• Does not include variables in the models that are likely to
be influential colliders or intermediates on the causal
pathway.
And any of the following:
• The potential for bias to explain some of the results is
high based on an inability to rule out residual
confounding, such as a lack of demonstration that key
confounders of the exposure-outcome relationships were
considered;
• Descriptive information on key confounders (e.g., their
relationship relative to the outcomes and exposure levels)
is not presented; or
• Strategy of evaluating confounding is unclear or is not
recommended (e.g., only based on statistical significance
criteria or stepwise regression [forward or backward
elimination]).
Critically I • Includes variables in the models that are colliders and/or
Deficient intermediates in the causal pathway, indicating that
substantial bias is likely from this adjustment; or
• Substantial confounding is likely present and not
accounted for, such that all of the results were most likely
due to bias.
A-91
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Is confounding of the effect of the exposure likely?
• If confounders not considered by design or in the analysis
(e.g., only simple correlations presented).
Notes: ADHD = attention deficit hyperactivity disorder.
A-92
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.7.1.6 Analysis
Information relevant to evaluation of analysis includes, but is not limited to: the extent (and if applicable, treatment) of missing data
for exposure, outcome, and confounders; approach to modeling; classification of exposure and outcome variables (continuous vs.
categorical); testing of assumptions; sample size for specific analyses; and relevant sensitivity analyses (Table A-22).
Table A-22. Study Quality Evaluation Considerations for Analysis
Core Question: Does the analysis strategy and presentation convey the necessary familiarity with the data and assumptions?
Prompting Questions
Follow-Up Questions
Suggested Considerations
Are missing outcome, exposure, and covariate
data recognized, and if necessary, accounted for
in the analysis?
Does the analysis appropriately consider variable
distributions and modeling assumptions?
Does the analysis appropriately consider
subgroups of interest (e.g., based on variability in
exposure level or duration or susceptibility)?
Is an appropriate analysis used for the study
design?
Is effect modification considered, based on
considerations developed a priori?
Does the study include additional analyses
addressing potential biases or limitations
(i.e., sensitivity analyses)?
If there is a concern about
the potential for bias, what
is the predicted direction or
distortion of the bias on the
effect estimate (if there is
enough information)?
' Use of an optimal characterization of the outcome variable.
' Quantitative results presented (effect estimates and
confidence limits or variability in estimates (e.g., standard
error, standard deviation); i.e., not presented only as a
p-value or "significant"/"not significant").
' Descriptive information about outcome and exposure
provided (where applicable).
' Amount of missing data noted and addressed appropriately
(discussion of selection issues—missing at random vs.
differential).
' Where applicable, for exposure, includes LOD (and
percentage below the LOD), and decision to use log
transformation.
' Includes analyses that address robustness of findings,
e.g., examination of exposure-response (explicit
consideration of nonlinear possibilities, quadratic, spline,
or threshold/ceiling effects included, when feasible);
relevant sensitivity analyses; effect modification examined
based only on a priori rationale with sufficient numbers.
' No deficiencies in analysis evident. Discussion of some
details may be absent (e.g., examination of outliers).
' Same as Good, except:
' Descriptive information about exposure provided (where
applicable) but may be incomplete; might not have
discussed missing data, cut points, or shape of distribution.
' Includes analyses that address robustness of findings
(examples in Good), but some important analyses are not
performed.
A-93
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Does the analysis strategy and presentation convey the necessary familiarity with the data and assumptions?
Deficient
• Descriptive information about exposure levels not provided
(where applicable).
• Effect estimate and p-value presented, without standard
error or confidence interval (where applicable).
• Results presented as statistically "significant'T'not
significant."
Critically
• Results of analyses of effect modification examined
Deficient
without clear a priori rationale and without providing
main/principal effects (e.g., presentation only of
statistically significant interactions that were not
hypothesis driven).
• Analysis methods are not appropriate for design or data of
the study.
Notes: LOD = limit of detection.
A-94
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.7.1.7 Selective Reporting
This domain concerns the potential for misleading results that can arise from selective reporting (e.g., of only a subset of the measures
or analyses that were conducted). The concept of selective reporting involves the selection of results from among multiple outcome
measures, multiple analyses, or different subgroups, based on the direction or magnitude of these results (e.g., presenting "positive"
results) (Table A-23).
Table A-23. Study Quality Evaluation Considerations for Selective Reporting
Core Question: Is there reason to be concerned about selective reporting?
Prompting Questions
Were results provided for all the primary
analyses described in the methods section?
Is there appropriate justification for restricting
the amount and type of results that are shown?
Are only statistically significant results
presented?
Follow-Up Questions
If there is a concern about
the potential for bias, what
is the predicted direction or
distortion of the bias on the
effect estimate (if there is
enough information)?
Suggested Considerations
Adequate • The results reported by study authors are consistent with
the primary and secondary analyses described in a
registered protocol or methods paper
OR
• The authors described their primary (and secondary)
analyses in the methods section and results were reported
for all primary analyses.
Deficient • Concerns were raised based on previous publications, a
methods paper, or a registered protocol indicating that
analyses were planned or conducted that were not reported,
or that hypotheses originally considered to be secondary
were represented as primary in the reviewed paper.
• Only subgroup analyses were reported; results for the entire
group were omitted without any justification (e.g., to
address effect measure modification).
• Of the PECO-relevant outcomes examined, only
statistically significant results were reported.
A-95
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.7.1.8 Study Sensitivity
The aim of evaluation of this domain is to determine if there are features of the study that affect its ability to detect a true association
(Table A-24). Some of the study features that can affect study sensitivity may have already been included in the outcome, exposure, or
other categories, such as the validity of a method used to ascertain an outcome, the ability to characterize exposure in a relevant time
period for the outcome under consideration, selection of affected individuals out of the study population, or inappropriate inclusion of
intermediaries in a model.
Other features may not have been addressed, and so should be included here. Examples include the exposure range (e.g., the contrast
between the "low" and "high" exposure groups within a study), the level or duration of exposure, and the length of follow-up. In some
cases (for very rare outcomes), sample size or number of observed cases may also be considered within this "sensitivity" category.
Table A-24. Study Quality Evaluation Considerations for Study Sensitivity
Core Question: Is there a concern that sensitivity of the study is not adequate to detect an effect?
Prompting Questions Follow-Up Questions
Is the exposure range/contrast adequate to detect - Adequate
associations that are present?
Was the appropriate (at risk) population included?
Was the length of follow-up adequate? Is the
time/age of outcome ascertainment optimal given
the interval of exposure and the health outcome?
Are there other aspects related to risk of bias or
otherwise that raise concerns about sensitivity?
Deficient
Suggested Considerations
• The range of exposure levels provides adequate variability
to evaluate primary hypotheses in study.
• The population was exposed to levels expected to have an
impact on response.
• The study population was sensitive to the development of
the outcomes of interest (e.g., ages, life stage, sex).
• The timing of outcome ascertainment was appropriate
given expected latency for outcome development
(i.e., adequate follow-up interval).
• The main effects and stratified analyses were fairly precise
(relatively small confidence bounds)
• The study was adequately powered to observe an effect.
Consider sample size, precision (e.g., width of confidence
intervals), anticipated power, exposure ranges and
contrasts.
• No other concerns raised regarding study sensitivity.
• Concerns were raised about the issues described for
Adequate that are expected to notably decrease the
sensitivity of the study to detect associations for the
outcome.
A-96
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.7.1.9 Overall Confidence
Table A-25. Evaluation Considerations for Overall Study Confidence - Overall Confidence, Epidemiological Studies
Provide judgement and rationale for each endpoint or groups of endpoints. The overall confidence rating considers the likely impact of the noted
concerns (i.e., limitations or uncertainties) in reporting, bias and sensitivity on the results. Evaluation Core Question: Considering the identified
strengths and limitations, what is the overall confidence rating for the endpoint(s)/outcome(s) of interest?
Prompting Questions
Suggested Considerations
High
Confidence
For each endpoint/outcome or grouping of
endpoints/outcomes in a study:
Were concerns (i.e., limitations or uncertainties)
related to the reporting quality, risk of bias, or
sensitivity identified?
If yes, what is their expected impact on the overall
interpretation of the reliability and validity of the
study results, including (when possible) Low
interpretations of impacts on the magnitude or Confidence
direction of the reported effects?
• No notable concerns are identified (e.g., most or all domains rated Good).
NOTE: Reviewers should mark studies that are
rated lower than high confidence only due to low
sensitivity (i.e., bias towards the null) for
additional consideration during evidence
synthesis. If the study is otherwise well-conducted
and an effect is obser\>ed, the confidence may be
increased.
Medium • Some concerns are identified but expected to have minimal impact on the interpretation
Confidence of the results, (e.g., most domains rated Adequate or Good; may include studies with
Deficient ratings if concerns are not expected to strongly impact the magnitude or
direction of the results). Any important concerns should be carried forward to evidence
synthesis.
• Identified concerns are expected to significantly impact on the study results or their
interpretation (e.g., generally. Deficient ratings for one or more domains). The concerns
leading to this confidence judgment must be carried forward to evidence synthesis (see
note).
Uninformative
> Serious flaw(s) that make the study results unusable for informing hazard identification
(e.g., generally. Critically Deficient rating in any domain; many Deficient ratings).
Uninformative studies are not considered further in the synthesis and integration of
evidence.
A-97
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1.7.2 Study Quality Evaluation for Animal Toxicological Studies
As noted in the IRIS Handbook, the approach to evaluating study quality for animal
toxicological studies considers study design and experimental conduct in the context of reporting
quality, risk of bias, and study sensitivity. As outlined in Section A. 1.7 of this appendix, study
quality evaluations are performed for a set of established domains, and core and prompting
questions are provided for each domain to guide the reviewer. Each domain is assigned a score
of Good, Adequate, Deficient, Not Reported or Critically Deficient, and rationales to support
the scores are developed. Once all domains are evaluated, a confidence rating of High, Medium,
or Low confidence or Uninformative is assigned for each endpoint/outcome from the study.
The tables in the following sections describe the core and prompting questions and
considerations for assessing each domain during animal toxicological study quality evaluation.
Tables within each section also provide example evaluations for each domain.
A-98
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l.7.2.1 Reporting Quality
Evaluation of this domain is focused on ascertaining whether the study reports enough information to enable evaluation of the study
(Table A-26).
Table A-26. Study Quality Evaluation Considerations for Reporting Quality
Core Question: Does the study report information for evaluating the design and conduct of the study for the endpoint(s)/outcome(s) of interest?
Prompting Questions Suggested Considerations Example Answers
Does the study report the following?
Good
• Minimal concern for selection bias based • Good. Important information is provided
on description of recruitment process for test species, strain, sex, age, exposure
Critical information necessary to ncrform
(e.g., selection of comparison population, methods, experimental design endpoint
study evaluation:
population-based random sample evaluations and the presentation of
• Species; test article name; levels and duration
selection, recruitment from sampling results.
of exposure; route (e.g., oral; inhalation);
frame including current and previous • The authors report that "the study was
qualitative or quantitative results for at least
employees) such that study participants conducted in compliance with the OECD
one endpoint of interest
were unlikely to differ from a larger guidelines for Good Laboratory Practice
cohort based on recruitment or enrollment [c(81) 30 (Final)]".
Imnortant information for evaluating the studv
methods (or data provided to confirm a
methods:
lack of difference)
• Test animal: strain, sex, source, and general
• Exclusion and inclusion criteria specified
husbandry procedures
and would not be likely to induce bias.
• Exposure methods: source, purity, method of
• Participation rate is reported at all steps
administration
of study (e.g., initial enrollment, follow-
• Experimental design: frequency of exposure.
up, selection into analysis sample). If rate
animal age and lifestage during exposure and at
is not high, there is appropriate rationale
endpoint/outcome evaluation
for why it is unlikely to be related to
• Endpoint evaluation methods: assays or
exposure (e.g., comparison between
procedures used to measure the
participants and nonparticipants or other
endpoints/outcomes of interest
available information indicates
differential selection is not likely).
• Comparison groups are similar with
respect to factors expected to influence
exposure-outcome relationship
(confounders, effect measure modifiers).
A-99
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Does the study report information for evaluating the design and conduct of the study for the endpoint(s)/outcome(s) of interest?
Note:
• Reviewers should reach out to authors to obtain
missing information when studies are
considered key for hazard evaluation and/or
dose-response.
• This domain is limited to reporting. Other
aspects of the exposure methods, experimental
design, and endpoint evaluation methods are
evaluated using the domains related to risk of
bias and study sensitivity.
Adequate • Enough of a description of the
recruitment process (i.e., recruitment
strategy, participant selection or case
ascertainment) to be comfortable that
there is no serious risk of bias.
• Inclusion and exclusion criteria specified
and would not induce bias.
• Participation rate is incompletely reported
for some steps of the study, but available
information indicates participation is
unlikely to be related to exposure.
• Comparison groups are largely similar
with respect to factors expected to
influence exposure-outcome relationship
(confounders, effect measure modifiers)
or these are mostly accounted for in the
study analysis.
Deficient
Adequate. All critical information is
reported but some important information
is missing. Specifically, it is unclear what
strain of rats was used.
> Little information on recruitment process,
selection strategy, sampling framework
and/or participation OR aspects of these
processes raises the likelihood of bias
(e.g., healthy worker effect, survivor
bias).
Example: Enrollment of "cases" from a
specific clinic setting (e.g., diagnosed
autism), which could be biased by
referral practices and ser\>ices
availability, without consideration of
similar selection forces affecting
recruitment of controls.
• Deficient. All critical information is
reported, but some important information
is missing that makes additional study
evaluation and interpretation of the results
difficult. Specifically, it is not reported
(and cannot be inferred) what age/life
stage the animals were at outcome
evaluation.
A-100
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Does the study report information for evaluating the design and conduct of the study for the endpoint(s)/outcome(s) of interest?
Critically
Deficient
• Aspects of the processes for recruitment,
selection strategy, sampling framework,
or participation result in concern that the
likelihood of selection bias is high
(e.g., convenience sample with no
information about recruitment and
selection, cases and controls are recruited
from different sources with different
likelihood of exposure, recruitment
materials stated outcome of interest and
potential participants are aware of or are
concerned about specific exposures).
• Convenience sample, and recruitment and
selection not described.
• Case report, case series, or other study
designs lacking a comparison group
(these should be excluded if they do not
meet assessment PECO criteria).
• Example 1: Critically Deficient. Critical
information is missing. Authors did not
report the duration of the exposure or the
results (qualitative or quantitative).
• Example 2: Critically Deficient. Critical
information is missing. The study reports
animals were exposed to per-and
polyfluoroalkyl substances (PFAS), but
the specific chemicals tested were not
provided.
Notes: OECD = Organisation for Economic Co-operation and Development.
For the Reporting domain, the Deficient rating was used as a flag to potentially reach out to study authors to obtain missing critical information (e.g., blinding, randomization) that
may impact the overall confidence rating of the study (e.g., from b confidence to Medium confidence). A Deficient rating does not necessarily relegate the study to Low
confidence, but it is an indicator that obtaining information from the study authors may change the overall confidence rating. EPA could then judge if it was necessary to contact
the study authors. If the study received a Deficient rating for this domain and correspondence with the study authors could potentially increase the confidence, a statement was
added to indicate that obtaining information from the study authors could impact the confidence.
If EPA followed up with authors to obtain missing information, the study details page was updated to note that the authors were contacted and provided the corresponding details.
A-101
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l.7.2.2 Selection and Performance - Allocation
Table A-27. Study Quality Evaluation Considerations for Selection and Performance - Allocation
Core Question: Were animals assigned to experimental groups using a method that minimizes selection bias?
Prompting Questions
Suggested Considerations
Example Answers
For each study:
• Did each animal or litter have an equal chance
of being assigned to any experimental group
(i.e., random allocation)?
• Is the allocation method described?
• Aside from randomization, were any steps
taken to balance variables across experimental
groups during allocation?
• Experimental groups were randomized
and any specific randomization procedure
was described or inferable (e.g.,
computer-generated scheme). [Note that
normalization is not the same as
randomization (see response for
Adequate').]
• Good. The study authors report that "Fifty
males and fifty females were randomly
assigned to groups by a computer-
generated weight-ordered distribution
such that individual body weights did not
exceed + 20% of the mean weight for
each sex."
• Authors report that groups were
randomized but do not describe the
specific procedure used (e.g., 'animals
were randomized'). Alternatively, authors
used a non-random method to control for
important modifying factors across
experimental groups (e.g., body weight
normalization).
• Example 1: Adequate. Randomization
was not performed. However,
normalization procedures that balance
important variables across groups were
performed. Specifically, the authors state
that animals were "allocated into groups
with similar distributions in body
weight."
• Example 2: Adequate. The study authors
state that "animals were randomly
distributed to exposure groups."
However, the specific randomization
method used was not described.
• Example 3: Adequate. Randomization
was not explicitly reported. However, the
study was performed according to OECD
416 and EPA OPPT 870.3800 guidelines
which both specify randomization,
although the specific methods of
randomization used in the current study
could not be inferred. OECD 416
guidelines state "animals should be
A-102
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Were animals assigned to experimental groups using a method that minimizes selection bias?
randomly assigned to the control and
treated groups (stratification by body
weight is recommended)." The EPA
OPPT 870.3800 guidelines state "animals
should be randomly assigned to the
control and treatment groups, in a manner
which results in comparable mean body
weight values among all groups."
• Example 4: Adequate. The study authors
state that "Animals were randomized by
weight into treatment groups," and do not
present the specific randomization
procedural details.
Not Reported
(Interpreted as
Deficient)
• No indication of randomization of groups
or other methods (e.g., normalization) to
control for important modifying factors
across experimental groups.
• Not reported (interpreted as Deficient).
The authors did not indicate
randomization or other normalization
procedures for balancing important
variables across groups.
Critically
Deficient
• Bias in the animal allocations was
reported or inferable.
• Critically Deficient. There is direct
evidence that animals were allocated to
treatment groups in a subjective way,
involving the judgment of the
investigator. Specifically, the study
authors report "the heavier dams were
assigned to the higher dose groups to
reduce toxicity from [chemical]"; dam
weight is an important variable for these
developmental outcomes.
Notes: OECD = Organisation for Economic Co-operation and Development; OPPT = Office of Pollution Prevention and Toxics.
A-103
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.7.2.3 Selection and Performance - Observational Bias/Blinding
Table A-28. Study Quality Evaluation Considerations for Selection and Performance - Observational Bias/Blinding
Core Question: Did the study implement measures to reduce observational bias?
Prompting Questions Suggested Considerations Example Answers
For each endpoint/outcome or grouping of
Good
• Measures to reduce observational bias • Examnlc 1: Good. Histooatholoev:
endpoints/outcomes in a study:
were described (e.g., blinding to conceal Although the study did not indicate
treatment groups during endpoint blinding, blinding during the initial
Does the study report blinding or other
evaluation; consensus-based evaluations evaluation of tissues for initial or non-
methods/procedures for reducing observational
of histopathology lesions3). targeted evaluations is generally not
bias?
recommended as masked evaluation can
make the task of separating treatment-
If not, did the study use a design or approach for
related changes from normal variation
which such procedures can be inferred?
more difficult and may result in subtle
lesions being overlooked {Crissman,
What is the expected impact of failure to
2004, 51763}. The study did include a
implement (or report implementation) of these
secondary evaluation by a pathology
methods/procedures on results?
working group (PWG) review on coded
pathology slides which minimized the
potential for observational bias.
• Examiile 2: Good. Orsan weishts. FOB.
motor activity, swim maze and
histooatholoev: Authors reported that the
investigators were blinded to the animal
treatment group during evaluation for all
outcome measures. Although blinding is
not recommended for initial or non-
targeted evaluations {Crissman, 2004,
51763}, this study evaluated prespecified
outcomes in targeted evaluations for
which blinding is appropriate (cell counts
in the CA3 region of the hippocampus).
Adequate
• Methods for reducins observational bias • Adeauate. Histooatholoev measures:
(e.g., blinding) can be inferred or were Authors report "lesions were counted by
reported but described incompletely. 2 observers in a blinded fashion"
although it should be noted that blinding
during the initial evaluation of tissues is
generally not recommended for initial or
A-104
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Did the study implement measures to reduce observational bias?
non-targeted evaluations as masked
evaluation can make the task of
separating treatment-related changes from
normal variation more difficult and may
result in subtle lesions being overlooked
{Crissman. 2004, 51763}.
Not Reported
• Measures to reduce observational bias • Example 1: Not reported (interpreted as
(Interpreted as
were not described. Adeauate). Bodv and orsan weiehts.
Adequate)
• The potential concern for bias was developmental landmarks, and hormone
mitigated based on use of measures: Authors did not indicate
automated/computer driven systems, whether investigators were blinded during
standard laboratory kits, relatively simple, outcome assessment. Potential concern
objective measures (e.g., body or tissue for bias was mitigated for these endpoints
weight), or screening-level evaluations of which were measured using
histopathology. automated/computer driven systems.
standard laboratory kits, relatively simple.
objective measures (e.g., body or tissue
weight).
• Example 2: Not reported (interpreted as
Adeauate). Histonatholoev: Blindins
during the initial evaluation of tissues is
generally not recommended as masked
evaluation can make the task of
separating treatment-related changes from
normal variation more difficult and may
result in subtle lesions being overlooked
{Crissman, 2004, 51763}. Histopathology
was evaluated by an independent
laboratory (Toxicology Pathology
Associates Little Rock, Arkansas, John
Pletcher, D.V.M.., DACPV). No
subsequent steps to minimize the
potential for observational bias were
reported (i.e., conducting a secondary
targeted blinded review, independent
prospective or retrospective peer-review.
formation of a pathology working group).
A-105
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Did the study implement measures to reduce observational bias?
• Example 3: Not reported (interpreted as
Adequate). Fetal evaluation for
malformations: Blinding during initial
evaluation of fetuses is typically not
conducted as masked evaluation can
make the task of separating treatment-
related changes from normal
developmental variation more difficult
and may result in subtle developmental
anomalies being overlooked. Fetal
evaluations were conducted in accordance
with regulatory test guideline
recommendations, using standardized
nomenclature. No subsequent steps to
minimize the potential for observational
bias were reported (e.g., conducting a
secondary targeted blinded review, or an
independent prospective or retrospective
peer-review).
• Not reported (interpreted as Deficient).
Neurobehavior (auditory and visual
sensory reactivity): Procedural methods
addressing observational bias were not
described for these endpoints, which were
measured using highly subjective
methods (i.e., it appears that investigators
measured reactivity using manually
operated timers).
• Critically Deficient Neurobehavior after
restraint stress: There is direct evidence
of observational bias in testing methods.
Specifically, the study reported that, to
minimize stress from changing
investigators across trials, one
investigator consistently stressed control
mice each day for 30 minutes and
subsequently tested behaviors, while a
separate investigator conducted stress and
Not Reported • Measures to reduce observational bias
(Interpreted as were not described.
Deficient) • The potential impact on the results is
major (e.g., outcome measures are highly
subjective).
Critically I • Strong evidence for observational bias
Deficient that could have impacted results.
A-106
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Did the study implement measures to reduce observational bias?
behavioral testing in treated mice. There
was no mention of blinding of
investigators.
Notes: FOB = functional observed battery.
a For non-targeted or screening-level histopathology outcomes often used in guideline studies, blinding during the initial evaluation of tissues is generally not recommended as
masked evaluation can make 'the task of separating treatment-related changes from normal variation more difficult' and 'there is concern that masked review during the initial
evaluation may result in missing subtle lesions.' Generally, blinded evaluations are recommended for targeted secondary review of specific tissues or in instances when there is a
pre-defined set of outcomes that is known or predicted to occur {Crissman, 2004, 51763}.
A-107
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.7.2.4 Confounding/Variable Control
Table A-29. Study Quality Evaluation Considerations for Confounding/Variable Control
Core Question: Are variables with the potential to confound or modify results controlled for and consistent across all experimental groups?
Prompting Questions
Suggested Considerations
Example Answers
For each study:
Are there differences across the treatment groups
(e.g., co-exposures, vehicle, diet, palatability,
husbandry, health status, etc.) that could bias the
results?
If differences are identified, to what extent are
they expected to impact the results?
• Outside of the exposure of interest,
variables that are likely to confound or
modify results appear to be controlled for
and consistent across experimental
groups.
• Good. Based on the study report, vehicle
(deionized water with 2% tween 80) and
husbandry practices were inferred to be
the same in controls and treatment
groups. The experimental conditions
described provided no indication of
concern for uncontrolled variables or
different practices across groups.
' Some concern that variables that were
likely to confound or modify results were
uncontrolled or inconsistent across groups
but are expected to have a minimal
impact on the results.
' Example 1 (oral): Adequate. Honnone
measurements: Authors did not use a soy-
free diet. Soy-based rodent feeds contain
phytoestrogens that may act as a
confounder for endocrine-related
measures. Since this study includes
relatively high doses (100 and
1500 mg/kg-d) the concern is minimal.
' Example 2 (inhalation): Adequate.
Behavior, immunological responses, and
hormonal changes: control rats did not
appear to receive chamber air exposures
(they were left in their home cages). As
this might introduce a difference in
stressors across groups, this difference is
interpreted as a possible confounder for
measures shown to be sensitive to stress,
although the impact of this limitation on
the results is expected to be minimal.
A-108
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Are variables with the potential to confound or modify results controlled for and consistent across all experimental groups?
Deficient • Notable concern that potentially • Deficient. Dams in the medium and high
confounding variables were uncontrolled exposure groups (1500 and 15,000 ppm,
or inconsistent across groups and are respectively) showed significantly lower
expected to substantially impact the consumption of the treated food
results. throughout the exposure period
(gestation) that increased to control levels
after the exposure ended. Addition of the
test chemical may have affected the
palatability of the food and reduced food
intake during gestation may have
significantly impacted the developmental
outcomes in the pups.
• Critically Deficient. The study did not
include a vehicle-only control group, and,
given the high concentration of DMSO
required to solubilize the test article in
other experiments using a similar
exposure design, this is interpreted as
likely to be a significant driver of any
observed effects.
ppm = parts per million; DMSO = dimethyl sulfoxide.
Critically I • Confounding variables were presumed to
Deficient be uncontrolled or inconsistent across
groups, and are expected to be a primary
driver of the results.
A-109
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l. 7.2.5 Reporting and Attrition Bios
Table A-30. Study Quality Evaluation Considerations for Selective Reporting and Attrition - Reporting and Attrition Bias
Core Question: Did the study report results for all prespecified outcomes and tested animals?
Prompting Questions
Suggested Considerations
Example Answers
For each study:
Selective reporting bias:
Are all results presented for endpoints/outcomes
described in the methods (see note)?
Attrition bias:
Are all animals accounted for in the results?
If there are discrepancies, do authors provide an
explanation (e.g., death or unscheduled sacrifice
during the study)?
If unexplained results omissions and/or attrition
are identified, what is the expected impact on the
interpretation of the results?
NOTE: This domain does not consider the
appropriateness of the analysis/results
presentation. This aspect of study quality is
evaluated in another domain.
• Quantitative or qualitative results were
reported for all prespecified outcomes
(explicitly stated or inferred), exposure
groups and evaluation timepoints. Data
not reported in the primary article is
available from supplemental material. If
results omissions or animal attrition are
identified, the authors provide an
explanation and these are not expected to
impact the interpretation of the results.
• Good. Animal loss was reported (the
authors treated 10 rats/sex/dose group and
noted one death in a high-dose male rat at
day 85 of study). All endpoints described
in methods were reported qualitatively or
quantitatively.
' Quantitative or qualitative results are
reported for most prespecified outcomes
(explicitly stated or inferred), exposure
groups and evaluation
timepoints. Omissions and/or attrition are
not explained but are not expected to
significantly impact the interpretation of
the results.
' Adequate. Animal loss occurred and was
reported (see below), but these are not
expected to significantly impact the
interpretation of the results. All endpoints
described in methods were reported
qualitatively or quantitatively.
¦ "In the high dose (1000 mg/kg-day)
group no male animals were able to
complete the entire study; whereas all
male rats exposed at other doses
completed the 4-week experiment. In the
female group, 1 rat was removed in the
250 mg/kg-day group at day 25, 1 rat in
the 500 mg/kg-day was removed at day
21 and 8 rats in the 1000 mg/kg/day
group were removed between days 16 and
27 of the experiment." Justification for
removals was provided by the study
authors.
A-110
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Did the study report results for all prespecified outcomes and tested animals?
Deficient
' Quantitative or qualitative results are
missing for many prespecified outcomes
(explicitly stated or inferred), exposure
groups and evaluation timepoints and/or
high animal attrition; omissions and/or
attrition are not explained and may
significantly impact the interpretation of
the results.
' Example 1: Deficient. Unaccounted for
loss of animals was difficult to assess
because the study authors do not provide
a clear description of the number of
animals per exposure group or the
selection of animals for outcome analysis.
Table 1 states there were 8 animals used
in experiment 1 and 6 animals used in
experiments 2 and 3. The figures and
tables report data for varying numbers of
animals (from 4 to 8), but the authors do
not provide a description of the approach
used to sample animals for each outcome.
1 Example 2: Deficient. Although the
authors indicated that "the liver, kidneys,
and spleen were weighed and processed
for routine histopathology at study
termination", qualitative or quantitative
findings were not reported for liver or
kidney weights, nor for liver, kidney, or
spleen histopathology ("spleen weights"
were described as unchanged during the
description of changes in cultured splenic
immune cells).
Critically I • Extensive results omission and/or animal
Deficient attrition are identified and prevents
comparisons of results across treatment
groups.
Critically Deficient. None of the animals
in the high and medium dose groups
survived and there was high mortality
(>75%) in the low dose group.
A-lll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l. 7.2.6 Exposure Methods Sensitivity - Chemical Administration and Characterization
Table A-31. Study Quality Evaluation Considerations for Exposure Methods Sensitivity - Chemical Administration and
Characterization
Core Question: Did the study adequately characterize exposure to the chemical of interest and the exposure administration methods?
Prompting Questions
Suggested Considerations
Example Answers
For each study:
Does the study report the source and purity and/or
composition (e.g., identity and percent distribution |
of different isomers) of the chemical? If not, can
the purity and/or composition be obtained from
the supplier (e.g., as reported on the website)
Was independent analytical verification of the test
article purity and composition performed?
Did the authors take steps to ensure the reported
exposure levels were accurate?
For inhalation studies: were target concentrations
confirmed using reliable analytical measurements
in chamber air?
For oral studies: if necessary, based on
consideration of chemical-specific knowledge
(e.g., instability in solution; volatility) and/or
exposure design (e.g., the frequency and duration
of exposure), were chemical concentrations in the
dosing solutions or diet analytically confirmed?
Are there concerns about the methods used to
administer the chemical (e.g., inhalation chamber
type, gavage volume, etc.)?
NOTE: Consideration of the appropriateness of
the route of exposure is not evaluated at the
• Chemical administration and
characterization are complete (i.e.,
source, purity, and analytical verification
of the test article are provided). There are
no concerns about the composition,
stability, or purity of the administered
chemical, or the specific methods of
administration. For inhalation studies,
chemical concentrations in the exposure
chambers are verified using reliable
analytical methods.
' Example 1 (oral): Good. Source (3M)
and purity (98%) are described, and the
authors provided verification using
analytical methods (GC/MS). Addressing
concerns about known instability in
solution for this chemical, the authors
verified the dosing solutions twice weekly
over the course of the experiment.
Animals were exposed via gavage with
all dose groups receiving the same
volume.
' Example 2 (inhalation): Good. Source
(3M) and purity (98%) of the test article
are described. All animals were
transferred to dynamic inhalation
exposure chambers for the exposures. The
concentration of the test chemical in the
air was continuously monitored from the
animals' breathing zone throughout the 6-
hour exposure periods and mean daily
average concentrations and variability
were reported.
' Some uncertainties in the chemical
administration and characterization are
identified but these are expected to have
minimal impact on interpretation of the
results (e.g., source and vendor- reported
purity are presented, but not
independently verified; purity of the test
article is sub-optimal but not concerning;
For inhalation studies, actual exposure
' Example 1 (oral): Adequate. Purity
(98%) is described, but source is missing.
Purity is assumed to be vendor reported
because independent analytical
verification of the purity is not described.
Authors were contacted to try to obtain
the vendor information however they did
not respond. Stability assessments were
not necessary because fresh dosing
solutions were prepared daily.
A-112
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Did the study adequately characterize exposure to the chemical of interest and the exposure administration methods?
individual study level. Relevance and utility of the
routes of exposure are considered in the PECO
criteria for study inclusion and during evidence
synthesis.
concentrations are missing or verified
with less reliable methods).
' Example 2 (inhalation): Adequate.
Source (3M) and purity (98%) of the test
article are described. All animals were
transferred to dynamic inhalation
exposure chambers for the exposures. The
nominal/target concentrations of the test
chemical were not verified by analytical
measurements of the chamber air.
• Uncertainties in the exposure
characterization are identified and
expected to substantially impact the
results (e.g., source of the test article is
not reported; levels of impurities are
substantial or concerning; deficient
administration methods, such as use of
static inhalation chambers or a gavage
volume considered too large for the
species and/or lifestage at exposure).
' Example 1 (oral): Deficient. Test
chemical supplied by the chemical
manufacturer. Purity and isomeric
composition are not described and could
not be obtained from manufacturer's
website. Analytical verification of the test
article's purity and composition was not
provided, and the stability of chemical in
the diet across the 1-year exposure period
does not appear to have been assessed.
1 Example 2 (inhalation): Deficient.
Source (3M) and vendor-reported purity
are described, although these were not
independently verified. The animals
appear to have been exposed in static
(i.e., without dynamic airflow) chambers;
this is not interpreted as a critical
deficiency due to the relatively short (2-
hour) durations of daily exposure.
Critically I • Uncertainties in the exposure
Deficient characterization are identified and there is
reasonable certainty that the results are
largely attributable to factors other than
exposure to the chemical of interest (e.g.,
identified impurities are expected to be a
primary driver of the results).
• Example 1 (oral): Critically Deficient.
The test article contains large amounts of
a known impurity [specify] that has
previously been shown to cause the
outcome(s) of interest. Based on the doses
tested (and inferences regarding the
administered doses of the impurity), this
is likely to be a significant driver of any
observed effects.
• Example 2 (inhalation): Critically
Deficient. Dams were exposed in static
A-113
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Did the study adequately characterize exposure to the chemical of interest and the exposure administration methods?
Notes: GC/MS = gas chromatography mass spectrometry.
chambers during gestation, and there was
evidence of overt toxicity (i.e., gasping)
throughout the 12-hr daily exposures at
all tested concentrations. This is likely to
be a substantial driver of any observed
developmental effects.
A-114
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l.7.2.7 Exposure Methods Sensitivity - Exposure Timing, Frequency, and Duration
Table A-32. Study Quality Evaluation Considerations for Exposure Methods Sensitivity - Exposure Timing, Frequency, and
Duration
Core Question: Was the timing, frequency, and duration of exposure sensitive for the endpoint(s)/outcome(s) of interest?
Prompting Questions
Suggested Considerations
Example Answers
For each endpoint/outcome or grouping of
endpoints/outcomes in a study:
Does the exposure period include the critical
window of sensitivity?
Was the duration and frequency of exposure
sensitive for detecting the endpoint of interest?
The duration and frequency of the
exposure was sensitive and the exposure
included the critical window of sensitivity
(if known).
• Example 1: Good. Study uses a standard
OECD short-term (28-day) study design
to examine toxicological effects that are
routinely evaluated in this testing
guideline.
• Example 2: Good. The experimental
design and exposure period were
appropriate for evaluation of potential
male reproductive and developmental
effects. The experiment was designed to
evaluate reproductive and developmental
outcomes and followed recommendations
in {OECD, 2001, 3421602} and {U.S.
EPA, 1998, 2229410} guidelines.
• The duration and frequency of the
exposure was sensitive and the exposure
covered most of the critical window of
sensitivity (if known).
• Adequate. The study does not include the
full developmental window of exposure
most informative to evaluating potential
effects on androgen-dependent
development of male reproductive organs.
Specifically, the study exposed rats from
GD 18-GD 21, whereas the critical
window for the development of these
endpoints (i.e., cryptorchidism; testes and
seminal vesicle weights; and male
reproductive organ histopathology)
begins on GD 15, and peaks around GD
17 (NRC 2008 [635834]; Scott et al 2009
[673313]) in rats. The incomplete
coverage of this critical window in this
study is expected to result in a minor bias
towards the null.
A-115
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Was the timing, frequency, and duration of exposure sensitive for the endpoint(s)/outcome(s) of interest?
• The duration and/or frequency of the • Deficient. The experimental design is not
exposure is not sensitive and did not considered appropriate for evaluation of
include the majority of the critical male fertility. Male rats were exposed for
window of sensitivity (if known). These chemicalXfor 1 week and fertility was
limitations are expected to bias the results assessed on week 2 of the study. This
towards the null. design is considered deficient because in
most rodent species "damage to
spennatogonial stem cells will not appear
in samples from the cauda epididymis or
in ejaculates for 8 to 14 weeks" {U.S.
EPA, 1996, 30019}.
Critically
• The exposure design was not sensitive • Critically Deficient. The experimental
Deficient
and is expected to strongly bias the results design is not appropriate for evaluation of
towards the null. The rationale should cancer endpoints. Animals were
indicate the specific concern(s). necropsied and tissues evaluated for the
presence of tumors and/or neoplasms 4
weeks after only a 28-day exposure
period. Notably, because this critical
deficiency is due to insensitivity.
depending on other identified limitations.
the utility of this study will depend on
whether effects were observed in the
study (i.e., if tumors were observed, this
study could be adjusted to a higher
rating).
Notes: OECD = Organisation for Economic Co-operation and Development; OPPT = Office of Pollution Prevention and Toxics; GD = gestation day.
A-116
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l. 7.2.8 Outcome Measures and Results Display - Endpoint Sensitivity and Specificity
Table A-33. Study Quality Evaluation Considerations for Outcome Measures and Results Display - Endpoint Sensitivity and
Specificity
Core Question: Are the procedures sensitive and specific for evaluating the endpoint(s)/outcome(s) of interest?
Prompting Questions Suggested Considerations Example Answers
For each endpoint/outcome or grouping of
Good
• Example 1: Good. Lipid/Lipoproteins:
endpoints/outcomes in a study:
There are no notable concerns about
aspects of the procedures, or for the
Are there concerns regarding the specificity and
timing of these evaluations. Study authors
validity of the protocols?
used standard methodology (i.e.,
commercial kits) appropriate for use in
Are there serious concerns regarding the sample
adult liver tissue samples.
size (see note)?
• Examiile 2: Good. Orsan weisht. bodv
weights, and hormone measures: no
Are there concerns regarding the timing of the
concerns regarding the specificity and
endpoint assessment?
validity of the protocols and measures
were identified. Study authors used
NOTE: Sample size alone is not a reason to
standard methodology for evaluating
conclude an individual study is critically deficient.
organ and body weights. Thyroid
hormones were measured using
commercial electrochemiluminescence-
immunoassay methods, and the known
diurnal variation in these measures was
accounted for during blood collection.
Adequate
• Examiile 1: Adeauate. Histooatholoev:
Tissues were fixed in 10% neutral
buffered formalin, trimmed, sectioned (5
microns) and embedded and stained with
H&E. Evaluations included 12 tissues
from all animals in the control and
highest dose groups. Although not
explicitly stated, it is inferred that tissues
from animals in the low- and mid-dose
groups would have been evaluated if
significant increases in lesion incidence
were observed at the highest dose. This
practice is consistent with NTP pathology
A-117
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Are the procedures sensitive and specific for evaluating the endpoint(s)/outcome(s) of interest?
guidelines (ref) and is expected to be of
minimal concern unless effects are
observed at the high dose. Additionally,
the report did not provide information on
sampling (e.g., # sections
evaluated/tissue, sections evaluated at x
micron or section intervals). Together, the
missing study details introduce some
concern for potential insensitivity.
• Example 2: Adequate. Clinical
chemistry: Some concern was raised
regarding the procedural methods, as no
information was provided on the
diagnostic kits and, for some of the
specific measures (i.e., those without
specific data reported), it is unclear
whether serum or plasma was analyzed.
• Example 1: Deficient. Histopathology
(testis): Concerns regarding the method
used to preserve testis for histological
analysis: 10% formalin. For evaluation of
histopathological effects in the testis,
conventional immersion fixation in
buffered formalin is not recommended as
it gives very poor penetration of fixative
and may result in artifacts (Haschek (ed)
et al 2009 [3987435]; Foley et al 2001
[PMID: 11215684]).
• Example 2: Deficient. Nipple retention:
Concerns for insensitivity were raised due
to the timing of endpoint evaluation.
Specifically, the authors examined nipple
retention in rats at PND 9, whereas this
endpoint is more appropriately evaluated
around PNDs 12-14.
• Example 3: Deficient. Motor activity:
Concerns were raised regarding the small
sample sizes used to evaluate these
A-118
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Are the procedures sensitive and specific for evaluating the endpoint(s)/outcome(s) of interest?
outcomes. Specifically, the authors tested
4 animals (sex not specified, but assumed
males) per group. Ideally, it is preferable
to have more than 10 animals/sex/group
for this type of evaluation, according to
OECD guidelines.
Critically
Deficient
- • Criticallv Deficient. fEndooint namel:
[Assay X] has been shown to be
unreliable for evaluating [endpoint of
interest]. Currently best practice is to use
[Assay Y] for this endpoint.
Notes: NTP = National Toxicology Program; PND = postnatal day; OECD = Organisation for Economic Co-operation and Development.
A-119
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.l.7.2.9 Outcome Measures and Results Display - Results Presentation
Table A-34. Study Quality Evaluation Considerations for Outcome Measures and Results Display - Results Presentation
Core Question: Are the results presented in a way that makes the data usable and transparent?
Prompting Questions
Suggested Considerations Example Answers
For each endpoint/outcome or grouping of
Good
• Good. There are no notable concerns
endpoints/outcomes in a study:
about the way the results are analyzed or
presented.
Does the level of detail allow for an informed
Adequate
• Examiile 1: Adeauate. Reproductive
interpretation of the results?
orsan weiehts. hormone measures: results
Are the data analyzed, compared, or presented in a
are presented graphically; however, the
authors do not clarify whether error bars
way that is inappropriate or misleading?
correspond to SD or SE.
• Examiile 2: Adeauate. Developmental
effects: the studv failed to report
information on potential maternal
toxicity; however, all tested doses other
than the highest dose are not expected to
cause overt toxicity in adults, reducing
the level of concern.
• Examiile 3: Adeauate. Anoeenital
distance (AGD): The authors reported
AGD without adjusting for body weight.
which is preferred (Daston 1998
[3393032]). However, because the study
also provided body weight data.
approximation was possible, limiting
concern.
Deficient
• Examiile 1: Deficient. Histopatholoev:
Incidence and severity of individual
effects was unclear, as only scores across
multiple, disparate pathological endpoints
were reported.
• Examiile 2: Deficient. Behavior
(neuromuscular function and dexterity):
Performance on the rotarod was presented
as incidence of falling off the rod within
an arbitrary time, rather than as time
A-120
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Are the results presented in a way that makes the data usable and transparent?
spent on the rod (the preferred metric).
This dichotomization of continuous data
without sound justification is expected to
strongly bias the results towards
observing an effect.
• Exami)le 3: Deficient. Brain weisht:
Authors presented only relative brain
weights, and absolute weights could not
be calculated. The adult CNS is highly
protected, and absolute brain weight data
are preferred [include reference].
• Exami)le 4: Deficient. Birth outcomes:
Data on pup viability, weights, and
malformations were reported as pup
averages, without addressing potential
litter effects.
Critically
- • Critically Deficient. Endooint name: The
Deficient
study presents the results for this endpoint
in both a table and figure; however, the
data do not match (e.g., mean ± SE
reported for the control group is 2.3 ± 0.5
in the table and 1.9 ± 0.2 in the figure).
This reporting discrepancy could not be
resolved from the information provided in
the study and study authors did not
respond to queries for clarification.
A-121
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.7.2.10 Overall Confidence
The overall confidence rating considers the likely impact of the noted concerns (i.e., limitations or uncertainties) in reporting, bias and
sensitivity on the results (Table A-35).
Table A-35. Evaluation Considerations for Overall Study Confidence - Overall Confidence, Animal Toxicological Studies
Provide judgement and rationale for each endpoint or groups of endpoints. The overall confidence rating considers the likely impact of the noted
concerns (i.e., limitations or uncertainties) in reporting, bias and sensitivity on the results. Evaluation Core Question: Considering the identified
strengths and limitations, what is the overall confidence rating for the endpoint(s)/outcome(s) of interest?
Prompting Questions
Suggested Considerations
Example Answers
For each endpoint/outcome or grouping of
endpoints/outcomes in a study:
Were concerns (i.e., limitations or uncertainties)
related to the reporting quality, risk of bias, or
sensitivity identified?
If yes, what is their expected impact on the overall
interpretation of the reliability and validity of the
study results, including (when possible)
interpretations of impacts on the magnitude or
direction of the reported effects?
NOTE: Reviewers should mark studies that are
rated lower than high confidence only due to low
sensitivity (i.e., bias towards the null) for
additional consideration during evidence
synthesis. If the study is otherwise well-conducted
and an effect is obser\>ed, the confidence may be
increased.
High • No notable concerns are identified (e.g..
Confidence most or all domains rated Good).
• High Confidence. Reproductive and
developmental effects other than
behavior: The study was well-designed
for the evaluation of reproductive and
developmental toxicity induced by
chemical exposure. The study applied
established approaches,
recommendations, and best practices, and
employed an appropriate exposure design
for these endpoints. Evidence was
presented clearly and transparently.
Medium • Some concerns are identified but
Confidence expected to have minimal impact on the
interpretation of the results, (e.g., most
domains rated Adequate or Good; may
include studies with Deficient ratings if
concerns are not expected to strongly
impact the magnitude or direction of the
results). Any important concerns should
be carried forward to evidence synthesis.
• Example 1: Medium Confidence.
Developmental effects: The study was
adequately designed for the evaluation of
developmental toxicity. Although the
authors failed to describe randomized
allocation of animals to exposure groups
and some concerns were raised regarding
the sensitivity (i.e., timing) and sample
sizes (i.e., n=6 litters/group) used for the
evaluation of potential effects on male
reproductive system development with
gestational exposure, these limitations are
expected to have a minimal impact on the
results.
• Example 2: Medium Confidence.
Histopathology: The study authors did not
A-122
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Provide judgement and rationale for each endpoint or groups of endpoints. The overall confidence rating considers the likely impact of the noted
concerns (i.e., limitations or uncertainties) in reporting, bias and sensitivity on the results. Evaluation Core Question: Considering the identified
strengths and limitations, what is the overall confidence rating for the endpoint(s)/outcome(s) of interest?
report information on the severity of
histological effects for which this is
routinely provided. The authors also
failed to describe use of methods to
reduce potential observational bias.
Low
• Identified concerns are expected to • Example 1: Low Confidence.
Confidence
sienificantlv impact on the studv results Developmental effects: Substantial
or their interpretation (e.g., generally, concerns were raised regarding
Deficient ratings for one or more quantitative analyses without addressing
domains). The concerns leading to this potential litter effects. Other significant
confidence judgment must be carried limitations included incomplete data
forward to evidence synthesis (see note). presentation (sample sizes for outcome
assessment were unclear; no information
on maternal toxicity was provided), and
methods for selection of animals for
outcome assessment.
• Example 2: Low Confidence. Behavioral
measures: The cursory cage-side
observations of activity are considered
insensitive and non-specific methods for
detecting motor effects, with a strong bias
towards the null.
Uninformative
• Serious flaw(s) that make the study • Example 1: Uninformative. Critical
results unusable for informing hazard information was not reported.
identification (e.g., generally. Critically Specifically, the study authors did not
Deficient rating in any domain; many report the duration of the exposure or the
Deficient ratings). Uninformative studies results (qualitative or quantitative). Given
are not considered further in the synthesis this critical deficiency, the other domains
and integration of evidence. were not evaluated.
• Example 2: Uninformative. Concerns
were raised over the lack of information
on test animal strain and allocation, and
chemical source/purity. The lack of
information on blinding or other methods
to reduce observational blinding is also of
significant concern for the endpoints of
A-123
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Provide judgement and rationale for each endpoint or groups of endpoints. The overall confidence rating considers the likely impact of the noted
concerns (i.e., limitations or uncertainties) in reporting, bias and sensitivity on the results. Evaluation Core Question: Considering the identified
strengths and limitations, what is the overall confidence rating for the endpoint(s)/outcome(s) of interest?
interest (i.e., follicle counts, ova counts,
and evaluation of estrous cyclicity).
Finally, concerns were also raised over
the apparent self-plagiarism in similar
chromium studies published in 1996 by
this group of authors. Taken together, this
combination of limitations resulted in an
interpretation that the results were
unreliable.
• Example 3: Uninformative. Sperm
Measures: Issues were identified with the
methods used to prepare samples for
analysis, which are likely to introduce
artifacts. Concerns were also raised
regarding results presentation (i.e., lack of
group variability), missing information on
sample sizes and loss of animals, and a
lack of information on the timing of these
evaluations. Taken together, the
evaluation of this endpoint was
considered uninformative.
A-124
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.8 Data Extraction for Epidemiological Studies
All epidemiological studies identified as PECO-relevant after full-text screening were considered
eligible for data extraction. As noted in the IRIS Handbook {U.S. EPA, 2022, 10476098}, during
data extraction, relevant results from each study are extracted to facilitate organization,
visualization, comparison, and analysis of findings and results. Data from PECO-relevant
epidemiological studies published prior to 2016 (i.e., from the 2016 HESD and the 2021 ATSDR
Toxicological Profile for Perfluoroalkyls) or identified in the updated literature searches were
extracted if they received a medium or high confidence study quality evaluation rating. In cases
where data was limited (e.g., thyroid cancer) or when there was a notable effect, results from low
confidence studies were extracted. Studies evaluated as being uninformative were not considered
further and therefore did not undergo data extraction. Extraction was targeted towards the five
main health outcomes recommended by the SAB (i.e., cancer, cardiovascular, developmental,
hepatic, and immune). Results from main analyses were extracted, and age- and sex-stratified
analyses were extracted if available. Results from other stratified and sensitivity analyses were
extracted on a case-by-case basis (e.g., medication use status for cardiovascular outcomes).
Data extraction of epidemiological studies was carried out using a set of structured forms in
DistillerSR. Studies slated for extraction were pre-screened by an expert epidemiologist who
identified the relevant results to be extracted. Data extraction was performed by one reviewer
and then independently verified by at least one other reviewer for quality control. Any conflicts
or discrepancies related to data extraction were resolved by discussion and confirmation within
the evaluation team.
Table A-36 outlines the content of the DistillerSR forms that were populated during data
extraction of epidemiological studies, including the extraction questions or prompts and response
options.
Table A-36. DistillerSR Form for Data Extraction of Epidemiological Studies
Question/Prompt Response Options Suggested Considerations
1
Has this study been
QC'd?
[Select one]
• No (select if doing data
extraction)
• Yes, no corrections needed
• Yes, corrections were
needed and completed
during QC (please list any
major revisions, e.g.,
incomplete responses,
NOEL/LOEL incorrect,
etc.)
• Study is not PECO-relevant
(please specify why)
2
Reference (short form)
e.g., Smith et al. (1978)
[Free-text]
• Enter author information; use the format
specified in the Distiller form.
3
Population
[Select one]
• General population, adults
and children
• General population, adults
• Do not select "pregnant women" if pregnant
women are only included as part of a general
population sample.
A-125
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt Response Options Suggested Considerations
• General population, • When exposure is measured in cord blood
children and and outcome in children, the study
adolescents <18 years population would be "children."
• Occupational
• Pregnant women
• Occupational/general
population, adults
• Other
4 Population Summary - • Briefly describe the study population (e.g.,
[Free-text] women undergoing fertility treatment,
NHANES adults 18+). Try to capture
anything outside a typical general population
sample. Keep it brief - does not need to be in
full sentences.
• For studies of mother-child cohorts, when
exposure is in maternal blood and outcome is
evaluated in children, use "pregnant women
and their children."
For example, if any of these (non-exhaustive)
scenarios apply, capture them in this field:
• Known potential forPFAS exposure (e.g.,
contamination event/lawsuit)
• Follow-up timing
• Participants are drawn from a specific
population, such as people with a specific
health condition, narrow age range within
"adults" and "children" (e.g., infants,
seniors), specific environments (e.g., assisted
living facility, daycare, farmers), etc.
5
Study Design
[Select one]
• Cohort
• Case-control
• Cross-sectional
• Ecological
• Controlled trial
• Other
• Nested caste-control
• Cross-sectional and
prospective analyses
• Cohort and cross-sectional
• Case-control and cross-
sectional
• See Appendix A. 1.8.1 for different types of
study design.
• Note: Third trimester samples with outcome
measured at birth should be classified as
cohort studies.
• Cohort studies reporting prospective and
cross-sectional analyses should be classified
as Cohort and cross-sectional.
• Case-control studies reporting cross-sectional
analyses among the whole study population
or within cases or controls should be
classified as Case-control and cross-
sectional.
6
Study Name (if
applicable)
[Free-text]
• Only use the name of an official study or
cohort. Leave blank if there is no name.
7
Country (or Countries)
[Free-text]
-
• Use full names such as "United States" (not
US).
8
Year of Data
List which year(s) the data
came from.
[Free-text]
• For prospective cohort studies that only state
the period the population was recruited (e.g.,
2012-2015) and mention the outcomes were
assessed at follow-up (e.g., state "5 years
later" but do not provide dates), extract
A-126
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
"recruitment 2012-2015, outcome assessed
at 5-year follow-up."
9 Exposure Measurement
• Biomonitoring
[Select all that apply]
• Air
• Food
• Drinking water
• Occupational (use in cases
where exposure is based on
factors such as job function,
place in building where
people worked, job
exposure matrices)
• Modeled
• Questionnaire
• Direct administration - oral
• Direct administration -
inhalation
• Other
10 If "biomonitoring" was
• Blood
• For biomonitoring matrix, if PFAS is
selected, indicate the
• Serum
measured in serum, select serum (and not
matrix.
• Plasma
also blood). Only select blood if something
[Select all that apply]
• Maternal blood
more specific is not specified (e.g., cord
• Cord blood
blood, maternal blood, plasma, serum).
• Urine
• Feces
• Breast milk
• Hair
• Saliva
• Nails
• Teeth
• Semen
• Cerebrospinal fluid
• Exhaled breath
• Other
• Glucose
• Maternal serum
• Amniotic fluid
• Maternal Plasma
11 Quantitative Data Extraction (Sub-Forms)
11.1 Health Effect Category
• Cancer
• See Appendix A. 1.6.5.1 for what kind of
[Select one]
• Cardiovascular
health outcomes are grouped under which
• Dermal
health effect category. Please create a
• Developmental
separate form for each outcome.
• Endocrine
• Gastrointestinal
• Hematologic
• Hepatic
• Immune
• Metabolic
A-127
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
• Musculoskeletal/Connective
Tissue
• Nervous
• Ocular
• Reproductive, female
• Reproductive, male
• Respiratory
• Renal
• Other
11.2
Measured
Outcome/Endpoint
[Free-text]
• Describe the measured outcome/endpoint and
start with most relevant word (e.g., "glucose
concentration in serum" preferred to "serum
glucose").
• Provide units in parentheses if relevant and
readily available.
• If the outcome is log transformed, please note
it here:
• Weight (ln-grams)
• Triglyceride (logio-mg/dL)
• Some outcomes are dichotomous (e.g., high
blood pressure, high cholesterol, etc.),
indicate the outcome definition in
parentheses. For example:
• High cholesterol (> 5.0 mg/dL)
11.3
If developmental, when
was the outcome
measured?
[Select all that apply]
• < 2 years of age
• > 2-5 years of age
• > 5 years of age
11.4
PFAS
[Select one]
• PFOA
• PFOS
-
11.5
For neurodevelopmental
outcomes, when was
PFAS exposure
measured?
[Select all that apply]
• Participants
were < 6 months of age
• Participants
were > 6 months of age
11.6
Sub-population
[Free-text]
• If relevant, specify sub-group within the
study (e.g., sex, age group, age at outcome
and/or exposure measurement).
• Leave blank if not applicable.
11.7
N
[Free-text]
• N should be for everyone in the analysis, not
just one exposure/comparison group.
However, if extracting results for specific
population subgroups (age category, gender-
specific) and if reported, the N should reflect
the number of participants in that specific
sub-group (e.g., number of boys in the male-
specific result extracted).
11.8
Exposure Levels
[Free-text]
• Exposure level should be for everyone in the
analysis, not just one comparison group.
• Ideally extract median and the 25th-75th
percentile range for PFAS being extracted.
A-128
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
The following format is preferred:
median = xx (units) (25th-75th percentile:
xx-xx).
> Provide labels and units (e.g.,
median = xx (units) (range: min-max: xx-
xx)).
> If median is not available, please extract
other measures of distribution, such as mean
or geometric mean, range, other percentiles.
> Extract levels for the overall study
population. If only available by subgroups,
specify which subgroup.
1 Example:
1 Males: median =6.4 ng/mL (25th-
75th percentile: 3.6-9.2 ng/mL); Females:
median = 5.8 ng/mL (25th-75th percentile:
3.1-8.3 ng/mL)
> Note: sometimes manuscripts will incorrectly
use IQR rather than 25th-75th percentile.
The IQR is the difference between the 75th
and the 25th percentile, so it should be a
single number, not a range. If a range is
labeled IQR, please use "25th-75th
percentile."
11.9 % with Negligible
Exposure (e.g., below the
LOD)
[Free-text]
< Number of samples below LOD/LOQ; do not
include the percent sign.
> Leave blank if not reported.
11.10 Description of the Effect
Estimate, including
Comparison Group if
applicable
[Free-text]
• Describe the effect estimate, including
comparison group if applicable.
• Brief description of the effect estimate:
describe the comparison being made (e.g.,
beta regression coefficient for IQR increase;
OR for Q2 vs. Ql). Make sure to specify unit
change for continuous measures (e.g., 1 ln-
unit, IQR change, SD increase).
• Use ln() over log() for natural log
transformations. If not In, specify log (base)
(e.g., logioor log(10)).
Good Examples/Formatting:
• regression coefficient (per l-log2 ng/mL
increase in PFOA)
• OR (per 1-ln ng/mL increase in estimated
plasma PFOS)
• OR (for Q2 vs. Ql)
• OR [for Q2 (0.83-1.4 ng/mL) vs. Ql
(< 0.83 ng/mL)]
• OR [fortertile 2 (0.83-1.4 ng/mL) vs. tertile
1 (< 0.83 ng/mL)]
Bad Examples/Formatting:
• beta coefficient
• linear regression coefficient (standard error)
with one unit increase in log-PFC in adults
A-129
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
11.11 Rank this Comparison
-
• For standalone result of unit change, leave
Group by Exposure
blank.
[Free-text]
• If results are presented for quantiles of
exposure, the comparison group for Q2 to Ql
would be ranked as 1, while Q3 to Ql would
be ranked as 2.
11.12 Effect Estimate Type
• Odds Ratio (OR)
• If the effect estimate is a regression
[Select one]
• Relative Risk Ratio (RR)
coefficient (a beta or (3), select from the menu
• Absolute Risk %
"Regression Coefficient" rather than "Beta
• Beta Coefficient (b)
Coefficient."
• Beta Coefficient
• If PFOS/PFOA was the outcome of interest
(standardized)
(e.g., study looked at the impact of a disease
• Standardized Mortality
on PFOS/PFOA level), please still extract the
Ratio (SMR)
data but make a note under the Results
• Standardized Incidence
Comments (11.19).
Ratio (SIR)
• Incidence Risk Ratio (IRR)
• Absolute Risk
Reduction/Risk Difference
(ARR or RD)
• Hazard Ratio (HR)
• Comparison of Means
• Incidence Rate Ratio
• Comparison of Means
• Spearman's Correlation
Coefficient
• Correlation Coefficient
• Percent Incidence
• Regression Coefficient
• Proportionate Mortality
Ratio (PMR)
• Mean Difference
• Percent Difference
• Percent Change
• Benchmark Dose (BMD)
• Mean
• Geometric Mean
• Least Square Means (LSM)
• Geometric Mean Ratio
• Fecundability Ratio
• Adjusted r2
• Mean Ratio
• Prevalence Ratio (PR)
11.13 Effect Estimate - • Only report the effect estimate from the
[Free-text] adjusted model. If there are multiple
adjustment sets, use the final model.
• Do not extract the reference group (1) for
results comparing exposure levels (i.e.,
extract OR (for Q2 vs Ql), but don't extract
the OR of 1 for the reference group Ql).
A-130
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
11.14 CI LCL: Confidence - -
Interval - Lower
Confidence Limit
[Free-text]
11.15 CI UCL: Confidence
Internal - Upper
Confidence Limit
[Free-text]
11.16 SD or SE
[Free-text]
• Enter the SD or SE if reported for the effect
estimate.
• Leave blank if not reported.
11.17 p-value
[Free-text]
• Enter the quantitative p-value if available
(e.g., "0.0001" or "<0.001")
o If the study/table only indicates that p-
value is not significant, enter "ns" for not
significant,
o If the p-value is not reported or does not
apply to the estimate being reported,
leave blank,
o If table footnote mentioned "*p<0.05" for
the results with *, then enter < 0.05. If
results do not have a * and no p-value
was reported, then leave blank,
o If the p-value is not reported and
text/methods mention significance level is
0.05, and:
¦ the text mentioned the specific result
is statistically significant, then
enter <0.05 (and make a note in the
Results Comments (11.19) which
page is this from).
¦ the text mentioned a result as not
statistically significant, then enter
"ns" (and make a note in the Results
Comments (11.19) which page is this
from).
• Make sure the p-value reported corresponds
to the regression coefficient being extracted.
Authors will occasionally report p-values for
other things such as the model fit.
• Other types of p-values such as interaction p-
values or trend p-values are reported, these
can be placed in Results Comments (11.19).
11.18 Covariates in Model - • If there are multiple adjustment sets, list
[Free-text] covariates in the final model, but make a note
in the comment field on the main form (14).
that additional adjustment sets were available
for sensitivity analyses.
• List just the covariates, no need to add
"adjusted for..."
• Example: age, gender, race, SES
11.19 Results Comments - • Enter the location of the extracted data (e.g.,
[Free-text] "Table 3" or "in-text p. 650").
A-131
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
• Enter any relevant p-values, such as
interaction p-values or trend p-values.
• Enter any additional details on the outcome
measurement or definition.
12
Select PFOS or PFOA if
it was measured in the
study but not analyzed
with health effects.
• PFOS
• PFOA
13
Correlations across the
included PFAS presented
in paper or supplement.
[Select one]
• Yes
• No
• Note whether the main manuscript or the
supplemental material present a table or text
describing the (Spearman) correlation
coefficients between concentrations of PFAS
included in the paper.
14
Comments
Include brief description of
results provided in
supplemental materials but
not extracted (e.g.,
stratified analyses,
sensitivity analyses).
[Free-text]
• Briefly mention if effect modification is
analyzed but not extracted (e.g., stratified
analyses by race, by BMI categories, etc.).
Note: Stratification by sex and age should
always be extracted.
• Do not need to specify how values below the
LOD were handled.
• If data is presented by sub-group/strata
(e.g., race) in the supplemental material, just
note that here. Note: Stratification by sex
and age should always be extracted.
• Briefly, describe any other supplemental
results (e.g., sensitivity analyses, etc.) here;
no need to list all confounders other models
adjusted for.
• Any outcome definitions if study specific
(e.g., how was elevated ALT defined in a
study reporting ORs of elevated ALT).
Notes: QC = quality control; NOAEL = no-observed-adverse-effect level; LOAEL = lowest-observed-adverse-effect level;
IQR = interquartile range; LOD = limit of detection; LOQ = limit of quantification; SES = socioeconomic status; BMI = body
mass index; ALT = alanine transaminase.
A. 1.8.1 Epidemiological Study Design Definitions
Epidemiological studies with cross-sectional, cohort, case-control, ecological, or controlled trial
study designs were included. The study design definitions shown in Table A-37 were used
throughout full-text screening and data extraction for epidemiological studies.
Table A-37. Epidemiological Study Design Definitions
Study Design Description
Cross-sectional Exposure and outcome are examined at the same point in time in a defined study
population. Cannot determine if exposure came before or after outcome.
Cohort A group of people is examined over time to observe a health outcome. Everyone
belongs to the same population (e.g., general U.S. population; an occupational group;
cancer survivors). All cohort studies (prospective or retrospective) consider exposure
data from before the occurrence of the health outcome.
A-132
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Design Description
Case-control Cases (people with the health outcome) and controls (people without the health
outcome) are selected at the start of a study. Exposure is determined and compared
between the two groups. A case-control study can be nested within a cohort.
Ecological The unit of observation is at the group level (e.g., zip code; census tract), rather than the
individual level. Ecological studies are often used to measure prevalence and incidence
of disease. Cannot make inferences about an individual's risk based on an ecological
study.
Controlled Trial Exposure is assigned to subject and then outcome is measured.
A-133
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.9 Data Extraction for Animal Toxicological Studies
All animal toxicological studies identified as PECO-relevant after full-text screening in DistillerSR were eligible for data extraction.
As noted in the IRIS Handbook {U.S. EPA, 2022, 10476098}, during data extraction, relevant results from each study are extracted to
facilitate organization, visualization, comparison, and analysis of findings and results. PECO-relevant animal toxicological studies that
received a medium or high confidence study quality evaluation rating were extracted.
Data extraction was carried out using a set of structured forms in HAWC (Table A-38). Studies slated for extraction were pre-screened
by an expert toxicologist who identified the relevant results. Extraction was performed by one reviewer and then independently
verified by at least one other reviewer for quality control. Any conflicts or discrepancies were resolved by discussion and confirmation
with a third reviewer.
Table A-38. HAWC Form Fields for Data Extraction of Animal Toxicological Studies
Questions/Prompts and Options Suggested Considerations
1 Experiment
1.1 NameField • Name should be short and simple. For example,'28-Day Oral' '2-Year Drinking Water','1-Week
[Free-text] Inhalation'.
• Reproductive/developmental if appropriate, then route of exposure (oral/inhalation), not number of
generations or acute/short-term/sub-chronic/chronic.
• If a study includes multiple experiments (e.g., multiple species, varied exposure durations), create separate
experiments for each.
• For reproductive and/or developmental studies, select 'reproductive' or 'developmental' as appropriate
(recognizing that a study may contain both reproductive and developmental endpoints, but is typically defined
as one or the other based on design).
• In general, use reproductive when the study begins treatment prior to mating and continues through birth and
in some cases through a second generation. These studies will typically evaluate reproductive outcomes in the
dams (e.g., copulation and fertility indices, numbers of corpora lutea and implantation sites, pre- and post-
implantation loss). Use developmental when the exposure occurs during gestation and dams are sacrificed
prior to birth. These studies are typically focused on the pups and evaluate viability, developmental
milestones, and other growth and developmental effects in pups and primarily they are looking for
abnormalities in the pups.
• If reproductive or developmental are selected, indicate if there are data for more than one generation.
1.3 Chemical Name Field • Enter the preferred name of the chemical (i.e., PFOA or PFOS).
[Free-text] • Refer to the PECO statement in for a list of synonyms for each chemical.
1.2 Type Field
[Select one]
A-134
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Questions/Prompts and Options
Suggested Considerations
1.4
Chemical Identifier (CAS) Field
[Free-text]
• Be sure to include the dashes in the CAS number.
• The CAS number for the chemical can be found in the PECO statement if they are not listed in the paper.
1.5
Chemical Source Field
[Free-text]
• If the chemical source is not provided by the authors, add in "Not Reported" to this field.
1.6
Chemical Purity Fields
[Checkbox]
• As a default, the 'Chemical purity available?' box will be checked. If the box is checked, entries for 'Purity
qualifier' and 'Chemical purity (%)' are required.
• Uncheck this box if chemical purity information is not available.
2
Animal Group
2.1
Name Field
[Free-text]
• Name should include sex, common strain name, and species (e.g., Male Sprague Dawley Rats).
• For reproductive or developmental studies, include the generation before sex in title (e.g., F1 Male Sprague
Dawley Rats orPO Female C57 Mice).
• If a study combines male and female subjects into one group, use "Male and Female" (e.g., Male and Female
Sprague Dawley Rats).
• If gender is unclear, do not mention (e.g., Sprague Dawley Rats).
• Use the plural form for species (e.g., Rats, Mice).
2.2
Animal Source and Husbandry Field
[Free-text]
• Copy and paste details directly from the paper using quotation marks.
• If the authors do not provide the animal source, add in "Not Reported" to this field.
• For multigenerational reproductive or developmental studies, the animal group dosed might be the parental
(or P0) group. For example, a P0 female rat may be dosed during pregnancy and/or lactation, and
developmental effects are then measured in offspring—or Fi animals.
• For a multigenerational study, specify the 'Generation'.
3
Add Dosing Regime
3.1
Exposure Duration (Days) Field
[Free-text]
• Decimals are allowed, so a 4h single day study can be represented as 0.17 days. However, decimals are likely
not needed for the PFOA/PFOS project since acute studies are not PECO relevant.
3.2
Exposure Duration (Text) Field
[Free-text]
• For all time units, use the following abbreviations: year = yr; month = mo; week = wk; day = d; hour = hr;
minute = min; second = sec.
• Eliminate unnecessary space between length of time and unit (i.e., "2wk" instead of "2 wk").
3.3
Description Field
[Free-text]
• Include dosing description from materials and methods. Be sure to use quotation marks around all text
directly copied/pasted from the paper.
• Include any information on how dosing solutions were prepared.
• Summarize any results the authors present on analytical work conducted to confirm dose, stability, and
purity.
A-135
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Questions/Prompts and Options Suggested Considerations
3.4 Dose-Groups Field • Dose groups should be listed lowest to highest (dose group 1 = 0 mg/kg-d).
[Free-text] • For visualization purposes dose units need to be in mg/kg-d. For studies that provide the units, please use
those for extraction purposes.
• For dietary or drinking water studies, if they provide BOTH concentration of the dose formulation (e.g., ppm)
AND doses as mg/kg-d, please extract both.
• For dietary or drinking water studies that ONLY provide the dose concentration, enter the dose concentrations
as reported in the study and then utilize the conversions spreadsheet to convert the dosage into mg/kg-day
(note that mg/kg body weight/day is the same as mg/kg-d so you just need to use the mg/kg-d).
• If PFOA/PFOS are administered as salts and the doses are presented as salts of PFOA/PFOS, please contact
senior-level extractors before using the conversion spreadsheet.
• If converting doses, add in "Data extractor calculated [PFOS/PFOA] equivalent doses for mg/kg-day" into the
"Description" box.
• When defining the dosing regime for a multigenerational experiment, creating a new dosing regime may not
be needed; instead specify the existing dosing regime of the Po (dosed during gestation and/or lactation).
• A new dosing regime may be needed if offspring were exposed after weaning and, if applicable, acknowledge
parental exposure in the 'Description' field on the 'Dosing regime' page.
• If the authors provide internal measurements of PFOS/PFOA in any tissue, add this information in as an
additional dose group using the mean tissue levels as the value and the tissue as part of the dose units (e.g.,
mg/kg bone, ppm brain).
4
Endpoints (General)
4.1
Endpoint Name Field
[Free-text]
• Name should not include descriptive information captured in other fields within HAWC such as sex, strain,
species, duration, route, etc.
• Include common abbreviation in parenthesis if applicable.
• Endpoint detail should be added after main endpoint, ex. "Body Weight, Fetal" NOT "Fetal Body Weight".
• In general, specific endpoint names are used except for general categories such as 'Clinical Observations' or
histopathology (e.g., 'Kidney Histopathology'), which may comprise a number of observational endpoints.
• Examples: Liver Weight, Relative; Triiodothyronine (T3)
4.2
System Field
[Free-text]
• Represents the appropriate system for the endpoint.
• Examples: Hepatic; Endocrine
4.3
Organ (and Tissue) Field
[Free-text]
• Represents the appropriate organ or tissue for the endpoint.
• Examples: Liver; Thyroid
4.4
Effect and Effect Subtype Fields
[Free-text]
• Represents the appropriate system for the endpoint.
• Examples: Hepatic; Endocrine
4.5
Observation Time Fields
[Free-text]
• The 'Observation time' text field is included in visualizations and should be filled in; the 'Observation time'
numeric field and 'Observation time units' can be left blank.
A-136
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Questions/Prompts and Options
Suggested Considerations
• For all time units, use the following abbreviations: year = yr; month = mo; week = wk; day = d; hour = hr
• Eliminate unnecessary space between length of time and unit (i.e., "2wk" instead of "2 wk").
• Example: 2yr; 6hr; 45d; 90min
• For developmental and reproductive studies, specify observation time in terms of development (e.g., GD 16,
PNDo).
4.6
Values Estimated Field
[Free-text]
• If data was extracted from a figure into HAWC using a measured ruler, check this box.
• For data requiring a digital ruler, use the WebPlotDigitizer tool: https://apps.automeris.io/wpd/.
• If there are multiple time points, extract only the latest time point (i.e., end of treatment) or if the last time
point is not significant and an earlier time point is, extract the earlier time point (this information should be
provided in the data to extract instructions, but this is the general rule in case there are no instructions
provided).
• Provide additional information in the results comment box to make note of what happened at other timepoints
that were not extracted.
4.7
Litter Effects Field
[Free-text]
• If the experiment type has been identified as either 'reproductive' or 'developmental', the 'Litter effects' will
be required, and a choice other than 'not applicable' must be selected.
4.8
Dataset Type Field
[Free-text]
• Select the appropriate dataset type for the endpoint. In general, 'Dataset type' is continuous except for
incidence data, which is dichotomous.
4.9
NOAEL and LOAEL Fields
[Free-text]
• Be sure to enter the significance level (e.g., 0.05) for significant results as well as NOAEL/LOAEL.
• The NOAEL is the highest dose at which there was not an observed toxic or adverse effect. If the LOAEL is
the lowest (non-control) dose, then NOAEL should be , not 0.
• The LOAEL is the lowest dose at which there was an observed toxic or adverse effect. These fields are
critical to the visualizations. If there is no LOAEL, leave as .
• In cases where the study authors did not conduct statistical tests, use the study authors conclusions to indicate
where effects occur. Just make sure to note in the results comments that these were based on author
conclusions and no statistical testing was conducted.
4.10 Statistical Test Field
[Free-text]
• If the statistical test is not provided in the study, add "Not Reported" to the text field.
4.11
Results Notes Field
[Free-text]
• If needed, copy and paste details into this field using quotation marks. Although the methods text field can
describe all methods used, results comments should be more endpoint specific.
5
Endpoint (Dummy Variables)
Data to be extracted using dummy
variables for the following reasons:
• Results that are qualitatively
discussed in the text, but actual data
are not provided.
• For endpoints for which no quantitative data are provided, create the endpoint as described above with the
exceptions below.
• 'Dataset type' is dichotomous or continuous based on the data type if there were data available.
• For 'Response units,' use whatever units correspond to the effect for which you are creating the dummy
variable (e.g., 'incidence' for histopathology observations, 'grams' for body weight)
• Under 'Dose-response data', fill in with a dummy variable. Use 0 to indicate no change from control, a 1 to
indicate an increase from control and a -1 to indicate a decrease from the control.
A-137
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Questions/Prompts and Options
Suggested Considerations
• For instances where study authors
specify that only the significant
effects are described - and certain
endpoints are then not discussed -
assume that no change occurred in
these endpoints. Create dummy
variables for all endpoints stated to be
measured with the assumption if they
are not discussed they were not
significant and make sure to
document this in the results comments
field.
• If an endpoint is discussed in the
methods, but there is no mention at all
in the results (even to indicate that
only significant effects were
reported), then create an endpoint
only and do not extract any data. In
this case, uncheck the 'data reported'
and 'data extracted' boxes on the
endpoint page.
• Organs/tissues that were examined for
histopathological changes, but no
changes were noted.
• Clinical observations in which
multiple clinical signs or general
observations are grouped together.
' 'Significance Level' should be populated if the author indicates significance. Otherwise, 'Significance Level'
is left blank.
' Multiple clinical observations can be grouped together into a single endpoint.
' Example: create an endpoint for clinical observations and add dummy variables to indicate no effect.
' If a single endpoint called "Clinical Observation," create the dummy variables above using all 0 with nothing
tagged as significant.
¦ Or if there was an effect, still create a single endpoint called "Clinical Observation" and then put a 1 at the
dose where the effects were observed and then in the results comment field indicate the effects that were
observed. This would be common in reproductive and developmental studies; indicate if there were "Clinical
Observations in Dams" and where they occurred but didn't want to have a separate endpoint for each
observation.
' Example: for any organ listed but not specified any lesions to extract, create a histopathology endpoint and
create a dummy variable to indicate no treatment-related effect.
' Create an endpoint for each organ (e.g., Liver Histopathology, Kidney Histopathology, Uterus
Histopathology), and create the dummy variables described above using all 0 with nothing tagged as
significant.
¦ Whenever using dummy variables instead of actual data, make sure to note in the results comment text box
that the data are dummy variables using the standard language given in the instructions in HAWC under the
'Results notes' box.
Notes: NOAEL = no-observed-adverse-effect level; LOAEL = lowest-observed-adverse-effect level; CAS = Chemical Abstracts Service.
A-138
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.10 Evidence Synthesis and Integration
For the purposes of this assessment, evidence synthesis and integration are considered distinct
but related processes. For each assessed health effect, the evidence syntheses provide a summary
discussion of each body of evidence considered in the review, considering the conclusions from
the individual study quality evaluations. Syntheses of the evidence for human and animal health
effects are based primarily on studies of high and medium confidence; low confidence results
were given less weight compared to high or medium confidence results during evidence synthesis
and integration. However, in certain instances (i.e., for health outcomes for which few or no
studies with higher confidence are available), low confidence studies might be used to help
evaluate consistency, or if the study designs of the low confidence studies address notable
uncertainties in the set of high or medium confidence studies on a given health effect.
The available human and animal evidence pertaining to the potential health effects of PFOS were
synthesized separately, and a summary discussion of the available evidence was developed for
each evidence stream. Available mechanistic evidence was also considered in the development
of each synthesis. Strength-of-evidence judgments were made for each health outcome within
each evidence stream (i.e., human or animal) using standard terminology (i.e., robust, moderate,
slight, indeterminate) and definitions according to the framework described in the IRIS
Handbook and outlined in Table A-39 and Table A-40.
Following evidence synthesis, the evidence for human and animals were integrated for each
health outcome. Integrated judgements were drawn across all lines of evidence for each assessed
health outcome as to whether and to what extent the evidence supports that exposure to PFOS
has the potential to be hazardous to humans. The evidence integration provided a summary of the
causal interpretations from the available studies, as well as mechanistic evidence and other
supplemental information. Mechanistic evidence was organized by signaling pathway or other
categories (e.g., key characteristics of carcinogens) as relevant to each outcome. The integrated
judgments are developed through structured review of the evidence against an established set of
considerations for causality. These considerations include risk of bias, sensitivity, consistency,
strength (effect magnitude) and precision, biological gradient/dose-response, coherence, and
mechanistic evidence related to biological plausibility. During evidence integration, a structured
and documented process was used, as follows:
• Summarize human and animal health effect studies in parallel but separately, using the set
of considerations for causality first introduced by Austin Bradford Hill {Hill, 1965,
71664} and relevant mechanistic evidence (or mode of action (MOA) understanding).
• Identify strength of the human and animal health evidence in light of inferences across
evidence streams.
• Summarize judgment as to whether the available evidence base for each potential health
outcome as a whole indicates that PFOS exposure has the potential to cause adverse health
effects in humans (see Table A-41) ("evidence demonstrates," "evidence indicates
(likely)," "evidence suggests," "evidence is inadequate," or "strong evidence supports no
effect").
The decision points within the structured evidence integration process are summarized in an
evidence profile table for each assessed health effect.
A-139
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table A-39. Framework for Strength-of-Evidence Judgments for Epidemiological Studies3
Strength-of-
Evidence
Judgment
Description
Robust A set of high- or medium-confidence studies reporting an association between the exposure and
(©©©) the health outcome, with reasonable confidence that alternative explanations, including chance,
bias, and confounding, can be ruled out across studies. The set of studies is primarily consistent,
with reasonable explanations when results differ; and an exposure response gradient is
demonstrated. Supporting evidence, such as associations with biologically related endpoints in
human studies (coherence) or large estimates of risk or severity of the response, may help to rule
out alternative explanations. Similarly, mechanistic evidence from exposed humans may serve to
address uncertainties relating to exposure-response, temporality, coherence, and biological
plausibility (i.e., providing evidence consistent with an explanation for how exposure could cause
the health effect based on current biological knowledge) such that the totality of human evidence
supports this judgment.
Moderate • Multiple studies showing generally consistent findings, including at least one high or medium
(©©O) confidence study and supporting evidence, but with some residual uncertainty due to potential
chance, bias, or confounding (e.g., effect estimates of low magnitude or small effect sizes
given what is known about the endpoint; uninterpretable patterns with respect to exposure
levels). Associations with related endpoints, including mechanistic evidence from exposed
humans, can address uncertainties relating to exposure response, temporality, coherence, and
biological plausibility, and any conflicting evidence is not from a comparable body of higher
confidence, sensitive studies
• A single high- or medium-confidence study demonstrating an effect with one or more factors
that increase evidence strength, such as: a large magnitude or severity of the effect, a dose-
response gradient, unique exposure or outcome scenarios (e.g., a natural experiment), or
supporting coherent evidence, including mechanistic evidence from exposed humans. There
are no comparable studies of similar confidence and sensitivity providing conflicting evidence,
or if there are, the differences can be reasonably explained (e.g., by the population or exposure
levels studied)
Slight One or more studies reporting an association between exposure and the health outcome, where
(©OO) considerable uncertainty exists:
• A body of evidence, including scenarios with one or more high or medium confidence studies
reporting an association between exposure and the health outcome, where either (1) conflicting
evidence exists in studies of similar confidence and sensitivity (including mechanistic evidence
contradicting the biological plausibility of the reported effects), a (2) a single study without a
factor that increases evidence strength (factors described in moderate), OR (3) considerable
methodological uncertainties remain across the body of evidence (typically related to exposure
or outcome ascertainment, including temporality), AND there is no supporting coherent
evidence that increases the overall evidence strength.
• A set of only low confidence studies that are largely consistent.
• Strong mechanistic evidence in well-conducted studies of exposed humans (medium or high
confidence) or human cells, in the absence of other substantive data, where an informed
evaluation has determined that the data are reliable for assessing the health effect of interest
and the mechanistic events have been reasonably linked to the development of that health
effect.
Indeterminate • No studies in humans or well-conducted studies of human cells.
(OOO) • Situations when the evidence is highly inconsistent and primarily of low confidence.
• May include situations with medium or high confidence studies, but unexplained heterogeneity
exists (in studies of similar confidence and sensitivity), and there are additional outstanding
concerns such as effect estimates of low magnitude, uninterpretable patterns with respect to
exposure levels, or uncertainties or methodological limitations that result in an inability to
discern effects from exposure.
A-140
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Strength-of-
Evidence Description
Judgment
• A set of largely null studies that does not meet the criteria for compelling evidence of no
effect, including evidence bases with inadequate testing of susceptible populations and
lifestages.
Several -confidence studies showing null results (for example, an odds ratio of 1.0), ruling
out alternative explanations including chance, bias, and confounding with reasonable confidence.
Each of the studies should have used an optimal outcome and exposure assessment and adequate
sample size (specifically for higher exposure groups and for susceptible populations). The set as a
whole should include the full range of levels of exposures that human beings are known to
encounter, an evaluation of an exposure response gradient, and an examination of at-risk
populations and lifestages.
Notes:
a Table slightly adapted from Table 11-3 in the IRIS Handbook.
Table A-40. Framework for Strength-of-Evidence Judgments for Animal Toxicological
Studies"
Compelling
evidence of no
effect ( )
Strength-of-
Evidence
Judgment
Description
Robust A set of high- or medium-confidence studies with consistent findings of adverse or
(©©©) toxicologically significant effects across multiple laboratories, exposure routes, experimental
designs (e.g., a subchronic study and a two-generation study), or species; and the experiments
reasonably rule out the potential for nonspecific effects to have caused the effects of interest. Any
inconsistent evidence (evidence that cannot be reasonably explained based on study design or
differences in animal model) is from a set of experiments of lower confidence or sensitivity. To
reasonably rule out alternative explanations, multiple additional factors in the set of experiments
exist, such as: coherent effects across biologically related endpoints; an unusual magnitude of
effect, rarity, age at onset, or severity; a strong dose-response relationship; or consistent
observations across animal lifestages, sexes, or strains. Similarly, mechanistic evidence (e.g.,
precursor events linked to adverse outcomes) in animal models may exist to address uncertainties
in the evidence base such that the totality of animal evidence supports this judgment.
Moderate • At least one high- or medium-confidence study with supporting information increasing the
(©©O) strength of the evidence. Although the results are largely consistent, notable uncertainties
remain. However, in scenarios when inconsistent evidence or evidence indicating nonspecific
effects exist, it is not judged to reduce or discount the level of concern regarding the positive
findings, or it is not from a comparable body of higher confidence, sensitive studies. The
additional support provided includes either consistent effects across laboratories or species;
coherent effects across multiple related endpoints; an unusual magnitude of effect, rarity, age
at onset, or severity; a strong dose-response relationship; or consistent observations across
exposure scenarios (e.g., route, timing, duration), sexes, or animal strains. Mechanistic
evidence in animals may serve to provide this support or otherwise address residual
uncertainties.
• A single high or medium confidence experiment demonstrating an effect in the absence of
comparable experiment(s) of similar confidence and sensitivity providing conflicting evidence,
namely evidence that cannot be reasonably explained (e.g., by respective study designs or
differences in animal model).
Slight • Scenarios in which there is a signal of a possible effect, but the evidence is conflicting or
(©OO) weak:
• A body of evidence, including scenarios with one or more high or medium confidence
experiments reporting effects but without supporting or coherent evidence (see description in
moderate) that increases the overall evidence strength, where conflicting evidence exists from
A-141
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Strength-of-
Evidence Description
Judgment
a set of sensitive experiments of similar or higher confidence (including mechanistic evidence
contradicting the biological plausibility of the reported effects).
• A set of only low confidence experiments that are largely consistent.
• Strong mechanistic evidence in well-conducted studies of animals or animal cells, in the
absence of other substantive data, where an informed evaluation has determined the assays are
reliable for assessing the health effect of interest and the mechanistic events have been
reasonably linked to the development of that health effect.
• No animal studies or well-conducted studies of animal cells.
• The available models (not considering human relevance) or endpoints are not informative to
the hazard question under evaluation.
• The evidence is inconsistent and primarily of low confidence.
• May include situations with medium or high confidence studies, but there is unexplained
heterogeneity and additional concerns such as small effect sizes (given what is known about
the endpoint) or a lack of dose-dependence.
• A set of largely null studies that does not meet the criteria for compelling evidence of no
effect.
A set of high confidence experiments examining a reasonable spectrum of endpoints relevant to a
type of toxicity that demonstrate a lack of biologically significant effects across multiple species,
both sexes, and a broad range of exposure levels. The data are compelling in that the experiments
have examined the range of scenarios across which health effects in animals could be observed,
and an alternative explanation (e.g., inadequately controlled features of the studies' experimental
designs; inadequate sample sizes) for the observed lack of effects is not available. The
experiments were designed to specifically test for effects of interest, including suitable exposure
timing and duration, post exposure latency, and endpoint evaluation procedures, and to address
potentially susceptible populations and lifestages. Mechanistic data in animals (in vivo or in
vitro) that address the above considerations or that provide information supporting the lack of an
association between exposure and effect with reasonable confidence may provide additional
support such that the totality of evidence supports this judgment.
Notes:
a Table slightly adapted from Table 11-4 in the IRIS Handbook.
Table A-41. Evidence Integration Judgments for Characterizing Potential Human Health
Effects in the Evidence Integration"
Evidence
integration Explanation and example scenarios
judgment level
Evidence A strong evidence base demonstrating that [chemical] exposure causes [health effect] in humans
demonstrates • For when there is robust human evidence supporting an effect
• Could also be used when there is moderate human evidence and robust animal evidence if
there is strong mechanistic evidence that MOA(s) or key precursors identified in animals are
expected to occur and progress in humans
Evidence An evidence base that indicates that [chemical] exposure likely causes [health effect] in humans,
indicates although there may be outstanding questions or limitations.
(likely) • Used if there is robust animal evidence supporting an effect and slight or indeterminate human
evidence, or with moderate human evidence when strong mechanistic evidence is lacking
• Could also be used with moderate human evidence supporting an effect and slight or
indeterminate animal evidence, or with moderate animal evidence supporting an effect and
slight or indeterminate human evidence. In these scenarios, any uncertainties in the moderate
evidence are not sufficient to substantially reduce confidence in the reliability of the evidence,
Indeterminate
(OOO)
Compelling
evidence of no
effect ( )
A-142
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence
integration
judgment level
Explanation and example scenarios
Evidence
suggests
or mechanistic evidence in the slight or indeterminate evidence base (e.g., precursors) exists to
increase confidence in the reliability of the moderate evidence
A decision between "evidence indicates" and "evidence suggests" considers the extent to which
findings are coherent or biologically consistent across lines of evidence streams, and may
incorporate other supplemental evidence (e.g., structure-activity data; chemical class information)
An evidence base that suggests that [chemical] exposure may cause [health effect] in humans, but
there are very few studies that contributed to the evaluation, the evidence is weak or conflicting,
and/or the methodological conduct of the studies is poor.
• Used if there is slight human evidence and indeterminate or slight animal evidence
• Used with slight animal evidence and indeterminate or slight human evidence
• Could also be used with moderate human evidence and slight or indeterminate animal
evidence, or with moderate animal evidence and slight or indeterminate human evidence. In
these scenarios, there are outstanding issues regarding the moderate evidence that substantially
reduced confidence in the reliability of the evidence, or mechanistic evidence in the slight or
indeterminate evidence base (e.g., null results in well-conducted evaluations of precursors)
exists to decrease confidence in the reliability of the moderate evidence
• When there is general scientific understanding of mechanistic events that result in a health
effect, this judgment level could also be used if there is strong mechanistic evidence that is
sufficient to highlight potential human toxicity in the absence of informative conventional
studies in humans or in animals
This conveys either a lack of information or an inability to interpret the available evidence for
[health effect]. On an assessment-specific basis, a single use of this "evidence inadequate"
judgment might be used to characterize the evidence for multiple health effect categories.
• Used if there is indeterminate human and animal evidence
• Used if there is slight animal evidence and compelling evidence of no effect human evidence
• Could also be used with slight or robust animal evidence and indeterminate human evidence if
strong mechanistic information indicated that the animal evidence is unlikely to be relevant to
humans
Extensive evidence across a range of populations and exposure levels has identified no
effects/associations. This scenario requires a high degree of confidence in the conduct of
individual studies, including consideration of study sensitivity, and comprehensive assessments
of the endpoints and lifestages of exposure potentially relevant to the heath effect of interest.
• Used if there is compelling evidence of no effect in human studies and compelling evidence of
no effect or indeterminate animal evidence
• Also used if there is indeterminate human evidence and compelling evidence of no effect
animal evidence in models judged as relevant to humans
• Could also be used with compelling evidence of no effect in human studies and moderate or
robust animal evidence if strong mechanistic information indicated that the animal evidence is
unlikely to be relevant to humans
Evidence
inadequate13
Strong evidence
supports no
effect
Notes:
a Table adapted from Table 11-5 in the IRIS Handbook.
b An "evidence inadequate" judgment is not a determination that the chemical does not cause the indicated human health
effect(s), but rather an indication that the available evidence is insufficient to reach a judgment.
1.10.1 Epidemiological Studies Included from HESDs
For all non-priority health outcomes, epidemiological studies identified and reviewed in the 2016
HESD were included in summary paragraphs describing previously reached conclusions for each
health outcome. Study quality was considered but domain-based, structured study quality
A-143
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
evaluations were not performed for 2016 HESD studies. Inferences drawn from evidence in the
current literature search were compared to the results described from 2016 studies.
For the 5 main health outcomes (i.e., developmental, immune, hepatic, cardiovascular and
cancer), epidemiological studies identified and reviewed in the 2016 HESD and other pre-2016
assessments were included in the evidence synthesis, including discussion of study quality
considerations, according to the recommendations from the SAB. Inferences drawn from studies
included from the 2016 HESD were considered in drawing health effects conclusions.
The evidence integration was conducted following the guidance outlined in the "Systematic
Review Protocol for the PFBA, PFHxA, PFHxS, PFNA, and PFDA (anionic and acid forms)
IRIS Assessments" {U.S. EPA, 2020, 8642427}. Briefly, the evidence integration involved
evidence stream evaluation, including evaluation of the qualitative summaries on the strength of
evidence from studies in animals and humans, and inference across evidence streams. Across
evidence streams, human relevance of animal models and mechanistic evidence were considered.
The evidence integration involved an overall judgment on whether there was sufficient evidence
or insufficient evidence for each potential human health effect and an evidence basis rationale.
1.10.2 Epidemiological Studies Excluded from Synthesis
Some epidemiological studies were not included in the evidence synthesis narrative if they
included factors that could lead to overlapping results (e.g., overlapping NHANES studies).
Studies reporting results from the same cohort with the same health outcome were considered
overlapping evidence, and these studies were not discussed in the synthesis narrative to avoid
duplication or overrepresentation of results from the same group of participants. When
participants from the same cohort were included in more than one eligible study, the study with
the largest number of participants was included in the evidence synthesis narrative. In general, to
best gauge consistency and magnitude of reported associations, EPA largely focused on the most
accurate and most prevalent measures. In some cases, such as developmental outcomes, studies
on the same population providing more accurate outcome measures (e.g., birthweight and birth
length for fetal growth restriction) were given preference over studies providing less accurate
outcome measures (e.g., ponderal index for fetal growth restriction). Overlapping studies were
included in study quality figures.
Meta-analyses were considered during evidence integration as support of consistent effects
across studies. Details of the identified meta-analyses and assessment implications are
summarized in Section A.2.
A. 1.11 Dose-Response Assessment: Selecting Studies and
Quantitative Analysis
As noted in the IRIS Handbook, selection of studies and endpoints for dose-response assessment
involves judgments about the data that build from "judgments" and decisions made during earlier
steps of the systematic review and assessment process. EPA guidance and support documents
that describe data requirements and other considerations for dose-response modeling include
EPA's Benchmark Dose Technical Guidance {U.S. EPA, 2012, 1239433}, Review of the
Reference Dose and Reference Concentration Processes {U.S. EPA, 2002, 88824}, Guidelines
A-144
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
for Carcinogen Risk Assessment {U.S. EPA, 2005, 6324329}, and Supplemental Guidance for
Assessing Susceptibility from Early-Life Exposure to Carcinogens {U.S. EPA, 2005, 88823}.
Dose-response assessments are performed for both noncancer and cancer oral health hazards, if
supported by existing data. For noncancer hazards, an oral RfD will be derived when possible.
An RfD is an estimate, with uncertainty spanning perhaps an order of magnitude, of an exposure
to the human population (including susceptible subgroups) that is likely to be without an
appreciable risk of deleterious health effects over a lifetime {U.S. EPA, 2002, 88824}. Reference
values are not predictive risk values; that is, they provide no information about risks at higher or
lower exposure levels.
For cancer hazards, a CSF will be derived to estimate human cancer risk when low-dose linear
extrapolation for cancer effects is supported. A CSF is a plausible upper bound lifetime cancer
risk from chronic ingestion of a chemical per unit of mass consumed per unit body weight per
day (mg/kg-day). In contrast to RfDs, CSFs can be used in conjunction with exposure
information to predict cancer risk at a given dose.
The derivation of reference values will depend on the conclusions drawn during previous steps of
this protocol. Specifically, EPA will attempt dose-response assessments for noncancer outcomes
when the evidence integration judgements indicate stronger evidence of hazard (i.e., evidence
demonstrates and evidence indicates integration judgements). Quantitative analyses are generally
not attempted for other evidence integration conclusions. Similarly, EPA will attempt dose-
response assessments for cancer outcomes for chemicals that are classified as Carcinogenic or
Likely to be Carcinogenic to Humans. When there is Suggestive Evidence of Carcinogenic
Potential to Humans, EPA generally does not conduct dose-response assessment unless a well-
conducted study is available and a quantitative analysis is deemed useful.
A. 1.11.1 Study Selection
Selection of specific endpoints for toxicity value derivation is primarily a result of the evidence
integration and hazard characterization. Specific issues that may be considered for their potential
to affect the feasibility of dose-response modeling for individual data sets are described in more
detail inthq Benchmark Dose Technical Guidance {U.S. EPA, 2012, 1239433}. In general,
studies and endpoints that are most useful for dose-response analysis will generally have at least
one exposure level in the region of the dose-response curve near the benchmark response (BMR;
the response level to be used for deriving toxicity values) to minimize low-dose extrapolation.
Such studies will also have more exposure levels and larger sample sizes overall {U.S. EPA,
2012, 1239433}. These attributes support a more complete characterization of the shape of the
exposure-response curve and decrease the uncertainty in the associated exposure-response metric
(e.g., RfD) by reducing statistical uncertainty in the POD and minimizing the need for low-dose
extrapolation. Some important considerations include:
• human data are preferred over animal data to eliminate interspecies extrapolation
uncertainties,
• animal species known to respond similarly to humans are preferred over studies of other
species,
• high or medium confidence studies are preferred over low confidence studies,
A-145
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• chronic or subchronic studies, or studies encompassing a sensitive lifestage (i.e.,
gestational) are preferred for the derivation of chronic toxicity values over acute studies,
and
• studies with a design or analysis that addresses relevant confounding for a given outcome
are preferred.
The number of studies considered for toxicity value derivation will be reduced based on these
considerations and others described in EPA {2012, 1239433; 2022, 10476098}.
A.1.11.2 Conducting Dose-Response Assessments
Several EPA guidance and support documents provide background for the derivation of toxicity
values {U.S. EPA, 2002, 88824; U.S. EPA, 2005, 6324329; U.S. EPA, 2022, 10476098}. Steps
of the dose-response process include: 1) selecting BMR values; 2) dose characterization and
dose-response modeling, including conversion of administered doses to internal doses (animal
studies only) and conversion of PODs to human equivalence doses; 3) candidate toxicity value
development; 4) characterizing uncertainty; and 5) selection of final toxicity values.
The recommended EPA human health risk assessment (HHRA) approach described in EPA's A
Review of the Reference Dose and Reference Concentration Processes describes a multistep
approach to dose-response assessment, including analysis in the range of observation followed
by extrapolation to lower levels {U.S. EPA, 2002, 88824}. In this effort, EPA conducted a dose-
response assessment to define a POD and extrapolated from the POD to an RfD. For PFOS, EPA
performed benchmark dose (BMD) modeling of animal and human studies to refine the critical
effect POD in deriving the RfD. The BMD approach involves dose-response modeling to obtain
BMDs, i.e., dose levels corresponding to specific response levels near the low end of the
observable range of the data and the lower limit of the BMD (BMDLs) to serve as potential
PODs for deriving quantitative estimates below the range of observation {U.S. EPA, 2012,
1239433}. EPA used several approaches for dose-response modelling. EPA generally used the
publicly available Benchmark Dose Software (BMDS) program developed and maintained by
EPA (https://www.epa.eov/bmds). BMDS fits mathematical models to the data and determines
the dose (i.e., BMD) that corresponds to a pre-determined level of response (i.e., BMR).
Considerations for BMR selection are discussed in detail in EPA's Benchmark Dose Technical
Guidance {U.S. EPA, 2012, 1239433}. For the derivation of RfDs, the BMR selected should
correspond to a low or minimal level of response in a population for the outcome of interest and
is generally the same across assessments, though the BMR could change over time based on new
data or developments. The following general recommendations for BMR selection were
considered for this assessment:
• For dichotomous data (e.g., presence or absence), a BMR of 10% extra risk is generally
used for minimally adverse effects. Lower BMRs (5% or lower) can be selected for severe
or frank effects. For example, developmental effects are relatively serious effects, and
BMDs derived for these effects could use a 5% extra risk BMR. Developmental
malformations considered severe enough to lead to early mortality could use an even
lower BMR {U.S. EPA, 2012, 1239433; U.S. EPA, 2022, 10476098}.
• For continuous data, a BMR is ideally based on an established definition of biologic
significance in the effect of interest. In the absence of such a definition, a difference of
A-146
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
one standard deviation (SD) from the mean response of the control mean is often used and
one-half the standard deviation is used for more severe effects. Note that the standard
deviation used should reflect underlying variability in the outcome to the extent possible
separate from variability attributable to laboratory procedures, etc. {U.S. EPA, 2012,
1239433; U.S. EPA, 2022, 10476098}.
• For outcomes for which there is no accepted percent change that is considered adverse,
EPA used the hybrid approach to derive the BMR.
Deviations of these recommendations, if any, will be described in the assessment.
The preferred approach for dose estimation for dose-response modeling is PBPK modeling
because it can incorporate a wide range of relevant chemical-specific information, describe the
active agent more accurately, and provide a better basis for extrapolation to human equivalent
exposures. For animal studies, EPA used a pharmacokinetic model to make predictions of the
internal dose in laboratory animals used in toxicity studies or in humans based on the
administered dose used in the study (see PFOS MCLG main document for additional detail).
Concentrations of PFOS in blood are considered for all the internal dose-metrics. For animal
studies, this conversion would occur prior to BMD modeling.
If multiple studies are suitable for exposure-response modeling and if no single study is judged
to be appreciably better than the others for the purposes of deriving toxicity values, data or
results from multiple studies may be derived from different studies for comparison. For each
modeled response, a POD from the observed data will be estimated to mark the beginning of
extrapolation to lower doses. The POD is an estimated dose (expressed in human-equivalent
terms) near the lower end of the observed range without significant extrapolation to lower doses.
The POD will be used as the starting point for subsequent extrapolations and analyses. For
noncancer dose-response data not amenable to BMD modeling, a no-observed-adverse-effect
level (NOAEL) or lowest-observed-adverse-effect level (LOAEL) was used as the POD.
Subsequent to POD derivation, EPA used a pharmacokinetic model for human dosimetry to
estimate human equivalent doses (HEDs) from both animal and epidemiological studies. For the
human and animal endpoints of interests, serum concentration was identified, based on the
available data, as a suitable internal dosimetry target. The selected pharmacokinetic models are
discussed in Section 4 of the PFOS Main Document.
A. 1.12 Candidate Toxicity Value Derivation and Selection
For each noncancer data set analyzed for dose-response, reference values are estimated by
applying relevant adjustments to the point-of-departure human equivalent doses (PODheds) to
account for five possible areas of uncertainty and variability: human variation, extrapolation
from animals to humans, extrapolation to chronic exposure duration, the type of POD being used
for reference value derivation, and extrapolation to a minimal level of risk (if not observed in the
data set). The particular value for these adjustments is usually 10, 3, or 1, but different values
based on chemical-specific information may be applied if sufficient information exists in the
chemical database. The assessment discusses the scientific bases for estimating these data-based
adjustments and uncertainty factors (UFs). UFs used in this assessment were applied according
A-147
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
to methods described in EPA's Review of the Reference Dose and Reference Concentration
Processes {U.S. EPA, 2002, 88824}.
• Animal-to-human extrapolation: If animal results are used to make inferences about
humans, the toxicity value incorporates cross-species differences, which may arise from
differences in toxicokinetics or toxicodynamics. If a biologically based model adjusts
fully for toxicokinetic and toxicodynamic differences across species, this factor is not
used. Otherwise, if the POD is standardized to equivalent human terms or is based on
toxicokinetic or dosimetry modeling, a factor of 101/2 (rounded to 3) is applied to account
for the remaining uncertainty involving toxicokinetic and toxicodynamic differences.
• Human variation: The assessment accounts for variation in susceptibility across the human
population and the possibility that the available data may not be representative of
individuals who are most susceptible to the effect. If population-based data for the effect
or for characterizing the internal dose are available, the potential for data-based
adjustments for toxicodynamics or toxicokinetics is considered. Further, "when sufficient
data are available, an intraspecies UF either less than or greater than 10x may be justified
{U.S. EPA, 2002, 88824}. However, a reduction from the default (10) is only considered
in cases when there are dose-response data for the most susceptible population" {U.S.
EPA, 2002, 88824}. This factor is reduced only if the POD is derived or adjusted
specifically for susceptible individuals (not for a general population that includes both
susceptible and non-susceptible individuals) {U.S. EPA, 2002, 88824; U.S. EPA, 1991,
732120}. Otherwise, a factor of 10 is generally used to account for this variation.
• LOAEL to NOAEL: If a POD is based on an LOAEL or a BMDL associated with an
adverse effect level, the assessment must infer an exposure level where such effects are
not expected. This can be a matter of great uncertainty if there is no evidence available at
lower exposures. A factor of up to 10 is generally applied to extrapolate to a lower
exposure expected to be without appreciable effects. A factor other than 10 may be used
depending on the magnitude and nature of the response and the shape of the dose-response
curve.
• Subchronic-to-chronic exposure: If a chronic reference value is being developed, a POD is
based on subchronic evidence, the assessment considers whether lifetime exposure could
have effects at lower levels of exposure. A factor of up to 10 is applied when using
subchronic studies to make inferences about lifetime exposure. A factor other than 10 may
be used, depending on the duration of the studies and the nature of the response. This
factor may also be applied, albeit rarely, for developmental or reproductive effects if
exposure covered less than the full critical period.
• In addition to the adjustments above, if database deficiencies raise concern that further
studies might identify a more sensitive effect, organ system, or lifestage, the assessment
may apply a database UF {U.S. EPA, 2002, 88824; U.S. EPA, 1991, 732120}. The size of
the factor depends on the nature of the database deficiency. For example, EPA typically
follows the suggestion that a factor of 10 be applied if a prenatal toxicity study and a two-
generation reproduction study are both missing, and a factor of 101/2 (rounded to 3) if
either one or the other is missing. A database UF would still be applied if this type of
study were available but considered to be a low confidence study.
A-148
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The POD for a particular RfD is divided by the product of these factors. The RfD review
recommends that any composite factor that exceeds 3,000 represents excessive uncertainty and
recommends against relying on the associated RfD.
For each cancer data set analyzed for dose-response, the approach for extrapolation depends on
the MOA for carcinogenesis (i.e., linear or nonlinear). If the chemical causes cancer through a
mutagenic change to deoxyribonucleic acid (DNA), or if the MOA for causing cancer is not
known, this extrapolation is conducted by drawing a line from the POD to the origin (zero dose,
zero tumors). The slope of the line (Aresponse/Adose) gives the CSF which can be interpreted as
the risk per mg/kg/day. In addition, under the supplemental guidance {U.S. EPA, 2005, 88823},
affirmative determination of a mutagenic MOA (as opposed to defaulting to a mutagenic MOA
based on insufficient data or limited data indicating potential mutagenicity) determines if age-
dependent adjustment factors are applied in the quantification of risk to account for additional
sensitivity of children. A CSF is derived by dividing the BMR by the PODhed.
If the chemical is shown to cause cancer via a MOA that is not linear at low doses, and the
chemical does not demonstrate mutagenic or other activity consistent with linearity at low doses,
a nonlinear extrapolation is conducted. The 2005 guidelines state that "where tumors arise
through a nonlinear MOA, an oral RfD or inhalation reference concentration, or both, should be
developed in accordance with EPA's established practice of developing such values, taking into
consideration the factors summarized in the characterization of the POD" {U.S. EPA, 2005,
88823}.
The next step is to select an organ/system-specific toxicity value for each hazard (cancer and
noncancer) identified in the assessment. This selection can be based on the study confidence
considerations, the most sensitive outcome, a clustering of values, or a combination of such
factors; the rationale for the selection is presented in the assessment. Key considerations for
candidate value selection are described in the IRIS Handbook {U.S. EPA, 2022, 10476098} and
include: 1) the weight of evidence for the specific effect or health outcome; 2) study confidence;
3) sensitivity and basis of the POD; and 4) uncertainties in modeling or extrapolations. The value
selected as the organ/system-specific toxicity value is discussed in the assessment.
The selection of overall toxicity values for noncancer and cancer effects involves the study
preferences described above, consideration of overall toxicity, study confidence, and confidence
in each value, including the strength of various dose-response analyses and the possibility of
basing a more robust result on multiple data sets. The values selected as the overall RfD and CSF
are discussed in the assessment.
A-149
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.2 Meta-Analysis Table
Studies identified in title/abstract and full-text screening as assessments or records with no original data were considered supplemental
material. Meta-analysis studies were included among those secondary studies. Consideration of meta-analyses alongside original
epidemiology studies could lead to duplication of results and give greater weight to studies included in meta-analyses; therefore, meta-
analysis studies were summarized separately. For PFOS, 13 meta-analysis studies were identified and summarized below (Table
A-41).
Table A-42. Epidemiologic Meta-Analysis Studies Identified from Literature Review
Reference
Number of
Studies
Countries
Health Outcome
Results/Conclusions3
Verner et al.
(2015, 3150627)
Canada, Denmark,
Japan, Norway,
Taiwan, United
Kingdom, United
States
Developmental
Birthweight
• Pooled (3 per 1 ng/mL increase of PFOS in maternal or cord blood (6
studies): -5.0 g (-8.9, -1.1)
• Physiologically based pharmacokinetic model simulations suggest that the
association between PFAS levels and birthweight may be confounded by
changes in glomerular filtration rate and due to blood draw timing
Negri et al.
(2017, 3981320b)
Canada, China,
Denmark,
Germany,
Greenland, Japan,
13 Norway, Poland,
South Korea,
Taiwan, Ukraine,
United Kingdom,
United States
Developmental
Birthweight:
• Pooled (3 per 1 ng/mL increase in PFOS in maternal or cord blood (8
studies): -0.92 g (-3.4, 1.6), I2 = 74%
• Pooled (3 per 1-ln ng/mL increase in PFOS in maternal or cord blood (8
studies): -46.1 g (-80.3, -11.9), I2 = 25%
Dzierlenga et al.
(2020, 7643488)
29
Australia, Belgium,
Canada, China,
Denmark,
Greenland, Japan,
Norway, Poland,
South Korea, Spain,
Sweden, Taiwan,
Ukraine, United
Kingdom, United
States
Developmental
Birthweight:
• Pooled (3 per 1 ng/mL increase in PFOS in maternal or cord blood (29
studies): -3.22 g (-5.11, -1.33), I2 = 58.3%
• Pooled (3 per 1 ng/mL in PFOS sampled before or in early pregnancy (8
studies): -1.35 g (-2.33, -0.37), I2 = 5%
• Pooled (3 per 1 ng/mL in PFOS sampled in later pregnancy (21 studies): -
7.15 g (-10.93,-3.41), I2 = 55%
• Meta-regression modeling for timing of blood draw (early r.v. late) showed
that when drawn from before or early pregnancy, there was no significant
relationship between birthweight and PFOS: 0.59 g/ng/mL (-1.94, 3.11)
A-150
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Number of
Studies
Countries
Health Outcome
Results/Conclusions"
Cao et al. (2021,
9959525)
Korea, Spain,
Taiwan, United
States
Developmental
LBW:
• Pooled OR for PFOS in maternal blood (5 studies): 1.32 (1.09, 1.55),
I2 = 0.00%
• Stratified by region: positive association in United States (2 studies):
OR = 1.44(1.15, 1.72)
Deji et al. (2021,
7564388)
21
Brazil, Canada,
China, Denmark,
Norway, Spain,
United States
Developmental,
Female Reproductive
PTB°:
• Pooled OR (16 studies): 1.20 (1.04, 1.38), I2 = 54.3%
• Pooled OR (6 studies in in North America) = 1.09 (1.01, 1.19); I2 = 0%
Miscarriage:
» Pooled OR (6 studies): 1.01, 95% CI: 0.92, 1.10; I2 = 35.9%
Gao et al. (2021,
9959601)
Brazil, Canada,
China, Denmark,
29 Norway, Spain,
Sweden, United
States
Developmental,
Female Reproductive
Preeclampsia:
• Pooled OR per 1-log increase in PFOS (4 studies): 1.27 (1.06, 1.51)
PTB°:
• Pooled OR per 1 ng/mL increase in PFOS (8 studies): 1.01 (1.00-1.02)
GDM (7 studies), miscarriage (2 studies), pregnancy-induced hypertension
(2 studies), SGA (6 studies), LBW (2 studies): Associations not statistically
significant
Yang et al.
(2022, 10176603)
22
Belgium, Canada,
China, Denmark,
Netherlands,
Norway, Slovakia,
Spain, Sweden,
United States
Developmental
PTB:
• Pooled OR (14 studies): 1.54 (1.20, 1.98), I2 = 63.4%
o Significant associations between PFOS and PTB in America [5 studies,
pooled OR = 1.44(1.19, 1.76), I2 = 2.1%]
o Significant associations for PFOS in maternal blood sampled in lst-2nd
trimester and in 3rd trimester to delivery, and for maternal blood sample
type overall
Miscarriage:
• Pooled OR (5 studies): 1.10 (0.93, 1.32), I2 = 0%
SGA:
• Pooled OR (9 studies): 1.22 (0.92, 1.61), I2 = 74.3%
o Significant associations for PFOS in cord blood at delivery [2 studies,
pooled OR = 2.51 (1.45, 4.34), I2 = 0.00%]
• Pooled OR (7 studies): 1.52 (1.19, 1.94), I2 = 19.1%
LBW
• Pooled OR (2 studies, U.S. only): 1.71 (1.19, 2.47), I2 = 0%
• Pooled OR for PFOS in maternal blood (6 studies): 1.48 (1.16, 1.90),
I2 = 22.9%
A-151
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Number of
Studies
Countries
Health Outcome
Results/Conclusions"
Costello et al.
(2022,
10285082b)
25
Asia (NOS),
Europe (NOS),
United States
Hepatic
ALT:
• In adults and adolescent, Cross-sectional (6 studies) weighted z-
score = 3.55, p < 0.001
o One longitudinal study reported positive associations
• ALT in children <12 years of age, GGT, AST, liver enzymes: associations
not statistically significant
Abdullah
Soheimi et al.
(2021, 9959584)
Canada, China,
Denmark, Italy,
29 Norway Spain,
Sweden, Taiwan,
United States
Cardiovascular (16
studies)
Serum Lipids (10
studies)
CVD:
• Strong evidence of association between serum PFOS and CVD risk (14
studies); z = 3.87, p < 0.0001,12 = 60.13%
CIMT:
• Inconsistent associations between serum PFOS and CIMT (2 studies)
• Consistent associations between serum PFOS and increased serum TC,
LDL, and triglyceride levels
GDM:
Metabolic (3 studies) • Inconsistent associations between serum PFOS and increased GDM in
pregnant mothers compared to non-pregnant mothers
Kim et al. (2018,
5079795)
12
Canada, China,
Korea, Japan,
Norway, Taiwan,
United States
Endocrine - Thyroid
Free T4:
• Pooled z-value (9 studies): 0.05 (0.03, 0.08), I2 = 0%
• More pronounced correlation between blood PFOS and free T4 in
intermediate exposure group (8-16 ng/mL): 0.07 (0.02, 0.11), I2 = 0%
• Association not statistically significant among subgroup of pregnant
women
• Total T4 (8 studies), Total T3 (8 studies), TSH (12 studies): Associations
not statistically significant
o Sensitivity analyses removed outlier for total T4 and total T3; total T4 z
value = -0.04 (-0.07, -0.01), I2 = 5%; total T3 z value = -0.04
(-0.06, -0.01),
Zare Jeddi et al.
(2021, 8347183)
Canada, China,
Croatia, Italy,
United States
Metabolic
Metabolic syndrome:
• Pooled OR: 0.94 (0.79, 1.10), I2 = 78.7%
Stratakis et al.
(2022, 10176437)
21
China, Demark,
Faroe Islands,
Greenland,
Netherlands,
Norway, Spain,
Metabolic
BMI z-score:
• In infancy (3 studies): Pooled (3 per unit increase in prenatal PFOS: -0.007
(-0.012, -0.003), I2 = 0%
• In childhood period (2-9 years) (10 studies): Pooled (3 per unit increase in
prenatal PFOS= 0.00 (-0.01, 0.01), I2 = 42.9%
A-152
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Number of
Studies
Countries
Health Outcome Results/Conclusions3
Sweden, Taiwan,
Waist circumference:
Ukraine, United
• In childhood (4 studies): Pooled (3 per unit increase in prenatal
Kingdom, United
PFOS = -0.06 (-0.19, 0.07), I2 = 20.5%
States
• Inconsistent associations between PFOA exposure and fat mass,
overweight risk
Qu et al. (2021,
9959569)
Denmark,
Greenland,
Norway, Poland,
Sweden, Ukraine,
United States
Neurodevelopmental
ADHD:
• Pooled OR: 1.01 (0.88, 1.14), I2 = 54.7%
• Subgroup analysis between children's blood and prevalence rate of ADHD
(2 studies), pooled OR: 1.05 (1.02, 1.08), I2 = 48.7%
• Subgroup analysis between PFOS exposure and prevalence rate of ADHD
in the United States (2 studies), OR: 1.05 (1.02, 1.08), I2 = 48.7%
Notes: LBW = low birth weight; OR = odds ratio; PTB = preterm birth; GDM = gestational diabetes mellitus; SGA = small for gestational age; ALT = alanine aminotransferase;
GGT = y-glutamyltransferase; AST = aspartate aminotransferase; CVD = cardiovascular disease; CIMT = carotid artery intima-media thickness (mm); TC = total cholesterol;
LDL = low density lipoproteins; T4 = thyroxine; T3 = triiodothyronine; TSH = thyroid stimulating hormone; BMt = body mass index; ADHD = attention deficit-hyperactivity
disorder.
a Results reported as effect estimate and 95% confidence interval (CI) unless otherwise stated.
b Toxicological study data included in these publications were not subject to meta-analysis.
c Preterm birth was defined as birth <37 weeks of gestation.
A-153
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A3 Studies Identified After Assessment Literature Cut-Off Date
Studies identified after the updated literature review (February 2022) did not undergo the systematic review protocol. Studies were
reviewed for major findings and how those findings may affect the assessment. For PFOS, 8 studies were identified after the updated
literature review and are summarized below (Table A-43).
Table A-43. Studies Identified After Updated Literature Review (Published or Identified After February 2022)
Reference
Major Findings
Assessment Implications
Ding et al. (2022,
10328874)
Cohort study of 1,058 midlife women initially free of hypertension from
the multiethnic and multiracial SWAN. Compared with the lowest
tertile, women in the highest tertile of baseline serum PFOS
concentrations had adjusted HRs of 1.42 (95% CI: 1.19, 1.68) (p-
trend = 0.01). In the mixture analysis, women in the highest tertile of
overall PFAS concentrations had a hazard ratio of 1.71 (95% CI: 1.15,
2.54; p-trend=0.008), compared with those in the lowest tertile.
PFOS might be associated with increased risk of hypertension
in women. Possible mixture effects with hypertension in
women. No change.
Feng et al. (2022,
10328872)
Case-cohort study within the Dongfeng-Tongji cohort, including
incident breast cancer cases (n = 226) and a random sub-cohort
(n = 990). No association with PFOS. Quantile g-computation analysis
observed a 19% increased incident risk of breast cancer along with each
simultaneous quartile increase in all ln-transformed PFCA
concentrations (HR = 1.19, 95% CI: 1.01, 1.41), withPFOA accounting
for 56% of the positive effect.
No change.
Goodrich et al., 2022
(10369722)
Nested case-control study within the Multiethnic Cohort (MEC) Study,
including incident, non-viral hepatocellular carcinoma (HCC) cases
(n=50) and healthy controls (n=50). Significant increase in risk in in
those with high exposure (>85th percentile; >54.9 ug/L) r.v. low
exposure (<85th percentile; < 54.9 ug/L) (OR = 4.50, 95% CI: 1.20,
16.00).
PFOS may be associated with incident, non-viral HCC.
Contrasts findings in Eriksen et al., 2009 (2919344), however,
Eriksen et al., 2009 (2919344) did not specify cancer type or
etiology in their analysis.
Gui et al., 2022 Meta-analysis of 23 studies, pooled change in birthweight per 1-
(10365824) In ng/mL increase in PFOS (unadjusted for gestational
age/unstandardized birth weight): -34.88 g (95% CI: -52.53, -17.24),
I2= 66.1%. Significant effects observed for birth length and ponderal
index. No associations observed for preterm birth, low birth weight or
small for gestational age. Subgroup analyses were included, by fetal
gender, time of blood sample collection, blood sample type and whether
adjusted for GA/parity, study design, and geographic region. Included
assessment of risk of bias for studies included in the meta-analyses.
Supports an association between PFOS and birth weight, birth
length and ponderal index. Similar conclusions as previous
meta-analyses.
A-154
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Major Findings
Assessment Implications
Jiang et al. (2022,
10328207)
Luo et al., 2022
(10273290)
Velarde et al. (2022,
9956482)
Wen et al. (2022,
10328873)
Zhang et al., 2022
(9944433)
Meta-analysis of 8 studies across 8 countries. No association between
PFOS and breast cancer risk (OR = 1.01; 95% CI: 0.87, 1.17),
I2 = 99.8%.
Prospective study in the Danish National Birth Cohort, 656 children.
Prenatal exposure to PFOS was not associated with facial features
(measures of palpebral fissure length, philtrum groove, and upper-lip
thickness) in children at age 5.
Case-control study of 150 Filipino women (75 breast cancer cases and No change.
75 controls). Serum PFOS levels were significantly higher in cases than
on controls. PFOS was positively but not statistically significant
associated with breast cancer risk across quartiles of exposure after
adjusting for potential confounders. Positive significant association
observed in crude models only in the highest quartile of PFOS.
Population -based cohort study of 11,747 participants from 1999-2014
NHANES followed up to December 2015. PFOS was statistically
significantly associated with an increased risk in all-cause (OR = 1.57;
95% CI: 1.22, 2.07), heart disease (OR = 1.65; 95% CI: 1.09, 2.57) or
cancer mortality (OR = 1.75; 95% CI: 1.10, 2.83), but only in the
highest tertile (>17.1 ng/mL) compared to the lowest tertile
(<7.9 ng/mL).
Prospective cohort study (the Shanghai Birth Cohort) of 2,395 mother-
infant pairs. Prenatal PFOS exposure measured in early pregnancy
(median, 15 gestational weeks) was not associated with infant length,
weight, and head circumference at birth, 42 days, 6 months, and 12
months.
No change. Serious methodological limitations warrant
cautious interpretation of results from this publication.
No change.
No change.
No change.
Notes: SWAN = Study of Women's Health Across the Nation; HR = hazard ratio; OR = odds ratio; NHANES = National Health and Nutrition Examination Survey.
A-155
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix B. Detailed Toxicokinetics
B.l Absorption
B.l.l Cellular Uptake
Lipid binding may influence PFOS accumulation in various cell types relevant to absorption as
well as distribution. Sanchez-Garcia et al. (2018, 4234856) compared PFOA and PFOS in their
ability to accumulate and be retained in cells including lung epithelial cells (NCI-H292). Cellular
accumulation and retention of PFOS was observed in lung cells at levels higher than
azithromycin-dihydrate (AZI), a lysosomotropic cationic amphiphilic drug used as a reference
compound. In contrast, PFOA only accumulated to very low levels (Table B-l). Phospholipid
binding was assessed by measuring the relative affinity for a phosphatidylcholine (PC)-coated
column at pH 7.4 to calculate a chromatographic index (CHIIAM7.4). Lipid binding (LogD7.4)
was determined by measuring the relative affinity of compounds for a C18-coated liquid
chromatography column at pH 7.4. LogP values obtained from the PubChem database were used
as a comparative lipophilicity measure. Phospholipophilicity correlated (r2 = 0.75) to cellular
accumulation better than other lipophilicity measures. The extent to which PFOS
phospholipophilicity influences absorption through the GI tract, lungs, or skin is unknown.
Table B-l. Cellular Accumulation and Retention Relative to Lipophilicity and
Phospholipidicity as Reported by Sanchez-Garcia et al. (2018, 4234856)
Cellular Accumulation and Retention
Lipophilicity
Chemical
Accumulation in Retention in
Lung Epithelium (% Lung Epithelium
AZI)
Phospholipid Binding
(CHIIAM7.4)
Lipid Binding
(LogD7.4)
LogP
PFOS
313 ±101* 26 ±4
39 ±3*
2.33 ±0.11*
5
PFOA
15 ±3 ND
29 ± 1
1.29 ±0.02
4.9
Notes: AZI = azithromycin-dihydrate; ND = not determined.
* Statistically significant at p < 0.05 from PFOA.
The study by Sanchez-Garcia et al. (2018, 4234856) raises the possibility of passive update of
PFOS into cells. This is consistent with observations that cells transfected with vector only,
could take up PFOS, albeit at lower levels than cells transfected with PFOS-specific
transporters (discussed further in Section B.4.2.1). Ebert et al. (2020, 6505873) determined
membrane/water partition coefficients (Kmem/w) for PFOS and examined passible permeation
into cells by measuring the passive anionic permeability (Pion) through planar lipid bilayers.
Membrane permeability and partition coefficients were predicted using an approach
developed to model molecules in micellar systems and biomembranes {COSMOmic and
related tools, Klamt, 2008, 9641966}. The predicted log (Kmem/w/[L/kg]) for PFOS was 4.69,
similar to the experimentally determined value of 4.89 ± 0.30. Kmem/wvalues increase with
increasing chain length, reflecting increased surface area for van der Waals interactions. The
authors observed that perfluoroalkanesulfonic acids (PFSAs) adsorb about 1.2 log units more
strongly to the membrane than perfluorocarboxylates (PFCAs) with the same number of
perfluorinated carbons. Permeability showed the same chain-length dependence as Kmem/w
B-l
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
values. The predicted anionic permeability (log Pi0n/[cm/s]) for PFOS ranged from -4.74 to
-3.58, considered high enough to explain observed cellular uptake by passive diffusion in the
absence of active uptake processes. The extent to which passive uptake influences absorption
in vivo remains to be determined.
B.1.2 Oral Exposure
Chang et al. (2012, 1289832) administered a single oral dose of 4.2 mg/kg of PFOS-14C in
solution to three male Sprague-Dawley rats. At 48 hours after dosing, only 9.08 ± 0.51% of the
total PFOS-14C dose was recovered across digestive tract, feces, or urine, while the carcass
retained 94.2 ± 5.1%, indicating that the PFOS was largely absorbed.
B.1.3 in ho lotion Exposure
An acute median lethal concentration (LCso) study in rats indicates that PFOS absorption occurs
after inhalation exposures; however, pharmacokinetic data were not included in the published
report {Rusch, 1979, 7561179}. The analytical methods for measuring PFOS in animals were
limited at the time the study was conducted. More recent data on PFOS absorption following
inhalation exposure are not available.
B.1.4 Dermal Exposure
The literature contains no studies on the dermal absorption of PFOS.
B.1.5 Developmental Exposure
The literature contains no studies on PFOS absorption following developmental exposure.
Additional information on PFOS distribution during reproduction and development is found in
Section B.2.3.
B.1.6 Bioavailability
Toxicokinetic parameters informing absorption were derived by comparing oral to intravenous
(IV) dosing in rats {Kim, 2016, 3749289}. Sprague-Dawley rats were administered 2 mg/kg by
either the IV or oral route. Urine and feces were collected weekly, and blood was collected at 10
time points over the first day and then up to 70 days after exposure. In contrast to the sex
differences observed for PFOA, the time to reach the maximum PFOS plasma concentration
(Tmax) following oral exposure was similar in males and females (10.8 hr and 11.5 hr,
respectively). In a similar study {Huang, 2019, 5387170}, male and female Sprague-Dawley rats
were administered a single dose of 2 mg/kg by IV injection or a single dose of 2 mg/kg or
20 mg/kg by oral gavage and observed from 5 minutes to 20 weeks after dosing. The maximal
plasma concentrations (Cmax) were similar for oral gavage and IV administration of 2 mg/kg,
and Tmax values were consistent with those observed by Kim and colleagues (14.3 hr and 12.2 hr
in males and females, respectively).
The results from these studies are compared in Table B-2. Both studies found very high
(> 100%) bioavailability in rats (calculated by dividing the dose-adjusted gavage area under the
curve (AUC) by the IV AUC). Huang and colleagues speculate that the > 100% bioavailability
B-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
after oral dosing is due to enterohepatic circulation that occurs after gavage but not IV
administration. The Tmax values ranged from 10.8 to 14.3 hours and was slightly longer in the
Huang study for both males and females. Neither bioavailability nor Tmax exhibited sex-specific
differences. However, Huang et al. did observe slightly higher Cmax concentrations in females
relative to males.
Table B-2. PFOS Parameters from Toxicokinetic Studies Informing Bioavailability in
Sprague-Dawley Rats
Study
Dose (mg/kg)
Route
Sex
Cmax (ng/mL)a
Tmax (hours)b
Kim et al. (2016,
2
Oral
Male
6.71 ±0.30
10.8 ±0.96
3749289)
IV
Male
5.23 ±0.24
NA
Oral
Female
6.66 ±0.29
11.52 ± 1.2
IV
Female
5.69 ±0.33
NA
Huang et al. (2019,
2
Oral
Male
5.00 ±5.00
14.3 ±2.7
5387170)
IV
Male
5.00 ±5.00
NA
Oral
Female
10.00 ±5.00
12.2 ±5.2
IV
Female
5.00 ±5.00
NA
Notes: Cmax=maximum serum concentration, IV = intravenous, NA = not applicable, Tmax = time to Cmax.
a Converted published Cmax (mM) to Cmax ((ig/mL) for Huang et al. (2019, 5387170).
b Converted published Tmax (days) to Tmax (hours) for Kim et al. (2016, 3749289).
B.2 Distribution
B.2.1 Protein Binding
Kerstner-Wood et al. (2003, 4771364) examined the in vitro protein binding of PFOS in rat,
monkey, and human plasma at concentrations of 1 ppm to 500 ppm and found that PFOS was
bound to plasma protein in all three species. When incubated with separate human-derived
plasma protein fractions, PFOS was highly bound (99.8%) to albumin and showed affinity for
low-density lipoproteins (95.6%) with some binding to alpha-globulins (59.4%) and gamma-
globulins (24.P/o). Low levels of binding to alpha-2-macroglobulin and transferrin were
measured when the protein concentrations were approximately 10%> of physiological
concentration.
Zhang et al. (2009, 2919350) conducted an in vitro study using equilibrium dialysis,
fluorophotometry, isothermal titration calorimetry, and circular dichroism to characterize
interactions between PFOS with serum albumin and DNA. The authors reported that serum
albumin could bind up to 45 moles of PFOS/mole of protein and 0.36 moles/base pair of DNA.
The binding ratio increased with increasing PFOS concentrations and decreasing solution pH.
The authors concluded that the interactions between serum albumin and PFOS were the results of
surface electrostatic interactions between the sulfonate functional group and the positively
charged side chains of lysine and arginine. Hydrogen binding interactions between the negative
dipoles (fluorine) of the PFOS carbon-fluorine bonds could also play a role in the noncovalent
bonding of PFOS with serum albumin.
Chen and Guo (2009, 1280480) investigated the binding of PFOS to human serum albumin using
site-specific fluorescence and found that PFOS induced fluorescence quenching indicative of
B-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
binding. A binding constant of 2.2 x 1041VT1 and a binding ratio of PFOS to human albumin of
14 moles PFOS/mole albumin were calculated. Fluorescence displacement measurements were
used to study the interaction between PFOS and two high-affinity drug binding sites on human
serum albumin known as Sudlow's drug Site I and Site II. The findings indicated that PFOS has
binding sites that are similar to those identified for fatty acids.
Salvalaglio et al. (2010, 2919252) used molecular modeling to determine the structure and
energy of PFOS binding sites for human serum albumin. The binding sites impacted were ones
identified as human serum albumin fatty acid binding sites. The most populated albumin binding
site for PFOS was dominated by van der Waals interactions. The PFOS binding site with the
highest energy (-8.8 kcal/mole) was located near the tip of the tryptophan 214 binding site, and
the maximum number of ligands that could bind to human serum albumin for PFOS was 11.
D'Alessandro et al. (2013, 5084740) used electrospray ionization mass spectrometry to evaluate
PFOS binding to bovine serum albumin. Using this approach, the maximum number of PFOS
binding sites was estimated as 11, but the data on collision-induced PFOS removal was more
consistent with 7 binding sites. This study also showed that PFOS competes with ibuprofen for
its site when the PFOS:ibuprofen ratio is > 0.5 moles: 1 mole. In addition, when the binding site
is occupied by PFOS, ibuprofen is unable to bind. Zhang et al. (2009, 2919350) conducted a
similar study of the impact of PFOS on the ability of serum albumin to bind vitamin B2
(riboflavin) and found that, under normal physiological conditions, PFOS decreased the binding
ratio of serum albumin for riboflavin in vitro. These data suggest that PFOS can alter the
pharmacokinetics and pharmacodynamics of medicinal and natural substances that share a
common site on albumin.
Beesoon and Martin (2015, 2850292) examined differences in the binding of linear and branched
chain isomers of PFOS to calf serum albumin and human serum proteins. The linear PFOS
molecule was found to bind more strongly to calf serum albumin than the branched chain
isomers. When arranged in order of increasing binding, the order was 3m < 4m < lm < 5m < 6m
(iso) < linear. In the isomer-specific binding to spiked total human serum protein, the lm
branched PFOS isomer bound most strongly and the 4m branched PFOS isomer the least.
Liu et al. (2017, 3856708) used spectroscopy, molecular modeling, and calorimetry techniques to
evaluate the mechanism by which PFOS interacts with human serum albumin through hydrogen
bonds and electrostatic interactions. PFOS binding to albumin is a spontaneous exothermic
process driven by electrostatic interactions. This study observed that the backbone and secondary
structure of albumin did not significantly change after exposure to PFOS; however, results
suggest the interaction with PFOS changed the local structure around the esterase active site. A
molecular docking study indicated that PFOS binds to the active center Arg 410 residue in
albumin. This corresponded to a 28.6% decrease in esterase activity. By examining multiple
PFAS, esterase activity of albumin was found to decrease with the shortening of the carbon chain
and the authors suggest this may correlate with toxicity.
Sheng et al. (2020, 6565171) measured uptake of PFOS in human placental choriocarcinoma
(JAr) cells in the presence or absence of human serum albumin for 48 hours. PFOS
concentrations in the culture medium decreased by 21.4%, 78.1%>, and 92.8% with the addition
of 0.5 |iM, 10 |iM, and 200 |iM albumin, respectively. This result supports a paradigm in which
binding of albumin to PFOS in the culture medium blocked their entrance into the cells. The
B-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
binding affinity (Kd) of PFOS to human serum albumin was calculated to be 30.7 |iM, Using a
limited proteolysis technique, the authors identified the core albumin peptides that bind to PFOS
as residues 189-457.
Binding to albumin and other serum proteins may affect transfer of PFOS from maternal blood to
the fetus. Gao et al. (2019, 5387135) correlated placental transfer with experimentally measured
dissociation constants (Kd) to human serum binding proteins, serum albumin, and L-FABP. For
PFOS, Kd values were calculated to be 49 ± 8 |iM for serum binding proteins, 38 ± 5 |iM for
albumin, and 81 ± 7 |iM for L-FABP. These Kd values significantly correlated with placental
transfer efficiencies measured in 132 maternal blood-cord blood pairs from subjects in Beijing,
China, suggesting serum and binding proteins, especially albumin, play an important role in
placental transfer efficiency. The authors suggested that lower cord blood albumin levels
compared to maternal blood albumin levels may set up a competition for PFOS binding on either
side of the placenta.
Since there is effectively a competition between PFOS binding in maternal serum vs. cord blood,
lower cord blood albumin levels compared to maternal blood albumin levels are likely to reduce
transfer from maternal serum across the placenta. Consistent with this hypothesis, Pan et al.
(2017, 3981900) found that the concentration of cord serum albumin was associated with higher
transfer efficiencies (increase of 4.1% (CI: 2.7, 5.4) per 1 g/L albumin). However, maternal
serum albumin concentration was associated with reduced transfer efficiency (decrease of 3.4%
(CI: -5.0, -1.8) per 1 g/L albumin). Because albumin cannot cross the placental barrier, the
authors speculate that binding of PFOS to maternal serum albumin can reduce the free PFOS
available to move across the barrier through passive diffusion. Similarly, higher fetal albumin
levels will lead to less free PFOS in cord blood, which may facilitate the rate of placental transfer
via passive diffusion.
PFOS also binds to intracellular proteins. Luebker et al. (2002, 1291067), Zhang et al. (2013,
5081488), and Yang et al. (2020, 6356370) conducted in vitro studies that examined the binding
of PFOS and other PFAS to the liver fatty acid binding protein (L-FABP). L-FABP is an
intracellular lipid carrier protein that reversibly binds long-chain fatty acids, phospholipids, and
an assortment of peroxisome proliferators {Erol, 2004, 5212239} and constitutes 2%-5% of the
cytosolic protein in hepatocytes. Luebker et al. (2002, 1291067) evaluated the ability of
perfluorinated chemicals to displace a fluorescent substrate from L-FABP and reported that
PFOA exhibited some binding to L-FABP, but its binding potential was about 50% less than that
of PFOS and far less than that of oleic acid. Zhang et al. (2013, 5081488) cloned the human L-
FABP gene and used it to produce purified protein for evaluation of the binding of PFOA and
PFOS. The median inhibiting concentration (IC50) values for PFOA and PFOS were
9.0 ± 0.7 |imol and 3.3 ± 0.1 |imol, respectively, suggesting that PFOA has a lower binding
affinity than PFOS. PFOA was bound to the carrier protein in a 1:1 ratio, and the interaction was
mediated by electrostatic interactions and hydrogen binding with the fatty acid binding site.
Using size-exclusion column coelution and nontarget analysis to identify additional PFAS
ligands from contaminated environmental sources, Yang et al. (2020, 6356370) also found that
that both polar and hydrophobic interactions are crucial for binding affinities to L-FABP for
PFOA and PFOS.
A computational modeling approach that combined molecular docking and molecular dynamics
simulation techniques was used to estimate the relative binding of affinity of PFOS for human
B-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
and rat L-FABP {Cheng, 2018, 5024207}. The authors found that predicted free energies
correlated well with binding affinities measured in 3 previous studies {Woodcroft, 2010,
2919284; Zhang, 2013, 5081488; Sheng, 2018, 4199441}. Key residues contributing to free
binding energies (AGbind) for L-FABP include ARG 122, SER 124, and ILE 52 (human) and
TYR 120, ARG 122, ILE 60, and ILE 53 (rat).
B.2.2 Tissue Distribution
B. 2.2.1 Human Studies
Human blood is a known site of PFOS accumulation. A recent example measured PFAS in blood
samples from 344 Wilmington, NC residents (289 adults and 55 children) exposed to
contaminated drinking water from release of PFAS chemicals into the Cape Fear River between
1980 and 2017. The mean serum PFOS concentration was 9.4 ng/mL in adults and 5.1 ng/mL in
children {Kotlarz, 2020, 6833715}.
PFOS accumulation in blood impacts distribution to various tissues and organs, but few studies
have examined PFOS partitioning to human blood fractions. Forsthuber et al., (2020, 6311640)
measured the distribution of PFOS in blood fractions including plasma, albumin, and lipoprotein
fractions (e.g., very low-density lipoproteins (VLDL), low-density lipoproteins (LDL), and high-
density lipoproteins (HDL)). Blood from four young healthy volunteers (two women, two men,
23-31 years old) were separated into fractions using size fractionation (for proteins) and serial
ultracentrifugation. Results found that albumin was the most important carrier for PFOS with
4.3 ± 2.2 ng/mL present in this fraction. In contrast, the amount of PFOS associated with VLDL,
LDL and HDL fractions was below the limit of quantification (LOQ), 0.1 ±0.1 ng/mL, and
0.16 ± 0.06 ng/mL, respectively.
Jin et al. (2016, 3859825) analyzed 60 blood samples from a Chinese population, and three
whole blood samples from an exposed Canadian family to investigate the partitioning of PFAS
of different chain lengths and their major isomers between human blood and plasma. Increasing
chain length for PFAS correlated with an increased mass fraction in human plasma (Fp) from C6
(mean 0.24) to CI 1 (0.87). The PFOS plasma:whole blood ratio in the Jin et al. (2016, 3859825)
study was lower (1.5 ± 0.42) compared to the mean plasma:whole blood (2.2-2.3) {Ehresman,
2007, 1429928} and serum:whole blood (1.2-2.3) {Karrmen, 2006, 2159543; Hanssen, 2013,
3859848} ratios previously reported. Linear isomers of PFOS had lower mean Fp than their
corresponding total branched isomers. In blood samples obtained from three highly exposed
Canadian subjects, the highest levels of PFOS were measured in plasma (0.14 ng/mL) compared
to red blood cells (RBCs, 0.04 ng/mL) and in washed RBCs (0.04 ng/mL). The authors
suggested that these values could be used as more accurate conversion factors when converting
concentrations between whole blood and plasma.
Fractionation to blood fractions was also examined in 61 male and female participants from
Oslo, Norway in 2013-2014 {Poothong., 2017, 4239163}. The median relative PFAS
compositions in serum, plasma, and whole blood were dominated by PFOS, followed by PFOA
(representing 60%-70% of blood PFAS), relative to the other 23 PFAS chemicals analyzed.
Median PFOS concentrations in plasma, serum, and whole blood were 5.24 ng/mL, 4.77 ng/mL,
and 2.85 ng/mL, respectively. Similar to other studies, PFOS preferentially accumulated in
plasma relative to serum and whole blood; this result suggests that the common practice of
B-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
multiplying by a factor of 2 to convert the concentrations in whole blood to serum or plasma will
not provide accurate estimates for PFOS.
B.2.2.1.1 Distribution in Tissues
No clinical studies are available that examined tissue distribution in humans following
administration of a controlled dose of PFOS. However, samples collected in biomonitoring and
epidemiological studies provide data showing distribution of PFOS.
In humans, PFOS distributes primarily to the liver and blood. Olsen et al. (2003, 3005572)
sampled both liver and serum from cadavers for PFOS and found a good correlation between
samples from the same subject. There were no sex- or age group-specific differences in PFOS
concentrations. In another study, Karrman et al. (2010, 2732071) identified PFOS in postmortem
liver samples (n = 12; 6 males, 6 females, aged 27-79 years) with a mean concentration of
26.6 ng/g tissue.
Perez et al. (2013, 2325349) collected tissue samples (liver, kidney, brain, lung, and bone) in the
first 24 hours after death from 20 adult subjects (aged 28-83 years) who had been living in
Catalonia, Spain. PFOS was present in 90% of the samples but could be quantified in only 20%
(median 1.9 ng/g). PFOS accumulated primarily in the liver (104 ng/g), kidney (75.6 ng/g), and
lung (29.1 ng/g), and brain (4.9 ng/g), with levels below the limit of detection (LOD) in the bone.
PFOS also accumulates in follicular fluid. Kang et al. (2020, 6356899) measured 6.82 ng/mL in
follicular fluid samples from 28 women undergoing oocyte retrieval for in vitro fertilization
procedures. A positive correlation was found between paired serum and follicular fluid samples
for PFOS (r2 = 0.78, p < 0.001), though PFOA correlations were even stronger (r2 = 0.93,
p < 0.001). Exposure of oocytes to PFOS raise the possibility of reproductive toxicity in humans.
Stein et al. (2012, 1332468) compared PFAS levels in paired samples of maternal serum and
amniotic fluid from 28 females in their second trimester of pregnancy. PFOS was detected in all
serum samples (0.0036-0.0287 |ig/mL) and in nine amniotic fluid samples (0.0002 |ig/mL-
0.0018 |ig/mL), The Spearman correlation coefficient between the serum and amniotic fluid
levels was 0.76 (p = 0.01), indicating a direct relationship between PFOS levels in blood and
amniotic fluid. The median ratio of maternal serum:amniotic fluid concentration was 25.5.
Two studies examined accumulation of PFOS in cerebrospinal fluid and serum {Harada, 2007,
2919450; Wang, 2018, 5080654}. In both studies, PFOS levels in cerebrospinal fluid were two
orders of magnitude lower than in the serum. These results indicate that PFOS does not easily
cross the adult blood-brain barrier.
PFOS has been detected in both umbilical cord blood and breast milk indicating that maternal
transfer occurs {Apelberg, 2007, 1290900; Von Ehrenstein, 2009, 194805; Volkel, 2008,
3103448}. Karrman et al. (2010, 2732071) identified PFOS in breast milk samples from healthy
females (n = 10; aged 30-39 years), and the levels in milk (mean 0.12 ng/mL) were low
compared to levels in the liver.
B.2.2.2 Animal Studies
Studies of tissue distribution are available for several species of animals including non-human
primates, rats, and, to a lesser extent, mice. Studies of non-human primates indicate that levels of
B-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOS in serum accumulate in a dose-dependent manner. While data are limited on liver
accumulation of PFOS in monkeys, PFOS accumulation in the liver appears to be similar to that
of serum, if not slightly lower. Several rodent studies identified the liver as a major site of
accumulation, and that PFOS distributes to a wide range of tissues including kidney, heart, and
lungs, and spleen. Interestingly, PFOS has been measured in moderate quantities in both the
brain and testicles of rodents, indicating that it does cross the blood-brain barrier and blood-testis
barrier. While monkeys had nearly a 1:1 liver to serum ratio, rodent models were observed to
contain accumulate far more PFOS in liver than serum.
B.2.2.2.1 Non-Human Primates
Two long-term studies in monkeys examined PFOS accumulation in the serum and liver. Seacat
et al. (2002, 757853) administered 0 mg/kg/day, 0.03 mg/kg/day, 0.15 mg/kg/day, or
0.75 mg/kg/day PFOS orally in a capsule by intragastric intubation to young-adult to adult
cynomolgus monkeys for 26 weeks. Serum and tissues were collected at necropsy. The dosing
was followed by a 52-week recovery period in 2 animals in the control, 0.15 mg/kg/day, and
0.75 mg/kg/day groups. Serum PFOS measurements demonstrated a linear increase with dosing
duration in the 0.03 mg/kg/day and 0.15 mg/kg/day groups and a non-linear increase in the
0.75 mg/kg/day group. Levels in the high-dose group appeared to plateau after about 100 days
(14 weeks) but began to decline sometime after week 37. The average percent of the cumulative
dose of PFOS in the liver at the end of treatment ranged from 4.4% to 8.7% with no difference
by dose group or sex. At the two lower doses, serum levels were comparable in the males and
females, whereas at 0.75 mg/kg/day, levels were generally elevated in the males compared to
females. Only the highest dose group appeared to reach a serum steady state at Week 16. In the
0.03 mg/kg/day groups, the serum levels continued to increase temporally until Week 27 when
serum sampling stopped for that cohort. Once dosing ceased, serum levels declined in all animals
that continued in the study.
In the second study conducted in cynomolgus monkeys {Chang, 2017, 3981378}, animals were
given PFOS doses to reach target serum concentrations of 70 |ag/m L or 100 |ag/m L that were
chosen to match levels of the medium- and high-dose groups from Seacat et al. (2002, 757853).
The control group (n = 6/sex) was dosed with vehicle, the low-dose group (n = 6/sex) received a
single dose of 9 mg/kg PFOS on day 106 of the study, and the high-dose group (n = 4-6/sex)
received 3 separate PFOS doses (11-17.2 mg/kg) on days 43, 288, and 358. Measurements of
serum PFOS indicate that male and female monkeys reached the target dose of 70 jag/m L and
100 |ig/mL on day 113 and 50, respectively. Male and female animals in the high dose group
reached peak PFOS serum levels of 160 |ag/mL -165 |ag/mL on day 365. Consistent with the
previous study, no sex differences were found. At the end of the experiment, the animals were
reported to have a 1:1 PFOS livenserum ratio, while the previous Seacat et al. (2002, 757853)
study reported a ratio closer to 2:1. Chang et al. (2017, 3981378) attributed these differences in
findings to the dosing approaches and regimens used in the two studies (gelatin capsule vs.
gastric intubation).
B.2.2.2.2 Rats
Numerous studies have been performed on models of PFOS distribution in rats. These studies
range from acute (hours) to longer-term studies (20 weeks) and include various levels of dosing.
Distribution is measured primarily in serum, liver, and lungs, but approaches were used to
measure brain distribution as well.
B-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Martin et al. (2007, 758419) administered PFOS (10 mg/kg/day) to adult male Sprague-Dawley
rats for 1, 3, or 5 days by gavage and determined the liver and serum levels. Mean liver PFOS
levels were 83 ± 5 jag/g, 229 ± 10 (J,g/g, and 401 ± 21 j_ig/g after 1, 3, or 5 daily doses,
respectively. Mean serum concentrations were 23 ± 2.8 j_ig/g and 87.7 ±4.1 [j,g/mL after 1 and
3 days of dosing, respectively. Day 5 serum levels were not available through the publication.
This study observed a liver: serum ratio of nearly 3:1.
In another acute study performed by Yu et al. (2011, 1294541), female Wistar rats were
administered doses of PFOS (0, 0.2, 1.0, or 3.0 mg/kg/day) dissolved in 0.5% Tween 20 for
5 consecutive days. Blood and bile were collected 24 hours after the last dose was given. Data
indicate that there is a linear dose-dependent increase in both serum and bile, which likely
reflects levels in liver.
A 28-day toxicity study by NTP exemplifies patterns of PFOS accumulation in blood and liver
{NTP, 2019, 5400978}. Male and female Sprague-Dawley rats were administered daily doses of
0 mg/kg/day, 0.312 mg/kg/day, 0.625 mg/kg/day, 1.25 mg/kg/day, 2.5 mg/kg/day, or
5 mg/kg/day of PFOS by oral gavage. Plasma and liver concentrations were analyzed
approximately 24 hours after the last dose. A dose-dependent increase in plasma concentrations
of PFOS was observed in both males and females. In contrast to studies with PFOA, plasma
PFOS concentrations in females were generally similar to males, and dose-normalized plasma
concentrations ([iM/mmol/kg/day) in males and females were within 1.5-fold across the dose
groups. The lowest dose-normalized concentration was observed in the highest dose group in
both sexes. In males, PFOS concentrations in plasma were 23.73 ± 1.11 |ig/mL and
318.2 ± 8.87 |ig/mL at the lowest and highest doses, respectively. In females, these values were
30.53 ± 0.92 |ig/mL and 413.56 ± 8.07 |ig/mL at the lowest and highest doses, respectively.
However, there were quantifiable levels of PFOS in female controls that were 562 times lower
than the lowest dose administered and required caution in interpreting these findings.
Concentrations in livers of males increased with increasing dose, but when normalized with
dose, there was a steady decrease as dose increased. This corresponded with a decreasing
liver:plasma ratio as dose increased. Liver:plasma ratios, measured only in males, were
3.76 ± 0.24 at the lowest dose and 2.74 ± 0.08 at the highest dose.
Additional studies have been performed that expand on PFOS dosing, time of treatment, and
organ distribution. Cui et al. (2009, 757868) delivered 5 or 20 mg/kg/day of PFOS via oral
gavage to 3-month old Sprague-Dawley rats. At the end of dosing (28 days), serum and organ
concentrations were measured (Table B-3). No blood samples were available at the
20 mg/kg/day dose due to animal deaths in this group. The liver appeared to have by far the
highest concentration of PFOS at both 5 mg/kg/day and 20 mg/kg/day. Levels in the heart were
approximately half the concentration observed in liver followed by the kidney, serum, and lungs.
Of the organs examined, testicles and spleen exhibited the lowest PFOS levels. Of note was the
differential accumulation by organ and dose. For liver, kidney, and heart, 2-3-fold increases in
PFOS concentrations were observed between the low and high doses even though the high dose
was 4 times higher than the low dose. Interestingly, the brain and lungs were most susceptible to
the increase in dose by accumulating 10- and 5-fold more PFOS, respectively.
B-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-3. Concentrations of PFOS in Various Tissues of Male Sprague-Dawley Rats
Exposed to PFOS by Gavage for 28 Days as Reported by Cui et al. (2009, 757868)
Tissue3
0 mg/kg/day
5 mg/kg/day
20 mg/kg/day
Blood (ng/mL)
ND
72.0 ±25.7
No sampleb
Liver (|ig/g)
ND
345 ± 40
648 ± 17
Kidney (|ig/g)
ND
93.9 ± 13.6
248 ± 26
Lung (|ig/g)
ND
46.6 ± 17.8
228 ±122
Heart (|ig/g)
ND
168 ± 17
497 ± 64
Spleen (|ig/g)
ND
38.5 ± 11.8
167 ± 64
Testicle (|ig/g)
ND
39.5 ± 10.0
127 ± 11
Brain (|ig/g)
ND
13.6 ± 1.0
146 ± 34
Notes: PFOS = perfluorooctane sulfonate; ND = not detected.
a Data are presented as mean ± standard deviation.
b Animal deaths in this group precluded blood measurements.
In a similar study conducted by Curran et al. (2008, 757871), male and female Sprague-Dawley
rats were administered 0 mg/kg/day, 2 mg/kg/day, 20 mg/kg/day, 50 mg/kg/day, or
100 mg/kg/day via feed for 28 days (Table B-4). The highest PFOS concentration was found in
the liver at all doses, accounting for 70%-80% of total distribution measured in males and 65%-
80% of total distribution in females. The spleen and heart also contained notable levels of PFOS,
however, accumulation in the heart was approximately 25% less than the amount in spleen.
PFOS in animal livers followed a linear dose-dependent distribution between 2 mg/kg/day and
20 mg/kg/day; however, this linearity was lost between the 20 mg/kg/day, 50 mg/kg/day, and
100 mg/kg/day dose escalation. This could be due to an increase in excretion or changes in
distribution to other organs that were not measured in this study. No consistent differences
between the sexes were found, however, female rats generally had higher levels of PFOS in the
heart and spleen at all doses.
B-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-4. Concentrations of PFOS in Various Tissues of Male and Female Sprague-Dawley Rats Exposed to PFOS by Feed
for 28 Days as Reported by Curran et al. (2008, 757871)
0 mg/kg/day 2 mg/kg/day 20 mg/kg/day 50 mg/kg/day 100 mg/kg/day
Parameter
Males Females Males Females Males Females Males Females Males Females
PFOS consumption 0 0 0.14 ±0.02 0.15 ±0.02 1.33 ±0.24 1.43 ±0.24 3.21 ±0.57 3.73 ±0.57 6.34 ± 1.35 7.58 ±0.68
(mg/kg bw/day)
Spleen 0.27 ± 0.36 2.08 ± 4.17 6.07 ± 1.85 7.94±3.76 45.27±2.16 70.03±36.66 122.51 ± 7.83 139.45 ± 15.44 230.73 ± 11.47 294.96±26.66
Og/g)
Heart 0.10±0.14 1.42 ±2.91 4.67 ± 1.73 6.54±3.07 33.00±3.44 54.65±30.89 90.28±4.95 107.53 ±6.24 154.13 ± 11.78 214.45±17.58
Og/g)
Seram 0.47 ± 0.27 0.95 ± 0.51 0.95±0.13 1.50 ±0.23 13.45 ± 1.48 15.40 ± 1.56 20.93±2.36 31.93 ±3.59 29.88±3.53 43.20±3.95
Og/g)
Liver 0.79 ±0.49 0.89 ±0.44 48.28 ±5.81 43.44 ±6.79 560.23 ± 104.43 716.55 ± 59.15 856.90 ± 353.83 596.75 ± 158.01 1030.40 ± 162.80 1008.59 ± 49.41
Og/g)
Liver: Serum 2.04 ± 1.39 1.30 ± 1.32 51.34 ± 9.20 29.99 ± 8.11 42.10±9.20 46.81 ±5.26 41.42 ± 16.95 20.23±7.50 35.23 ±8.50 23.48 ±1.98
Ratio
Notes:
a Data are presented as mean ± standard deviation.
B-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Iwabuchi et al. (2017, 3859701) exposed male Wistar rats to PFOS in drinking water at
0 |ig/kg/day, 0.077 |ig/kg/day, 0.38 |ig/kg/day, or 1.8 |ig/kg/day for 1 or 3 months. Animals were
necropsied at the end of the 1- or 3-month study, and serum, whole blood, and organ levels of
PFOS were measured (Table B-5). Similar to previous studies, the liver was found to contain the
highest levels of PFOS; however, distribution to other organs (kidney, spleen, and heart) and
serum were remarkably lower when compared to other studies.
Table B-5. Distribution of PFOS in Male Wistar Rats Exposed via Drinking Water for 1 or
3 Months as Reported by Iwabuchi et al. (2017, 3859701)
1-Month Exposure
3-Month Exposure
Tissue3
0.077
0.38
1.8
0.077
0.38
1.8
fig/kg/day
fig/kg/day
fig/kg/day
fig/kg/day
fig/kg/day
fig/kg/day
Brain (|ig/kg)
0.95
0.14
0.081
0.35
0.3
0.43
Heart (|ig/kg)
0.17
0.23
0.12
0.6
0.57
0.7
Liver (|ig/kg)
44
45
25
110
100
100
Spleen (|ig/kg)
0.366
0.36
0.21
0.96
0.91
1.3
Kidney (|ig/kg)
1.1
1.1
0.57
3.6
2.6
3.5
Whole Blood (|ig/L)
0.69
0.77
0.46
1.5
1.4
2.1
Serum (|ig/L)
1.1
1.3
0.73
2.7
2.5
3.1
Notes:
a Data are presented as mean values.
A combined chronic toxicity/carcinogenicity good laboratory practice (GLP) study was
performed in male and female Sprague-Dawley Crl:CD (SD)IGS BR rats administered 0 ppm,
0.5 ppm, 2 ppm, 5 ppm, or 20 ppm PFOS (equivalent to 0 mg/kg/day, 0.018-0.023 mg/kg/day,
0.072-0.099 mg/kg/day, 0.184-0.247 mg/kg/day, and 0.765-1.1 mg/kg/day, respectively) for
104 weeks {Thomford, 2002, 5029075; Butenhoff, 2012, 1276144}. A recovery group was
administered the test substance at 20 ppm for 52 weeks and observed until necropsy at
106 weeks. Serum and liver samples were obtained during and at the end of the study to
determine the concentration of PFOS (Table B-6). The findings were in opposition to the
Iwabuchi et al. (2017, 3859701) study as dose-dependent increases in the PFOS level in the
serum and liver were observed in both male and female rats, with values slightly higher in
females after the 5 ppm and 20 ppm doses.
Table B-6. PFOS Levels in the Serum and Liver of Male and Female Sprague-Dawley Rats
Exposed to PFOS in Feed for 2 Years as Reported by Thomford (2002, 5029075)
Timepoint
0 ppm
0.5 ppm
2 ppm
5 ppm
20 ppm
(weeks)
Males Females
Males Females
Males Females
Males
Females
Males Females
Serum PFOS levels (jig/mL)
0
< LOQa 0.0259
0.907 1.61
4.33 6.62
7.57
12.6
41.8 54.0
14
< LOQb 2.67
4.04 6.96
17.1 27.3
43.9
64.4
148 223
53
0.0249 0.395
-
-
-
-
146 220
105
0.0118 0.0836
1.31 4.35
7.60
22.5
75.0
69.3 233
B-12
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Timepoint
0 ppm
0.5 ppm
2 ppm
5 ppm
20 ppm
(weeks)
Males
Females
Males
Females
Males Females
Males Females
Males
Females
o
O
-
-
-
-
-
-
2.42
9.51
Liver PFOS levels (jig/g)
0
0.104
0.107
11.0
8.71
31.3 25.0
47.6 83.0
282
373
10
0.459
12.0
23.8
19.2
74.0 69.2
358 370
568
635
53
0.635
0.932
-
-
-
-
435
560
105
0.114
0.185
7.83
12.9
26.4
70.5 131
189
381
o
O
-
-
-
-
-
-
3.12
12.9
Notes: LOQ = limit of quantification.
aLOQ = 0.00910 pg/mL.
bLOQ = 0.0457 pg/mL
c Samples were obtained from the recovery group administered 20 ppm for 52 weeks and then observed until necropsy at
106 weeks.
B.2.2.2.3 Mice
Few studies have evaluated PFOS exposure in mice. Findings within these studies focus
primarily on serum and liver concentrations after dosing. Lai et al (2018, 5080641) observed that
distribution from serum to liver exhibited dose-dependency after long-term (7 weeks) PFOS
administration in female CD-I mice. At the lower dose (0.3 mg/kg/day), liver and serum
concentrations were similar (32,942 ng/g and 33,781 ng/g, respectively). At the higher dose
(3 mg/kg/day) liver concentrations were higher (503,817 ng/g) than those observed in serum
(109,526 ng/g).
Bogdanska et al. (2011, 2919253) performed a radioisotope distribution study in adult C57BL/6
male mice using 35S-PFOS feed at a low and high dose for 1, 3, and 5 days. Doses were
equivalent to 0.031 mg/kg/day in the low-dose group and 23 mg/kg/day in the high-dose group.
At both doses and at all timepoints, the liver contained the highest amount of PFOS. At the low
dose, the liver PFOS level relative to blood concentration increased with time, whereas at the
high dose, the ratio plateaued after 3 days. The autoradiography indicated that the distribution
within the liver did not appear to favor one area to a greater extent than any other. The liver
contained 40-50% of the recovered PFOS at the high dose. The authors hypothesized that this
could possibly reflect high levels of binding to tissue proteins. After the liver, lungs accumulated
PFOS at the next highest level in the high dose group. Distribution was fairly uniform with some
favoring of specific surface areas. The tissue:blood ratio for the lung was greater than that for all
other tissues except the liver. The lowest PFOS levels were in the brain and fat deposits. Levels
for the kidney roughly equaled those values observed in the blood at both concentrations and all
timepoints. For the bone measurements, a whole-body autoradiogram of a mouse 48 hours after a
single oral dose of 35S-PFOS (12.5 mg/kg) indicated that most PFOS was found in the bone
marrow and not the calcified bone.
Recently, the spatial distribution of PFOS in the kidney was investigated using imaging mass
spectrometry (IMS) based on matrix-assisted laser desorption/ionization (MALDI) {Yang, 2019,
5387049}. This methodology can provide spatial information (defined as pixel-to-pixel) with a
unique mass to charge ratio (m/z) for a specified compound in the same tissue section without
extra labeling. The authors first determined that a-Cyano-4-hydroxycinnamic acid (CHCA) was
the optimal matrix for detection of PFOS. Next, male BALB/c mice were administered PFOS by
B-13
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
oral gavage at 10 mg/kg/day for 14 days, at which time kidneys were harvested and frozen.
Continued tissue sections were cut. One section was used for the analysis by MALDI-IMS while
the other two sections were homogenized and used to quantitate PFOS using HPLC-MS/MS. The
average concentration of two sections in the PFOS-exposed kidney was 2.56 ± 0.193 (J,g/mL,
almost 1,000-fold higher than the 3.25 ± 0.274 ng/mL measured in control sections. PFOS was
mainly distributed in the kidney cortex region, which was consistent with the PFOS-induced
glomerular atrophy observed in hematoxylin and eosin-stained sections. The authors conclude
that the average concentration of the whole kidney fails to reflect the spatial accumulation of
PFOS within the kidney, which can be measured and correlated to pathogenetic changes using
MALDI-IMS.
In an immunotoxicity study conducted by Qazi et al. (2009, 1937260), C57BL/6 male mice were
administered diets with 0% to 0.02% PFOS for 10 days and PFOS levels in serum were
measured. The authors found that PFOS levels in the serum increased as the dietary level of
PFOS increased. While this study does not assess PFOS levels over time, it does demonstrate
dose-dependent increases in serum concentrations.
Wimsatt et al. (2016, 3981396) dosed male (0 mg/kg, 10 mg/kg, 50 mg/kg, or 200 mg/kg single
dose) and female (0 mg/kg, 20 mg/kg, or 250 mg/kg single dose) mice with PFOS via drinking
water. After 8 weeks for males and 9 weeks for females, serum PFOS levels were found to be
dose-dependent.
Similar to rats {Cui, 2019, 757868}, PFOS exposure is found to cross the blood-brain barrier. In
Yu et al. (2019, 5918598), male ICR mice were dosed with 0 mg/kg/day, 0.25 mg/kg/day,
2.5 mg/kg/day, 25 mg/kg/day, or 50 mg/kg/day for 28 days via oral gavage, and measurements
of PFOS in serum and in brain deposits were collected. Mean serum PFOS levels were
approximately 0 |ig/mL, 5 |ig/mL, 40 |ig/mL, 240 |ig/mL, and 300 |ag/mL and PFOS levels in
the brain were approximately 0 jug/g, 2 jug/g, 5 jug/g, 30 jug/g, and 70 |ig/g for the 0 mg/kg/day,
0.25 mg/kg/day, 2.5 mg/kg/day, 25 mg/kg/day, and 50 mg/kg/day dose groups, respectively.
These data indicated that PFOS levels in serum and in brain deposits are dose-dependent and that
brain levels were much lower (100-fold less than that observed in blood and liver). These authors
also conducted in vitro studies showing that PFOS significantly decreased the expression of tight
junction-related proteins (e.g., ZO-1, Claudin-5, Claudin-11, Occludin) in endothelial cells.
These finding suggest that exposure to PFOS may also disrupt the blood-brain barrier, that in
turn could lead to increased accumulation of PFOS in brain.
Qui et al. (2013, 2850956) exposed ICR mice orally to PFOS at 0 mg/kg/day, 0.25 mg/kg/day,
2.5 mg/kg/day, 25 mg/kg/day, or 50 mg/kg/day for 28 days via gavage and examined the
testicular deposition of PFOS. The study found a positive correlation between the linear dose
dependent increases in serum concentration and testicle deposition, indicating that PFOS can
cross the blood-testis barrier in mice.
B.2.2.3 Tissue Transporters
PFOS entry from serum into tissues appears to be controlled by several families of membrane
transporters based on PFOA studies. Yu et al. (2011, 1294541) administered PFOS to rats and
extracted the messenger ribonucleic acids (mRNAs) for OATpl, OATp2, and MRP2 from the
liver to determine if changes in expression of transport molecules correlated with hepatic uptake.
B-14
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Female Wistar rats were administered PFOS at 0 mg/kg/day, 0.2 mg/kg/day, 1 mg/kg/day, or
3 mg/kg/day via gavage for 5 consecutive days. Blood, bile, and liver tissue were collected
24 hours after the last dose. Exposure to 3.0 mg/kg/day of PFOS increased hepatic OATp2
mRNA expression (1.43-fold) while MRP2 was increased approximately 1.80-fold and 1.69-fold
in the 1 mg/kg/day and 3 mg/kg/day groups, respectively. No effect with treatment was observed
on OATpl.
Transporters responsible for PFOS transport across the placenta are not well understood. Kummu
et al. (2015, 3789332) used placentas donated from healthy mothers to investigate the role of
OAT4 and ATP-binding cassette transporter G2 (ABCG2) proteins. Using an ex vivo perfusion
system, the authors administered concentrations of PFOA and PFOS (1,000 ng/mL) by perfusing
through the maternal circulation. The fetal: maternal ratios of PFOA and PFOS were 0.20 ± 0.04
and 0.26 ± 0.09, which corresponded to transfer index percentages (TI%) of 12.9 ± 1.5% and
14.4 ± 3.9%, respectively. Immunoblot analysis of OAT4 and AGCG2 in perfused placentas
indicated a linear negative correlation between the expression of OAT4 protein (but not ABCG2)
and PFOA (r2 = 0.92, p = 0.043) and PFOS (r2 = 0.99, p = 0.007) transfer at 120 min. The
authors speculated that OAT4 may play a role in decreasing placental passage of PFAS and
intrauterine exposure to these compounds; however, the low number of placentas examined and
lack of direct evidence for uptake via OAT4 indicates further studies are needed to understand
what, if any, role transporters play in placental transfer of PFOA and PFOS.
To further elucidate the role of placental transporters in facilitating the transfer of maternal PFAS
into the fetus, Li et al. (2020, 6505874) compared gene expression of selected transporters in
preterm and full-term placentas and determined whether the differences in expression could
influence the transplacental transfer efficiencies (TTEs). The authors selected nine placental
genes with known xenobiotic activity on the maternal side of the placenta: organic
cation/carnitine transporter 2 (OCTN2), reduced folate carrier 1 (RFC-1), equilibrative
nucleoside transporter (ENT1), folate receptor alpha (FRa), heme carrier protein 1 (PCFT),
serotonin transporter (SERT), p-glycoprotein (MDR1), multi-drug resistance-associated protein 2
(MRP2), and breast cancer resistance protein (BCRP). MDR1 expression levels were
significantly associated with TTEs of branched PFOS and iso-PFOS, (3+4+5)m-PFOS, but not
linear PFOS or PFOA. MRP2 expression was associated with total PFOS, linear PFOS, branched
PFOS, and iso-PFOS, (3+4+5)m-PFOS, but not PFOA. BCRP expression levels did not
significantly change with PFOA or PFOS. Interestingly, the pattern of expression of MDR1,
MRP2 and BCRP were only observed in full-term placentas. Preterm placentas showed
significant expression levels of ENT1, FRa, and SERT and were associated with lm-PFOS and
iso-PFOS. Thus, the expression of transporters and TTEs appear to differ between preterm and
full-term placentas. Authors noted that the three transporters that were significantly associated
with PFOS (MDR1, MRP2, and BCRP) are also ATP-binding cassette (ABC) transporters,
which play a protective role for the placenta tissue and the fetus by effluxing xenobiotics across
the placental barrier thereby reducing exposure to PFOS. It is unclear why there were no
correlations with PFOA although this may be related to the fact that gene expression associations
with TTE were not confirmed using protein expression data of the candidate genes.
More research is needed to explain how different transporters respond to PFAS and whether
physiochemical properties such as chain length and branching may influence the substrate
binding capacity of these transplacental transporters.
B-15
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.2.3 Distribution during Reproduction and Development
The availability of distribution data from pregnant females plus animal pups and neonates is a
strength of the PFOS pharmacokinetic database because it helps to identify those tissues
receiving the highest concentration of PFOS during development. For this reason, the
information on tissue levels during reproduction and development are presented separately from
those that are representative of other life stages.
B. 2.3.1 Human Studies
Zhang et al. (2013, 3859792) recruited 32 pregnant females (aged 21-39 years; gestational
period 35-47 weeks) from Tianjin, China, for a study to examine the distribution of PFOS
between maternal blood, cord blood, the placenta, and amniotic fluid. Samples were collected at
time of delivery (31 maternal whole blood samples, 30 cord blood samples, 29 amniotic fluid
samples, and 29 placentas). The maternal blood contained variable levels of 10 PFAS, and the
mean maternal blood concentration was highest for PFOS (14.6 ng/mL), followed by PFOA
(3.35 ng/mL). In both cases, the mean was greater than the median, indicating a distribution
skewed toward the higher concentrations. PFOS was found in all fluids/tissues sampled. It was
transferred to the amniotic fluid to a lesser extent than PFOA based on their relative proportions
in the maternal blood and cord blood (21% vs. 58%, respectively). Compared to the mean PFOS
value in maternal blood, the mean levels in the cord blood, placenta, and amniotic fluid were
21%), 56%o, and 0.14%> of the mean levels in the mother's blood, respectively. The correlation
coefficients between the maternal PFOS blood levels and placenta, cord blood, and amniotic
fluid levels ranged from 0.7 to 0.9 (p < 0.001).
B.2.3.1.1 Partitioning to Placenta
The placenta serves as an important link between the mother and the growing fetus throughout
gestation. It forms a physiological barrier that facilitates the exchange of nutrients, gases,
xenobiotics, and several biological components between maternal and fetal circulation. Several
PFAS compounds including PFOA and PFOS have been identified in amniotic fluid, cord blood,
and fetal tissue, indicating that these chemicals cross the transplacental barrier and influence
PFAS distribution to the fetus and elimination during pregnancy.
The role of the placenta in facilitating the transport of PFAS compounds to the fetal
compartment during gestation is informed by the ratio of placental concentration and matched
maternal serum concentration, or RPM. RPM is a quantitative measure of the placenta's ability
to retain or accumulate compounds. To determine the transplacental transfer of PFOS, Chen et
al. (2017, 3859806; 2017, 3981340) examined the distribution of PFAS in maternal serum, cord
serum, and placentas from 32 pregnant women and their matched infants in Wuhan, China. Mean
maternal age for the population was 27.1 years, with average pre-pregnancy BMI of 20.4 and
gestational age of 38.9 weeks. In Chen et al. (2017, 3859806), mean concentrations of total
PFOS in the placentas, cord serum, and maternal serum were 2.842 ng/g, 3.668 ng/mL, and
8.670 ng/mL, respectively, and the mean RPM was 0.330. The PFOS concentrations in all three
matrices from Chen et al. (2017, 3981340) followed a similar pattern, however, the PFOS
accumulation in the placenta was approximately 14.5%> less in Chen et al. (2017, 3981340) than
in Chen et al. (2017, 3859806).
B-16
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Zhang et al. (2013, 3859792) (described above) recorded mean PFOS concentrations of
8.18 ng/g in the placenta, 3.09 ng/mL in cord blood, and 14.6 ng/mL in maternal blood. These
concentrations were significantly higher than the PFOA concentrations in all three
compartments. Based on RPM, 59% of maternal PFOS is accumulated in the placenta. This
study and the Chen et al. (2017, 3859806; 2017, 3981340) studies had similar maternal
characteristics (sample size, geographical location (China), gestational age, maternal age), yet
placental PFOS accumulation significantly varied across studies, ranging from 4.8% to 59%.
One distinguishing characteristic that may account for increased PFOS accumulation in Zhang et
al. (2013, 3859792) is parity. About 82% of the mothers in Zhang et al. (2013, 3859792) were
primiparous whereas only 46.8% were primiparous in Chen et al. (2017, 3859806; 2017,
3981340), which may explain the higher PFOS concentrations in maternal serum and placenta
found in the Zhang et al. (2013, 3859792) study. Primiparous mothers also tend to have higher
levels of PFAS in breast milk than women who have had multiple children {Lee, 2013,
3983576}, adding to the evidence that pregnancy and lactation durations are critical for PFAS
distribution.
Mamsen et al. (2019, 5080595) demonstrated that factors such as gestational age can affect
PFOS concentrations in maternal serum and placentas. Using a linear graph of normalized
percentage placenta accumulation as a function of gestational age, the authors observed a steady
increase of placenta accumulation of PFOS during gestation days 50 to 300, with male and
female placentas showing similar trends. However, accumulation was significantly higher in
males than in females. Authors estimated a placenta PFOS accumulation rate of 0.13% increase
per day during gestation.
Zhang et al. (2015, 2851103) determined that branched PFOS makes up 18% of total PFOS in
placenta, suggesting that branched and linear PFOS accumulate in the placenta at different
proportions. Among branched isomers of the same compound, RPM seemed to differ by
functional groups and branching. Particularly, RPM of branched PFOS isomers seem to increase
as the branching points away from the sulfonate group: iso-PFOS < 4m-PFOS < (3+5)m-
PFOS < lm-PFOS. In contrast, the RPM of PFHxS showed a different pattern: branched
PFHxS < linear PFHxS {Chen, 2017, 3859806}. Moreover, RPM of linear and branched PFOA
(3m-PFOA) did not significantly differ from each other. The variation in RPM between the
branched isomers of PFOS, PFHxS, PFOA and their corresponding linear isomers suggest that
their capacity to accumulate in the placenta is partly influenced by structure, functional group,
and isomerization.
Umbilical cord blood is a known tissue for PFOS distribution during pregnancy. Kato et al.
(2014, 2851230) collected blood samples from 71 mothers and their infants in a prospective birth
cohort in the Cincinnati, Ohio metropolitan area. They quantified PFAS in maternal blood at 16
weeks of gestation and at delivery, evaluated the correlation between maternal PFAS levels in
maternal serum and matched cord blood. Maternal serum levels at 16 weeks of gestation and at
the time of delivery were higher for PFOS (12.7 |ig/L and 8.50 |ig/L, respectively) than PFOA
(4.8 |ig/L and 3.3 |ig/L, respectively). Authors reported a positive correlation between maternal
serum PFOS levels during gestation and cord serum (correlation coefficient = 0.87). Similarly,
the correlation between maternal serum at the time of delivery and cord serum was also positive
(correlation coefficient = 0.82).
B-17
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Porpora et al. (2013, 2150057) quantified PFOS levels in maternal serum and cord blood from 38
mother-infant pairs in Rome, Italy. The women were Italian Caucasian between the ages of 26
and 45 (mean age, 34.5 years). The average gestational age for participants in this study was
39 weeks. Maternal and cord serum PFOS concentrations were 3.2 ng/g and 1.4 ng/g,
respectively. A strong positive correlation was observed between maternal and cord serum
concentrations (r = 0.74, p < 0.001). These values suggest a cord to maternal serum ratio of 0.44.
Fromme et al. (2010, 1290877) measured PFOS in mothers and infants in Munich, Germany.
Maternal blood was sampled during pregnancy, at delivery, and 6 months after delivery in
mothers aged 21-43 years. PFOS was also measured in cord blood and in infant blood at 6 and
19 months after birth. Maternal PFOS serum concentrations ranged from 0.8 to 9.4 |ig/L (38
samples) and cord serum concentrations ranged from 0.3 to 2.8 |ig/L (33 samples). The cord to
maternal serum mean ratio was 0.3.
Wang et al. (2019, 5083694) measured the levels of 10 PFAS chemicals, including PFOS, in
paired maternal and umbilical cord serum from a prospective birth cohort in Shandong, China.
PFOS was detected in all maternal and umbilical cord serum samples with a geometric mean of
4.25 ng/mL (range of 0.55 ng/mL-29.85 ng/mL) in maternal serum and 1.33 ng/mL (range
0.12 ng/mL-5.89 ng/mL) in cord serum. PFOS concentrations in maternal serum were strongly
correlated to concentrations in cord blood (r = 0.745).
Linear and branched PFOS have been detected in both maternal and cord serum {Cai, 2020,
6318671; Li, 2020, 6505874}. Branched PFOS levels in cord blood are consistently lower than
linear PFOS levels. Branched PFOS isomers contributed approximately 19.5% of total PFOS in
cord blood {Cai, 2020, 6318671}. Similarly, Li et al. (2020, 6505874) showed that branched
PFOS makes up 17% of total PFOS in cord blood from preterm births and 19.2% from full-term
births (Table B-7). Together, these studies suggest that branched PFOS is likely less
accumulative in cord blood than linear isomers. It is worth noting that other factors, such as
differential binding affinities in serum and type of chemical exposure (branched vs. linear
PFOS), may also influence the proportions in serum.
Similar to PFOA, differential TTEs were observed for linear PFOS isomers. Cai et al. (2020,
6318671) found an 8% increase in branched PFOS accumulation compared to linear PFOS
isomers. Similarly, Li et al. (2020, 6505874) showed a 6% increase in branched PFOS
accumulation compared to linear PFOS isomers. Zhao et al. (2017, 3856461) observed higher
TTEs for lm, 4m, 3+5m, and m2 compared to n-PFOS. Moreover, the TTEs of branched PFOS
isomers increased as the branching point moved closer to the sulfonate moiety. Together, these
findings indicate that branched isomers of PFOS transfer more efficiently from maternal blood to
cord blood compared to linear isomers.
In summary, these studies suggest that maternal serum levels of PFOS is positively correlated
with cord blood and is a direct determinant of in utero exposure regardless of gestational age or
location of exposure. Maternal serum PFOS levels are consistently higher than cord serum levels
across all studies. PFOS concentrations in both maternal and cord serum varied substantially
across studies, and factors such as exposure sources, parity, and other maternal demographics
may account for these variations. For example, in Eryasa et al. (2019, 5412430), authors noted
that seafood diet (including high consumption of pilot whale) and consumer products as main
sources of exposure. This may likely explain why maternal and cord serum PFOS concentrations
B-18
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
are higher than all other studies listed in Table B-7. Additionally, linear PFOS are detected at
higher frequency and at higher levels in blood than branched PFOS but are less transferable
across compartments from maternal serum to cord serum.
B-19
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-7. PFOS concentrations in Human Cord Blood, Maternal Blood, and Transplacental Transfer Ratios (RCM)
Study
Number of
„ , , -w- « . Mean Gestational PFOS
Country, Cohort Maternal-Iniant . . . ,h ,,
pajrsa Age (weeks) Measurement
Cord Serum
(ng/mL)c
Maternal Serum
(ng/mL)c
Cord: Maternal
serum ratios
(Rcm)"
Manzano-Salgado
Sabadell and 53 NR total PFOS
1.86
6.99
0.30
et al. (2015,
Valencia, Spain
3448674)
Note: Serum concentrations reported as p50. whereas geometric mean concentrations were used by authors to calculate cord:maternal serum
ratios. Reported concentrations from 66 maternal plasma samples, and 66 cord blood samples, and 53 maternal serum samples.
Chen et al. (2017,
Wuhan, China 32 38.9 ±1.6 total PFOS
3.67 ±2.51
8.67 ±5.27
0.431
3981340)and
Chen et al. (2017,
3859806)
n-PFOS
iso-PFOS
2.713
0.203
6.971
0.49
0.384
0.388
(3+5)m-PFOS
0.506
0.466
0.684
4m-PFOS
1.8
0.157
0.695
lm-PFOS
0.226
0.136
0.835
Cariou et al.
(2015, 3859840)
Note: PFOS detected in 100% of maternal and cord samples except for m-PFOS in cord samples, where the detection rate of 96.87%. PFOS
isomers were reported in Chen (2017, 3981340) and total PFOS was reported in Chen (2017, 3859806).
Toulouse, France 94 NR total PFOS 1.28 3.67 0.38
Note: Concentrations represent mean values from 100 pairs. Semi-quantified values below LOD were taken into account for mean
calculation.
Eryasa et al.
Faroese Birth
100
39.9 ± 1.3
total PFOS
9.5 (6.34-13.89)
23.8 (15.8-36.9)
0.38e
(2019, 5412430)
Cohort, Denmark
(cohort 3)
n-PFOS
5.98 (3.97-8.71)
15.6 (10.5-22.96)
0.37
branched PFOS
3.50 (2.38-4.94)
8.15(5.22-12.58)
0.42
Faroese Birth
51
39.7 ± 1.1
total PFOS
3.09 (2.31-4.42)
8.82 (6.94-11.6)
0.36e
Cohort, Denmark
n-PFOS
1.89 (1.46-2.84)
5.55 (4.16-7.45)
0.35
(cohort 5)
branched PFOS
1.17 (0.88-1.73)
3.18(2.35-4.33)
0.37
Note: Cohort 3 included 100 singleton births from 1999 to 2001 and Cohort 5 included 51 singleton births from 2008 to 2005. Both cohorts
had the same source of exposure and are similar in maternal characteristics. Ratios were reported as median p50. Serum concentrations
reported here geometric mean and interquartile ranges(IQR).
Cai et al. (2020,
6318671)
Maoming Birth
Cohort, China
424
39.3 ± 1.1
total PFOS
linear PFOS
branched PFOS
2.66 ±4.80
2.14 ±4.42
0.52 ±0.49
6.71 ± 19.57
5.62 ± 17.33
1.09 ±2.35
0.51
0.5
0.58
B-20
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Country, Cohort
Number of
Maternal-Infant
Pairs"
Mean Gestational
Age (weeks)b
PFOS
Measurement
Cord Serum
(ng/mL)c
Maternal Serum
(ng/mL)c
Cord: Maternal
serum ratios
(Rcm)"
Note: Values represented as mean concentrations ± SD. Ratios were calculated from matched maternal and infant pairs for which all cord
blood samples were >LOD. Percent detect rates were 100% for total PFOS, 99.76% for linear PFOS, and 99.53% for branched PFOS.
Li et al. (2020,
Maoming Birth
86 33.8 ±3.0
total PFOS
1.93
5.87
0.32
6505874)
Cohort, China
linear PFOS
1.6
4.85
0.3
(pre-term infants)
branched PFOS
0.33
1.01
0.36
iso-PFOS
0.08
0.35
0.26
(3+4+5)m-PFOS
0.2
0.57
0.35
lm-PFOS
0.06
0.09
0.65
Maoming Birth
187 39.5 ±1.1
total PFOS
2.6
4.44
0.58
Cohort, China
linear PFOS
2.1
3.76
0.57
(full-term infants)
branched PFOS
0.5
0.68
0.68
iso-PFOS
0.11
0.2
0.51
(3+4+5)m-PFOS
0.32
0.41
0.73
lm-PFOS
0.08
0.07
1.07
Note: 273 mother-infant pairs were analyzed, including 86 preterm deliveries and 187 full-term deliveries. Only PFAS substances
quantifiable in >50% of maternal and cord sera are included in generating mean concentration values.
Li et al. (2020,
Beijing, China
112 39.0 ±1.2
total PFOS
2.31
6.74
0.482
6506038)
Note: PFOA detection rate was 97.44% in maternal serum and 95.73% in cord serum. For PFOS, 112 of 117 matched cord and maternal
serum samples were used to generate Rcm.
Wangetal. (2019,
Shandong, China
369 39.4 ±1.3
total PFOS
1.33
4.25
0.30
5083694)
Note: PFOS detected in 100% of maternal and cord samples.
Panetal. (2017,
Wuhan, China
100 39.4 ±1.3
total PFOS
4.33
12.7
0.34
3981900)
Note: Maternal blood collected in third trimester (38.4 ±1.6 weeks) used for RCm
calculation and PFOS was detected in 100% of maternal
and cord samples.
Zhao et al. (2017,
People's Hospital
63 39.3 ±0.82
n-PFOS
3.86
16.8
0.21
3856461)
of Hong'an
59 39.3 ±0.82
iso-PFOS
0.229
1.08
0.22
County, China
63 39.3 ±0.82
3+5m-PFOS
0.417
1.44
0.29
38 39.3 ±0.82
4m-PFOS
0.142
0.536
0.51
61 39.3 ±0.82
lm-PFOS
0.716
1.25
0.48
19 39.3 ±0.82
m2-PFOS
0.043
0.099
0.3
63 39.3 ±0.82
total-PFOS
5.41
21.2
0.22
B-21
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Country, Cohort
Number of
Maternal-Infant
Pairs"
Mean Gestational
Age (weeks)b
PFOS
Measurement
Cord Serum
(ng/mL)c
Maternal Serum
(ng/mL)c
Cord: Maternal
serum ratios
(Rcm)"
Note: Authors reported that samples < LOD were not included in RCM analysis. Mean ratios reported for matched pairs.
Beeson et al.
Chemicals, Health
20
NR
total PFOS
1.8
5.5
0.33
(2011,2050293)
and Pregnancy
20
NR
n-PFOS
NR
NR
0.33
(CHirP) cohort,
20
NR
Iso-PFOS
NR
NR
0.36
Vancouver,
Canada
20
NR
5m-PFOS
NR
NR
0.53
20
NR
4m-PFOS
NR
NR
0.53
20
NR
3m-PFOS
NR
NR
0.67
20
NR
lm-PFOS
NR
NR
0.87
Note: Ratios were derived from PFOA concentrations in cord serum at delivery by maternal serum concentration at 15 weeks of gestation
for each mother-cord pair
Fei et al. (20071,
Danish National
50
40.06 ± 1.57
total PFOS
11.0 ±4.7
35.3 ± 13.0
0.29
1005775)
Birth Cohort,
maternal blood
obtained in first
trimester
Danish National
50
40.06 ± 1.57
total PFOS
11.0 ±4.7
29.9 ± 11.0
0.34
Birth Cohort,
maternal blood
obtained in second
trimester
Note: First trimester samples collected between gestation weeks 4 and 14. Timing of second trimester blood collection was not reported.
Ratios and concentrations were generated from blood samples collected from 50 randomly selected matched maternal-cord pairs that met
study criteria (from a total of = 80,678 maternal participants in the cohort).
Hanssen et al.
(2010, 2919297)
Johannesburg.
South Africa
NR
total PFOS
0.7
71 maternal
samples, 58 cord
samples
Note: Authors did not specify if matched maternal and cord blood samples were used to derive ratios.
1.6
0.45
Inoue et al. (2004,
2994839)
Kim et al. (2011,
1424975)
Hokkaido, Japan 15 39.7 ±1.05 total PFOS 1.6 -5.3 4.9 - 17.6 0.32
Note: Authors collected maternal and cord blood from 15 matched pairs. Authors report individual concentrations, but not mean
concentrations for this population.
Seoul, Cheongju 44 maternal 39 ±1.6 total PFOS 1.26 (0.81-1.82) 2.93 (2.0-4.36) 0.48
and Gumi, South samples, 43 cord
Korea samples
B-22
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Number of
„ , , -w- « . Mean Gestational PFOS
Country, Cohort Maternal-Iniant . . . ,h ,,
p . a Age (weeks) Measurement
Cord Serum
(ng/mL)c
Maternal Serum
(ng/mL)c
Cord: Maternal
serum ratios
(Rcm)"
Note: Median serum concentrations reported. Values in parentheses are 25-75% IQRs
Fromme et al.
(2010, 1290877)
Germany 38 maternal NR total PFOS
samples, 33 cord
samples
Note: Maternal and cord blood samples taken at time of delivery.
1
2.9
0.3
Needham et al.
(2011, 1312781)
Faroe Islands 12 NR total PFOS 6.6
Note: Serum concentrations reported as median values, RCMs reported as arithmetic means.
19.7
0.34
Liu et al. (2011,
2919240)
Jinhu, China 50 (all) NR total PFOS
26 (males infants) NR total PFOS
24 (female infants) NR total PFOS
Note: Maternal samples collected in the first weeks after delivery.
1.686
NR
NR
3.184
NR
NR
0.57
0.55
0.58
Midasch et al.
(2007, 1290901)
NR 11 NR total PFOS 7.3
Note: Serum concentrations reported as median values, RCMs reported as arithmetic means
13
0.6
Verner et al.
(2015, 3299692)
NA NA NA NA NA NA 0.45
Note: Authors developed a two-compartment, two-generation pharmacokinetic model of prenatal and postnatal exposure to PFOA and
PFOS. Rcms applied in model were derived from an average of ratios reported in Aylward et al. (2014, 2920555).
Notes: IQR = interquartile range; LOD = level of detection; NA = not applicable, NR = not reported; SD = standard deviation.
a Number represents number of matched pairs used for RCM calculation unless otherwise noted in comments.
b Gestational age reported as mean ± SD, represents gestational age at the time of cord blood sampling (delivery) and may not be the same as age at the time of maternal blood
sampling.
c Concentrations in cord or maternal samples are reported as means with or without SD or IQR unless otherwise noted in comments. Note that several studies, the mean serum
concentrations may be derived from more subjects than values used for RCM calculation, which typically included only matched pairs for which both cord and maternal serum
concentrations were above the limit of detection.
dData are presented as a ratio of cord serum to maternal serum concentrations unless otherwise noted in comments.
B-23
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.2.3.1.2 Partitioning to Amniotic Fluid
Zhang et al (2013, 3859792) measured the levels of 11 PFAS chemicals in maternal blood, cord
blood and placenta. All 11 PFAS were detected in their respective biological tissues at different
concentrations. The mean concentration ratio between amniotic fluid and maternal blood
(AF:MB) was higher in PFOA (0.13) than in PFOS (0.0014). Similarly, the mean concentration
ratio between amniotic fluid and cord blood (AF:CB) was higher in PFOA (0.023) than in PFOS
(0.0065). Authors attributed the differences in ratios between the two compartments to the
solubility of PFOS and PFOA and their respective binding protein binding capacities in the two
matrices. PFOA is highly soluble in water relative to PFOS (solubilities of 3.4 g/L and 0.68 g/L,
respectively). Since amniotic fluid is 94% water, the solubility properties may account for the
observation that the PFOA concentration (0.044 ng/mL) was twice as much as PFOS
(0.02 ng/mL) in this matrix.
Table B-8 presents means or medians and ranges of measured and estimated PFOS
concentrations in maternal blood from recent studies (2013 to present) that also measured fetal
indicators of exposure (cord blood, placenta, and/or amniotic fluid). These studies demonstrate
the variability of PFOS accumulation in these tissues across geographic regions. Maternal serum
values ranged from 0.062 ng/mL in Rome, Italy {Porpora, 2013, 2150057} to 183 ng/mL in
Hubei, China {Zhao, 2017, 5085130}. Cord serum values ranged from < LOD in Wuhan, China
{Chen, 2017, 3859806} and Toulouse, France {Cariou, 2015, 3859840} to 13.89 ng/mL in Faroe
Islands, Denmark {Eryasa, 2019, 5412430}. Fewer studies measured PFOS in placentas and
amniotic fluid. Placenta values were lower than maternal and cord blood values and ranged from
0.06 ng/g in Wuhan, China {Chen, 2017, 3981340} to 21.4 ng/g in Tianjin, China {Zhang, 2013,
3859792}. Only two studies from Tianjin, China measured PFOS in amniotic fluid, which
showed lower levels than those observed in other matrices. Values ranged from < LOD {Zhang,
2014, 2850251} to 0.121 ng/mL {Zhang, 2013, 3859792}. The very wide concentration ranges
observed across these geographic locations and matrices highlight the challenges of comparing
partitioning of PFOS from mother to fetus across studies.
In addition to geographic variation, inter-individual variability likely plays an important role in
the range of concentrations observed in maternal and fetal tissues and matrices. Variability was
examined by Brochot et al. (2019, 5381552) using a PBPK model calibrated in a population
framework to provide quantitative estimates for the PFOA and PFOS placental transfers in
humans. The measured values of maternal plasma:cord serum inputted in their model were, on
average, close to 1 but showed a variability of close to tenfold. The measured transfer rates of
PFOA and PFOS used were also quite variable, indicating that PFOA crosses the placental
barrier at a 3-times higher rate than PFOS. The coefficients of variation of the maximal transfer
rate across subjects were estimated at 75% for PFOA and 55% for PFOS, Variation was also
observed in the ranking of PFOA and PFOS when comparing exposure levels to fetal indicators
of exposure. Maternal daily intake estimates were then used as inputs to the PBPK model to
simulate the fetal exposure in several target organs over the whole pregnancy. The PFOA and
PFOS fetal plasma concentrations are quite similar at the end of pregnancy for the whole cohort.
This similarity was also predicted for brain, but not in kidneys and liver. When examined at the
individual level, the ranking of PFOA and PFOS exposure exhibited a wide range of variability.
Interestingly, the model estimated that approximately one-third of the population has levels of
one compound always higher than levels of the other compound, whereas the remaining two-
thirds exhibited different patterns of accumulation for PFOA and PFOS. The majority, however,
B-24
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
were predicted to accumulate PFOA at higher levels than PFOS levels for most of the fetal
indicators of exposure. The authors concluded that differences in fetal exposure are not predicted
by the measurement of the maternal concentration during pregnancy.
Table B-8. Summary of PFOS Concentrations in Human Maternal Blood, Cord Blood,
Placenta and Amniotic Fluid Studies
Study (Location
Maternal Blood
Cord Blood
Infant
Placenta
Amniotic Fluid
of Study)
Blood
Porpora et al.
Maternal serum
Cord serum
NR
NR
NR
(2013, 2150057)
Mean: 3.2 ng/g
Mean: 1.4 ng/g
(Rome, Italy)
Median: 2.9 ng/g
Median: 1.1
Range: 0.062-13 ng/g
Range: 0.23-3.7 ng/g
Zhang et al.
NR
NR
NR
Mean: 8.18 ng/g
Mean: 0.020 ng/mL
(2014, 2850251)
Median:
Median: < LOQ
(Tianjin, China)
7.32 ng/g
ng/mL
Yang et al.
Maternal serum
Cord serum
NR
NR
NR
(2016, 3858535)
Mean: 3.10 ng/mL
Mean: 1.41 ng/mL
(Jiangsu, China)
SD: 1.44 ng/mL
SD: 0.93 ng/mL
Median 2.98 ng/mL
Median: 1.23 ng/mL
Range: 0.76-
Range: 0.25-
9.47 ng/mL
5.60 ng/mL
Manzano-
Maternal plasma
Cord serum
NR
NR
NR
Salgado et al.
Median: 6.18 ng/mL
Median: 1.86 ng/mL
(2015, 3448674)
Range: 1.46-
Range: 0.53-
(Sabadell and
38.58 ng/mL
4.71 ng/mL
Valencia, Spain)
IQR: 4.44-
IQR: 1.40-3.07 ng/mL
12.63 ng/mL
Maternal serum
Median: 6.99 ng/mL
Range: 1.17-
23.14 ng/mL
IQR: 4.47-
11.12 ng/mL
Chen et al. (2017, Mean: 8.670 ng/mL,
Mean: 0.331 ng/mL,
NR
Mean:
NR
3859806)
Range: 1.72-
Range: LOD-
0.216 ng/mL,
(Wuhan, China)
22.857 ng/mL
1.070 ng/mL
range: LOD-
0.531 ng/g
Chen et al. (2017, Maternal serum
Cord serum
NR
Mean: 0.42 ng/g
NR
3859806)
Mean: 8.670 ng/mL
Mean: 3.67 ng/mL
SD: 0.30 ng/g
(Wuhan, China)
SD: 5.27 ng/mL
SD: 2.51 ng/mL
Median:
Median: 7.01 ng/mL
Median: 3.64 ng/mL
0.35 ng/g range:
Range: 1.72-
Range: 0.54-
0.06-0.1.38 ng/g
22.9 ng/mL
12.7 ng/mL
Panetal. (2017,
Maternal serum
Cord serum
NR
NR
NR
3981900)
T1
Mean: 4.38 ng/mL
(Wuhan, China)3
Mean: 14.1 ng/mL
Median: 4.38 ng/mL
Median: 14.23 ng/mL
IQR: 2.68-6.19 ng/mL
IQR: 7.99-
21.68 ng/mL
Maternal serum
T")
NR
NR
NR
1Z
Mean: 13.0 ng/mL
B-25
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study (Location
Maternal Blood
Cord Blood
Infant
Placenta
Amniotic Fluid
of Study)
Blood
Median: 13.20 ng/mL
IQR: 7.62-
20.38 ng/mL
Maternal serum
T"3
NR
NR
NR
X j
Mean: 12.7 ng/mL
Median: 12.32 ng/mL
IQR: 7.61-
20.03 ng/mL
Caserta et al.
Mean: 1.54 ng/mL
Mean: 1.75 ng/mL
NR
NR
NR
(2018, 4728855)
SD: 1.28 ng/mL
SD: 1.70 ng/mL
(Rome, Italy)
Range: 0.018-
Range: 0.018-
4.7 ng/mL
6.00 ng/mL
Wang et al.
Maternal serum
Cord serum
NR
NR
NR
(2019, 5083694)
GM: 4.25 ng/mL
Mean: 1.33 ng/mL
(Shandong,
Median: 4.55 ng/mL
Median: 1.39 ng/mL
China)
Range: 0.55-
Range: 0.12-
29.85 ng/mL
5.89 ng/mL
Zhao et al. (2017,
Maternal blood
Cord Blood
NR
NR
NR
3856461)
Mean: 21.2 ng/mL
Mean: 5.41 ng/mL
(Hong'an, China)
Median: 6.59 ng/mL
Median: 1.35 ng/mL
Range: 1.51—
Range: 0.346-
582 ng/mL
183 ng/mL
Brochot et al.
Group 1 mean
Mean: 2.08 ± 1.00
NR
NR
NR
(2019, 5381552)
(plasma): 7.14 ± 5.35
Range: 0.53-
(INMA
(0.69-38.58) ng/mL
4.71 ng/mL
Prospective birth
Group 2 mean
cohort, Spain)b
(plasma): 5.70 ± 3.45
(0.26-25.98) ng/mL
Gao et al. (2019,
Mean: 4.64 ng/mL
Mean: 2.35 ng/mL
NR
NR
NR
5387135)
median: 4.07 ng/mL
Median: 1.8 ng/mL
(Beijing, China)
range: 0.07-
Range: 0.04-
22.6 ng/mL
8.01 ng/mL
Eryasa et al.
Cohort 3
Cohort 3
NR
NR
NR
(2019, 5412430)
Maternal serum
Cord serum:
(Faroese Birth
Mean: 23.8 ng/mL
Mean: 9.50 ng/mL
Cohort,
SD: 1.2 ng/mL
SD: 0.49 ng/mL
Denmark)0
IQR: 15.8-
IQR: 6.34-
36.9 ng/mL
13.89 ng/mL
Whole cord blood:
Mean: 4.90 ng/mL
SD: 0.26 ng/mL
IQR: 3.33-6.94 ng/mL
Cohort 5
Cohort 5
NR
NR
NR
mean: 8.82 ng/mL
Cord serum:
SD: 0.51 ng/mL
mean: 3.09 ng/mL
IQR: 6.94-
SD: 0.22 ng/mL
11.6 ng/mL
IQR: 2.31-4.42 ng/mL
Whole cord blood:
mean: 1.60 ng/mL
B-26
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study (Location
Maternal Blood
Cord Blood
Infant
Placenta
Amniotic Fluid
of Study)
Blood
SD: 0.11 ng/mL
IQR: 1.18-2.32 ng/mL
Cai et al. (2020,
Maternal serum
Cord serum
NR
NR
NR
6318671)
Mean: 6.71 ng/mL
Mean: 2.66 ng/mL
(Maoming Birth
SD: 19.57 ng/mL
SD: 4.80 ng/mL
Cohort, China)
Median: 4.32 ng/mL
Median: 1.93 ng/mL
IQR: 2.94-
IQR: 1.23-2.66 ng/mL
6.34 ng/mL
Li et al. (2020,
Total PFOS:
Total PFOS:
NR
NR
NR
6505874)
Preterm delivery:
Preterm delivery:
(Maoming Birth
Mean: 5.87 ng/mL
Mean: 1.93 ng/mL
Cohort, China)d
Median: 3.53 ng/mL
Median: 1.47 ng/mL
IQR: 2.36-5.93
IQR: 0.83-1.97
Full-term delivery:
Full-term delivery:
Mean: 4.44 ng/mL
Mean: 2.60 ng/mL
Median: 3.54 ng/mL
Median: 2.08 ng/mL
IQR 2.25-5.98
IQR 1.28-3.06
Li et al. (2020,
Mean: 6.74 ng/mL
Mean: 2.31 ng/mL
NR
NR
NR
6506038)
(95% CI: 6.27, 8.95)
(95% CI: 2.9, 3.4)
(Maoming Birth
Median: 5.99 ng/mL
Median: 1.65 ng/mL
Cohort, China)
Zhang et al.
Mean: 14.6 ng/mL
Mean: 3.09 ng/mL
NR
Mean: 8.18 ng/g
Mean: 0.020 ng/mL
(2013, 2639569)
RSD: 4.98
RSD: 1.84
RSD: 3.03
RSD: 0.032
(Tiajin, China)
Range: 7.39-
Range: 0.14-
Range: 3.25-
Range:
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study (Location
Maternal Blood
Cord Blood
Infant
Placenta
Amniotic Fluid
of Study)
Blood
range: < 0.08-
0.89 ng/mL
Mamsen et al.
Mean: 8.2 ng/g,
NR
NR
Mean: 1.3 ng/
NR
(2017, 3858487)
Range: 2.5-16.7 ng/g
Range: 0.3-
(Denmark)
3.1 ng/g
Mamsen et al.
Tl serum
NR
NR
Mean: 1.43 ng/g
NR
(2019, 5080595)
Mean: 8.14 ng/mL
SD: 0.63 ng/g
(Denmark)3
SD: 3.82 ng/mL
Median:
Median: 6.76 ng/mL
1.35 ng/g
Range: 2.49-
Range: 0.65-
16.66 ng/mL
3.09 ng/g
T2 serum
NR
NR
Mean: 1.23 ng/g
NR
Mean: 3.87 ng/mL
SD: 0.60 ng/g
SD: 1.99 ng/mL
Median:
Median: 3.43 ng/mL
1.08 ng/g
Range: 1.04-
Range: 0.63-
8.19 ng/mL
2.33 ng/g
T3 serum
NR
NR
Mean: 1.53 ng/g
NR
Mean: 3.58 ng/mL
SD: 0.90 ng/g
SD: 1.85 ng/mL
Median:
Median: 3.26 ng/mL
1.42 ng/g
Range: 1.07-
Range: 0.45-
9.66 ng/mL
3.87 ng/g
Kato et al. (2014,
Maternal Serum at
Cord serum at delivery
2851230)
16 weeks
Median: 3.50 |ig/L
(Ohio, USA)f
Median: 12.70 \ig!L
Maternal serum at
delivery
Median: 8.50 [ig/L
Notes: CI = confidence interval; INMA = INfancia y Medio Ambiente; IQR = interquartile range; LOD = limit of detection;
LOQ = limit of quantification; SD = standard deviation; NR = not reported; RSD =relative standard deviation; T1 = trimester 1;
T2 = trimester 2; T3 = trimester 3.
aPFOS was collected at different timepoints during gestation: first trimester (Tl), second trimester (T2) and third trimester (T3).
bBrochot et al., collected samples from women in 2 cohorts: Group 1 consist of 52 mother-child pairs that had available samples
of maternal blood during pregnancy and cord serum. Group 2 consist of 355 mothers who provided maternal blood during
pregnancy. Cord blood was not collected for the Group 2.
c Eryasa et al. (2019, 5412430) collected serum and whole blood from participants in two birth cohorts: Cohort 3 (100 Singleton
births from 1999 to 2001), and Cohort 5 (50 singleton birth from 2008 to 2005). Both cohorts had the same source of exposure
and are similar in maternal characteristics
dLi et al. (2020, 6505874) measured PFOS in matched maternal-cord serum pairs with pre-term deliveries and full-term
deliveries.
e Hanssen et al. (2013, 3859848) collected whole blood and plasma from women in 2 geographical locations: Norilsk (n = 7) and
Uzbekistan (n = 10). Cord blood and cord plasma from infants born to the Norilsk mothers only.
f Kato et al. (2014, 2851230) measured PFOS in 71 matched maternal and cord serum pairs. Maternal serum samples were
collected at 16 weeks of gestation and at the time of delivery
B.2.3.1.3 Distribution in Fetal Tissues
Mamsen et al. (2017, 3858487) measured the concentrations of 5 PFAS chemicals in human
fetuses, placentas, and maternal plasma from a cohort of 39 pregnant women in Denmark, who
legally terminated their pregnancies before gestational week 12 for reasons other than fetal
abnormality. The samples collected included 24 maternal blood, 34 placenta, and 108 fetal
B-28
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
organs. The participants were healthy women ages 18-46 years with an average BMI of 22.7.
About 51% of the mothers smoked during pregnancy at an average of 10 cigarettes per day or
were exposed to secondhand cigarette smoke for an average of 1.8 hours per day. Mean
concentrations of PFOS in maternal serum, placenta, and fetal organs were reported as 8.2 (2.5-
16.7) ng/g, 1.0 (0.3-2.6) ng/g, and 0.3 (0-0.7) ng/g, respectively. The concentrations of PFOS in
all three matrices were significantly higher than all four PFAS chemicals including PFOA. For
21 of the samples where all three specimens (maternal plasma, placenta, and fetal tissues) were
collected from the same women, the concentration of PFOS decreased from maternal serum to
fetal tissues as follows: maternal serum > placenta > fetal tissues. The relative concentration of
PFOS in the placenta was 14% of the concentrations found in maternal plasma and were further
reduced to 5% in fetal tissues. Although PFOS concentrations in all three matrices were higher
than the remaining PFAS chemicals, PFOS had the lowest relative concentrations in fetal tissues.
In general, a positive trend was observed between gestational age and fetal/maternal plasma
ratio. Although the gestational age reported in this study is short (37-68 days post conception),
the results suggest that PFOA and PFOS accumulate in the fetus and may potentially continue to
accumulate across gestation.
To determine whether PFOS accumulation in fetal organs changes across trimesters during
gestation, Mamsen et al. (2019, 5080595) quantified PFAS levels in embryos and fetuses at
gestational weeks 7-42 and serum from their matched maternal pairs. Like Mamsen et al. (2017,
3858487), participants were similar in age (18-46 years) and BMI (22.8 (first trimester)).
However, the smoking status of the women in this study was not reported and the majority of the
pregnancies were terminated due to intrauterine fetal death (IUFD) caused by placental
insufficiency and intrauterine growth restriction (58%), and infection (13%). A total of 78
pregnant women were enrolled in the study. Fetal tissues (placenta, liver, lung, heart, CNS, and
adipose) were collected from 38 first trimester pregnancies, 18 second trimester pregnancies, and
22 third trimester pregnancies. In all fetal tissues examined and across trimesters, PFOS
concentrations were highest compared to other PFAS. The concentration of PFAS in fetal tissues
fluctuated across trimesters and did not follow any particular trend. For example, PFOS
concentration in the liver was higher in the second trimester compared to the third trimester, and
lowest in the lung in the second trimester compared to the first and third trimesters. Interestingly,
PFOA concentration in the liver was also highest in the second trimester compared to the first
and third trimesters. Authors attributed this phenomenon to the unique architecture of the fetal
liver during early gestation when Authors attributed this phenomenon to the unique architecture
of the fetal liver during early gestation when oxygenated cord venous blood bypasses the liver
into the heart through the ductus venosus and is then delivered throughout the fetus. This pattern
of blood distribution changes between week 20 and 26 of gestation (late second trimester). The
amount of blood shunted from the liver is reduced from 60% to 30% in the second trimester
Pennati et al. (2003, 9642023). This reduction results in increased flow of cord blood through the
liver, thus increasing levels of PFOA and PFOS during the second trimester. Furthermore,
Mamsen et al. (2019, 5080595) observed that PFOA and PFOS levels were lowest in the CNS
than any of the tissues examined, suggesting that the CNS has less PFAS exposure and may be
protected by the blood brain barrier (BBB). When interpreting these results, it is important to
note that second and third trimester fetal tissues were obtained from patients with IUFD and may
not be comparable to normal pregnancies as the fetus died in utero of placental insufficiency and
intrauterine growth restriction. Placental insufficiency can potentially reduce the amount of
PFAS crossing the placenta. In addition, the PFAS exposure level in this cohort may vary due to
B-29
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
different geographical locations of the participants. The first trimester participants were from
Denmark and the second and third trimester participants came from Sweden.
B.2.3.1.4 Partitioning to Infants
Four studies shown in Table B-9 analyzed PFOS levels in maternal serum and levels in breast
milk and/or infant blood. Maternal and infant serum PFOS levels were substantially higher in
subjects in the United States exposed to contaminated drinking water {Mondal, 2014, 2850916}
compared to subjects analyzed in France, Denmark (Faroe Islands), or Sweden {Cariou, 2015,
3859840; Mogensen, 2015, 3859839; Gyllenhammar, 2018, 4778766}. In the Mondal study,
geometric mean (GM) maternal serum PFOS concentrations were lower in breastfeeding mothers
(11.63 ng/mL) vs. non-breastfeeding mothers (13.48 ng/mL). Conversely, breastfed infants had
higher GM serum PFOS (13.54 ng/mL) than infants who were never breastfed (12.65 ng/mL).
Cariou et al. (2015, 3859840) reported that PFOS levels in breastmilk were approximately 66-
fold lower relative to maternal serum and the ratio between breastmilk and maternal serum PFOS
was 0.38 ± 0.16 (n = 19). The authors noted that the transfer rates from serum to breastmilk of
PFAAs were lower compared to other lipophilic persistent organic pollutants such as
polychlorinated biphenyls. In this study, four PFAS compounds were analyzed (PFOA, PFOS,
PFNA, and PFHxS), and the individual patterns for these compounds exhibited important
interindividual variability. While PFOS was the main contributor in serum, PFOA and PFOS
were found to be the main contributors in breastmilk. Interestingly, while the number of
pregnancies was inversely correlated with maternal serum levels, after adjustment, the
correlation with parity did not reach significance for PFOS, although it did reach significance for
PFHxS.
Mogensen et al. (2015, 3859839) relied on maternal serum concentrations measured at 32 weeks
of pregnancy to assess prenatal exposure and measured concentrations in the serum of children at
11 and 18 months of age. They applied linear mixed models to estimate age-dependent serum
concentrations for up to 5 years after birth. The only other exposure source adjusted for in this
study was the eating whale meat by the infants. As shown in Table B-9, the increases in infant
blood PFOS concentrations over time, with the greatest increases found at the end of the
breastfeeding period, suggest that breastfeeding is the primary exposure source during infancy.
Gyllenhammar et al. (2018, 4778766) used multiple linear regression and general linear
model analysis to investigate associations between serum PFOS concentrations in 2-4-
month old infants and maternal PFOS concentrations close to delivery, duration of in utero
exposure (gestational age at delivery), duration of breastfeeding, and other parameters. The
authors examined PFAAs of various chain lengths and observed decreased strength of
association between maternal and infant concentrations with increased PFAA carbon chain
length among breastfed infants. PFOS showed the highest median in both infants and
mothers (order among measured PFAAs was
PFOS > PFOA > PFHxS > PFNA > PFDA > PFUnDA). The infant:maternal serum ratios
were similar for total, linear, and branched PFOS (0.69 (0.14-1.5), 0.66 (0.095-1.4), and
0.72 (0.19-1.7), respectively). Despite similar ratios, the authors observed that branched
PFOS isomer concentrations increased on average 1% per day of gestational age, whereas
linear isomer concentrations increased 0.75% per day of gestational age, supporting a higher
efficiency of placental transfer of branched as opposed to linear isomers during gestation.
B-30
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-9. Summary of Human PFOS Concentrations in Maternal Serum, Breast Milk, and Infant Serum
Study
Subjects
Maternal Blood
Breastmilk
Infant Blood
Mondal et al. (2014,
2850916)
Subjects were a subcohort of the C8
Science Panel Study (exposed to
contaminated drinking water in six
water districts near Parkersburg, West
Virginia) who had
a child <3.5 years of age and who
provided blood samples and reported
detailed information on breastfeeding
at the time of survey (633 mothers and
49 infants included). PFAA serum
concentrations were available for all
mothers and 8% (n = 49) of the
infants. Maternal and infant serum
concentrations were regressed on
duration of breastfeeding.
Maternal serum
Breastfed & not breastfed
GM: 12.33 ng/mL
95% CI: 11.77, 12.92
Breastfed:
GM: 11.63 ng/mL
95% CI: 10.98, 12.31
Not breastfed
GM: 13.48 ng/mL
95% CI: 12.45, 14.58
NR
Infant serum
Breastfed & not breastfed
GM: 13.21 ng/mL
95% CI: 11.17, 15.61
Breastfed
GM: 13.54 ng/mL
95% CI: 10.79, 17.00
Not breastfed
GM: 12.65 ng/mL
95% CI: 9.74, 16.43
Mogensen et al. (2015,
3859839)a
80 singleton children in Faroese birth
cohort born between 1997-2000. The
children were breastfed exclusively
for a median of 4.5 months, followed
by partial breastfeeding with
supplementary
baby food for a median of 4 months.
NR
NR
Birth: median: 6.0 ng/mL
(IQR 5.2,7.2)
11 months: median: 23.2 ng/mL
(IQR 14.9, 34.7)
18 months: median: 24.0 ng/mL
(IQR 20.2, 29.1)
60 months: median: 13.3 ng/mL
(IQR 10.6, 16.6)
Cariouetal. (2015,
3859840)
Female volunteers hospitalized
between June 2010 and January 2013
for planned caesarean delivery in
France. Maternal blood samples
(n = 100) were collected during
cesarean delivery and breast milk
samples (61) were collected between
the 4th and 5th day after delivery.
Maternal serum
Mean: 3.67 ng/mL
Median: 3.065 ng/mL
Range: 0.316-24.5 ng/mL
Mean: 0.040 ng/mL
Median: < LOQ
LOQ = 0.040 ng/mL
Range: < LOD-
0.376 ng/mL
NR
B-31
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Subjects
Maternal Blood
Breastmilk
Infant Blood
Gyllenhammar et al.
(2018, 4778766)
Primaparae mother/child pairs in
1996-1999 recruited in Sweden. 101
maternal and 107 infant samples were
available for PFAA analyses. Serum
concentrations were determined in
mothers 3 weeks after delivery and in
2-4-month old infants.
Maternal serum
Mean: 20 ng/g
SD: 8.9 ng/g
Median: 18 ng/g
Range: 7.7-61 ng/g
NR
Infant serum
Mean: 14 ng/g
SD: 6.7 ng/g
Median: 13 ng/g
Range: 2.2-44 ng/g
Notes: CI = confidence interval; GM = geometric mean; IQR = interquartile range; LOD = limit of detection; LOQ = limit of quantification; PFAA = perfluoroalkyl acid; NR = not
reported; SD = standard deviation.
a Neonatal serum-PFAS concentrations was calculated based on PFAS ratios between cord and maternal pregnancy serum concentrations previously estimated for the same cohort
(0.34 for PFOA) from Needham et al. (2011, 1312781).
B-32
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mondal et al. (2014, 2850916) also examined the change in maternal and infant PFOS levels
with duration of breastfeeding (Table B-10). Maternal serum concentrations decreased with
each month of breastfeeding (-3%; 95% CI: -5%, -2%) with the greatest decrease observed
after 12 months of breastfeeding (-39%). Correspondingly, the infant PFOS serum
concentrations increased by 4% (95% CI: 1%, 7%) with each month of breastfeeding. Using
mixed linear model regression (Table B-l 1), Mogensen et al. (2015, 3859839) calculated more
dramatic increases in infants during months with exclusive breastfeeding of 29.2% and 30.2%
per month at 18 and 60 months, respectively. Increases were less striking for months with partial
breastfeeding and small or none for months without breastfeeding. The Gyllenhammar et al.
(2018, 4778766) study included only five exclusively bottle-fed infants. In this group, they
observed a higher percentage of branched PFOS compared to exclusively breast-fed infants,
which may be the result of the higher efficiency of placental transfer of branched PFOS isomers
vs. linear isomers. Altogether, these findings support breastfeeding as the primary source of
infant PFOS accumulation and that distribution to the infant correlates with the length of
breastfeeding.
Table B-10. Percent Change in PFOS Ratios in Human Maternal Serum and Breast Milk
and Breast Milk and Infant Serum by Infant Age as Reported by Mondal et al. (2014,
2850916)
Infant Age
Maternal Serum: Breast Milk
Breastmilk: Infant Serum
< 6 months
-9% (-18%, 1%)
-31% (-53%, 1%)
7-12 months
-24% (-34%, -13%)
40% (-9%, 115%)
>12 months
-39% (-52%, -23%)
71% (9%, 167%)
Continuous (per month)
-3% (-3%, -2%)
4% (1%, 7%)
Table B-ll. Percent Change in Human PFOS Serum Concentration by Exclusive, Mixed or
No Breastfeeding Per Month as Reported by Mogensen et al. (2015, 3859839)
Breastfeeding
Status
Exclusive
Partial
None
Mixed Model up to 18 Months
Percent Change
29.2 (25.3, 33.1)
4.4(1.0, 7.8)
0.7 (-0.5, 1.9)
p-value
<0.0001
0.0108
0.2693
Mixed model up to 60 Months
Percent Change
30.2 (26.2, 34.3)
1 (-1.2, 3.2)
-0.9 (-1.2, -0.6)
p-value
<0.0001
0.3762
<0.0001
The contributions of placental transfer, breastfeeding, and ingestion of PFAA-contaminated
drinking water to early life PFOS levels in children were analyzed {Gyllenhammar, 2019,
5919402}. This study measured PFOS concentrations in children aged 4, 8, and 12 years (n = 57,
55, and 119, respectively) between 2008 and 2015 as part of the Persistent Organic Pollutants in
Uppsala Primiparas (POPUP) study in Sweden. Mixed linear regression (MLR) models were
used to ascertain associations with PFOS for these exposure sources. PFOS concentrations
increased 1.3% per unit (ng/g serum) of increase in the maternal serum level at delivery. PFOS
significantly increased 3.8% per month of nursing. Maternal serum and nursing duration showed
the strongest correlations in 4-year old children. PFOS increased 0.93% per month of cumulative
B-33
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
drinking water exposure. The authors suggested that, in addition to exposure in utero and
through lactation, drinking water with low-to-moderate PFOS contamination is an important
source of exposure for children.
B.2.3.2 Animal Studies
B. 2.3.2.1 Rots
To determine the dose-response curve for neonatal mortality in rat pups born to PFOS-exposed
dams and to investigate associated biochemical and pharmacokinetic parameters, 5 groups of 16
female Sprague-Dawley Crl:CD(SD)IGS VAF/Plus rats were administered 0, 0.1, 0.4, 1.6, or
3.2 mg PFOS/kg bw/day by oral gavage beginning 42 days prior to cohabitation and continuing
through gestation day (GD) 14 or GD 20 {Luebker, 2005, 1276160}. PFOS levels were analyzed
in serum, liver, urine, and feces samples in dams and fetuses as indicated in Table B-12. The
urine, feces, and liver of the control animals all contained PFOS at small concentrations. In
treated rats, the highest concentration of PFOS was in the liver. Serum levels in the dams for
each dose were consistent between GD 1 and GD 15, indicating achievement of steady state prior
to conception. The GD 21 levels in the dams had dropped below those observed earlier in the
pregnancy. Serum levels in the GD 21 fetuses were higher than those in the dams. In contrast,
PFOS levels in the livers of dams on GD 21 were about three times higher than in the fetuses.
Fecal excretion was greater than urinary excretion by the dams.
Table B-12. Liver, Serum, Urine, and Feces PFOS Concentrations in Pregnant Sprague-
Dawley Dams and Fetuses {Luebker, 2005,1276160}
Parameter
Dose
GD 1
GD 7
GD 15
GD 21
(mg/kg/day)
Dams
Dams
Dams
Dams
Fetuses
Serum3
0.1
8.90 ± 1.10
7.83 ± 1.11
8.81 ± 1.47
4.52 ± 1.15
9.08
0.4
40.7 ±4.46
40.9 ±5.89
41.4 ±4.80
26.2 ± 16.1
34.3
1.6
160 ± 12.5
154 ± 14.0
156 ±25.9
136 ±86.5
101
3.2
318 ± 21.1
306 ±32.1
275 ± 26.7
155 ±39.3
164
Liverb
0.1
0.4
1.6
3.2
-
-
-
29.2 ± 10.5
107 ± 22.7
388±167
610±142
7.92
30.6
86.5
230
Urine3
0.1
0.05 ±0.02
0.06 ±0.03
0.07 ±0.04
0.06 ±0.01
-
0.4
0.28 ±0.19
0.31 ±0.20
0.53 ±0.23
0.55 ±0.16
-
1.6
0.96 ±0.39
1.10 ±0.57
0.36 ±0.35
2.71 ±2.07
-
3.2
1.53 ±0.87
1.60 ±0.97
0.52 ±0.28
1.61 ±0.53
-
Feces'3
0.1
0.50 ±0.14
0.49 ±0.11
0.66 ±0.10
0.42 ±0.10
-
0.4
2.42 ±0.49
2.16 ±0.43
2.93 ±0.62
2.39 ± 1.21
-
1.6
10.3 ±3.01
9.20 ±2.68
11.1 ± 3.28
9.94 ±4.51
-
3.2
23.9 ±4.16
33.0 ± 10.0
29.5 ±8.92
20.1 ±4.21
—
Notes: GD = gestation day.
3 Data presented in mean ± standard deviation (|ig/mL)
bData presented in mean ± standard deviation (|ig/g)
B-34
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
This same study also included a subset of dams allowed to litter naturally and dosed through
lactation day (LD) 4. Liver and serum samples were collected from dams and pups on LD 5. In
this sampling, serum PFOS levels were similar between the dam and offspring, but the liver
values were now higher in the neonates than in the respective dams.
Twenty-five female Sprague-Dawley rats/group were administered 0 mg/kg/day, 0.1 mg/kg/day,
0.3 mg/kg/day, or 1.0 mg/kg/day potassium PFOS by gavage from GD 0 through PND 20. An
additional 10 mated females served as satellite rats to each of the four groups and were used to
collect additional blood and tissue samples. Further details from this study are provided in the
main document (See PFOS Main Document) as reported in Butenhoff et al. (2009, 757873).
Samples were taken from the dams, fetuses, and pups for serum and tissue PFOS concentrations
and the results were reported by Chang et al. (2009, 757876) (Table B-13).
B-35
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-13. Serum, Liver, and Brain Tissue PFOS Concentrations of Sprague-Dawley Dams and Offspring as Reported by
Chang et al. (2009, 757876)
Time
Dose
Serum PFOSa
Liver PFOSb
Brain PFOSb
(mg/kg)
Dam
Offspring
Dam
Offspring
Dam
Offspring
GD 20°
Control
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
On GD 20, PFOS concentrations in maternal serum, liver, and brain correlated with the daily
doses administered. Maternal liver-to-serum PFOS ratios ranged from 1.8 to 4.9, while the
maternal brain-to-serum ratios were 0.04 to 0.09 {Chang, 2009, 757876}. The concentrations in
the brains of fetuses was about 10 times higher than in their dams for all doses. Based on the
maternal and offspring data on GD 20, there is placental transfer of PFOS from rat dams to
developing fetuses. Serum values were approximately 1-2 times greater in the fetuses than in the
dams at GD 20. The concentration of PFOS in fetal liver was less than that of dams, and the
brain values were much higher; this is possibly due to the lack of development of the blood-brain
barrier at this stage of offspring development. PFOS serum concentrations in the offspring were
lower than those for the dams on postnatal day (PND) 4 and continued to drop through PND 72.
However, based on the concentrations still present in the neonate serum, lactational transfer of
PFOS was occurring. At PND 72, the males appeared to be eliminating PFOS more quickly as
the serum values were lower than those in the females; this difference was not observed at earlier
timepoints. In the liver, PFOS was the greatest in the offspring at PND 4 and decreased
significantly by PND 72. Liver values were similar at all timepoints between males and females.
On GD 20, the brain levels for the pups were tenfold higher than those for the dam. The levels in
pup brains gradually declined between PND 4 and PND 21.
Ishida et al. (2017, 3981472) also examined distribution to livers and brains in Wistar rat dams
and pups on PND 4. Tissue-to-plasma partition coefficients (Kps) for brain/plasma decreased
with increasing dose in dams (0.92 in dams at 1 mg/kg and 0.87 in dams at 2 mg/kg). In pups, the
brain/plasma Kp values were 0.447 and 0.408 at 1 mg/kg and 2 mg/kg, respectively.
Liver/plasma Kp values were 4.13 and 3.85 in dams and 3.30 and 2.07 in pups at the lower and
higher doses, respectively. Thus, the brain-plasma ratio of PFOS in pups is approximately 5
times higher than that in dams despite very similar liver/plasma ratios in pups and dams,
indicating an age-dependent accumulation of PFOS in the CNS.
In a study by Zeng et al. (2011, 1326732), 10 pregnant Sprague-Dawley rats/group were
administered 0 mg/kg/day, 0.1 mg/kg/day, 0.6 mg/kg/day, or 2.0 mg/kg/day of PFOS by oral
gavage in 0.5% Tween 80 from GD 2 to GD 21. On GD 21, dams were monitored for parturition,
and the day of delivery was designated PND 0. On PND 0, five pups/litter were sacrificed, and
the trunk blood, cortex, and hippocampus were collected for examination. The other pups were
randomly redistributed to dams within the dosage groups and allowed to nurse until PND 21,
when they were sacrificed with the same tissues collected as previously described. PFOS
concentrations in the hippocampus, cortex, and serum increased in a dose-dependent manner but
overall was lower in all tissues on PND 21 compared to PND 0 (Table B-14).
Table B-14. Serum, Hippocampus, and Cortex PFOS Concentrations of Sprague-Dawley
Rat Pups as Reported by Zeng et al. (2011,1326732)
Time
Dose (mg/kg/day)
Serum3
Hippocampusb
Cortexb
PND0
Control
ND
ND
ND
0.1
1.50 ±0.43*
0.63 ±0.19*
0.39 ±0.09*
0.6
24.60 ±3.02"
7.43 ± 1.62*
5.23 ± 1.58"
2.0
45.69 ±4.77"
17.44 ±4.12*
13.43 ±3.89**
PND21
Control
ND
ND
ND
0.1
0.37 ± 1.12*
0.25 ±0.14*
0.06 ±0.04*
B-37
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Time Dose (mg/kg/day)
Serum3
Hippocampus'"
Cortexb
0.6
1.86 ±0.35"
1.59 ±0.78"
1.03 ±0.59**
2.0
4.26 ± 1.73*"
6.09 ± 1.30"*
3.69 ±0.95***
Notes: ND = not detected; PND = postnatal day.
p < 0.05 compared with control in the same day.
p < 0.05 compared with 0.1 mg/kg group in the same day.
p < 0.05 compared with 0.6 mg/kg group in the same day.
a Data presented as mean ± standard deviation ((ig/mL).
bData presented as mean ± standard deviation (f-ig/g).
Sprague-Dawley rats were administered PFOS in 0.05% Tween (in deionized water) once daily
by gavage from GD 1 to GD 21 at 0 mg/kg/day, 0.1 mg/kg/day, or 2.0 mg/kg/day. There was a
postnatal decline in the serum and brain PFOS levels between PND 0 and PND 21. PFOS
concentrations were higher in the serum when compared to the lung in offspring on both PND 0
and PND 21 {Chen, 2012, 1276152} (Table B-15).
Table B-15. Serum and Lung PFOS Concentration of Sprague-Dawley Rat Pups {Chen,
2012,1276152}
Age
Dose (mg/kg/day)
Serum3
Lungb
PND 0
0.0
ND
ND
0.1
1.7 ±0.35*
0.92 ±0.04*
2.0
47.52 ±3.72*
22.4 ± 1.03*
PND 21
0.0
ND
ND
0.1
0.41 ± 0.11*
0.21 ±0.04*
2.0
4.46 ± 1.82"
3.16±0.11**
Notes: ND = not detected; PND = postnatal day.
*p < 0.05 compared with control.
** p < 0.01 compared with control.
a Data presented as mean ± standard deviation ((ig/mL).
bData presented as mean ± standard deviation (f-ig/g).
B.2.3.2.2 Mice
Borg et al. (2010, 2919287) administered a single dose of 12.5 mg/kg 35S-PFOS by intravenous
injection (n = 1) or gavage (n = 5) on GD 16 to C57B1/6 dams. Using whole-body
autoradiography and liquid scintillation, counting distribution of PFOS was determined for the
dams/fetuses (GD 18 and GD 20) and neonates (PND 1). Distribution of PFOS in the dams was
similar regardless of the route of exposure, with the highest levels in the liver and lungs at all
timepoints (liver and lung PFOS levels approximately 4 times and 2 times that of blood,
respectively). The distribution of PFOS in the kidneys was similar to blood and the amount in the
brain was lower than that of the blood. In the fetuses, the highest concentrations of PFOS were
found in the kidneys and liver. In the kidneys, the highest concentration of PFOS was observed
in the fetuses on GD 18 (3 times higher than maternal levels). In the fetuses on GD 18, values in
the lungs were similar to the maternal lungs, and this value increased by GD 20.
Accumulation in fetal liver was also observed C57BL/6 mice {Lai, 2017, 3981375}. In the
offspring at all timepoints, PFOS was homogeneously distributed in the liver at a level 2.5 times
higher than maternal blood and 1.7 times lower than maternal liver. In pups on PND 1, PFOS
B-38
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
was mostly concentrated in the lungs and liver. Pups on PND 1 had PFOS levels that were
3 times higher in the lungs compared to maternal blood with a heterogeneous distribution. In the
kidneys, the levels in pups on PND 1 were similar to their respective dams despite being higher
in fetuses on GD 18. Levels in the brain were similar at all timepoints in the offspring and higher
than in the maternal brain, likely due to an immature brain-blood barrier. Select data are
provided in Table B-16.
Table B-16. Concentration Ratios of 35S-PFOS Maternal Serum to Various Organs of
C57BL/6 Mouse Dams, Fetuses, and Pups {Lai, 2017, 3981375}
[35S-PFOS]orga„/[35S-PFOS]mater„al blood
Group
Liver*
Lungs"
Kidneys"
Brain3
Bloodb
(n = 6-8)
(n = 5-6)
(n = 3-6)
(n = 6-9)
(n = 1-6)
Dams
4.2" ± 0.7
2.0* ±0.4
0.9 ±0.1
0.2" ±0.05
1.0
Fetuses on GD 18
2.6" ± 0.8
2.1* ± 0.6
2.8**± 0.3
1.2 ±0.3
2.3
Fetuses on GD 20
2.4" ± 0.5
2.5** ±0.4
1.4 ±0.2
0.9 ±0.1
1.1 ±0.04
Pups on PND 1
2.4* ±0.4
3.0** ±0.5
1.0 ±0.5
0.9 ±0.2
1.7" ± 0.3
Notes: S-PFOS = S-radioisotope perfluorooctance sulfonic acid; GD = gestation day; PND = postnatal day.
'Statistically-significant (p <0.01) in comparison to maternal blood.
"Statistically-significant (p < 0.001) in comparison to maternal blood.
aData presented as mean ± standard deviation (|ig/g).
bData presented as mean ± standard deviation ((ig/mL).
Male and female KM mice were administered PFOS by subcutaneous injection one time on PND
7, PND 14, PND 21, PND 28, or PND 35 at concentrations of 0 mg/kg or 50 mg/kg {Liu, 2009,
757877}. Animals were killed 24 hours after treatment and the PFOS concentration levels
obtained. The percent distribution found in the blood, brain, and liver are provided in Table
B-17. The distribution shows that, beyond PND 14, the levels in the liver are approximately 2-
4 times greater than those found on PND 7.
Table B-17. Percent Distribution of PFOS in Male and Female KM Mice After 50 mg/kg
Subcutaneous Injection {Liu, 2009, 757877}
D\TT*
Males
Females
Blood3
Brainb
Liverb
Blood3
Brainb
Liverb
7
11.78 ±2.88
5.04 ± 1.49
14.84 ±4.01
10.77 ± 1.16
4.17 ± 1.17
16.23 ±4.84
14
13.78 ± 1.52
1.61 ±0.80**
26.50 ±7.36
12.31 ±2.24
3.26 ±0.58
26.30 ±4.54
21
9.85 ±2.74
2.40 ± 0.60**
51.35 ± 11.06"
12.37 ±3.80
2.14 ±0.38"
51.48 ±3.44**
28
9.89 ±2.94
0.85 ±0.19**
63.39 ± 19.78"
12.16 ±2.32
2.10 ±0.73"
51.05 ± 10.59**
35
13.33 ±0.89
1.02 ±0.28**
73.68 ±6.86"
11.54 ± 1.28
0.90 ±0.23"
69.92 ± 18.52**
Notes: PFOS = perfluorooctance sulfonic acid; PND = postnatal day
"Statistically significant from PND 7 (p < 0.01).
aData presented as mean percentage ± standard deviation ((ig/mL).
bData presented as mean percentage ± standard deviation (|ig/g).
B-39
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.2.4 Volume of Distribution
B. 2.4.1 Human Studies
None of the available studies provide data for calibration of the volume of distribution (Vd) of
PFOS in humans. However, several researchers have attempted to characterize PFOS exposure
and intake in humans {Thompson, 2010, 2919278; Egeghy, 2011, 723765} through
pharmacokinetic modeling. In the models discussed below, Vd was defined as the total amount of
PFOS in the body divided by the blood or serum concentration.
Both research groups defined a Vd for humans using a simple, first-order, one-compartment
pharmacokinetic model {Thompson, 2010, 2919278; Egeghy, 2011, 723765}. The models
developed were designed to estimate intakes of PFOS by young children and adults {Egeghy,
2011, 723765} and the general population of urban areas on the east coast of Australia
{Thompson, 2010, 2919278}. In both models, the Vd was calibrated using human serum
concentration and exposure data from NHANES, and it was assumed that most PFOS intake was
from contaminated drinking water. Thus, the value for Vd was calibrated so that model prediction
of elevated blood levels of PFOS matched those seen in the study population.
Thompson et al. (2010, 2919278) adjusted the Vd for PFOS (230 mL/kg) based on the calibrated
PFOA data by 35% in accordance with the differences in PFOA and PFOS volumes of
distribution calculated by Andersen et al. (2006, 818501). The original Andersen et al. (2006,
818501) model was developed from oral data in monkeys and optimized a Vd of 220 mL/kg for
PFOS and 140 mL/kg for PFOA. Thus, the Vd in monkeys for PFOS was approximately 35%
greater than that for PFOA in the optimized models. Therefore, Thompson et al. (2010, 2919278)
used a Vd of 230 mL/kg for humans in their model.
Egeghy and Lorber (2011, 723765) used high and low bounding estimates of 3,000 mL/kg and
200 mL/kg for Vd since data in humans were not available. The two separate estimates of Vd
were used in a first-order, one-compartment model to estimate a range of intakes of PFOA. They
concluded that the Vd was likely closer to the lower value based on a comparison of predicted
modeled intake with estimates of intakes based on exposure pathway analyses. Use of the lower
value gave a modeled intake prediction similar to that obtained by a forward-modeled median
intake based on an exposure assessment. The authors concluded that the lower value of
200 mL/kg was appropriate for their analysis.
Both of the models described above used a Vd calibrated from actual human data on serum
measurements and intake estimates. A calibration parameter obtained from human studies, where
constant intake was assumed and blood levels were measured, is considered a more robust
estimate for Vd than that optimized within a model developed from animal data.
The application of Vd values used in several modelling studies are shown in Table B-18. A single
value of 239 mL/Kg has been uniformly applied for most PFOS studies. Gomis et al. (2017,
3981280) used a Vd of 235 mL/kg by averaging of Vd values estimated for both humans and
animals. Vd values may be influenced by differences in distribution between males and females,
between pregnant and non-pregnant females, and across serum, plasma, and whole blood.
B-40
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table B-18. Summary of PFOS Volume of Distribution Values Assigned in Human Studies
AUC or
Mean/Median
Notes and
Study
Population
Sex
Compartment
Vd
Concentration
Measured in
Compartment
(ng/mL)
Considerations; Was
Steady State
Achieved?
Zhang et
Adult
Males
Whole blood
230 mL/kg
Mean: 12.8; GM: 8.62
Steady state assumed.
al. (2015,
2851103)
and
females
Pregnant,
adult
Females
Whole blood
230 mL/kg
Mean: 14.7; GM: 13.4
Steady state not
assumed due to variable
PFAS levels during
pregnancy.
Worley et
al. (2017,
3859800)
>12 years
Males
and
Females
Blood (2016)
230 mL/kg
bodyweight
Mean: 23.4 (18.5,
28.4)
>12 years
Males
and
Females
Blood (2010)
230 mL/kg
bodyweight
Mean: 39.8 (30.9,
48.9)
Fu et al.
Adult,
Males
Serum
230 mL/kg
Mean: 5624; median:
-
(2016,
3859819)
occupational
and
females
1725
Zhang et
Adults
Males
Serum and
230 mL/kg
Mean: 31
-
al. (2013,
and
whole blood
3859849)
Females
Gomis et
Humans and
Males
Serum
235 mLe/kg Reports an average of
Authors note that due to
al. (2017,
Animals
and
human and animal Vd
declining values in U.S.
3981280)
Females
values
and Australian
populations, steady state
was not achieved.
Notes: AUC = area under the curve; GM = geometric mean; Vd = volume of distribution.
B.2.4.2 Animal Studies
The Chang et al. (2012, 1289832) series of pharmacokinetic studies on rats, mice, and monkeys
described above, included Vd calculations. Values for all species were calculated following a
single oral or IV dose of PFOS. Based on these studies, the authors concluded that the VdS for
monkeys, rats, and mice are likely in the range of 200 mL/kg-300 mL/kg.
Two recent studies in rats {Kim, 2016, 3749289; Huang, 2019, 5387170} measured
toxicokinetic parameters including Vd (Table B-19). In the Kim et al. (2016, 3749289) study, Vd
values were calculated as Dose x AUMC/(AUC0-co)2, where AUMC is the area under the first
moment curve. Rats were dosed with 2 mg/kg PFOS by both oral and IV routes. Vd values were
higher after oral administration (382.55 ± 17.59 mL/kg in males and 351.50 ± 19.20 mL/kg in
females) compared with the IV administration (279.81 ± 16.71 mL/kg in males and
288.97 ± 15.59 mL/kg in females), but results between the sexes were similar. While organ-
specific Vd values were not determined, only the liver exhibited a partition coefficient (Pc)
greater than 1, and the liver Pc in males was significantly higher than the Pc in females
(2.63 ± 0.04 and 2.04 ± 0.03, respectively). This observation may contribute to the slightly lower
VdS observed after IV administration in males relative to females. Pcs in other tissues were 1
B-41
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(kidney, lung) or 2 (heart, spleen), lower that those observed in the liver for both males and
females.
Huang et al. (2019, 5387170) calculated the apparent volume of central (Vi) and peripheral (V2)
distribution in rats using standard equations {Gabrielsson, 2000, 9642135}. In this study, a two-
compartment model was the best fit for male rats for both IV and gavage routes of administration
and females dosed by the IV route whereas a one-compartment model was the best fit for female
rats dosed by oral gavage. As detailed in Table B-19, males and females were administered the
same dose (2 mg/kg) used by Kim et al. (2016, 3749289). In males, Vd values by the IV route
were 417 ± 31 mL/kg and 264 ±71 mL/kg in the central and peripheral compartments,
respectively. Interestingly, it was the Vd in the peripheral compartment that was most similar to
that observed by Kim et al. (2016, 3749289). Vd values in females after IV administration were
lower than that observed in males in both the central and peripheral compartments
(297 ± 43 mL/kg, and 124 ± 62 mL/kg, respectively). For the oral route, striking sex differences
were noted between the central and peripheral compartments. While Vd values were quite similar
in males (244 mL/kg-280 mL/kg) for both compartments, they were notably higher in the central
compartment (222 ± 84 mL/kg) compared to the peripheral compartment (93.4 ± 93 mL/kg) in
females.
In a third study {Iwabuchi, 2017, 3859701}, PFOS was administered to male Wistar rats as a
single bolus dose (BD) and Vd was measured as BD/elimination rate constant (ke) x plasma
concentration (AUC). Vd values were calculated for whole blood, serum, and several tissues. The
Vd of whole blood was much higher than that observed for serum (2.5 kg tissue volume/g bw and
0.96 kg tissue volume/kg bw, respectively). Organ Vd values were highest in the brain (7.9 kg
tissue volume/kg bw), heart (4.5 kg tissue volume/kg bw) and spleen (2.8 kg tissue
volume/kg bw). VdS were lower by 1 (kidney) or 2 (liver) orders of magnitude. Interestingly, for
this analysis of PFOS, the body organs behaved as an assortment of independent one-
compartments, with a longer elimination half-life in liver than serum in the elimination phase.
B-42
-------
DRAFT FOR PUBLIC COMMENT
Table B-19. Summary of PFOS Volume of Distribution in Rats
MARCH 2023
, Method of Vd
Study , , _L.
Calculation
Route
Dose
Species
Age
Sex Vd
Compartmen
t
AUC or mean/median
concentration measured
in compartment
Cmax
Steady state
considerations
Kim et Dose x
IV
2 mg/kg
Sprague-
8-12
Males 382.55 ± 17.59 m Blood Plasma
AUC: 216.47 ±8.63 \ig
5.23 ±0.24 (ig/mL
NR
al. (2016, AUMC/(AUC0-
Dawley
weeks
1/kg
day/mL
3749289) <»)2
Females 351.50 ± 19.20 m Blood Plasma
AUC: 203.60 ± 8.42 \ig
5.69 ±0.33 (ig/mL
NR
1/kg
day/mL
Oral
2 mg/kg
Sprague-
8-12
Males 279.81 ± 16.71 m Blood plasma
AUC: 272.69 ±20.39 \ig
6.71 ±0.30 ng/mL
NR
Dawley
weeks
1/kg
day/mL
Females 288.97 ± 15.59 m Blood Plasma
AUC: 234.61 ± 10.05 \ig
6.66 ± 0.29 |ig/mL
NR
1/kg
day/mL
Huang et Standard
IV
2 mg/kg
Sprague-
8 weeks
Males 417 ± 31 ml/kg
Central
AUC: 7.32 ±0.42 |iM-hr
0.01 ±0.01 mM
NR
al. (2019, equations
Dawley
264 ± 71 ml/kg
Peripheral
AUC: 7.32 ±0.42 nM-hr
0.01 ±0.01 mM
NR
5387170) {Gabrielsson,
2000,9642135}
Females 297 ± 43 ml/kg
Central
AUC: 10.72 ±0.78 nM-hr
0.01 ±0.01 mM
NR
124 ± 62 ml/kg
Peripheral
AUC: 10.72 ±0.78 nM-hr
0.01 ±0.01 mM
NR
Oral
2 mg/kg
Sprague-
8 weeks
Males 280 ± 48 ml/kg
Central
AUC: 9.86 ±0.74 nM-hr
0.01 ±0.01 mM
NR
Dawley
244 ±81 ml/kg
Peripheral
AUC: 9.86 ±0.74 nM-hr
0.01 ±0.01 mM
NR
Females 222 ± 84 ml/kg
Central
AUC: 17.74 ± 1.02 nM-hr
0.02 ±0.01 mM
NR
93.4 ± 93 ml/kg
Peripheral
AUC: 17.74 ± 1.02 nM-hr
0.02 ±0.01 mM
NR
2 mg/kg
Sprague-
8 weeks
Males 176 ± 27 ml/kg
Central
AUC: 58.18 ±3.00 nM-hr
0.11 ±0.01 mM
NR
(x5d)
Dawley
123 ± 42 ml/kg
Peripheral
AUC: 58.18 ±3.00 nM-hr
0.11 ±0.01 mM
NR
Females 136 ± 25 ml/kg
Central
AUC: 89.18 ±5.00 nM-hr
0.14 ± 0.02 mM
NR
86.3 ±37.3 ml/kg
Peripheral
AUC: 89.18 ±5.00 nM-hr
0.14 ± 0.02 mM
NR
20 mg/k
g
Sprague-
Dawley
8 weeks
Males 34.6 ± 4.8 ml/kg
43.9 ±7.7 ml/kg
Central
Peripheral
AUC: 149.76 ± 10.60 |iM-
hr
AUC: 149.76 ± 10.60 |iM-
hr
AUC:
0.21± 0.03 (iM-hr
AUC:
0.21± 0.03 nM-hr
NR
NR
Females 27.9 ± 4.7 ml/kg
Central
AUC: 213.94 ± 16.00 |iM-
hr
AUC:
0.27 ± 0.03 nM-hr
NR
27.5 ± 6.5 ml/kg
Peripheral
AUC: 213.94 ± 16.00 |iM-
hr
AUC:
0.27 ± 0.03 nM-hr
NR
Iwabuchi Dose / eliminati Oral
100 ng/k Wistar
7-9
Males 7.9 kg tissue
Brain
180 fj.g/kg tissue volume -
9.17 fj.g/kg tissue
NR
et al. on rate constant
g
weeks at
volume/kg BW
day
volume
(2017, (ke) x plasma
start of
4.5 kg tissue
Heart
380 ng/kg tissue volume -
27.7 ng/kg tissue
NR
3859701) concentration
exposure
volume/kg BW
day
volume
(AUC).
0.043 kg tissue
volume/kg BW
Liver
240000 ng/kg tissue
volume - day
2730 ng/kg tissue
volume
NR
B-43
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Method of Vd j^oute Dose Species Age
J Calculation 1
Sex Vd
Compartmen
t
AUC or mean/median
concentration measured
in compartment
Cmax
Steady state
considerations
2.8 kg tissue
Spleen
650 ng/kg tissue volume -
46.9 ng/kg tissue
NR
volume/kg BW
day
volume
0.85 kg tissue
Kidney
2300 ng/kg tissue volume -
197 ng/kg tissue
NR
volume/kg BW
day
volume
2.5 kg tissue
Whole blood
1800 ng/kg tissue volume -
52.6 ng/kg tissue
NR
volume/kg BW
day
volume
0.96 kg tissue
Serum
2200 ng/kg tissue volume -
127 ng/kg tissue
NR
volume/kg BW
day
volume
Notes: AUMC = area under the first moment curve; AUC = area under the curve; BW = body weight; Cmax = Maximum concentration achieved; IV = intravenous; NR = not
reported Vd = volume of distribution.
B-44
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Unlike the sex differences observed in rats, Vd calculations were similar in male and female
monkeys as shown in Error! Not a valid bookmark self-reference. {Chang, 2017, 3981378}.
Young adult cynomolgus monkeys (Macaca fascicularis) (6 per sex) were sham-dosed with
vehicle, a single dose of PFOS (9 mg/kg, low dose group), or 3 separate PFOS doses (11 mg/kg-
17.2 mg/kg, high dose group). Blood samples were drawn from all monkeys prior to, during, and
after PFOS administration for up to 1 year. Toxicokinetic parameters were determined using a
noncompartmental analysis. At the lower dose, a Vd of 127 mL/kg was calculated for both males
and females. At the higher dose, the Vd in males was calculated to be 135 mL/kg. Vd was slightly
higher in females (141 mL/kg).
Table B-20. Pharmacokinetic Parameters After Acute PFOS Exposure in Cynomolgus
Monkeys" {Chang, 2017, 3981378}
9 mg/kg 14 mg/kg
Parameter
Male Female Male Female
T1/2 (day)
124 ±3.89
102 ±29.2
117 ± 17.2
102 ±45.6
Kei (l/day)
0.00559 ±0.000175
0.00729 ± 0.00223
0.00605 ±0.000951
0.00757 ± 0.00270
CI (mL/day/kg)
0.712 ±0.0812
0.897 ±0.196
0.816 ±0.111
1.06 ±0.510
Vd (mL/kg)
127 ± 10.9
127 ± 18.9
135 ±6.69
141 ±38.5
AUC/dose
(ng/day/mL/mL/kg)
271,333 ±21,733
265,200 ± 15,057
249,667 ± 14,468
220,333 ±9,019
Notes: AUC/dose = area under the curve per dose; CI = clearance; Kei = elimination rate per day; T1/2 = half-life (time);
Vd=volume of distribution;.
a Data presented in mean ± standard deviation.
B3 Metabolism
The literature contains no studies on the metabolism of PFOS. It appears that PFOS is not further
metabolized once absorbed. Several studies investigating PFOA found no evidence of
metabolism {U.S. EPA, 2016, 3603279}, and it is likely that PFOS is similarly resistant to
metabolism in humans, primates, and rodents.
B.4 Excretion
B.4.1 Urinary and Fecal Excretion
B.4.1.1 Human Studies
Three major studies highlight the urinary excretion of PFOS in humans. Zhang et al. (2015,
2851103) derived estimates for PFOS's urinary excretion rate using paired urine and blood
samples from 54 adults (29 male, 25 female, ages 22-62) in the general population and 27
pregnant females (ages 21-39) in Tainjin, China. Urinary excretion was calculated by
multiplying PFOS concentration in first-draw morning urine samples by the predicted urinary
volume (1.6 L/day for males and 1.2 L/day for females). PFOS was detected in the blood
samples for all participants but only for 48% of the urine samples from the general population
(mostly males) and 11% of samples from the pregnant females. Total daily PFOS intake was
modeled for the general population with a geometric mean of 89.2 ng/day, resulting in an
estimated daily urinary excretion rate of 16% of the estimated total daily intake for PFOS. There
B-45
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
was no significant difference in excretion rate between males and females, but a significantly
(p = 0.015) higher rate among the younger adults. Nonpregnant females aged 21-50 had a higher
urine:blood ratio than those age 51-61 (0.0018 and 0.0006, respectively). A lower urine:blood
ratio was found in pregnant females compared to nonpregnant females (0.0004 and 0.0013,
respectively), suggesting the placenta and cord blood as possible elimination pathways.
Zhang et al. (2013, 3859849) measured renal clearance of PFOS in 86 paired blood and morning
urine samples from healthy volunteers in Hebei province, China. The calculated median renal
clearance rates of 0.044 mL/kg/day in young women and 0.024 mL/kg/day in men and older
women for total PFOS. The authors also observed that major branched PFOS isomers were more
efficiently excreted than the corresponding linear isomer.
In a later study, Fu et al. (2016, 3859819) determined renal clearance of PFOS, PFOA and
PFHxS in 302 occupational workers (213 male, 89 female) from one of the largest producers of
PFOS-related compounds in China. Paired serum and urine samples were collected. Mean and
median urine concentrations for PFOS among all workers were 4.4 ng/mL and 1.2 ng/mL,
respectively; in serum, the mean and median concentrations PFOS were 5624 and 1725 ng/mL,
respectively. The correlation coefficient of PFOS concentrations in paired serum and urine
samples of 0.72 was found to be highly statistically significant (p < 0.01), suggesting that urine
concentrations could serve as effective bioindicators for PFOS exposure in occupational settings.
Daily renal clearance was calculated for each PFAA as follows:
Urine FAA Concentrations Daily x Daily urine excretion volume
Serum PFAA concentrations x Body weight
Urine excretion volumes were assigned as 1.4 L/day and 1.2 L/day for males and females,
respectively), and body weight as reported in questionnaires. The daily renal clearance was the
highest for PFOA (GM 0.067 mL/kg/day) and lowest for PFOS (GM 0.010 mL/kg/day). Sex did
impact PFOS daily renal clearance values, which were significantly lower in males compared to
females (p < 0.01).
Fu and colleagues noted their half-life estimates are the shortest values ever, suggesting that the
overall elimination potential of PFAAs might have been underestimated. The shorter half-life
values presented could suggest that pathways other than renal clearance play important roles in
elimination of PFAAs in humans. Another possibility is that the apparent half-lives of PFAAs
calculated through annual decline rates could be affected by the high ongoing levels of exposure.
B.4.1.2 Animal Studies
In a study by Chang et al. (2012, 1289832), three Sprague-Dawley rats/sex/timepoint were
administered 14C-PFOS as the potassium salt, one time by oral gavage at a dose of 4.2 mg/kg.
Urine and feces were collected after 24 and 48 hours. The amounts recovered in urine and feces
were approximately equivalent at each time point: 1.57% and 1.55%, respectively, at 24 hours
and 2.52% and 3.24%, respectively, at 48 hours.
Further investigation by Kim and colleagues measured the amounts of unchanged PFOS excreted
into the urine and the feces of male and female Sprague Dawley rats with a single dose of
2 mg/kg by oral or intravenous administration {Kim, 2016, 3749289}. After dosing, urine and
feces were measured weekly throughout the 70-day study period. The highest concentrations
B-46
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
were found in urine under all conditions. In males, the levels detected in urine (76.13 ± 16.83 |ig)
and feces (61.65 ± 7.29 |ig) were similar after oral administration. After intravenous dosing,
urine levels in males (103.04 ± 21.56 jug) were more than 2-fold higher than fecal levels
(43.73 ± 5.29 |ig). Females also excreted higher levels in urine compared to feces by both dosing
routes. After oral administration, urine and fecal levels were 95.42 ± 22.14 |ig and
53.29 ± 8.64 |ig, respectively. Similar values in urine (88.29 ± 14.91 |ig) and feces
(48.37 ± 4.98 |ig) were measured after intravenous dosing. The similar concentrations in urine
and feces translated to similar half-life estimates for PFOS (26.44 and 28.70 days in males and
23.50 and 24.80 days in females by the oral and intravenous routes).
Another study evaluated repeat dosing in ten male Sprague-Dawley rats (~9 weeks old)/group
which were administered 0 mg/kg/day, 5 mg/kg/day, or 20 mg/kg/day PFOS by gavage once
daily for 4 weeks {Cui, 2010, 2919335}. Urine and feces were collected for 24 hour intervals on
the day prior to treatment (day 0), and days 1, 3, 5, 7, 19, 14, 18, 21, 24, and 28. Both dose
groups exhibited increased excretion over time, with greater excretion rates in the urine. No
notable difference in excretion between the dose groups remained after accounting for decreased
food intake and mortality in the high dose group.
Another study {Gao, 2015, 2851191} compared concentrations in urine and feces of male and
female Wistar rats. A mixture of PFOA/PFNA/PFOS were administered to rats by drinking water
for 90 days, with each compound at doses of 0 mg/L, 0.05 mg/L, 0.5 mg/L, and 5 mg/L. While
the focus of this study was measuring concentrations in the hair of animals (discussed below
under Other Routes of Excretion), the authors measured concentrations of each PFAA in urine
and feces samples by collecting excreta in standard metabolism cages overnight for 24 h
intervals on day 84 (week 12). The intake for each compound was calculated as the drinking
volume multiplied by water concentration of 0.05 mg/L, 0.5 mg/L, and 5 mg/L. In contrast to
observations by others, there were far higher levels of PFOS in feces compared to urine for both
males and females. However, this trend was also observed among female Crl:CD(SD)IGS
VAF/Plus rats by Luebker et al. (2005, 1276160), in which five groups of 16 dams each were
administered 0 mg, 0.1 mg, 0.4 mg, 1.6 mg, or 3.2 mg PFOS/kg bw/day by oral gavage
beginning 42 days prior to cohabitation and continuing through GD 14 or GD 20. Urine and
feces were collected overnight from dams on the eve of cohabitation day 1 and during GDs 6-7,
GDs 14-15, and GDs 20-21. The concentrations in the feces were consistently about 5 times
greater than in the urine It is unclear whether the higher levels of PFOS in feces reflects rat strain
or dose differences among the various studies or is driven by differential excretion pathways in
rats exposed to a mixture of PFAAs.
In summary, limited evidence supports excretion through the fecal route in both animals and
humans. Most studies indicate excretion by the fecal route is substantially lower than that
observed by the urinary route. There are sex-specific differences in excretion of PFOS through
feces. Excretion through the fecal route appears to be more efficient in males compared to
females. Also, in male rats, fecal and urinary concentrations were similar after oral but not
intravenous dosing. Finally, exposures to mixtures of PFNAs suggests that PFOS in the context
of a mixture may be preferentially excreted through the fecal route. The extent to which
resorption by hepatic and enteric routes impacts fecal excretion has not been established in either
humans or animals.
B-47
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.4.2 Physiological and Mechanistic Factors Impacting
Excretion
B. 4.2.1 Renal Resorption
Urinary excretion is the major route of elimination for PFOS. Excretion through urine is
impacted by saturable renal resorption of PFOS from the glomerular filtrate via transporters in
the kidney tubules.
Urinary excretion of PFOS in humans is also impacted by the isomeric composition of the
mixture present in blood and the sex/age of the individuals. The half-lives of the branched chain
PFOS isomers are shorter than those for the linear molecule, an indication that renal resorption is
less likely with the branched chains.
Zhang et al. (2013, 3859849) determined half-lives for PFOS isomers based on paired serum
samples and early morning urine samples collected from healthy volunteers in two large Chinese
cities. Half-lives were determined using a one compartment model and an assumption of first
order clearance. The mean half-life values for the six branched chain isomers of PFOS were
lower than the value for the linear chain with the exception of the 1-methyl heptane sulfonate,
suggesting that resorption transporters may favor uptake of the linear chain and 1-methyl
branched chain over the other isomers.
B.4.2.2 Enterohepatic Resorption
Early evidence of enterohepatic resorption of PFOS was revealed by Johnson and colleagues
(1984, 5085553), who demonstrated that cholestyramine (CSM) treatment increased mean
cumulative carbon-14 elimination in feces by 9.5-fold for male CD rats administered 3.4 mg/kg
[i4C]pfos. CSM is a bile acid sequestrant, and its facilitation of PFOS GI clearance suggests
enterohepatic circulation.
Evidence of enterohepatic excretion and potential resorption in humans includes Harada et al.
(2007, 2919450), in which serum and bile samples from patients (2 male and 2 female; aged 63-
76) undergoing gallstone surgery exhibited higher PFOS levels in the bile than in the serum,
suggesting bile as a route of excretion. The biliary resorption rate was 0.97, which could
contribute to the long half-life in humans. Method of exposure to PFOS was unknown.
Biliary excretion in humans and the potential for resorption from bile discharged to the GI tract
is supported by the Genuis et al. (2010, 2583643) self-study of the potential for CSM to lower
the levels of PFAS in blood. This was a case report and sole example of excretion analyzed after
inhalation PFOS exposure. A 51-year old exposed through carpet treated with soil/dirt repellants
presented with elevated serum levels of perfluorinated compounds including PFOS. After
treatment with CSM for 1 week (ingested 4 g/day, three times a day), PFOS serum levels
decreased from 23 ng/g serum to 14.4 ng/g serum. Additionally, the stool concentration of PFOS
was increased from undetectable before treatment (LOD = 0.5 ng/g) to 9.06 ng/g and 7.94 ng/g
in the weeks after treatment, suggesting that it may help with removing PFOS that gains access
to the GI tract with bile.
Table B-21 summarizes enterohepatic transporters identified in liver hepatocytes and intestinal
enterocytes in humans and rats by Zhao and colleagues (2015, 3856550; 2017, 3856461} and
B-48
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
suggests that PFOS is a substrate of both sodium-dependent and -independent enterohepatic
transporters involved in recirculation of bile acids. For these in vitro studies, the authors used
transformed ovary (CHO) and kidney (HEK293) cells stably or transiently transfected with
cDNA constructs encoding for the transporters as well as CHO Flp-In cells expressing human
OATP2B.Wild-type CHO cells and HEK293 cells transfected with vector only were used as
controls. With the exception of rat ASBT, PFOS was demonstrated to be a substrate for all
transporters as well as OSTalpha/beta.
Binding efficiency to the enterohepatic transporters was chain-length dependent. Sodium-
taurocholate cotransporting polypeptide (NTCP) transported PFSAs with decreasing affinity but
increasing capacity as the chain length increased {Zhao, 2015, 3856550}. The opposite trend
was seen for OATP-mediated uptake {Zhao, 2017, 3856461}. For these 5 OATPs, PFOS was
transported with the highest affinity compared to transport of PFBS and PFHxS. The authors
suggest that transport efficiency generally increased with the increase in chain length, and that
this may, at least in part, account for the shorter half-lives of short chained vs. long chained
perfluoroalkyl sulfonates. While these in vitro studies demonstrate that PFOS is a substrate of
enterohepatic transporters found in the livers and intestines of humans and rats, it is as unknown
whether and to what extent these transporters function in vivo.
Table B-21. Enterohepatic Transporters of PFOS
Human Transporters
Rat Transporters
Organ
Cell type
Liver
Hepatocyte
Intestine
Enterocyte
Liver
Hepatocyte
Intestine
Enterocyte
Sodium-dependent
{Zhao, 2015, 3856550}
NTCP
ASBT
NTCP
Sodium-independent 0ATP1B13
{Zhao, 2017, 3856461} OATP1B33
OATP2B13
0ATP1A13
OATP1B2
OATP1A5
OATP2B1
OATP2B13
OATP2B1
Notes: ASBT = human apical sodium-dependent bile salt transporter; NTCP = Na+/taurocholate cotransporting polypeptide;
OATP = organic anion transporting polypeptide.
a Transporter examined in transfection studies; PFOS also shown to be a substrate of these transporters in HEK293 cells
transiently transfected with cDNA constructs encoding these transporters {Zhao, 2015, 3856550} {Zhao, 2017, 3856461}.
B.4.3 Maternal elimination through lactation and fetal
partitioning
PFOS can readily pass from mothers to their fetuses during gestation and through breast milk
during lactation. In conjunction with elimination through menstruation discussed in Section
B.4.4, females may eliminate PFOS through routes not available to males. The total daily
elimination of PFOS in pregnant females was estimated to be 30.1 ng/day, higher than the
11.4 ng/day for PFOA {Zhang, 2014, 2850251}. The ratio of branched:total PFOS isomers in
cord blood was 0.27 and was statistically greater in cord blood compared to maternal blood and
B-49
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
placenta. These finding suggests branched PFOS isomers may transfer to the fetus more readily
than linear forms.
The distribution of PFOS from maternal serum to the fetus and infants is discussed in detail
above (Section B.2.3). A study by Zhang et al. (2013, 3859792) exemplifies the routes and
amounts of PFOS eliminated by pregnant females. Paired maternal whole blood and cord blood
samples were analyzed from 32 females from Tianjin, China. The maternal blood concentration
of PFOS was 14.6 ng/mL. The mean levels in the cord blood, placenta, and amniotic fluid were
21%, 56%, and 0.1%, respectively, of those in the mother's blood. Although levels in amniotic
fluid correlated to maternal blood for PFOA, the correlation was poor for PFOS. Nevertheless, in
addition to cord blood, placenta and amniotic fluids are additional potentially substantial routes
of elimination in pregnant females. Blood loss during childbirth could be another source of
excretion.
The elimination of PFOS in pregnant women corresponds to an increase in concentrations in the
placenta. Mamsen et al. (2019, 5080595) observed an increase in PFOS accumulation from
gestational age 50 to 300 days, with male placentas showing higher levels of than female
placentas. The authors estimated a placenta PFOS accumulation rate of 0.13% increase per day
during gestation.
Mamsen and colleagues (2017, 3858487) measured placental samples and fetal organs in relation
to maternal plasma levels of 5 PFAS in 39 Danish women who underwent legal termination of
pregnancy before gestational week 12 {Mamsen, 2017, 3858487}. All PFAS were transferred
from mother to fetus with different efficiencies and a significant positive correlation was
observed for fetal age (exposure duration) and for fetal:maternal plasma ratios for all PFAS
compounds. Fetal organ levels of PFOS were lower than maternal blood. The average
concentration of PFOS was 0.6 ng/g in fetal organs compared to 1.3 ng/g in placenta and
8.2 ng/g in maternal plasma. Increasing fetal PFOS levels with fetal age suggest that the rate of
elimination of PFOS from mother to fetus may increase through the gestational period.
The same group {Mamsen, 2019, 5080595} measured PFOS accumulation in fetal tissues
across the 3 trimesters from 78 pregnant women who underwent elective pregnancy
terminations and from cases of intrauterine fetal death. Fetal tissues (placenta, liver, lung,
heart, central nervous system (CNS) and adipose) were collected for 38 first trimester
pregnancies, 18 second trimester pregnancies and 22 third trimester pregnancies. PFOS was
above LOQ in 100% of maternal serum samples, in 93% of placenta samples and 76% of
fetal organs. In general, the concentrations of PFOS in fetal tissue increased from first
trimester to third trimester except for liver and heart which showed highest levels in the
second trimester compared to the third trimester. Analysis of the placenta: serum ratio of
PFOA revealed a higher ratio in male fetuses than in female fetuses, but unlike PFOA, the
difference between the sexes did not reach statistical significance. These studies support the
placenta and fetus as important routes of PFOS elimination in pregnant women.
Underscoring the importance of pregnancy as a life-stage when excretion is altered, Zhang et al.
(2015, 2857764) observed that the partitioning ratio of PFOS concentrations between urine and
whole blood in pregnant women (0.0004) was significantly lower (p = 0.025) than the ratios
found in non-pregnant women (0.0013) and may be affected by the increase in blood volume
during pregnancy {Pritchard, 1965, 9641812}.
B-50
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
After birth, women can also eliminate PFOS via lactation. Tao and colleagues measured 45
human breast milk samples collected in 2004 from Massachusetts and PFOS (mean 131 ng/L)
and PFOA (mean 43.8 ng/L) were the predominant PFAS compounds measured {Tao, 2008,
1290895}. Elimination through breast was more recently measured in 293 samples collected
from 127 mothers in the Children's Health and Environmental Chemicals in Korea (CHECK)
Cohort {Lee, 2017, 3983576}. Results were stratified by age, parity, body mass, delivery
method, and infant sex. The median PFOS concentrations in breast milk across all samples was
47.4 ng/L (range of 36.4 ng/L-63.8 ng/L) and the median concentration for all PFAS chemicals
measured was 151 ng/L (range of 105 ng/L-212 ng/L). Pooled breast milk samples were
measured to follow the time course of PFOS in breast milk after birth. Concentrations in breast
milk measured 30 days after birth were significantly higher than those measured prior to 7 days
after birth. These findings are contrast with results of Thomsen et al. (2010, 759807) that
reported that breast milk levels of PFOA and PFOS decreased by 7% and 3.1%, respectively,
during the first month after birth. Demographic factors, maternal diets, sample sizes, the
lactational periods measured may account for these discrepancies.
Lee and colleagues also observed that parity impacts PFOA levels in breastmilk. While
primiparous mothers showed higher levels of PFOA in breast milk to mothers giving birth to
more than 1 child (p < 0.05), levels of PFOS were not significantly different between these two
groups. In contrast, another study of a Slovakian cohort, multivariable models estimated that
parous women had 40% lower PFOS (95% CI: —56%, —17%) concentrations in colostrum
compared with nulliparous women {Jusko, 2016, 3981718}. The geometric mean concentration
in was 35.3 ng/L for PFOS and 32.8 ng/L for PFOA. These findings are also consistent with
higher PFOS levels (p < 0.001) in second trimester maternal serum (18.1 ± 10.9 ng/mL) than
maternal serum levels at delivery (16.2 ± 10.4 ng/mL), which were higher than the levels found
in cord serum (7.3 ± 5.8 ng/mL; p <0.001) {Monroy, 2008, 2349575}. In this study, samples
were measured in 101 pregnant women at 24-28 weeks of pregnancy, at delivery, and in
umbilical cord blood.
PFOS was also measured in maternal serum, cord serum and breast milk from 102 female
volunteers hospitalized between June 2010 and January 2013 for planned caesarean delivery in
Tolouse, France {Cariou, 2015, 3859840}. Mean PFOS concentrations were 3.67, 1.38 and
0.040 in maternal serum, cord serum and breast milk respectively (compared to 1.22, 0.9191 and
0.041 ng/mL for PFOA). The observed ratios of cord and maternal serum for PFOS was 0.38 in
this study, much lower than the ratio of 0.78 for PFOA. However, the ratio between breast milk
and maternal serum was 0.038 ± 0.016 (essentially the same as measured for PFOA). Thus,
PFOS exhibits a low transfer from maternal blood to cord blood and a 10-fold lower transfer
from maternal blood to breast milk.
In summary, partitioning to the fetus and breast milk represent important routes of elimination in
humans, and may account for some of the differences observed for blood and urinary levels of
PFOS by sex and age.
BAA Other routes of elimination
Wong et al. (2014, 2851239) looked at the role of menstrual blood as an excretory pathway to
explain the shorter half-life of PFOS in females than males. They fit a population-based
pharmacokinetic model to six cross-sectional National Health and Nutrition Examination Survey
B-51
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(NHANES) data sets (1999-2012) for males and females. They concluded that menstruation
could account for about 30% of the elimination half-life difference between females and males.
Wong et al. (2014, 2851239) did not account for other possible loss pathways of PFOS that are
unique to women of reproductive age such as the amount of blood loss in child delivery,
amniotic fluid, breast feeding. Verner and Longnecker (2015, 2850226) suggested a need to
consider the non-blood portion of the menstrual fluid and its albumin content in the Wong et al.
(2014, 2851239) estimate for the menstrual fluid volume. A yearly estimate for serum loss of
868 mL/year by Verner and Longnecker (2015, 2850226) compared to the 432 mL/year estimate
of Wong et al. (2014, 2851239) suggests that the menstrual fluid loss can account for > 30% of
the difference in the elimination half-life between females and males.
Two earlier studies supported an association between increased serum concentrations of PFOA
and PFOS and early menopause {Knox, 2011, 1402395; Taylor, 2014, 2850915}. However, a re-
analysis of this data {Ruark, 2017, 3981395} suggested that this association could be explained
by reversed causality and more specifically, that pharmacokinetic bias could account for the
observed association with epidemiological data. Also challenging the assumption that this is due
to menstruation, Singer et al. (2018, 5079732) failed to find evidence of associations between
menstrual cycle length and PFAS concentrations.
Furthermore, Lorber et al. (2015, 2851157) compared individuals who had undergone blood
removal treatments for medical reasons to menstruating females. Measurements showed lower
PFOA and PFOS concentrations in the groups experiencing regular blood loss. Estimated
concentrations based on a one-compartment model were consistent with measured serum
concentrations. Overall, this study provides data and modeling that support the initial hypothesis
that ongoing blood loss explains lower PFAA concentrations in humans. These authors
suggested that factors other than blood loss, such as exposure to or disposition of PFOA/PFOS,
may also help explain the differences in elimination rates between males and females. Curiously,
studies providing direct measurements of PFOS in menstrual blood were not identified.
However, for PFOS to be selectively retained from the blood lost through menstruation would
require a specific mechanism for that process and no such mechanism has been demonstrated or
proposed.
Hair has been demonstrated as a route of elimination in animals {Gao, 2015, 2851191}. Adult
male and female Wistar rats were exposed via drinking water to 0 mg/L, 0.05 mg/L, 0.5 mg/L,
and 5 mg/L of PFOS, PFNA, and PFOA for 90 days. At the end of the exposure period, dorsal
hair samples were collected, washed twice in Triton buffer to remove external contaminants and
alkaline digested to extract PFAS. PFOS was detected in hair samples of all the treatment
groups, suggesting a potential route of elimination. Hair from male and female rats contained
PFOS concentration ranged from 20.3 ng/g to 2086 ng/g in 0.05 mg/L and 5 mg/L treatment
groups, respectively. Notably, the PFOS concentration in hair was significantly higher than the
levels of PFOA (3.31 ng/g-444 ng/g) and PFNA (14.2 ng/g-1,604 ng/g) at 0.05 mg/L to 5 mg/L
doses. Unlike PFOA and PFNA which showed a sexually dimorphic pattern, where male rats
have significantly higher hair concentrations than female rats, hair PFOS levels were lower in
males of the 0.05 mg/L group than females of the same dose group and there were no significant
differences in hair PFOS concentrations between males and females of the 0.5 mg/L and 5 mg/L
dose.
B-52
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Gao et al. (2015, 2851191) also measured the composition of the mixture excreted in in urine,
feces, and hair after administration of 0.5 mg/L or 0.05 mg/L of a mixture of PFAS (Table B-22).
At the lower dose of 0.05 mg/L, PFOS was the dominant constituent in urine of males and made
up a smaller proportion of total mixture excreted in hair but not feces. In females however,
PFOA was the predominant constituent excreted in urine, but made up the minority constituent
excreted in feces and especially in hair. These findings underscore the impact of mixtures and
sex on PFOA excretion.
Table B-22. Estimated Percentage of the Sum of PFOS, PFNA, and PFOA in Excreta and
Serum of Male and Female Wistar Rats" as Reported by Gao et al. (2015, 2851191)
Sex
PFAA
Serum
Urine
Feces
Hair
Males
PFOS
24.6
89.0
20.8
30.0
PFNA
59.9
11.0
53.0
45.4
PFOA
15.6
ND
26.1
24.6
Females
PFOS
89.0
ND
62.4
78.0
PFNA
11.0
38.9
21.7
18.0
PFOA
ND
61.1
16.1
4.2
Notes: ND = not detected; PFNA = perfluorononanoic acid.
aData is presented in % total perfluoroalkyl acids administered. Animals exposed to 0.05 mg/L (in Gao, 2015,2851191)
A single case report study {Genuis, 2010, 2583643} examined PFOS excretion through sweat.
PFOS was measured in sweat as well as urine and stool from a single male subject exposed to
perfluorinated chemicals via inhalation exposure and subjected to treatment with bile
sequestrants. With the exception of PFHxS, no other PFAS chemicals, including PFOS, were
detected in sweat.
Thus far, no single study has conducted a comparative analysis of elimination of PFOS through
all possible routes of excretion. A comprehensive analysis stratified by age and sex would be
necessary to advance the understanding PFOS excretion by all possible routes, and to establish
factors that influence the proportion of PFOS excreted through urine vs. other excreta matrices.
B.4.5 Half life Do to
B.4.5.1 Overview
We recognize that in general a half-life represents elimination by all routes, which includes
metabolism for other chemicals, but because PFOA/PFOS are not metabolized, it can be
interpreted for excretion (after correction for body weight (BW) changes). The calculated values
of PFOS half-lives reported in the literature vary considerably, which poses challenges in
predicting both the routes and rates of excretion. Several interrelated physiological and
mechanistic factors impacting excretion are summarized here:
1. The capacity of PFOS to be reabsorbed via renal and enterohepatic routes of excretion
and binding affinities to relevant transporters including organic anion transporters
(OATs), organic anion transporting polypeptides (OATPs), MRPs, and sodium-
dependent transporters involved in bile acid transport including NTCP and the apical
sodium-dependent bile acid transporter (ASBT). Exposures to high levels of PFOS under
B-53
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
acute conditions (e.g., contaminated drinking water) or in occupational settings may
result in saturation of resorption transporters and increased excretion.
2. Binding affinity to serum proteins limiting the concentration of the unbound fraction
available for resorption through renal or enterohepatic transporters. Moreover, binding to
serum proteins may limit passive diffusion of perfluorinated chemicals across the
placental barrier.
3. Phospholipid lipid binding affinity (phospholipidphilicity), which can further reduce the
unbound fraction of PFOS as well as uptake into cells. As reported by Sanchez Garcia
and colleagues, phospholipophilicity shows the highest correlation to cellular
accumulation data compared to other measures of lipophilicity. Also, phospholipid
binding affinity could distinguish between high and low accumulating compounds as well
as half-life measures {Sanchez Garcia, 2018, 4234856}.
4. Chain length and branching. The half-lives of the branched-chain PFOS isomers are
shorter than those for the linear molecule, an indication that renal resorption is less likely
with the branched chains. Interactions with transporters also vary by chain length.
5. Exposure to mixtures of perfluorinated compounds with differential binding affinities to
transporters, serum binding proteins and phospholipids could impact both the rate and
route of PFOS excretion.
6. Sex and species can influence both the rate and route excretion. First, several elimination
pathways are specific to females including menstruation, pregnancy, and lactation.
Second, sex-specific hormones can impact expression of transporters involved in
resorption. Furthermore, elimination half-lives vary dramatically by species, with much
longer half-lives calculated in humans compared to animals.
Differences between species were observed in studies determining the elimination half-life (T1/2)
of PFOS in rats, mice, monkeys, and humans. Sex differences in rats do not appear to be as
dramatic for PFOS as they are for PFOA {Loccisano, 2012, 1289830; Loccisano, 2012,
1289833}.
B. 4.5.2 Human Studies
Blood sampling was performed on retirees from the 3M plant in Decatur, Alabama where PFOS
was produced. These samples were taken approximately every 6 months over a 5-year period to
predict the half-life of PFOS. Results ranged from approximately 4 years to 8.67 years {3M,
2000, 8568548; 3M, 2002, 6574114}. Both of these studies exhibited some deficiencies in
sample collection and methods.
More recently, Olsen et al. (2007, 1429952) obtained samples from 26 retired fluorochemical
production workers (24 males and 2 females) from the 3M plant in Decatur, Alabama to
determine the half-life of PFOS. Periodic serum samples (total of 7-8 samples per person) were
collected over a period of 5 years, stored at -80 °C, and at the end of the study, High-
performance liquid chromatography/mass spectrometry was used to analyze the samples. The
study took place from 1998 to 2004. The mean number of years worked at the plant was 31 years
(range: 20-36 years), the mean age of the participants at the initial blood sampling was 61 years
B-54
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(range: 55-75 years), and the average number of years retired was 2.6 years (range: 0.4-
11.5 years). The initial arithmetic mean serum concentration of PFOS was 0.799 |ag/m L (range:
0.145-3.490 |ig/mL), and when samples were taken at the end of the study the mean serum
concentration was 0.403 |ag/mL (range: 0.037 |ig/mL-l ,740 |ig/mL). Semi-log graphs of
concentration vs. time for each of the 26 individuals were created, and individual serum
elimination half- lives were determined using first-order elimination. The arithmetic and
geometric mean serum elimination half-lives of PFOS were 5.4 years (95% confidence interval
(CI): 3.9, 6.9 years) and 4.8 years (95% CI: 4.1, 5.4 years), respectively.
The rate of serum PFOS decline was measured in residents of two communities exposed to
contaminated municipal drinking water contaminated in Bleking County, Sweden in 2013 {Li,
2018, 4238434}. A biomonitoring program ensued between 2014 and 2016 for residents exposed
to contaminated water and an unexposed community. A subset of residents (age range of 15-
50 years) were included in a panel study to estimate PFOS half-lives. Drinking water PFOS
levels were 8000 ng/L prior to closure of the waterworks facility and 27 ng/L in the unexposed
community. The mean serum levels among the 106 participants 6 months after the end of
exposure was 387 ± 259 ng/mL. The average decrease in PFOA was 20% of its previous value
each year. The excretion rate constant after the end of exposure was 0.20 (95% CI: 0.19, 0.22)
and was significantly higher in females (0.22) than males (0.15). The mean half-life was
3.4 years and was also significantly shorter in females (3.1 years) than in males (4.6 years).
There was a high level of inter-individual variation in half-lives.
Fu et al. (2016, 3859819) determined the half-life of PFOS in 302 occupational workers (213
male and 89 female) from one of the largest producers of PFOS-related compounds in China.
The half-lives of PFAAs in workers were estimated by daily clearance rates and annual decline
rates of PFAAs in serum by a first-order model based on fasting blood and urine samples
collected over a period of five years. Mean and median serum concentrations for PFOS among
all workers were 5,624 ng/mL and 1,725 ng/mL, respectively, whereas in urine, mean and
median PFOS were 4.4 and 1.2 ng/mL. Fu et al. calculated that the renal clearance rate for PFOS
ranged from 5.0 x 10 5 mL/kg/day to 0.54 mL/kg/day (Geometric mean of 0.010 mL/kg/day).
Half-lives were calculated by Ln2/k using two approaches. In the first approach, k was defined
as Cltotai/Vd, where Vd stands for the volume of distribution of PFAAs in the human body and
Cltotai represents the total daily PFAAs clearance in the human body. Cltotai was defined as renal
clearance for men and women older than 50, and as the sum of menstrual and renal clearance in
young women. Vd was set to 230 mL/kg for PFOS. In the second approach, k was defined as the
average annual decline rates of PFAAs in workers who participated in this study.
The half-life of PFOS estimated using daily clearance rate of all workers had a geometric mean
and median value of 32.6 and 21.6 years, respectively. However, when measured by annual
decline rate, the half-life of PFOS was estimated to be 1.9 years. The GM values of the half-life
of PFOS for men here was 60.9 years and 8.0 years for women. The authors suggest that half-
lives estimated by the limited clearance route information could be considered as the upper limits
for PFAAs and that the unrealistically long half-lives determined using urine clearance values
may indicate that other clearance play important roles in elimination of PFAAs in humans
including fecal elimination. Another possibility is that the apparent half-lives of PFAAs
calculated through annual decline rates could be affected by the high ongoing levels of exposure.
B-55
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Calculated half-lives of PFOS were much longer than for PFOA. The authors postulate
differential accumulation kinetics of the pollutants and suggest that PFOS reaches a steady-state
much faster than PFHxS and PFOA in humans. The longer half-life estimates for PFOS
compared to PFOA may also reflect its stronger affinity for serum albumin as reported
previously {Salvalaglio, 2010, 2919252}. Other factors impacting half-lives could include higher
enterohepatic and renal reabsorption rates of PFOS relative to PFOA. The authors conclude that
the shorter half-lives of PFHxS and PFOS estimated by annual decline compared to those
estimated by daily clearance rates suggest that other important elimination pathways operate to
remove PFOS and might have been underestimated.
Worley and colleagues (2017, 3859800) calculated PFOS half-lives in community members (age
12-years old or older) living near a PFAS manufacturer in Alabama that had discharged waste
into a local wastewater treatment plant. Sewage sludge from this plant was applied to local
agricultural fields. In 2010, ATSDR collected blood samples from subjects and followed up with
blood and urine measurements in 2016. Biological half-lives were estimated for PFOA and
PFOS using a one-compartment pharmacokinetic model. Geometric mean serum PFOS
concentrations were significantly higher in subjects in both 2010 (39.8 (J,g/L) and 2016 (23.4
(j,g/L) relative to national averages reported by NHANES (9.32 [j,g/L in 2009-2010 and
4.99 ^g/L in 2013-2014).
Half-lives for PFOA and PFOS were estimated to be 3.9 and 3.3 years, respectively. When Vd
and intake values were altered by ± 20%, half-life values varied by several months (half-life
estimates for PFOA and PFOS ranged from 3.5-4.1, and 3.0-3.6 years). The authors suggest
these parameters have a significant impact on half-life estimates.
Xu et al. (2020, 6781357) estimated the half-life of PFAS by sampling urine (4 times) and blood
(5 times) from 26 airport employees between 2 weeks to 5 months after the end of a 2-month
exposure to PF AS-contaminated drinking water. The levels of PFOS in the airport's
contaminated water were 62 ng/L (0.062 ng/mL) for linear PFOS and 64 ng/L (0.064 ng/mL) for
branched PFOS. Specific gravity adjusted urine concentrations for PFOS were generally below
detectable limits for linear and branched forms of PFOS with respective ranges of
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1.04 years in Xu et al. (2020, 6781357) to 60.9 years in Fu et al. (2016, 3859819). These
comparisons support principles suggested by the broader literature. First, sex-related differences
with males exhibiting somewhat longer half-lives compared to (especially females of
reproductive age) may relate, at least in part, to menstruation as routes of elimination {Zhang,
2013, 3859849}. Second, blood and urine concentrations varied by several orders of magnitude
across these 4 studies. This variability in serum and urine concentrations may reflect the role of
non-urinary routes of PFOS excretion; the variability in concentrations may also reflect the
difficulty in measuring renal resorption. Finally, only two studies estimated PFOS intake in
subjects {Xu, 2020, 6781357; Worley, 2017, 3859800}. Altogether, there is insufficient data to
correlate PFOS intake measurements to serum/plasma and urine concentrations. These factors, as
well as age and health status of subjects, likely contribute to the variability in PFOS half-life
estimates in humans.
B-57
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-23. Summary of PFOS Concentration in Blood and Urine in Relation to Half-life values in Humans
Study
Number of
Subjects
Age Range
Primary
Exposure
Route
Intake
Plasma/Serum
Concentrations
Urinary
Concentrations
Estimated Half
Life (y)
Considerations
Xu et al.
26
22-
Oral
Drinking
Linear PFOS:
Linear PFOS:
Linear PFOS: 2.91,
• 1 woman was
(2020,
19 Males
62 years
water at
Median: 10 ng/mL
mean < LOD-
lm-PFOS: 1.27
previously
6781357)
7 Females
airport
(4.1-24 ng/mL)
0.084 ng/mL
3/4/5m-PFOS: 1.09
pregnant 2018
62 ng/|iL
2/6m-PFOS:
Median: < LOD
2/6m-PFOS: 1.04
during
(linear)
Median: 2.1 ng/mL
2/6m-PFOS
sampling year
64 ng/|iL
(0.57-8.1 ng/mL)
mean: < LOD-
• PFOS also
(branched)
1.6 ng/mL,
measured in the
130 ng/^L
Median: < LOD
private well of
Total
(not creatinine
adjusted)
one airport
employee living
near the airport
(PFOS
concentration in
well was lower
than the airport at
1.9 ng/|iL linear
and 0.24 ng/|iL
branched)
Worley et
153 (2010)
2010: Oral Drinking
2010
Not determined due
3.9 (2010)
• PFOS was
al. (2017,
63 Males
Mean 52.0 water
Geometric mean
to high proportion
3.3 (2016)
detected in
3859800)
90 Females
2016:
39.8 ng/mL
of < LOD samples
45.7% of
Mean 62.6
(30.9-48.9, 95% CI)
(creatinine adjusted)
samples. LOD
45 (2016)
2016
was 0.02 |ig/L
22 Males
Geometric mean
• Estimate intake
23 Females
23.4
(18.5-28.4, 95% CI)
rate for PFOS
was 6 ng/h,
based on PFOS
drinking water
concentration of
0.12 ng/L,
Volume of
distribution of
PFOS was
reported as
B-58
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Number of
Subjects
Age Range
Primary
Exposure Intake
Route
Plasma/Serum
Concentrations
Urinary
Concentrations
Estimated Half
Life (y)
Considerations
230 ml/kg body
weight.
• Clearance rate
was not reported
Fu et al.
302
Males: 19-
Occupational NR
Mean 5,624 ng/mL
Mean: 4.4 ng/mL,
Male (n= 136): GM
• Urinary samples
(2016,
213 Males
65, median
(assuming oral
Median
Median 1.2 ng/mL
60.9
were only taken
3859819)
89 Females
41
and inhalation
1,725 ng/mL
(not creatinine
Females (n= 71):
from 274
Females:
but not directly
(50.3-
adjusted)
GM 8.0
participants
19-50,
addressed in
118,000 ng/mL)
Overall (n = 207):
while there were
Median 37
study)
GM 32.6
serum samples
for every
participant
• For half -life
calculation for
females,
menstrual
clearance was
added to renal
clearance
• PFOS clearance
rate
0.017 mL/kg-day
Zhang et al.
86
22-68
Unspecified NR
Mean 21 ng/mL
Mean 47 ng/g
Young females: 6.2
• All participants
(2013,
47 Males
(Oral likely,
Median 19 ng/mL
creatinine
Males and older
had paired
3859849)
37 Females
Shijazhuang is a
capital city and
Handan is an
industrial city)
(1.4-180 ng/mL)
Branched
Median 28 ng/g
creatinine
(range 2.8-232 ng/g
creatinine)
females: 27
(whole
blood/serum and
urine)
• For young
females
menstrual
clearance was
estimated and
added to renal
clearance.
• Renal clearance
rate for total
B-59
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
, Number of . „ „ T , . Plasma/Serum Urinary Estimated Half „ .,
Study „ , Age Range Exposure Intake „ , ,. „ ^ . ... . , Considerations
Subjects Route Concentrations Concentrations Lite (y)
PFOS: mean
0.050 mg/kg/day
(young females),
0.037 mg/kg/day
(males and older
females)
Notes: CI = confidence interval; GM = geometric mean; LOD = limit of detection; NR = not reported.
B-60
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
All human PFOS half-life values identified in the literature review are provided in Table B-24. A
prominent feature of this data includes a very wide range of values ranging from less than 1 year
in a single male child of 16 years of age {Genuis, 2014, 2851045} to up to 60.9 years for males
occupationally exposed in a plant in China {Fu, 2016, 3859819}. Second, with one exception
{Genuis, 2014, 2851045}, half-lives estimated for males are longer than those estimated for
females. Third, studies that stratified by ages show an age-related increase in half-life values
{Genuis, 2014, 2851045}{Zhang, 2013, 2639569}. Fourth, linear isomers exhibit longer half-
lives than branched isomers {Zhang, 2013, 3859849}.
Gomis et al. (2017, 3981280) estimated half-lives in the general populations in the U.S. and in
Australia and reported a range of 3.3 to 5.4 years. Olsen et al. (2012, 1578499) estimated a
similar value in blood samples from Red Cross volunteer donors of 4.3 years. Interestingly, these
values were also in line with the half-life (2.3 y) estimated for subjects likely exposed to
contaminated drinking water in West Virginia and Ohio {Bartell, 2010, 379025}. Other studies
of subjects exposed to contaminated drinking water in Sweden {Li, 2017, 4238434} estimated
half-lives of 3.1 (for females) to 4.6 years (for males). Among the highest values are those for
occupationally exposed workers that ranged from 8.67 years (retired workers from a PFOS
production plant in Decatur, Alabama) to 60.9 years for workers in Hubei province, China.
While most studies were conducted in adults and/or adolescents, at least one study examined
PFOS half-lives in newborns {Spliethoff, 2008, 2919368}. Whole blood was collected as dried
spots on filter paper from almost all infants born in the United States. One hundred and ten of the
Newborn Screening Programs (NSPs) collected in the state of New York from infants born
between 1997 and 2007 were analyzed for PFOS. The analytical methods were validated by
using freshly drawn blood from healthy adult volunteers. The mean whole blood concentration
for PFOS ranged from 0.00081 |ag/m L to 0.00241 |ig/mL. The study grouped the blood spots by
two different time-points; those collected in 1999-2000 and in 2003-2004, which corresponded
to the intervals reported by NHANES. The PFOS concentrations decreased with a mean value of
0.00243 |ig/mL reported in 1999-2000 and 0.00174 |ig/mL in 2003-2004. The study authors
determined the half-life of PFOS using the regression slopes for natural log blood concentrations
vs the year 2000 and after. The calculated half-life for PFOS was 4.1 years.
B-61
-------
DRAFT FOR PUBLIC COMMENT
Table B-24. Summary of Human PFOS Half-Life Values
MARCH 2023
Study
Number of Subjects
Age Range
Estimated Half-Life (y)
Subjects
3M (2002, 6574114)
9
61 (55-64)
8.67 ±6.12
Retirees from the 3M plant in Decatur,
7 Males
(range: 2.29-21.3)
Alabama where PFOS was produced. Derived
2 Females
from 4 measurements over 18-month time
period from November of 1998 to May of 2000.
Bartell et al. (2010,
200
54.5 ± 15
2.3
Subjects were a subcohort of the C8
379025)
100 Males
Health Project, conducted in 2005-2006, who
100 Females
had lived in at least one of six affected water
districts near the DuPont Washington Works
plant.
Fu et al. (2016,
302
Males: 19-65, median 41
Based on daily clearance rate
Occupationally exposed subjects working in
3859819)
213 Males
Females: 19-50, median 37
Male (n = 136): GM60.9
one of the largest fluorochemical plants
89 Females
Females (n = 71): GM8.0
(Henxin Chemical Plant) in Yingcheng, Hubei
Overall (n=207): GM32.6
province, China
Based on annual decline rate
Overall (n=207): GM 1.9
Genuis et al. (2014,
53 Father
16-53
Father: 1.14
A family (6 patients) identified to have elevated
2851045)
47 Mother
Mother: 1.93
serum concentrations of PFAAs, likely through
22 1st Male Child
First Male child: 0.65
repeated commercial spraying of their home
19 2nd Female child
2nd Female child: 1.03
carpets with stain-repellants. Patients were
17 3rd Male child
3rd Male child: 0.78
treated by intermittent phlebotomy over a 4- to
16 4th Male child 3
4th Male child: 0.61
5-year period.
Glynn etal. (2012,
413 women
19-41
8.2
Primiparous women 3 weeks after delivery in
1578498)
Uppsala County, Sweden 1996-2010 (the
POPUP study (Persistent Organic Pollutants in
Uppsala Primiparas)
Gomis et al. (2017,
Australia: A total of
12+ (USA)
Australian Men: 4.9
Population based model using Australian
3981280)
24-84 pools per survey
< 16- > 60 (Australia)
American Men: 3.8
biomonitoring studies from Toms et al. (2014,
containing between 30-
Australian women: 5
2009) and NHANES from the U.S.. A total of
100 individual
American women: 3.3
24-84 pools per survey were obtained, with
samples.
each pool containing between 30 (2007) and up
USA: 2,000 individuals
to 100 individual samples (2003, 2009 and
were sampled
2011)
throughout the USA
Study reports intrinsic elimination half-lives.
B-62
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Number of Subjects
Age Range
Estimated Half-Life (y)
Subjects
Li etal. (2017,
4238434)
50
Males: 20
Females 30
15-50
Males: 4.6
Females: 3.1
Subjects in Ronneby, Sweden, exposed to
contaminated water through a municipal water
source.
Olsen et al. (2007,
1429952)
26
24 Males
2 Females
55-75
5.4
Retirees from the 3M plant in Decatur,
Alabama where PFOS was produced.
Olsen et al. (2012,
1578499)
600
Males: 300
Females: 300
5 age groups (20-29, 30-
39, 40-49, 50-59, 60-69)
4.3
Six American Red Cross adult blood donor
centers each provided 100 plasma samples for
analysis of 11 PFAA concentrations in 2010: 10
samples per every 10-year age interval (20-29,
30-39, 40-49, 50-59, and 60-69) for each sex.
The six American Red Cross blood donor
centers represented the following areas: Boston,
Massachusetts; Charlotte, North Carolina;
Hagerstown, Maryland; Los Angeles,
California; Minneapolis-St. Paul, Minnesota;
and Portland, Oregon
Splitehoff et al. (2008,
2919368)
240
Newborn infant (1-2 days)
4.1
New York State newborn screening program
blood spot specimens from newborn infants
Wong et al. (2014,
2851239)
Approx. 2,000 per
dataset (6 datasets)
Males and Females
Analyzed Separately
Eight age groups (age 12-
19, 20-29, 30-39, 40-
49, 50-59, 60-69, 70-79,
80+)
Males: 4.7
Females: 3.7
Females (accounting for rate of
menstrual blood loss): 4.0
Population based pharmacokinetic model
(Ritter) to six cross-sectional data sets from
1999 to 2012 from U.S. NHANES. Data from
age-stratified biomonitoring data for PFOS
extracted from U.S. NHANES from the years
1999-2000, 2003-2004, 2005-2006, 2007-
2008, 2009-2010, and 2011-2012
Worley et al. (2017,
3859800)
153 (2010)
63 Males
90 Females
45 (2016)
22 Males
23 Females
2010: mean 52.0
2016: mean 62.6
3.9 (2010)
3.3 (2016)
Residentially exposed population from
Lawrence, Morgan and Limestone Counties,
Alabama recruited by ATSDR
Xu et al. (2020,
6315709)
26
19 Males
7 Females
22-62 years
Linear PFOS: 2.91
lm-PFOS: 1.27
3/4/5m-PFOS: 1.09
2/6m-PFOS: 1.04
Subjects in Arvidsjaur, Sweden exposed to
contaminated drinking water occupationally
(working at the airport) and through residential
drinking water
B-63
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Number of Subjects Age Range
Estimated Half-Life (y)
Subjects
Yeungetal. (2013,
420 20-29
Munster: 4.3
Residents of Munster and Halle, Germany;
2850973)
Munster: 270
Halle: 4.8
samples collected between 1982 and 2009 in
Halle: 150
Munster and between 1995 and 2009 in Halle.
Zhang et al. (2013,
86 22-68
XPFOS
Healthy volunteers in Shijiazhuang and
3859849)
47 Males
Young females: 6.2
Handan, Hebei province, China, in April-May
37 Females
males and older females: 27
2010
n-PFOS
young females: 6.7
males and older females: 34
iso-PFOS
young females: 5.9
males and older females: 24
lm-PFOS
young females: 10
males and older females: 90
4m-PFOS
young females: 5.8 y
males and older females: 27
3 + 5m-PFOS
young females: 5 y
males and older females: 21
£m2-PFOS
young females: 5.1
males and older females: 14
Notes: ATSDR = Agency for Toxic Substances and Disease Registry; GM = geometric mean; NHANES = National Health and Nutrition Examination Survey;
PROS = perfluorooctane sulfonic acid; PFAA = Per- and polyfluoroalkyl acids.
B-64
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.4.5.3 Animal Studies
B.4.53.1 Non-Human Primates
In the study by Chang et al. (2012, 1289832), three male and three female monkeys were
administered a single IV dose of PFOS of 2 mg/kg and followed for 161 days. All monkeys were
observed twice daily for clinical signs, and body weights were obtained weekly. Urine and serum
samples were taken throughout the study. There was no indication that elimination was different
from males vs. females. Serum elimination half-lives ranged 122-146 days in male monkeys and
88-138 days in females. Mean values are shown in Table B-25. The Vd values suggest that
distribution was predominately extracellular.
In a second primate study, Seacat et al. (2002, 757853) administered 0, 0.03, 0.15, or
0.75 mg/kg/day potassium PFOS orally in a capsule by intragastric intubation to 6 young-adult to
adult cynomolgus monkeys/sex/group, except for the 0.03 mg/kg/day group which had 4/sex,
daily for 26 weeks (182 days) in a GLP study. Two monkeys/sex/group in the control, 0.15, and
0.75 mg/kg/day groups were monitored for 1 year after the end of the treatment period for
reversible or delayed toxicity effects. The elimination half-life for potassium PFOS in monkeys
was estimated from the elimination curves as approximately 200 days. This value is consistent
with that reported by Chang et al. (2012, 1289832) above.
B.4.5.3.2 Rats and mice
Half-lives rodents are very short relative to those observed in humans and primates (Table B-25).
In mice, Chang et al. (2012, 1289832) measured slightly higher half-lives in males (36-43 days)
compared to females (30-38 days). Ranges in mice were similar to those observed in rats.
Two recent studies evaluated toxicokinetic parameters informing half-lives in rats {Huang, 2019,
5387170}{Kim, 2016, 3749289}. In the Kim study, Sprague-Dawley rats were administered
2 mg/kg PFOS by either the IV or oral route. Urine and feces were collected weekly, and blood
was collected at 10 time points over the first day and then up to 70 days after exposure. Half-
lives in females and males were similar. In females, half-lives of 23.50 ± 1.75 and
24.80 ±1.52 days were estimated after oral and IV dosing, respectively. In males, values were
slightly longer (26.44 ± 2.77 and 28.70 ± 1.85 after oral and IV dosing, respectively).
In a similar study {Huang, 2019, 5387170}, male and female Sprague-Dawley rats were
administered a single dose of 2 mg/kg by IV injection or a single dose of 2 mg/kg or 20 mg/kg
by oral gavage and observed from 5 minutes to 20 weeks after dosing. After IV administration of
2 mg/kg, the overall half-life was 22 and 23 days in males and females, respectively days.
Similar values were obtained after a single gavage administration of the same dose (19.9 days in
males and 28.4 days in females) and after repeated dosing by oral gavage (19.0 in males and 21.1
in females. Half-lives in females administered the higher dose of 20 mg/kg were slightly longer
(18 days) than in males (14.5 days) and were slightly longer after repeated administration (19.0
and 21.1. days in males and females, respectively). Half-life values in the terminal elimination
phase were much longer than those measured in the initial elimination phase.
B-65
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-25. Summary of Animal PFOS Half-Life Values Identified in the Literature Review
Study
Species and Strain
Exposure
Route
Age or
Lifestage
Sex/Half-Life
Approach
Dose
Estimated Half
Life3
Chang et al. (2012,
1289832)
Cynomolgus Monkey IV
NR
Male
Female
Oral
4-6 years Male
Female
2 mg/kg and followed for
161 days
2 mg/kg and followed for
161 days
9 mg/kg
14 mg/kg
9 mg/kg
14 mg/kg
132 ±7
110± 15
124 ±3.89
117 ± 17.2
102 ±29.2
102 ±45.6
Seacat et al. (2002,
757853)
Cynomolgus Monkey Oral
Young-adult to Male
adult
Female
0.15 mg/kg
0.75 mg/kg
-200
-200
Chang et al. (2012,
1289832)
Mice, CD-I
Oral
8-10 weeks
Male
Female
1 mg/kg, followed for
20 weeks
20 mg/kg, followed for
20 weeks
1 mg/kg, followed for
20 weeks
20 mg/kg, followed for
20 weeks
42.81
36.42
37.80
30.45
Benskin et al. (2009, Rat, Sprague-Dawley Oral
1274133)
Adult (429 g) male
0.4 mg/kg PFOS (0.27 mg/kg
n-PFOS)
n-PFOS: 33.7
iso-PFOS: 23.4
5m-PFOS: 24.4
4m-PFOS: 23.1
3m-PFOS: 33.8
lm-PFOS: 102
tb-PFOS: 19.6
B7-PFOS: 15.4
B8-PFOS: 11.3
B9-PFOS11.1
Chang et al. (2012,
1289832)
Rat, Sprague-Dawley IV
8-10 weeks
Male
2 mg/kg, followed for 24 hr
7.99 ±4.94
Female (1 rat)
2 mg/kg, followed for 24 hr
5.62
Oral
8-10 weeks
Male
4.2 mg/kg, followed for 144 hr 8.23 ± 1.53
2 mg/kg, followed for 38.31 ± 2.32
10 weeks
B-66
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Species and Strain
Exposure
Route
Age or
Lifestage
Sex/Half-Life
Approach
Dose
Estimated Half
Life3
15 mg/kg, followed for
41.19 ± 2.01
10 weeks
Male (1 rat)
2 mg/kg, followed for 24 hr
3.1
Female
2 mg/kg, followed for 24 hr
1.94 ±0.13
2 mg/kg, followed for
62.30 ±2.09
10 weeks
15 mg/kg, followed for
71.13 ± 11.25
10 weeks
Huang et al. (2019,
Rat, Sprague-Dawley IV 8 weeks
Male - Overall
2 mg/kg
22.0 ±2.1
5387170)
elimination half-life
Male - initial phase
2 mg/kg
4.6 ±2.7
Male - terminal phase
2 mg/kg
39.7 ±4.4
Female - Overall
2 mg/kg
23.0 ±3.7
elimination half-life
Female - initial phase
2 mg/kg
0.3 ±0.3
Female - terminal
2 mg/kg
32.8 ±3.7
phase
Oral 8 weeks
Male - Overall
2 mg/kg
19.9 ±3.8
elimination half-life
2 (x5) mg/kg
19.0 ±3.2
20 mg/kg
14.5 ±2.1
Male - initial phase
2 mg/kg
3.1 ±2.4
2 (x5) mg/kg
0.3 ±0.1
20 mg/kg
4.0 ±2.9
Male - terminal phase
2 mg/kg
40.5 ±5.5
2 (x5) mg/kg
33.4 ±4.2
20 mg/kg
35.8 ±4.2
Female - Overall
2 mg/kg
28.4 ± 11.0
elimination half-life
2 (x5) mg/kg
21.1 ±4.3
20 mg/kg
18.0 ±3.1
Female - initial phase
2 mg/kg
0.8 ±2.1
2 (x5) mg/kg
0.3 ±0.2
20 mg/kg
2.2 ±3.0
B-67
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Species and Strain
Exposure
Route
Age or
Lifestage
Sex/Half-Life
Approach
Dose
Estimated Half
Life3
Female - terminal
2 mg/kg
40.7 ±3.5
phase
2 (x5) mg/kg
20 mg/kg
40.0 ±2.5
36.0 ±4.0
Kim et al. (2016,
Rat, Sprague-Dawley
IV
8-12 weeks
Male
2 mg/kg
28.70 ± 1.85
3749289)
Female
2 mg/kg
24.80 ± 1.52
Oral
8-12 weeks
Male
2 mg/kg
26.44 ± 2.77
Female
2 mg/kg
23.50 ± 1.75
Notes: IV = intravenous; NR = not reported.
a Data reported in mean days ± standard deviation.
B-68
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix C. Non-priority Health Systems
Evidence Synthesis and Integration
C.l Reproductive
EPA identified 60 epidemiological and 22 animal studies that investigated the association
between PFOS and reproductive effects. Of the epidemiological studies addressing male
reproductive endpoints, 2 were classified as high confidence, 15 as medium confidence, 6 as low
confidence, and 1 was considered uninformative (Section C.l.l). Of the epidemiological studies
addressing female reproductive endpoints, 5 were classified as high confidence, 24 as medium
confidence, 17 as low confidence, and 2 were considered uninformative (Section C.l.l). Of the
animal studies, 2 were classified as high confidence, 15 as medium confidence, 4 as low
confidence, and 1 was considered mixed (medium/low) (Section C.l.2). Studies may have
multiple judgments depending on the endpoint evaluated. Though low confidence studies are
considered qualitatively in this section, they were not considered quantitatively for the dose-
response assessment (See Main PFOS Document).
C.l.l Human Evidence Study Quality Evaluation and
Synthesis
C.l.l.l Male
C.l.l. 1.1 Introduction
The 2016 Health Advisory {U.S. EPA, 2016, 3982042} and HESD {U.S. EPA, 2016, 3603365}
reports identified limited evidence of effects of PFOS on reproductive effects in men and boys.
Analyses of male children in the C8 Health Project {Lopez-Espinosa, 2011, 1424973} suggested
an association between increasing PFOS exposure and delayed onset of puberty, defined by
measured testosterone levels (> 50 ng/dL testosterone and > 5 pg/mL free testosterone). The
effects of PFOS on semen quality parameters were mixed. In healthy, young Danish males
Joensen (2013, 2851244) observed significantly inverse associations with testosterone,
calculated free testosterone, free androgen index (FAI), and ratios of testosterone/luteinizing
hormone (LH), free testosterone/LH, and FAI/LH. Significant associations for semen quality
parameters were not observed among these young men. Regarding other studies examining
semen quality parameters, three studies {Buck Louis, 2015, 2851189; Joensen, 2009, 1405085;
Toft, 2012, 1332467} out of nine observed associations with morphologically abnormal sperm.
In a cross-sectional sample of military recruits (n = 105), Joensen (2009, 1405085) observed
significantly lower sperm counts in men with higher combined PFOS/PFOA exposure. A Texas-
and Michigan-based cohort (n = 462), the Longitudinal Investigation of Fertility and the
Environment (LIFE) study (Buck Louis, 2015, 2851189), observed limited evidence of the
effects of PFOS. Only one significant association was observed for a morphological parameter,
namely decreased percentage of sperm with coiled tails.
For this updated review, 23 studies3 (24 publications) report on the association between PFOS
and endocrine effects since the 2016 document. Eleven of the studies were in children and
3 Zhou, 2016, 3856472 and Zhou, 2017, 3858488 analyze participants from the same population using the same outcome.
C-l
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
adolescents {Di Nisio, 2019, 5080655; Ernst, 2019, 5080529; Goudarzi, 2017, 3981462; Itoh,
2016, 3981465; Jensen, 2020, 6311643; Lind, 2017, 3858512; Liu, 2020, 6569227; Lopez-
Espinosa, 2016, 3859832; Wang, 2019, 5080598; Zhou, 2016, 3856472; Zhou, 2017, 3858488},
one study was in pregnant women {Anand-Ivell, 2018, 4728675} and the remainder of the
publications were in the general population. Different study designs were utilized, including four
cohort studies {Ernst, 2019, 5080529; Goudarzi, 2017, 3981462; Itoh, 2016, 3981465; Jensen,
2020, 6311643}, one case-control study {Anand-Ivell, 2018, 4728675} with the remainder of the
studies following a cross-sectional design. All observational studies measured PFOS in blood
components (i.e., blood, plasma, or serum), however, PFOS was additionally measured in semen
for four studies {Cui, 2020, 6833614; Di Nisio, 2019, 5080655; Pan, 2019, 6315783; Song,
2018, 4220306} and amniotic fluid in one study {Anand-Ivell, 2018, 4728675}. The studies
were conducted in different study populations including populations from Australia, China,
Denmark, the Faroe Islands, Greenland, Italy, Japan, the Netherlands, Poland, Taiwan, Ukraine,
and the United States. There were several pairs of studies investigating the same population,
including the Biopersistent Organochlorines in Diet and Human Fertility (INUENDO) cohort
{Kvist, 2012, 2919170; Leter, 2014, 2967406), the Odense Child Cohort (OCC) (Lind, 2017,
3858512; Jensen, 2020, 6311643}, the Genetic and Biomarkers study for Childhood Asthma
(GBCA) {Zhou, 2016, 3856472; Zhou, 2017, 3858488}, and a cross-sectional sample of men
from an infertility clinic in Nanjing, China {Pan, 2019, 6315783; Cui, 2020, 6833614}. Two
studies assessed populations from related cohorts belonging to the Hokkaido Study on the
Environment and Children's Health {Itoh, 2016, 3981465; Goudarzi, 2017, 3981462}.
C.l.1.1.2 Study Quality
There are 24 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and male reproductive effects. Study quality evaluations for these 24
publications are shown in Figure C-l.
Of the 24 studies identified since the 2016 assessment, two studies were classified as high
confidence, 15 studies as medium confidence, six studies as low confidence, and one study
{Song, 2018, 4220306} was determined to be uninformative. Anand-Ivell, 2018, 4728675 was
considered low confidence for cryptorchidism and uninformative for amniotic fluid hormones.
Publications from the GBCA {Zhou, 2016, 3856472; Zhou, 2017, 3858488} were rated low
confidence because of concerns of selection bias and confounding. Cases and controls in Zhou,
2017, 3858488 were drawn from separate sources resulting in some concern for selection bias by
recruiting individuals from different catchment areas. One low confidence study {Di Nisio, 2019,
5080655} adjusted results only for age, resulting in concerns about potential for residual
confounding by socioeconomic status (SES). One National Health and Nutrition Examination
Survey (NHANES) study {Lewis, 2015, 3749030} did not adjust for the participant sampling
design in the analysis which contributed to a low confidence rating. Song, 2018, 4220306 only
reported bivariate correlations between exposure levels and semen parameters with no
accounting for potential confounders which contributed to the study being classified as
uninformative.
C-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Anand-lvell etal., 2018, 4728675-
Arbuckle et al., 2020, 6356900
Cui et al, 2020, 6833614-
Di Nisio et al., 2019, 5080656
Ernst et al., 2019, 5080529
Goudarzi et al., 2017, 3981462
Itoh etal., 2016, 3981465-
Jensen et a!.: 2020, 6311643
Kim etal., 2020, 6833596-
Kvist etal., 2012,2919170
Leteretal., 2014, 2967406-
Lewis et al., 2015, 3749030 -
Lind etal., 2017, 3858512
Liu et al., 2020, 6569227 -j +
Lopez-Espinosa et al., 2016, 3859832 -
Pan etal., 2019, 6315783-
Petersen et al., 2018, 5080277 -
Song etal., 2018, 4220306-
Tian et al„ 2019, 5390052-
Tsai etal., 2015, 2850160-
Wang et al., 2019, 5080598 -
Zhou etal., 2016, 3856472-
Zhou etal., 2017, 3858488-
van den Dungen etal., 2017, 5080340-
Legend
I Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
I Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
Figure C-l. Summary of Study Evaluation for Epidemiology Studies of PFQS and Male
Reproductive Effects
Interactive figure and additional study details available on HAWC.
C-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.1.1.1.3 Findings from Children and Adolescents
Sex hormone levels and related steroid hormone levels were examined in nine studies {Di Nisio,
2019, 5080655; Goudarzi, 2017, 3981462; Itoh, 2016, 3981465; Jensen, 2020, 6311643; Liu,
2020, 6569227; Lopez-Espinosa, 2016, 3859832; Wang, 2019, 5080598; Zhou, 2016, 3856472;
Zhou, 2017, 3858488} and three observed significant effects (Appendix D). A high confidence
prospective study on the Odense cohort {Jensen, 2020, 6311643; Lind, 2017, 3858512} did not
find evidence of effects on steroid hormones in the sex hormone metabolic pathway (e.g.,
dehydroepiandrosterone (DHEA), 17-hydroxyprogesterone (17-OHP)) in four-month-old male
infants. Similarly, a prospective cohort study {Goudarzi, 2017, 3981462} in boys from the
Hokkaido Study on the Environment and Children's Health reported no significant results with
steroid hormones in cord blood. However, a medium confidence study {Itoh, 2016, 3981465}
from a related cohort within the Hokkaido Study observed a significant positive association
(p = 0.033) for estradiol (E2). Increases in E2 potentially contributed to a significant decrease
(p = 0.002) in the testosterone-E2 ratio in male infants. Inverse associations were also observed
for progesterone (p = 0.043) and inhibin B (p < 0.001), and quartile analyses supported
significant trends for E2 (p-trend = 0.027), T/E2 (p-trend = 0.015), and inhibin B (p-
trend < 0.001) but did not support a significant trend for progesterone (p-trend = 0.231). A
medium confidence cross-sectional study (Lopez-Espinosa, 2016, 3859832) observed inverse
associations for E2 and total testosterone in children 6-9 years of age. Analyses by quartile of
exposure supported this trend for decreasing testosterone. A cross-sectional analysis in a medium
confidence study {Wang, 2019, 5080598} from China observed a positive association
(p < 0.001) for estriol (E3) in cord blood but did not find an association for E2.
Decreases in testosterone were seen in low confidence cross-sectional analyses {Zhou, 2016,
3856472; Zhou, 2017, 3858488} in children and adolescents (10-15 years of age) from the
GBCA in Taiwan. In boys, testosterone was observed to have a significant inverse association,
and a decreasing trend. No effects on E2 in boys were observed. A follow-up study {Zhou, 2017,
3858488} observed significant decreases in testosterone among children with asthma but not in
children without asthma. Sex stratified analyses for reproductive hormones were not conducted
in this study.
A cross-sectional study {Di Nisio, 2019, 5080655} in Italian high school students examined
associations between PFOS levels and possible risk factors for diseases of the male reproductive
system and observed significantly higher serum PFOS levels and testosterone (p < 0.001) in
exposed individuals compared to unexposed controls.
Pubertal development and semen parameters were examined in two studies {Di Nisio, 2019,
5080655; Ernst, 2019, 5080529} and effects were seen in one (Appendix D). One medium
confidence study {Ernst, 2019, 5080529} observed no associations between prenatal PFOS
exposure from first-trimester maternal serum samples and pubertal stages (i.e., Tanner stages)
and pubertal landmarks (e.g., acne, voice break, or first ejaculation). Comparisons of semen
analysis in Italian high school students {Di Nisio, 2019, 5080655} observed a reduced number of
sperm with normal morphology (p < 0.001) and a slight increase in semen pH (p = 0.005).
Anthropometric measurements of male reproductive organs were examined in four studies
{Arbuckle, 2020, 6356900; Di Nisio, 2019, 5080655; Lind, 2017, 3858512; Tian, 2019,
5390052} and three observed effects (Appendix D). A high confidence Danish study {Lind,
C-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
2017, 3858512} in children from the Odense cohort observed a significant positive association
with anoscrotal distance (AGDas) in the highest prenatal PFOS exposure group. Positive non-
significant associations were observed for anopenile distance (ADGap). Children from the
Shanghai-Minhang Birth Cohort Study {Tian, 2019, 5390052} were evaluated at birth,
six months, 12 months of age for changes in anogenital distance (AGD). At birth, significant
decreases in AGDas (p = 0.043) were observed in continuous analyses, and in the highest
quartile of exposure. Results were similar at six months of age. In contrast, associations were
positive and largely not significant at 12 months of age. However, a significant increase in
ADGap was observed among boys in the third quartile of exposure at 12 months. Results from a
medium confidence study {Arbuckle, 2020, 6356900} in children from the Maternal-Infant
Research on Environmental Chemicals (MIREC) cohort were inconsistent regarding the
relationship between prenatal PFOS exposure and AGD. Di Nisio et al. (2019, 5080655) reported
smaller AGD in exposed compared to unexposed adolescents (p = 0.019). Significant differences
(p < 0.001) were also observed for penile and testicular measurements among adolescents,
including smaller testicular volume, shorter penis length, and smaller penis circumference. A
smaller borderline significant pubis-to-floor distance was also observed (p = 0.064).
C.1.1.1.4 Findings from the General Adult Population
Serum sex hormones were examined in four studies {Cui, 2020, 6833614; Lewis, 2015,
3749030; Petersen, 2018, 5080277; Tsai, 2015, 2850160} and two observed effects (Appendix
D). Kmedium confidence study {Cui, 2020, 6833614} evaluated serum hormone concentrations
in men with fecundity issues and men from couples with female factor infertility. Serum and
semen PFOS were significantly correlated (Spearman's r = 0.793, p < 0.01). Total testosterone
and sex hormone binding globulin (SHBG) were inversely associated (p < 0.05) with serum and
semen PFOS. The total testosterone-LH ratio was negatively associated (p < 0.05) with semen
PFOS, and borderline significant with serum PFOS (p = 0.058). Results for total testosterone
remained among those 30 years old or younger after stratifying by age but were no longer
observed in men over 30 years of age. The pattern was similar for SHBG, but the association
with serum PFOS did not reach significance (p = 0.069). Analyses by quartile showed agreement
with the continuous regression analyses, indicating significant trends for total testosterone and
SHBG with serum and semen levels of PFOS. Kmedium confidence cross-sectional study
{Petersen, 2018, 5080277} on Faroese men observed a significant increase (p = 0.04) in
luteinizing hormone with increasing serum PFOS levels.
Semen characteristics and genomic effects in sperm were examined in five studies {Kvist, 2012,
2919170; Leter, 2014, 2967406; Pan, 2019, 6315783; Petersen, 2018, 5080277; Song, 2018,
4220306} and three observed effects (Appendix D). One medium confidence study {Kvist, 2012,
2919170} evaluating men from the INUENDO cohort from Greenland, Poland, or Ukraine
observed a significant positive association (p = 0.026) with the Y:X chromosome ratio in sperm
when pooling data across countries. This association was also observed in trend analyses for the
Greenland subset of the cohort but not in other country-specific analyses. Chromosomal changes
were further characterized in another INUENDO study {Leter, 2014, 2967406} using a sperm
DNA global methylation assay. Methylation of the Sata repeats, a non-transposonic repetitive
satellite DNA sequence generally found in or adjacent to every centromere, was significantly
increased (p < 0.05) in men from Ukraine, but no effect was observed in other INUENDO
communities or in the pooled analysis. Another method of analysis of sperm DNA methylation
utilized flow-cytometry to measure cell-by-cell methylated cytosines (% 5-mCs) by
C-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
immunodetection. A significant inverse relationship was observed among Polish men but was
not seen in other populations or the entire cohort. These results indicate hyper- and
hypomethylated states, respectively. Differences in results may be related to differences in each
method's approach.
A medium confidence cross-sectional study {Pan, 2019, 6315783} on a sample of men from
Nanjing, China, described above {Cui, 2020, 6833614}, investigated the effects ofPFOS on
semen characteristics. Two separate analyses were conducted, each using either serum or semen
as the biomonitoring matrix for PFOS exposure determination. In linear regression analyses
using semen PFOS exposure levels, significant positive associations (p < 0.05) were observed for
the sperm DNA fragmentation index (DFI)—a measure of the percentage of sperm with
damaged DNA. Significant inverse associations were observed for progressive motility, and
sperm straight-line velocity, suggesting an overall deleterious effect on sperm motility. No
significant associations were observed in analyses using serum PFOS levels.
C. 1.1.2 Female
C 1.1.2.1 Introduction
Reproductive health outcomes of interest in females vary with biological maturity over the life
course and by pregnancy status. Of interest across the life stages, reproductive hormone levels,
such as prolactin, follicle stimulating hormone (FSH), LH, testosterone, and E2, are commonly
examined as indicators of reproductive health. Additional reproductive health outcomes of
interest include timing of pubertal milestones among children and adolescents; fertility
indicators, impacts to menstruation, and occurrence of menopause among non-pregnant adult
females; and preeclampsia, gestational hypertension, pregnancy loss, and breastfeeding duration
among pregnant females.
The 2016 Health Assessment and Health Effects Support Document for PFOS {U.S. EPA, 2016,
3603365} concluded that there was suggestive evidence of an association with risk of gestational
hypertension or preeclampsia {Darrow, 2013, 2850966; Zhang, 2015, 2857764; Stein, 2009,
1290816}. There was generally consistent evidence of associations between serum PFOS and
reduced female fertility and fecundity {Bach, 2015, 3981738; Fei, 2009, 1291107; torgensen,
2014, 2851025; Velez, 2015, 2851037}. There were concerns over the possibility of reverse
causality explaining observed associations between PFOS exposure and various female
reproductive outcomes due to menstruation being a route ofPFOS excretion {Whitworth, 2012,
1332476}.
There are 48 studies (50 publications) that have investigated relationships between PFOS
exposure and female reproductive outcomes since the 2016 document {U.S. EPA, 2016,
3603365}. Among the 50 publications available for review, there were 20 cohort studies, 17
cross-sectional studies, and 13 case-control studies. 19 studies were conducted in adults, 6 were
conducted in children and adolescents, 13 were conducted in both adults and children, and 12
were conducted in pregnant women. Most studies used blood PFOS measures to assess exposure
while others used amniotic fluid and follicular fluid.
C.1.1.2.2 Study Quality
There are 48 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
C-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
association between PFOS and female reproductive effects. Study quality evaluations for these
48 studies are shown in Figure C-2 and Figure C-3.
Among the 48 publications available for review, 5 were classified as high confidence, 24 as
medium confidence, 16 as low confidence, and three were considered uninformative. Because
menstruation is a primary route of PFOS excretion for people who menstruate, reverse causality
was a specific concern for cross-sectional studies that measured blood PFOS and certain
reproductive hormones with known menstrual fluctuations without reporting sample collection
timing {Heffernan, 2018, 5079713; Zhang, 2018, 5079665}. Several low confidence studies
lacked an appropriate strategy for identifying potential confounders {McCoy, 2017, 3858475;
Zhou, 2017, 3859799} or failed to adjust for key confounders, such as age and SES {Heffernan,
2018, 5079713; Zhou, 2016, 3856472}. Low confidence studies had deficiencies in participant
selection {Zhang, 2018, 5079665; Bach, 2018, 5080557; Heffernan, 2018, 5079713}, exposure
measurement methods {Campbell, 2016, 3860110}, reliance on self-reporting for exposure,
outcome, or covariate information {Campbell, 2016, 3860110}, and small sample size
{Heffernan, 2018, 5079713; McCoy, 2017, 3858475}. Maekawa, 2017, 4238291 was considered
uninformative due to lack of information on participant selection, lack of adjustment in analyses
for key confounders. Lee, 2013, 3859850 was also considered uninformative due to lack of
consideration of key confounders in analyses. Arbuckle, 2013, 2152344 was considered
uninformative because PFOS was evaluated as the outcome and reproductive measures were
considered as predictors.
C-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C,1^ CM
Arbuckle et al., 2013, 2152344
Bach et al., 2015, 3981559-
Bach et al., 2018, 5080557 -
Bangma et al., 2020, 6833725 -
Borghese et al.. 2020, 6833656 -
Campbell et al.. 2016, 3860110
Caserta et al., 2013, 2000966
Caserta et al., 2013, 2001177 -
Crawford et al., 2017, 3859813
Ding etal., 2020, 6833612-
Donley et al., 2019, 5381537
Ernst etal., 2019, 5080529
Goudarzi et al., 2017, 3981462 -
Heffernan et al., 2018, 5079713 -
Huang et al., 2019, 5083564 -
Huo et al., 2020, 6505752^
Itoh et al., 2016, 3981465-
Jensen et al., 2020; 6311643 -
Kim et al., 2020, 6833596-
Lee etal., 2013, 3859850-
Lewis et al., 2015, 3749030 -
Liewet al., 2020, 6387285-
Liu etal , 2020, 6569227
Lopez-Espinosa et al., 2016, 3859832 -
Legend
^3 Good (metric) or High confidence (overall)
+ Adequate (metric) or Medium confidence (overall)
I - Deficient (metric) or Low confidence (overall)
H Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
Figure C-2. Summary of Study Evaluation for Epidemiology Studies of PFG
Reproductive Effects
Interactive figure and additional study details available on HAWC.
S and Female
C-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Louis etal., 2012, 1597490-
Lyngso et a I.,
Legend
I Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Q Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
Figure C-3. Summary of Study Evaluation for Epidemiology Studies of PFOS and Female
Reproductive Effects (Continued)
Interactive figure and additional study details available on HAWC.
C-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.l.1.23 Findings from Children and Adolescents
Two high confidence, eight medium confidence, and three low confidence studies assessed
relationships between PFOS exposure and female reproductive outcomes in children and
adolescents (Appendix D). Studies in infants primarily focused on reproductive hormone levels,
while studies in adolescents focused on reproductive hormone levels as well as pubertal
milestones.
Two high confidence {Yao, 2019, 5187556; Jensen, 2020, 6311643} and four medium
confidence studies {Itoh, 2016, 3981465; Liu, 2020, 6569227; Goudarzi, 2017, 3981462; Wang,
2019, 5080598} examined the effects of PFOS exposure on reproductive hormone levels in
female infants, reporting mixed results. Itoh, 2016, 3981465, a study of the Hokkaido birth
cohort, observed a significant negative association between maternal serum PFOS and
progesterone in cord blood (regression coefficient per unit change in PFOS (logio-
ng/mL) = -0.6; 95% CI: -0.9, -0.2) as well as prolactin in cord blood (regression coefficient per
unit change in PFOS (logio-ng/mL) = -0.5; 95% CI: -0.8, -0.2). A significant positive
association was observed between cord blood PFOS and E3 (regression coefficient per unit
increase in cord blood PFOS (logio-ng/mL) = 0.5; 95% CI: 0.3, 0.7) in another medium
confidence study {Wang, 2019, 5080598}. The two high confidence studies and four medium
confidence studies found no significant associations between maternal serum or cord blood
PFOS and reproductive hormones such as testosterone, the testosterone-to-estradiol ratio {Yao,
2019, 5187556}; E2, testosterone, SHBG, the testosterone-to-SHBG ratio {Itoh, 2016,
3981465}; 17-OHP, androstenedione, FSH, LH, DHEA, dehydroepiandrosterone sulfate
(DHEAS) {Jensen, 2020, 6311643}; 17-OHP, progesterone {Liu, 2020, 6569227};
androstenedione, DHEA {Goudarzi, 2017, 3981462}; P-E2, and estrone {Wang, 2019,
5080598}.
Three medium confidence {Lopez-Espinosa, 2016, 3859832; Maisonet, 2015, 3859841; Tsai,
2015, 2850160} and three low confidence {Lewis, 2015, 3749030; Zhou, 2016, 3856472; Zhou,
2017, 3858488} studies assessed the relationship between PFOS and reproductive hormone
levels in adolescent females. As part of the C8 Health Project, Lopez-Espinosa, 2016, 3859832
observed negative associations for total testosterone across serum PFOS quartiles and per unit
increase in serum PFOS among females 6-9 years old with high exposure (percent difference for
quartile 2 vs. quartile 1 = -1.1; 95% CI: -8.6, 7.1; percent difference for quartile 3 vs. quartile 1:
—7.8%; 95% CI: -15, -0.1; percent difference for quartile 4 vs. quartile 1: -11.1%; 95%
CL-18.2, -3.5; percent difference per unit increase in serum PFOS (ln-ng/mL) = -6.6%; 95%
CI: -10.1, -2.8). Maisonet, 2015, 3859841 found significantly increased serum testosterone
among 15-year-old females in the highest tertile of maternal serum PFOS during pregnancy
(beta: 0.18, 95% CI: 0.01, 0.35). No significant associations were observed for E2 {Lopez-
Espinosa, 2016, 3859832; Zhou, 2016, 3856472; Zhou, 2017, 3858488}, testosterone {Lewis,
2015, 3749030; Zhou, 2016, 3856472}, SHBG (Maisonet, 2015, 3859841; Tsai, 2015,
2850160}, or FSH {Tsai, 2015, 2850160}.
One medium confidence study drew data from the Danish National Birth Cohort (DNBC) to
examine the effects of prenatal PFOS exposure on pubertal milestones in female adolescents,
such as breast development (age at attainment of Tanner stages 2-5), pubic hair development
(age at attainment of Tanner stages 2-5), axillary hair development, and age at menarche in
adolescent girls {Ernst, 2019, 5080529}. Average age at attainment for all pubertal indicators
C-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
was significantly reduced across PFOS tertiles), while no other significant associations were
observed for breast development, age at menarche, axillary hair development, or pubic hair
development.
CI.1.2.4 Findings from Pregnant Women
One high confidence, five medium confidence studies, and one low confidence study examined
the relationship between PFOS exposure and preeclampsia (Appendix D). One medium
confidence study {Wikstrom, 2019, 5387145} reported significant positive associations between
serum PFOS and odds of preeclampsia in both continuous and quartile analyses (OR = 1.53;
95% CI: 1.07, 2.2; OR for PFOS highest vs. lowest quartile = 2.68; 95% CI: 1.17, 6.12). The
remaining five studies reported mixed non-significant associations {Borghese, 2020, 6833607;
Huang, 2019, 5083564; Rylander, 2020, 6833607; Huo, 2020, 6505752; Starling, 2014,
2446669}. Huo, 2020, 6505752, a high confidence cohort study of 3,220 pregnant women study
observed a non-significant reduction in odds of preeclampsia for women above the 80th
percentile for plasma PFOS compared to women in or below the 80th percentile and observed a
non-significant increase in odds of preeclampsia. In two medium confidence cohort studies, non-
significant positive associations were observed {Borghese, 2020, 6833656; Starling, 2014,
2446669}. Non-significant negative associations were observed in medium confidence case-
control {Rylander, 2020, 6833607} and cross-sectional {Huang, 2019, 5083564} studies. A low
confidence study found no association between median PFOS levels and hypertensive disorders
of pregnancy {Bangma, 2020, 6833725}.
One high confidence and two medium confidence studies examined the relationship between
PFOS exposure and gestational hypertension reporting non-significant mixed associations for
gestational hypertension and significant positive associations for blood pressure. Huo, 2020,
6505752, a high confidence cohort study of 3,220 pregnant women, observed a non-significant
negative association between plasma PFOS and odds of gestational hypertension. Borghese,
2020, 6833656, a medium confidence prospective cohort study, followed 1,708 women from
early pregnancy to delivery for gestational hypertension, preeclampsia, and changes in blood
pressure, measuring plasma PFOS once per trimester and again at delivery. Borghese, 2020,
6833656 observed a non-significant positive association between plasma PFOS and odds of
gestational hypertension. A significant positive association was reported for systolic blood
pressure (SBP) mmHg) per log2-(J,g/L increase PFOS at delivery (beta: 1.19, 95% CI 0.28, 2.1).
Significant positive associations were also observed in each trimester for diastolic blood pressure
(DBP) (mmHg) (beta for trimester 3: 0.66, 95 % CI 0.18, 1.14) but not at delivery. No
association between plasma PFOS levels and gestational hypertension was observed by Huang,
2019, 5083564.
Two medium confidence studies {Louis, 2016, 3858527; Liew, 2016, 6387285} and one low
confidence study {Jensen, 2015, 2850253} investigated the effect of PFOS exposure on
pregnancy loss and reported non-significant mixed results. In a cohort study of 501 couples,
Louis, 2016, 3858527 reported a non-significant, negative association between serum PFOS
levels and pregnancy loss during the first seven weeks of pregnancy. A case-control study nested
within the DNBC comparing 222 pregnancies ending in miscarriage to 218 pregnancies resulting
in live births observed non-significant positive associations across maternal plasma PFOS levels
for odds of miscarriage in both continuous and quartile analyses. Jensen, 2015, 2850253 also
C-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
reported non-significant positive associations for odds of miscarriage in both continuous and
tertile analysis.
Two medium confidence studies assessed the relationship between serum PFOS levels in
pregnancy and breastfeeding duration, with both reporting significant, inverse associations
between the two {Timmermann, 2017, 3981439; Romano, 2016, 3981728}. Using data from two
Faroese birth cohorts (n = 1,130), Timmermann, 2017, 3981439 observed a significant reduction
in total breastfeeding duration per doubling of maternal serum PFOS (regression coefficient per
doubling of serum PFOS (ng/mL) = -1.4; 95% CI: -2.1, -0.6) and a non-significant reduction in
exclusive breastfeeding duration per doubling of maternal serum PFOS (regression coefficient
per doubling of serum PFOS (ng/mL) = -0.3; 95% CI: -0.6, 0.1). These observations were
supported by a prospective birth cohort study of 336 women investigating the relationship
between serum PFOS levels during pregnancy and relative risk of breastfeeding termination at
three and six months postpartum {Romano, 2016, 3981728}. This study observed a positive
trend for relative risk of breastfeeding termination across maternal serum PFOS quartiles for
both time points. Relative risk for stopping breastfeeding by 3 months increased in maternal
serum PFOS quartiles 2, 3, and 4 compared to quartile 1, with a significant increase observed for
quartile 3 (relative risk for PFOS quartile 2 vs. 1 = 1.32; 95% CI: 0.97, 1.79; relative risk for
PFOS quartile 3 vs. quartile 1 = 1.39; 95% CI: 1.04, 1.88; relative risk for PFOS quartile 4 vs.
quartile 1 = 1.08; 95% CI: 0.79, 1.46). Relative risk for stopping breastfeeding by 6 months was
non-significantly increased in maternal serum PFOS quartiles 2, 3, and 4 compared to quartile 1
as well.
One high confidence study and one medium confidence study examined relationships between
PFOS exposure and female reproductive hormone levels in pregnant women. In a medium
confidence case-control study of 545 mother-infant pairs, Toft, 2016, 3102984 observed a
significant, positive association between PFOS in amniotic fluid and 17-OHP, with a significant
percent difference in the continuous analysis and a significant increase for tertile 3 compared to
tertile 1 (percent difference in median 17-OHP level per unit increase in amniotic fluid PFOS
(ln-ng/mL) = 0.15; 95% CI: 0.11, 0.2; percent difference in median 17-OHP for women in
amniotic fluid PFOS tertile 3 vs. tertile 1 = 18%; 95% CI: 11, 26). A significant, positive
association was also observed between amniotic fluid PFOS and androstenedione in the
continuous analysis and for tertile 3 compared to tertile 1 (percent difference in median
androstenedione level per unit increase in amniotic fluid PFOS (ln-ng/mL) = 0.15; 95% CI: 0.1,
0.21; percent difference in median androstenedione for women in amniotic fluid PFOS tertile 3
vs. tertile 1 = 17; 95% CI: 8, 25). Significant, positive associations across tertiles of PFOS were
observed for progesterone (percent difference per 1% increase in PFOS (ln-ng/mL) = 0.21; 95%
CI: 0.14, 0.29; percent difference for PFOS tertile 2 vs. 1 = 11%; 95% CI: 0, 23; percent
difference for PFOS tertile 3 vs. 1 = 22; 95% CI: 11, 34) and testosterone (percent difference per
1% increase in PFOS (ln-ng/mL) = 0.16; 95% CI: 0.09, 0.23; percent difference for PFOS tertile
2 vs. tertile 1 = 9%; 95% CI: -2, 20; percent difference for PFOS tertile 3 vs. tertile 1 = 18%;
95%) CI: 7, 29), but no association was observed for DHEAS. In a high confidence study, Mitro,
2020, 6833625, no significant association was observed between plasma PFOS during pregnancy
and SHBG levels three years postpartum.
C-12
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
One medium confidence study {Lyngs0, 2014, 2850920} examined the effects of serum PFOS
levels on pre-pregnancy menstruation. While evidence of increased odds of menstrual cycle
irregularity was reported, the association was not significant.
CI.1.2.5 Findings from the General Adult Population
Five medium confidence {Crawford, 2017, 3859813; Donley, 2019, 5381537; Kim, 2020,
6833596; Lum, 2017, 3858516; Wang, 2017, 3856459}, three low confidence studies { Bach,
2018, 5080557; McCoy, 2017, 3858475; Zhang, 2018, 5079665} and one uninformative study
{Arbuckle, 2013, 2152344} examined implications of PFOS exposure on female fertility,
reporting mixed results (Appendix D). Significant positive associations were reported in low
confidence studies, including for odds of premature ovarian insufficiency (POI) across plasma
PFOS quartiles {Zhang, 2018, 5079665} and for the fecundity ratio for parous women in plasma
PFOS quartiles {Bach, 2018, 5080557}. Non-significant positive associations were observed for
day-specific probability of pregnancy (Lum 2017, 3858516) and cycle and day-specific time to
pregnancy {Crawford, 2017, 3859813}. Associations with indicators of ovarian function were
largely non-significant, including no association observed between serum PFOS and anti-
Miillerian hormone (AMH) (Crawford, 2017, 3859813). Associations between maternal serum
PFOS during pregnancy and female adolescent AMH levels were also not observed {Donley,
2019, 5381537}. No significant associations were reported for infertility measures including
endometriosis-related infertility {Wang, 2017, 3856459}, and fertilization rate {Kim, 2020,
6833596}. Additionally, McCoy, 2017, 3858475 reported non-significant negative correlations
between PFOS in follicular fluid and blast conversion rate, fertilization rate, and follicle count.
No associations were observed for other outcomes related to menstrual cycles and gynecologic
pathologies, including menstrual cycle length {Lum, 2017, 3858516}, endometriosis, polycystic
ovary syndrome (PCOS), genital tract infections, and idiopathic infertility {Kim, 2020,
6833596}.
One high confidence study examined the relationship between PFOS exposure and age at natural
menopause: the Study of Women's Health Across the Nation (SWAN), a prospective cohort of
1,120 premenopausal women aged 45-56 {Ding, 2020, 6833612}. Significant, positive
associations were reported between serum Sm-PFOS and risk of natural menopause for women
in Sm-PFOS tertile 3 vs. tertile 1 (HR = 1.27; 95% CI: 1.01, 1.59) and between serum n-PFOS
and risk of natural menopause for women in n-PFOS tertile 3 vs. tertile 1 (HR = 1.26; 95% CI:
1.02, 1.57). Non-significant positive associations were observed for both Sm-PFOS and n-PFOS
when analyzed as a continuous variable and for women in tertile 2 vs. tertile 1.
One medium confidence {Tsai, 2015, 2850160} and five low confidence studies {Heffernan,
2018, 5079713; Lewis, 2015, 3749030; McCoy, 2017, 3858475; Petro, 2014, 2850178; Zhang,
2018, 5079665} reported associations between PFOS and female reproductive hormone levels in
non-pregnant adult women. Three low confidence studies reported significant mixed effects. In
women with and without PCOS, Heffernan, 2018, 5079713 observed significant negative
associations with FAI only in controls. McCoy, 2017, 3858475 observed a negative correlation
with plasma E2. In women with and without POI, Zhang, 2018, 5079665 observed significant
negative associations for E2 in both cases and controls and positive associations for FSH and
prolactin in cases only. No significant associations were observed for testosterone {Lewis, 2015,
3749030}; mean FSH and SHBG in young women (ages 12-30 years) {Tsai, 2015, 2850160};
C-13
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
testosterone, E2, and SHBG {Heffernan, 2018, 5079713}; E2 {Petro, 2014, 2850178}; or forLH
and testosterone {Zhang, 2018, 5079665}.
C.1.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 6 studies from the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} and 16 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOS HESD that investigated the association between PFOS and reproductive effects. Study
quality evaluations for these 22 studies are shown in Figure C-4.
C-14
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Alam et al., 2021, 9959508
Argus, 2000, 5080012
Butenhoff et al., 2009, 757873
Butenhoffetal., 2012, 1276144
Chen et al., 2018, 5080460 -
Lai etal., 2017, 3981773-
Lau et al., 2003, 757854-
Lee et al., 2015, 2851075
Lopez-Doval et al., 2015, 2848266
Luebker et al., 2005, 1276160
Luebker et al., 2005, 757857
NTP, 2019, 5400978
Qiu et al., 2013, 2850956
Qiu etal., 2016, 3981408
Qiu et al., 2020, 7276729
Qu etal., 2016, 3981454
Salgado et al., 2015, 3981583
Seacat et al., 2002, 757853
Thomford, 2002, 5432419
Zhang et al., 2019, 5918673
Zhang et al., 2020, 6315674
Zhong et al.,2016, 3748828
-
+
NR
++ ++ ++
+
+
NR
-
+
+
•
NR
NR
++
++
+
+
+
+
NR
+
+
+
+
NR
+
-
-
-
+
NR
+
+
+
+
-
NR
++ ++
-
-
+
NR
+
+
-
•
+
NR
+
-
-
-
NR
NR
+
+
+
Legend
| Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
NR] Not reported
* Multiple judgments exist
Figure C-4. Summary of Study Evaluation for Toxicology Studies of PFOS and
Reproductive Effects
Interactive figure and additional study details available on HAWC.
C-15
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Short-term, subchronic, chronic, and reproductive/developmental animal studies suggest that oral
exposure to PFOS can adversely affect the male and female reproductive systems. However, it is
not often clear whether the observed alterations reflect specific toxicity to the reproductive
system or if they result from concurrent systemic toxicity (i.e., reductions in body weight).
Effects observed in male rodents included alterations to hormone levels (prolactin, luteinizing
hormone, FSH, E2, and testosterone), as well as decreased testis weights, and decreased sperm
count. In female mice exposed to PFOS, effects on prolactin family hormones were observed.
Although effects were predominately seen in rodent species there were inconsistencies among
rats and mice. In cynomolgus monkeys no effects were noted in reproductive organ weights and
histopathology, although a decrease in male E2 levels was observed {Seacat, 2002, 757853}.
C. 1.2.1 Male and Female Fertility Parameters and Pregnancy
Outcomes
Male and female fertility parameters and pregnancy outcomes were evaluated in rodent and
rabbit species. Mating and fertility parameters, such as number of pregnancies per number of rats
that mated, number of days to inseminate, and number of matings during the first week of
cohabitation were unaffected by PFOS doses as high as 3.2 mg/kg/day in a two-generation
reproduction study in rats{Luebker, 2005, 1276160; Butenhoff, 2009, 757873}. Gestation and
fertility indices were unaffected in one- and two-generation rat reproduction studies {Luebker,
2005, 757857; Luebker, 2005, 1276160}; however, gestation length was significantly decreased
in a dose-dependent manner in dams exposed to > 0.8 mg/kg/day in the one-generation study
{Luebker, 2005, 757857} and in Po dams exposed to 3.2 mg/kg/day in the two-generation
study {Luebker, 2005, 1276160} (Figure C-5). Decreases in maternal bodyweight change were
noted in both studies {Luebker, 2005, 757857; Luebker, 2005, 1276160}(see PFOS Main
Document). In contrast, Butenhoff et al. (2009, 757873) reported no significant differences in
gestation length for rats treated with up to 1 mg/kg/day PFOS from GD 0 to PND 20. That study
also found no significant differences in the number of litters delivered or live litter size at birth
{Butenhoff, 2009, 757873}.
PFOS Reproductive KfTects - Rnl Gestation length
End point
Studv Name
Study Design
Observation Time
Animal Description
• No significa
t change A Significant increase Significant decrease I
Length of Gestation
Luebker et al.. 2005,757857
Luebker et al.. 2005.1276160
reproductive (80d (42d pre-mating. GD0-2
reproductive (42d prior mating-LD20)
. LD1-4)) GD0-21
GD23
Butenhoff et al.. 2009.757873
developmental (GD0-PND20)
0.5
1 1.5 2 2.5 3 3.5
Concentration (mg/kg/day)
Figure C-5. Gestation Length in Rats Following Exposure to PFOS
LD = lactation day; GD = gestation day; Po = parental generation.
Interactive figure and additional study details available on HAWC.
In mice, reproductive outcomes were examined in pregnant CD1 mice treated at 1.5, 3, and
6 mg/kg/day from GD 6-GD 18. Body weight and body weight change were significantly
reduced in dams given PFOS at 6 mg/kg/day in comparison to the controls {Fuentes, 2006,
757859}. The number of live and dead fetuses per litter and number of implantation sites were
not statistically significant even though high fetal mortality was observed in dams exposed to
PFOS at 6 mg/kg. Lastly, there was no observed effect on gravid uterine weight in pregnant CD1
mice on GD 18.
C-16
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
In a single study in New Zealand white rabbits, dams were administered 0 mg/kg/day,
0.1 mg/kg/day, 1.0 mg/kg/day, 2.5 mg/kg/day, or 3.75 mg/kg/day PFOS via intubation from GD
7 to GD 20 {Argus Research Laboratories, 2000, 5080012}. The number of rabbits pregnant at
the time of sacrifice (GD 29) decreased with increasing dose due to an increased incidence of
abortion with higher PFOS doses (see PFOS Main Document). Only 12/21 (57%) of dams that
became pregnant in the study from the 3.75 mg/kg/day dose group were pregnant on GD 29
compared to 100% pregnancy maintained in the 0 mg/kg/day, 0.1 mg/kg/day, and 1.0 mg/kg/day
groups and 94% pregnancy maintained in the 2.5 mg/kg/day group. Each individual doe that
aborted exhibited weight loss and severely reduced feed consumption. Overall, maternal body
weight gains were significantly reduced in the 1.0 mg/kg/day, 2.5 mg/kg/day, and
3.75 mg/kg/day groups {Argus Research Laboratories, 2000, 5080012}.
C. 1.2.2 Male Sperm Parameters
Sperm parameters were evaluated in studies of male rats and mice, with conflicting results
(Figure C-6). In a 28-day study conducted by NTP in which Sprague-Dawley rats, exposed to
PFOS for 28 days had no effect on spermatid headcount in the testis, sperm count in the
epididymis and cauda epididymis, or epididymal sperm motility in animals treated with
1.25 mg/kg/day to 5.0 mg/kg/day {NTP, 2019, 5400978}. In contrast, a general reduction in
epidydimal sperm count was observed in mice among studies of varying durations including two
4-week studies in ICR mice exposed to 2.5 mg/kg/day or 5 mg/kg/day, a 4-week study in ICR
mice exposed to 5 mg/kg/day and 10 mg/kg, a 5-week study in C57 mice exposed to
10 mg/kg/day, and CD-I pups on PND 63 exposed to 3 mg/kg/day during gestation {Qiu, 2013,
2850956; Qiu, 2016, 3981408; Qu, 2016, 3981454; Lai, 2017, 3981773; Qiu, 2020, 7276729}.
Qiu et al. (2016, 3981408) did not observe alterations in epididymis weight that may have
influenced epididymal sperm counts.
Endpoint
Epididymis Sperm Count
Study Name Study Design Observation Time
Lai etal., 2017, 3981773 developmental (GD1-17) PND63
Qiu et al., 2013, 2850956 reproductive (28d) 28d
Animal Description
F1 Mouse, CD-1 (
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
treated by gavage for 28 days. Conversely, in a subchronic study, Alam et al. (2021, 9959508)
observed significantly increased serum testosterone and progesterone levels in comparison to the
controls at 0.015 and 0.15 mg/kg via oral gavage for 60 days in Wistar rats. However, in a 28-
day study conducted by NTP (2019, 5400978), no effects on testosterone levels were noted in
male rats treated with up to 5 mg/kg/day. A 46% decrease relative to controls was also noted in
mice treated with 10 mg/kg/day for five weeks {Qu, 2016, 3981454}. Developmental studies in
mice showed a 31% decrease in testosterone at PND 63 in CD-I mice exposed to 3 mg/kg/day
throughout gestation {Lai, 2017, 3981773}. C57BL/6 mouse pups treated with 1 and
5 mg/kg/day showed 35% and 52% decreases, respectively, at postnatal week 4 (PNW 4) after
maternal oral exposure from GD 1 to GD 17 (significantly different in the 5 mg/kg/day group)
{Zhong, 2016, 3748828}. In the same study, 38% and 34% decreases were observed in the 1 and
5 mg/kg/day groups, respectively, at PNW 8, though only the response in the 1 mg/kg/day group
was statistically different from controls. Similarly, Qiu et al. (2020, 7276729) observed a
significant decrease in serum testosterone levels at 5, and 10 mg/kg/day in comparison to the
controls for four weeks in ICR mice. Cynomolgus monkeys treated up to 0.75 mg/kg/day for
182 days showed no statistically significant effects on testosterone levels {Seacat, 2002,
757853}.
PFOS Reproductive Effects - Testosterone Levels
Endpoint Study Name Study Design Observation Time Animal Description Dose (mg/kg/day) I O Statistically significant 0 Not statistically significant!—| 95% CI I
1
Testosterone Seacat et al., 2002, 757853 chronic (26wk)
Zhong etal., 2016, 3748S28 developmental (GD 1-17) PNW4
Lai et al.,2017, 3981773
Qiu etal., 2020, 7276729
Qu et al.,2016, 3981454
developmental (GD 1-17) PND63
short-term (4wk)
Monkey, Cynomolgus (,f\ N=4-6) 0
0.03
0.15
0.75
F1 Mouse. C57BL/6 (
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Interactive figure and additional study details available on HAWC.
The red dashed lines indicate a 100% increase or 100% decrease from the control response.
GD = gestation day; PND = postnatal day; PNW = post-natal week; Fi = first generation
Changes in E2 levels in males were noted in rats, mice, and cynomolgus monkeys across studies
of varying durations (Figure C-8); however, the direction of the change was not consistent across
the studies. In two studies from the same laboratory, following a 28-day exposure, Salgado et al.
(2015, 3981583) and Lopez-Doval et al. (2015, 2848266) noted decreases in E2 ranging from
13-19% in rats treated with 3.0 and 6.0 mg/kg/day and >1.0 mg/kg/day, respectively. Decreases
were similar across dose groups and were not dose dependent. In mice, subchronic exposure to
PFOS (35 days) at doses of 0.5 mg/kg/day and 10 mg/kg/day showed no statistically significant
effect on E2 levels, but there was a general increasing trend with increasing dose (5% and 10%
increase, respectively) {Qu, 2016, 3981454}. Male mouse pups exposed to 5.0 mg/kg/day from
GD 1 to GD 17 exhibited a 42% increase in serum E2 levels at PNW 4 {Zhong, 2016, 3748828}.
By PNW 8 the increase was no longer statistically significant but remained 28% higher than the
control group {Zhong, 2016, 3748828}. There was an apparent dose-dependent increase in
serum E2 at both PNW 4 and PNW 8. Conversely, no significant change or trend in serum E2
levels was observed in adult ICR male mice exposed to 0 mg/kg/day, 0.5 mg/kg/day,
5 mg/kg/day, and 10 mg/kg/day for four weeks {Qui, 2020, 7276729}. Seacat et al. (2002,
757853) observed a 97% decrease in serum E2 in male cynomolgus monkeys treated at
0.75 mg/kg/day for 182 days {Seacat, 2002, 757853}.
Endpoint Study Name Study Design Observation Time Animal Description Dose (mg/kg/day)
PFOS Reproductive Effects - Male Estradiol Levels
O Statistically significant 0 Not statistically significant!—I 95% Cl |
Estradiol Seacat et al., 2002. 757853 chronic (26wk) 182d Monkey. Cynomolgus (<•', N=4-6) 0
0.03
0.15
0.75
1
1
1
1
e
—#—
i
i
i
i
i
Zhong etal., 2016. 3748828 developmental (GD1-17) PNW4 F1 Mouse. C57BL/6 (
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Short-term exposure studies examining the effect of PFOS exposure on LH, FSH, and prolactin
levels in male rats were available (Figure C-9). Groups treated for 28 days with
doses > 0.5 mg/kg/day as well as 3.0 mg/kg/day and 6.0 mg/kg/day exhibited decreases in LH
(15%—30%) and prolactin (54%-78%), respectively {Lopez-Doval, 2014, 2850091; Lopez-
Doval, 2015, 2848266; Salgado, 2015, 3981583}. Additionally, increases ranging from 88-133%
in serum FSH levels were observed in all treated groups (0.5 mg/kg/day-6 mg/kg/day) when
compared to controls {Lopez-Doval, 2014, 2850091}. However, in a study by Qiu et al. (2020,
7276729), PFOS exposure did not significantly alter serum FSH and LH levels.
PFOS Reproductive Effects - Male LH and Prolactin Levels
Endpoint Study Name
Follicle Stimulating Hormone (FSH) Qiu et al., 2020. 7276729
Study Design Observation Time Animal Description Dose (mg/kg/day)
short-term (4wk) 4wk
Mouse, ICR (..', N=10)
Lcuteinizing Hormone (LH)
Qiuetai.. 2020, 7276729
short-term (4wk) 4wk
Mouse, ICR (J', N=10) 0
0.5
5
10
Lopez-Doval et al., 2015.2648266
short-term (28d) 29d
Rat. Sprague-Dawley (. \ N=15) 0
0.5
1
3
6
Prolactin (PRL)
Salgado et al., 2015, 3961583
short-term (28d) 28d
Rat. Sprague-Dawley (;';', N=7) 0
3
6
Statistically significant^ Not statistically significant |—|95%Cl|
-120 -100 -60 -60 -40 -20 0 20 40 60 80 100
Percent control response (%)
Figure C-9. Percent Change in LH and Prolactin Levels Relative to Controls in Male Rats
Following Exposure to PFOS
Interactive figure and additional study details available on HAWC.
The red dashed line indicates a 100% decrease from the control response.
C. 1.2.3.2 Females
Evidence that oral exposure to PFOS results in changes to levels of prolactin-family hormones in
female mice was noted in an investigation by Lee et al. (2015, 2851075) (Figure C-10). In this
study, the authors reported dose-dependent reductions in prolactin-family hormones, including
mouse placental lactogen (mPL-II) (46%-71%), mouse prolactin-like protein (mPLP)-Ca (20%-
53%>), and mPLP-K (30%>-57%>), in pregnant CD-I mice exposed to 0.5 mg/kg/day, 2.0
mg/kg/day, and 8 mg/kg/day PFOS from GD 11 to GD 16. Concurrent dose-dependent decreases
in bodyweight of 2%, 6%, and 21%, respectively, were also observed in these mice {Lee, 2015,
2851075}.
C-20
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOS Reproductive Effects - Prolactin Family Hormone Levels in Female Mice
Eridpoint
Mouse placental lactogen smPL Ml Lee
Study Name
etal ,2015 2?5inrs
Study Design
>l&.«?loomental iGD11 "0'i
Observation Time
GD1?
Animal Description
PC. Mouse CD-! i N=3t
Dose (mg/kg/day)
0
2
i<
"« Ftatistirolly significant 0 Nrt statistically signifran*
•
'IS" CI
9
S
t
Mouse prolactin-like protein jmPLPh-Ga Lea
developmental iGD11-1tj)
GD17
PC Mojsg CD-1 l , N=J|
0
i.) i.i
:¦>
8
:¦ ¦¦<::>
: -o i
» I
Mouse prolactin-like protein (mPLF}-K Lee
etal..2016.2851075
developmental (GD11-16)
GD17
P0 Mouse, CD-1 N«3>
0
0.5
2
8
f
~ !
! 0
¦ .
-80 -7Q -60 -50 -40 -30 -20 -10 0 10 20
Percent control response (%)
Figure C-10. Percent Change in Prolactin-Family Hormone Levels Relative to Controls in
Female Mice Following Exposure to PFOS
Interactive figure and additional study details available on HAWC.
GD = gestation day; Po = parental generation
In female cynomolgus monkeys treated with PFOS for 182 days, E2 levels decreased in a dose-
dependent manner relative to controls (decreases of 16%, 52%, and 73% in the 0.03 mg/kg/day,
0.15 mg/kg/day, and 0.75 mg/kg/day dose groups, respectively) {Seacat, 2002, 757853} (Figure
C-l 1). In the same study, testosterone levels were not affected in females in a dose-dependent or
statistically significant manner, though a decrease of 72% was observed in the 0.15 mg/kg/day
dose group {Seacat, 2002, 757853}. In contrast to female monkeys, evaluation of Fi female
mouse pups treated with 0.1 mg/kg/day, 1.0 mg/kg/day, or 5.0 mg/kg/day from GD 1 to GD 17
showed increases in E2 levels relative to the control at PNW4 (increases of 10%, 17%, and 8%,
respectively) andPNW8 (increases of 11%, 19%, and 12%, respectively), although statistical
significance was not achieved {Zhong, 2016, 3748828}. A dose-dependent decrease in
testosterone levels when compared to controls was noted at PNW4 in females (decreases of 18%,
26%), and 30% in the 0.1 mg/kg/day, 1 mg/kg/day, and 5 mg/kg/day groups, respectively), but
was not statistically significant {Zhong, 2016, 3748828}. In female rats exposed to PFOS for
28 days, testosterone levels were significantly increased with 1.25 mg/kg/day and 2.5 mg/kg/day
PFOS (increases of approximately 37% in both groups) but not in the 5 mg/kg/day dose group
{NTP, 2019, 5400978}.
C-21
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOS Reproductive Effects - Female Estradiol and Testosterone Levels
Endpolnt Study Name
Study Design Observation Time Animal Description Dose {mg/kg/day)
Estradiol Seacat et al.. 2002, 757853 chronic (26wk)
Zhong et al., 2016, 3748828 developmental (GD1-17) PNW4
Testosterone Seacat et al., 2002, 757853 chronic (26wk)
Zhong et al., 2016, 3748828 developmental (GD1-17) PNW4
NTP, 2019, 5400978 short-term {28d) 29d
Monkey, Cynomoigus (2, N=4-6) 0
0,03
0,15
0.75
F1 Mouse. C57BL/6 (2, N=12) 0
0.1
F1 Mouse, C57BL/6 (2. N=12) 0
Monkey. Cynomoigus N=4-6) 0
0.03
0.15
0.75
F1 Mouse, C57BL/6 (°, N=12) 0
0.1
F1 Mouse, C57BL/6{"'. N=12) 0
Rat, Sprague-Dawley (: , N=9-10) 0
0.312
0.625
1.25
2.5
-140-120-100 -80 -60 -40 -20 0 20 40 60
Percent control response (%)
100 120 140
Figure C-ll. Percent Change in Estradiol and Testosterone Levels Relative to Controls in
Female Rodents and Non-Human Primates Following Exposure to PFOS
Interactive figure and additional study details available on HAWC.
GD = gestation day; Po = parental generation; Fi = first generation; PNW = postnatal week; d = day; wk = week.
The red dashed lines indicate a 100% increase or 100% decrease from the control response.
C. 1.2.4 Estrous Cyclicity and Ovorion Function (Female) and
Reproductive System Development, Including Markers of Sexual
Maturation (Female and Male)
In females, a dose dependent increase in estrous cycle length was observed in rats treated with
0.625 mg/kg/day to 2.5 mg/kg/day over the course of 28-days (increased length of 0.4 days in the
2.5 mg/kg/day group compared to controls); however, this finding was not statistically
significant {NTP, 2019, 5400978}. Summary statistics indicated that the proportion of time spent
in each phase was unaffected, although Markov analysis indicated that females in all assessed
groups had an increased likelihood of transitioning to prolonged diestrus when compared to
controls. In the same study, the number of cycles was considered unaffected by treatment {NTP,
2019, 5400978}. In a two-generation reproduction study in rats, no significant effects were
observed on the number of estrous cycles of Po females treated with up to 3.2 mg/kg/day for 6
weeks prior to mating {Luebker, 2005, 1276160}.
No significant changes in the number or distribution of corpora lutea were noted in Po rats
exposed prior to mating and during gestation in the one- and two-generation reproductive
C-22
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
toxicity studies {Luebker, 2005, 757857; Luebker, 2005, 1276160}. Likewise, no changes in the
number of corpora lutea were seen in Po female rabbits exposed during gestation {Argus
Research Laboratories, 2000, 5080012}. Reproductive and developmental studies additionally
reported no impact of gestational PFOS exposure on the timing of preputial separation or vaginal
opening in rats {Luebker, 2005, 1276160; Lau, 2003, 757854; Butenhoff, 2009, 757873}.
C. 1.2.5 Reproductive Organ Weights and Histopathology
C. 1.2.5.1 Mole
Several studies investigated the effect of PFOS exposure on male reproductive organ weights.
No effects were noted in the absolute or relative epididymal and testes weights in rats treated up
to 5.0 mg/kg/day for 28 days {NTP, 2019, 5400978} or in absolute or relative testis weight in
rats exposed to 20 ppm in the diet for 53 weeks (equivalent to 0.984 mg/kg/day) {Butenhoff,
2012, 1276144}. In a sub-chronic study, no significant changes were observed in relative or
absolute testis weight upon exposure to PFOS at doses of 0.015 mg/kg/day and 0.5 mg/kg/day
for a duration of 60 days in Wistar rats {Alam, 2021, 9959508}. Effects in mice exposed to
PFOS were observed in a subchronic study in which significant decreases in absolute and
relative testis weights were noted in mice exposed to 10 mg/kg/day for 35 days {Qu, 2016,
3981454}. No effects were seen in relative epididymis or testis weights of mice treated up to
10 mg/kg/day for four weeks {Qiu, 2016, 3981408}, nor were any effects noted in the relative
testes weight of mouse pups treated from GD 1 to GD 17 {Lai, 2017, 3981773}. Similarly, no
significant changes in relative epididymis or testis weight were observed for ICR mice treated up
to 10 mg/kg/day for four weeks {Qiu, 2020, 7276729}. Male cynomolgus monkeys treated with
up to 0.75 mg/kg/day for 182 days showed no changes in absolute or relative epididymis or testis
weights {Seacat, 2002, 757853}.
Histopathological examination of rats following 28 days or 2 years of exposure revealed no
treatment-related changes in the testes, epididymis, seminal vesicle, or prostate {NTP, 2019,
5400978; Butenhoff, 2012, 1276144}. However, Lopez-Doval et al. (2014, 2850091) noted
edema around seminiferous tubules and malformed spermatids in male rats treated
at > 1 mg/kg/day with marked edema and loss and degeneration of the spermatozoids observed at
6 mg/kg/day following PFOS exposure up to 6 mg/kg/day for 28 days. The specific incidences of
histopathological findings were not reported in this study, and statistical analysis was not
conducted. In another study, subchronic exposure in rats revealed lesions including vacuolations
in spermatogonia, spermatocytes, and Ley dig cells, as well as exaggerated intracellular space and
disturbed germ cells in rats treated at 10 mg/kg/day; however, incidences of specific findings
were not reported, and statistical analyses were not conducted {Qu, 2016, 3981454}.
Relevant histopathological findings in a 28-day study in mice included Sertoli cell vacuolization
and derangement of the cell layers at 2.5 mg/kg/day, 25 mg/kg/day, and 50 mg/kg/day and
dislocated immature germ cells in seminiferous tubules at 50 mg/kg/day {Qiu, 2013, 2850956};
however, incidences of specific findings were not reported, and methods used for statistical
analysis are unclear. These findings were confirmed by observing the ultrastructure of
seminiferous epithelia by electron microscopy. In addition, PFOS was observed to disrupt the
blood-testis barrier in vitro and in vivo in two studies, suggesting that Sertoli cells in the testes
are a target for PFOS toxicity {Qiu, 2013, 2850956; Qiu, 2016, 3981408}. Along with
observations of reduced epididymal sperm count in these studies, these results collectively
C-23
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
suggest the potential for PFOS exposure to induce deterioration of the testis and impair
spermatogenesis in mice.
In a single study in cynomolgus monkeys, histopathology of the testes, prostate, and seminal
vesicles and cell proliferation in the testes were examined following exposure to PFOS for
182 days, however no differences were noted when compared to controls {Seacat, 2002,
757853}.
C. 1.2.5.2 Females
Female organ weight and histopathological data in rats were only available from the 28-day NTP
study {NTP, 2019, 5400978}. In females, relative and absolute uterus with cervix and vagina
weights in Sprague-Dawley rats were not affected following a 4-week exposure to PFOS at doses
up to 5 mg/kg/day. In addition, no treatment-related histopathological changes were observed in
the uterus or ovary {NTP, 2019, 5400978}. A chronic study in rats {Butenhoff, 2012, 1276144}
measured the weight of the uterus with cervix at the 53-week interim evaluation and evaluated
histopathology of the ovaries, uterus, vagina, and cervix after two years of exposure to
concentrations up to 20 ppm in the diet (equivalent to 1.251 mg/kg/day) and reported no
significant findings for those organs. Similarly, Seacat et al. (2002, 757853) did not report
alterations in ovary weight or uterine or vaginal histopathology in female cynomolgus monkeys
dosed with up to 0.75 mg/kg/day PFOS for 182 days. Effects on placental characteristics such as
weight and capacity, as well as histopathological effects were noted in rats and mice exposed to
PFOS during gestation (see PFOS Main Document).
C.1.3 Mechanistic Evidence
Mechanistic evidence linking PFOS exposure to adverse reproductive outcomes is discussed in
Sections 3.2.5, 3.3.4, and 3.4.1.2 of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365}. There
are 56 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD that investigated the mechanisms of action of PFOS that
lead to reproductive effects. A summary of these studies is shown in Figure C-12. Additional
mechanistic synthesis will not be conducted since evidence suggests but is not sufficient to infer
that PFOS may cause respiratory effects.
C-24
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mechanistic Pathway Animal Human In Vitro Grand Total
Angiogenic, Antiangiogenic, Vascular Tissue Remodeling 10 12
Big Data, Non-Targeted Analysis 4 15 9
Cell Growth, Differentiation, Proliferation, Or Viability 8 0 29
Cell Signaling Or Signal Transduction 9 1 27
Extracellular Matrix Or Molecules 2 0 2 4
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation 4 115
Hormone Function 10 1 13 23
Oxidative Stress 2 0 5 7
Xenobiotic Metabolism 2 0 4 6
Other 2 0 13
Not Applicable/Not Specified/Review Article 10 0 1
Grand Total 20 3 38 56
Figure C-12. Summary of Mechanistic Studies of PFOS and Reproductive Effects
Interactive figure and additional study details available on Tableau
C.1.4 Evidence Integration
C. 1.4.1 Reproductive Effects in Males
There is slight evidence for an association between PFOS exposure and male reproductive
effects based on inverse associations with testosterone in male children. Inverse associations
with testosterone were observed in two medium confidence studies in children, and one study
reported an inverse association for E2. Among low confidence studies, there was mixed evidence
for an association between PFOS exposure and testosterone in cross-sectional studies {Di Nisio,
2019, 5080655; Zhou, 2016, 3856472; Zhou, 2017, 3858488} in children and adolescents.
However, these mixed associations were observed in populations at different stages of pubertal
development. Results showed decreasing testosterone with increasing serum PFOS in children,
but increased testosterone with higher PFOS exposure levels in adolescents. In adolescents, there
were no effects on pubertal development, but associations were observed for penile
measurements, testicular measurements, and sperm parameters {Di Nisio, 2019, 5080655}.
Evidence was also inconsistent for AGD in infants. In adults, there was evidence in one study
{Cui, 2020, 6833614} of an inverse association between serum PFOS and testosterone, and these
associations were also observed using semen PFOS. Inverse associations were also seen for E2,
SHBG, and the total T/LH ratio. Regarding semen and sperm characteristics in adults,
associations were observed for several parameters in analyses of semen PFOS, including
increased sperm DNA fragmentation and decreased measures of sperm motility. Other results for
markers of genotoxic effects (e.g., sperm Y:X chromosome ratio, sperm DNA methylation) in
sperm were inconsistent.
The animal evidence for an association between PFOS exposure and reproductive toxicity in
males is slight based on several high or medium confidence studies of varying durations showing
that oral exposure to PFOS can affect the male reproductive system. However, many of the
C-25
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
observed reproductive effects (e.g., decreased E2 levels in male monkeys) occurred at doses that
also resulted in reduced body weight which can be confounding effects for reproductive
endpoints. Additionally, several of the observed effects were not consistent across species (e.g.,
sperm parameters, testis weight, E2 levels in males) which increases uncertainty about the
relevance of these effects to humans or potential differences in the MOA between species.
Several studies reported effects of PFOS exposure on male mouse and rat reproductive organ
histopathology {Qiu, 2013, 2850956; Qiu, 2016, 3981408; Qu, 2016, 3981454; Lopez-Doval,
2014, 2850091; Lopez-Doval, 2015, 2848266}. However, these studies did not report incidence
data which hinders further quantification or conclusions about these results. In male mice, these
histopathological alterations were accompanied by a reduction in epididymal sperm count,
though this effect was not observed in male rats. Although reductions in epididymal sperm
counts across mouse studies ranged from 25%-70% at the highest doses tested {Qiu, 2013,
2850956; Qiu, 2016, 3981408; Qu, 2016, 3981454; Lai, 2017, 3981773} and are consistent with
effects seen in humans, fertility may be normal in male rodents even with sperm reductions as
great as 90% {Gray, 1988,1332904}. Without further evidence of reduced fertility or quantitative
evidence of histopathological changes in the testes or epididymis, it is unclear whether
reductions in sperm counts can be considered adverse.
Similar uncertainties arise when linking the observed hormonal alterations with functional
reproductive consequences. Changes in LH, FSH, and prolactin were observed in male rats,
however, lack of histopathological and sperm parameter effects (specific to rats), as well as
inconsistent effects on testosterone levels, make it difficult to assess the relevance of these
changes. It is difficult to ascertain the magnitude of change in hormone levels that can be
considered adverse without concurrent supporting evidence of functional or histopathological
reproductive consequences.
C. 1.4.2 Evidence Integration Judgment
Overall, evidence suggests that PFOS exposure has the potential to cause reproductive effects in
males under relevant exposure circumstances (Table C-l). This conclusion is based primarily on
effects on inverse associations with testosterone in male children and adults, and decreased AGD
in children observed in studies in humans exposed to median PFOS ranging from 1.4 to
34.8 ng/mL. Although there is some evidence of negative effects of PFOS exposure on semen
and sperm characteristics in adults, there is considerable uncertainty in the results due to
inconsistency across studies and limited number of studies. For male reproductive toxicity, the
conclusion is based primarily on observed changes in hormonal parameters in adult rodents
following exposure to doses as low as 0.5 mg/kg/day PFOS. However, findings from animal
studies are similarly inconsistent as in epidemiological studies. In animal studies, there are
uncertainties in the adversity of the observed effects, a lack of quantifiable histopathological
evidence in reproductive organs, and inconsistencies in responses observed across studies and
species.
C-26
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-l. Evidence Profile Table for PFOS Reproductive Effects in Males
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.l .l)
Male reproductive
hormones
1 High confidence study
8 Medium confidence
studies
6 Low confidence studies
In children and
adolescents, inverse
associations for total
testosterone were
observed in two studies
(2/8), including a medium
confidence study reporting
a significant inverse trend.
One study reported higher
total testosterone levels
among highly exposed
adolescents but was of low
confidence. Findings for
estradiol in male children
were generally less
precise, however, one
medium confidence study
(1/6) observed a
significant, dose-
dependent increase in
estradiol, which was
accompanied by a
significant decrease in the
testosterone/estradiol
ratio. Findings for LH and
FSH were mixed, but
significantly increased LH
was observed in one low
confidence study, and
significantly decreased
FSH was observed in a
medium confidence study.
In adults, one study (1/3)
High and medium
confidence studies
Consistent direction
of effects for
testosterone levels
Low confidence studies
Lmprecision of most
findings
Potential for residual
confounding by SES
and smoking status
©OO
Slight
Evidence for male
reproductive effects is
based on several studies
reporting consistent and
coherent changes to sex
hormones. Effects on sex
hormones were supported
by adverse effects
observed for other
outcomes such as sperm
quality (i.e., sperm DFI
and HDS) and
anthropometric measures.
Uncertainties remain
regarding mixed results in
adults and imprecise
results in some medium
confidence studies. There
were also a limited
number of studies
evaluating certain
endpoints such as semen
parameters and pubertal
development.
©OO
Evidence Suggests
Primary basis:
Human evidence indicted
effects on inverse
associations with
testosterone in male
children and adults, and
decreased AGD in children
observed in studies in
humans exposed to median
PFOS. Although there is
some evidence of negative
effects of PFOS exposure
on semen and sperm
characteristics in adults,
there is considerable
uncertainty in the results
due to inconsistency across
studies and limited number
of studies. Animal evidence
indicated changes in
hormonal parameters in
adult rodents following
exposure to PFOS.
However, findings from
animal studies are similarly
inconsistent as in
epidemiological studies. In
animal studies, there are
uncertainties in the
adversity of the observed
effects, a lack of
C-27
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
observed significant
decreases in total
testosterone and the
testosterone/estradiol
ratio. Another medium
confidence study reported
a non-significant increase
in total testosterone, but
other results for
testosterone were
imprecise. One study
reported a non-significant
decrease in estradiol, and
one study reported a
significant increase in LH.
Findings for SHBG were
mixed.
Evidence Integration
Summary Judgment
quantifiable
histopathological evidence
in reproductive organs, and
inconsistencies in responses
observed across studies and
species.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
Semen parameters
4 Medium confidence
studies
1 Low confidence studies
One study examined
semen parameters in high
school students and
observed significant
increases in semen pH and
increased deficits in sperm
morphology. Semen
parameter findings in
adults were generally
consistent between
endpoints but did not
always indicate adverse
effects. Sperm count was
non-significantly
increased in two studies
(2/3), non-significant
positive associations were
observed for sperm
Medium confidence •
studies •
Consistent direction
of effects for most •
findings
Low confidence study
Lmprecision of most
findings
Lncoherence of
direction of effect for
adult semen parameters
Potential for residual
confounding by SES
and smoking status
C-28
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Evidence Integration
Studies and Summary and Key Factors that Increase Factors that Decrease Evidence Stream Summary Judgment
Interpretation Findings Certainty Certainty Judgment
concentration in three
studies (3/4), and semen
volume was reported to be
non-significantly
increased in two studies
(2/4). Adverse effects
were also observed,
including decreased
normal morphology (1/2),
increased sperm HDS, and
significantly increased
sperm DFI. Sperm HDS
and DFI are measures of
sperm chromatin integrity
and sperm DNA damage,
respectively.
Anthropometric
Three studies examined •
High and medium •
Low confidence study
measurements of male
measurements in male
confidence studies •
Lmprecision of some
reproductive organs
infants. Non-significant •
Consistent direction
findings
1 High confidence study
increases in AGD were
of effects •
Potential for residual
2 Medium confidence
observed in two studies •
Coherence of
confounding by SES
studies
(2/3), but findings were
findings
and smoking status
1 Low confidence study
not consistent across
timepoints. One study
examined anthropometric
measurements in male
high school students.
Adverse effects were
observed in adolescents
with higher exposure
levels, including smaller
testicular volume, shorter
penis length, and smaller
penis circumference.
C-29
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Male pubertal
development
1 Medium confidence
study
Findings for changes in
timing of pubertal
development were largely
non-significant. Study
authors reported earlier
onset of individual Tanner
stages (G2 and G5) and
earlier onset of voice
break, but none were
significant.
Medium confidence
study
Limited number of
studies examining
outcome
Evidence from In Vivo Animal Studies (Section C.1.2)
Male mating and
fertility
1 Medium confidence
study
No effects on male mating
or fertility parameters
were observed in a two-
generation reproduction
study in rats with exposure
beginning six weeks prior
to mating (1/1).
Medium confidence
study
Limited number of
studies examining
outcome
©OO
Slight
Evidence was based on 15
high and medium
confidence studies. There
were no observed effects
on mating or fertility in
the only available two-
generation reproduction
study; however, other
studies observed effects on
hormone levels, sperm
count, and testis weight
and histopathology. Some
of the reproductive effects
observed (e.g., decreased
testosterone and estradiol
levels) may be secondary
effects because they
occurred at doses that also
resulted in reduced body
weight. Additionally,
several of the observed
C-30
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Evidence Integration
Studies and Summary and Key Factors that Increase Factors that Decrease Evidence Stream Summary Judgment
Interpretation Findings Certainty Certainty Judgment
effects were not consistent
across species (e.g., sperm
parameters, testosterone
and estradiol levels) which
increases uncertainty
about the relevance of
these effects to humans or
potential differences in the
mode of action between
species. Studies reporting
alterations in testis
histopathology did not
report incidence data
which hinders conclusions
about these results. In
male mice, these
histopathological
alterations were
accompanied by a
reduction in epididymal
sperm count. Without
further evidence of
reduced fertility or
quantitative evidence of
histopathological changes
in the testis or epididymis,
it is unclear whether
reductions in sperm counts
can be considered adverse.
Changes in LH, FSH, and
prolactin were observed in
male rats; however, the
lack of histopathological
and sperm parameter
effects (specific to rats), as
well as inconsistent effects
C-31
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Evidence Integration
Summary Judgment
Studies and
Interpretation
Summary and Key Factors that Increase Factors that Decrease
Findings Certainty Certainty
Evidence Stream
Judgment
on testosterone levels,
make it difficult to assess
the relevance of these
changes.
Male reproductive
hormones
1 High confidence study
8 Medium confidence
studies
Alterations in testosterone
levels in male rats (3/8),
mice (4/8), and monkeys
(1/8) were inconsistent.
Reports of decreases (5/8),
increases (1/8), and no
change (2/8) in serum
testosterone were reported
following developmental,
short-term, and subchronic
exposure. Mixed effects
on serum estradiol
included decreased levels
in rats and monkeys (3/6),
increases (1/6) in mice, or
no effects (2/6). Short-
term studies in male
rodents reported no effect
onFSH (1/1), decreases in
LH (1/2), and decreases in
prolactin (1/1).
High and medium
confidence studies
Inconsistent direction
of effects across
species
Changes in body
weight may limit
ability to interpret
these responses
Sperm parameters
1 High confidence study
5 Medium confidence
studies
In mice, five short-term,
subchronic, or
developmental studies
observed dose-dependent
reductions in epididymal
sperm count (5/5).
However, in rats, no
effects on epididymal or
testicular sperm counts or
High and medium
confidence studies
Consistent direction
of effects within
species
Inconsistent direction
of effects across
species
C-32
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Evidence Integration
Studies and Summary and Key Factors that Increase Factors that Decrease Evidence Stream Summary Judgment
Interpretation Findings Certainty Certainty Judgment
epididymal sperm motility
were reported (1/1).
Male pubertal
No effects on age at
•
Medium confidence •
Limited number of
development
preputial separation were
studies
studies examining
3 Medium confidence
observed in reproductive
•
Consistent direction
outcome
studies
and developmental studies
in male rats (3/3).
of effects •
Organ weights
Most studies in rats, mice,
•
High and medium
No factors noted
2 High confidence
or monkeys found no
confidence studies
studies
effects on absolute or
•
7 Medium confidence
relative testis weight (8/9).
studies
One subchronic study in
mice observed decreases
in absolute and relative
testis weight (1/9) only at
the highest dose tested. No
effects on absolute or
relative epididymis weight
were observed (4/4).
Histopathology
Two high confidence
•
High and medium •
Inconsistent direction
2 High confidence
studies in rats and one
confidence studies
of effects across
studies
medium confidence study
species
4 Medium confidence
in monkeys found no
studies
histopathological changes
in the testes, prostate,
epididymides, or seminal
vesicles following short-
term or chronic exposure
(3/6). Three studies in
mice observed
histopathological changes
in the testes following 4-5
weeks of exposure (3/6).
These changes included
vacuolations in
C-33
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Studies and
Interpretation
Evidence Stream Summary and Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
spermatogonia,
spermatocytes, Ley dig
cells, and Sertoli cells, and
disturbed germ cell layers;
however, results were all
reported qualitatively
only.
Notes: LH = luteinizing hormone; FSH = follicle-stimulating hormones; SHBG = sex hormone binding globulin; SES = socioeconomic status; DFI = DNA fragmentation index;
HDS = high DNA stainability; DNA = deoxynucleic acid; AGD = anogenital distance; G2 = genital stage 2; G5 = genital stage 5.
C-34
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C. 1.4.3 Reproductive Effects in Females
There is slight evidence for an association between PFOS exposure and female reproductive
effects in humans based on observed increases in preeclampsia and gestational hypertension,
with most studies observing positive non-significant associations, in populations with high
exposure levels and at levels typical in the general population.
Epidemiological evidence of a relationship between PFOS exposure and female fertility is
mixed. Since the 2016 Health Assessment, nine studies have investigated associations between
PFOS exposure and fertility. While some studies reported more frequent or intense female
fertility problems with increasing PFOS exposure {Crawford, 2017, 3859813; McCoy, 2017,
3858475; Zhang, 2018, 5079665}, others found PFOS to be positively associated with female
fertility indicators {Lum, 2017, 3858516; Kim, 2020, 6833596; Bach, 2018, 5080557}, and some
did not observe any clear trends {Wang, 2017, 3856459}. Kim et al. (2020, 6833596) also
observed some non-significant, positive associations between follicular fluid PFOS and fertility
etiology factors for other gynecologic pathologies, including endometriosis, PCOS, genital tract
infections, and idiopathic infertility.
There is limited, consistent epidemiological evidence of an inverse association between serum
PFOS levels in pregnancy and breastfeeding duration. Timmermann et al. (2017, 3981439)
observed negative associations between PFOS exposure and exclusive and total breastfeeding
duration, while Romano et al. (2016, 3981728) observed increased relative risk of breastfeeding
termination with increasing PFOS exposure.
Human epidemiological evidence of a relationship between PFOS exposure and the female
reproductive milestones of age at menarche and menopause is mixed. In the 2016 Health
Assessment, Christensen et al. (2011, 1290803) observed a non-significant decreased adjusted
OR for earlier age at menarche for continuous prenatal PFOS exposure. Since the 2016 Health
Assessment, Ernst et al. (2019, 5080529) observed a significant inverse association between age
at attainment for overall puberty indicators and a non-significant inverse association for
continuous prenatal PFOS exposure and age at menarche. In the 2016 Health Assessment, Knox
et al. (2011, 1402395) observed significant increased odds of natural menopause across PFOS
quintiles for women ages 51-65 years in the C8 Health Project. Since the 2016 Health
Assessment, Ding et al. (2020, 6833612) observed significant, positive associations for serum
Sm-PFOS and n-PFOS and risk of natural menopause in women aged 45-56. However, findings
from studies concurrently assessing menstruation events and PFOS levels in blood must be
interpreted with caution due to potential reverse causality, as menstruation is a primary route of
PFOS excretion for people who menstruate.
Since the 2016 Health Assessment, 20 studies have assessed relationships between PFOS
exposure and various female reproductive hormones. 12 of these studies were conducted in
female infants and adolescents. Commonly assessed female reproductive hormones were 17-
OHP, DHEA, E2, FSH, SHBG, and testosterone. While most studies did not report significant
associations or consistent trends between PFOS exposure and these outcomes, Itoh et al. (2016,
3981465) observed significant negative associations for maternal serum PFOS and cord blood
prolactin and progesterone levels and Wang et al. (2019, 5080598) observed significant positive
associations for cord blood PFOS and cord blood estrone and E3. In pregnant women, Yao et al.
(2019, 5187556) observed significant, positive associations for cord blood PFOS and
C-35
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
testosterone and testosterone to E2 ratio and Toft et al. (2016, 3102984) observed significant,
positive trends in 17-OHP, androstenedione, progesterone, and testosterone across amniotic fluid
PFOS tertiles.
The recent epidemiological evidence is also suggestive of an association between PFOS and
preeclampsia and gestational hypertension, though there is conflicting evidence on altered
puberty onset and limited data suggesting reduced fertility and fecundity. The association are
inconsistent across reproductive hormone parameters, and it is difficult to assess the adversity of
these alterations.
The animal evidence for an association between PFOS exposure and female reproductive toxicity
is slight based on several high or medium confidence studies of varying durations showing that
oral exposure to PFOS can affect the female reproductive system. However, many of the
observed reproductive effects (e.g., decreased gestation length in female rats, decreased prolactin
levels in female mice) occurred at doses that also resulted in decreased gestational body weight
which can be confounding effects for reproductive endpoints.
Uncertainties arise when linking the observed hormonal alterations with functional reproductive
consequences. NTP (2019, 5400978) reported modest increases in testosterone concentrations
(37% increase) in female rats with PFOS doses of 1.25 mg/kg/day and 2.5 mg/kg/day, but not the
highest dose of 5 mg/kg/day. The response in the highest dose was confounded by decreased
body weight. The alterations in testosterone were accompanied by dose-dependent increases in
estrous cycle length, though this increase was not statistically significant and alterations in the
estrous cycle were not observed in a second study in female rats {Luebker, 2005, 1276160}. It is
difficult to ascertain the magnitude of change in hormone levels that can be considered adverse
without concurrent supporting evidence of functional or histopathological reproductive
consequences.
C. 1.4.4 Evidence Integration Judgment
Overall, evidence suggests that PFOS exposure has the potential to cause reproductive effects in
females under relevant exposure circumstances (Table C-2). This conclusion is based primarily
on effects on preeclampsia and gestational hypertension, female reproductive milestones, and
female reproductive hormonal outcomes observed in studies in humans exposed to median PFOS
ranging from 1.4 ng/mL to 34.8 ng/mL. There is considerable uncertainty in the results due to
inconsistency across studies and limited number of studies. For female reproductive toxicity, the
conclusion is based primarily on observed changes in hormonal parameters in adult rodents
following exposure to doses as low as 1.25 mg/kg/day PFOS. However, findings from animal
studies are similarly inconsistent as in epidemiological studies. In animal studies, there are
uncertainties in the adversity of the observed effects, a lack of quantifiable histopathological
evidence in reproductive organs, and inconsistencies in responses observed across studies and
species.
C-36
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-2. Evidence Profile Table for PFOS Reproductive Effects in Females
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.l.l)
Female reproductive
hormones
3 High confidence
studies
10 Medium confidence
studies
7 Low confidence studies
Results from assessment
of female reproductive
hormones were mixed. In
13 studies of female
children and adolescents,
7 studies reported
significant associations.
One medium confidence
study reported increased
El and E3 and an inverse
association with E2 (1/7).
Two other studies reported
increased E2 (2/7), and
one also reported
increased FSH (1/2).
Three studies, one high,
one medium, and one low
confidence, reported
increases in testosterone
(3/7). One medium
confidence study observed
inverse associations with
progesterone and prolactin
(1/7). Eight studies
examined adult women,
though many were low
confidence (5/8). Four
studies reported
significant effects (4/8).
Two low confidence
studies observed inverse
associations with E2 (2/4),
one medium study
High and medium
confidence studies
Coherence of
findings for
testosterone
Low confidence studies
Lnconsistent direction
of effects
Lmprecision of most
findings
Potential for selection
bias and residual
confounding by age
and SES
©OO
Slight
Evidence for female
reproductive effects is
based on several studies
reporting effects on sex
hormones and increased
odds of preeclampsia.
There was also evidence
for changes in age at
natural menopause.
Uncertainties remain
regarding mixed findings
in studies of sex
hormones, and a limited
number of studies
examining outcomes such
as female reproductive
milestones and
anthropometric
measurements.
®oo
Evidence Suggests
Primary basis:
Human evidence indicted
effects on preeclampsia and
gestational hypertension,
female reproductive
milestones, and female
reproductive hormonal
outcomes observed in
studies in humans exposed
to median PFOS. There is
considerable uncertainty in
the results due to
inconsistency across studies
and limited number of
studies. Animal evidence
indicated changes in
hormonal parameters in
adult rodents following
exposure to PFOS.
However, findings from
animal studies are similarly
inconsistent as in
epidemiological studies. In
animal studies, there are
uncertainties in the
adversity of the observed
effects, a lack of
quantifiable
histopathological evidence
in reproductive organs, and
inconsistencies in responses
C-37
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
observed increased
progesterone, testosterone,
and 17-OHP (1/4) and one
observed an inverse
association with free
androgen index (1/4).
Evidence Integration
Summary Judgment
observed across studies and
species.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
Preeclampsia and
gestational
hypertension
1 High confidence study
5 Medium confidence
studies
Six studies examined
preeclampsia in pregnant
women. One study
reported significant
positive results, while four
studies of medium and
high confidence reported
non-significant positive
associations with
preeclampsia. Three
studies reported inverse
associations (3/6).
Of the three studies
examining gestational
hypertension (3/6), two
reported inverse
associations but neither
reached significance (2/3).
After observing non-
significant increased odds
of gestational
hypertension, one medium
confidence study reported
significantly increased
DBP.
High and medium
confidence studies
Imprecision of all
findings
Inconsistent direction
of effects
Potential for reverse
causality
Female reproductive
milestones
Three studies examined
reproductive milestones
High and medium
confidence studies
Low confidence study
C-38
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
1 High confidence study
2 Medium confidence
studies
1 Low confidence study
related to menstruation,
two in adolescent
populations (2/3) and one
in an adult population
(1/3). Two studies, one
low confidence study in
adolescents (1/2) and one
medium confidence study
in adults (1/1), reported
non-significant increases
in long menstrual cycles.
A significant inverse
association was observed
among adolescents for
average age at attainment
for all pubertal indicators
(1/2). One high confidence
study reported significant
positive associations with
age at natural menopause.
Consistent direction •
of effects
Potential for residual
confounding by not
identifying
confounders
Limited number of
studies examining
specific outcomes
Fertility indicators
6 Medium confidence
studies
6 Low confidence studies
Examinations of fertility
indicators include
fecundability, fertilization
rate, and measures of
ovarian health, such as
anti-Mullerian hormone
levels or endometriosis.
Twelve studies evaluated
fertility indicators in non-
pregnant women with
mixed results. One
medium confidence study
reported significant
inverse associations with
endometriosis-driven
Medium confidence •
studies •
Low confidence studies
Lmprecision of most
findings
Potential for residual
confounding by not
identifying
confounders
Limited number of
studies examining
specific outcomes
C-39
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Evidence Integration
Studies and Summary and Key Factors that Increase Factors that Decrease Evidence Stream Summary Judgment
Interpretation Findings Certainty Certainty Judgment
infertility. In contrast, low
confidence studies
observed significantly
increased odds of
endometriosis (1/3) and
ovarian syndromes (2/3).
Other studies reported
non-significant positive
associations with
endometriosis (2/12).
Results from remaining
studies were inconsistent
and did not reach
significance.
Breastfeeding
Two medium confidence •
Medium confidence •
Limited number of
2 Medium confidence
cohort studies reported
studies
studies examining
studies
significant inverse •
Consistent direction
outcome
associations with
of effects
breastfeeding duration •
Precision of
(2/2).
findings
Anogenital distance
Two studies examined •
High and medium •
Limited number of
1 High confidence study
measures of anogenital
confidence studies
studies examining
1 Medium confidence
distance, including
outcome
study
anoclitoris and
•
Inconsistent direction
anofourchette distances, in
of effects
female infants. A high
confidence study reported
significant inverse
associations with
anoclitoris distance for the
highest exposure group
and in continuous
analysis. Results for
anofourchette distances
were inverse but not
C-40
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
significant. A medium
confidence study observed
non-significant mixed
results for both measures.
Evidence from In Vivo Animal Studies (Section C.1.2)
Female mating and
fertility
2 Medium confidence
studies
No effects on female
mating or fertility
parameters were observed
in one- and two-generation
reproduction studies in
rats with PFOS exposure
beginning six weeks prior
to mating (2/2).
Medium confidence •
studies
Consistent direction
of effects
Limited number of
studies examining
outcome
Female gestation length
3 Medium confidence
studies
Duration of gestation was •
slightly decreased in a
one-generation rat •
reproduction study and in
a two-generation rat study,
both with exposure
beginning six weeks prior
to mating (2/3). No effect
on gestation length was
observed in another rat
study with exposure
beginning on the first day
of gestation.
Medium confidence •
studies
Consistent direction
of effects •
Limited number of
studies examining
outcome
Small magnitude of
effect
Female reproductive
hormones
1 High confidence study
3 Medium confidence
studies
Significant alterations in •
female testosterone levels
were found (1/3). No •
significant changes in
serum E2 were found in
female monkeys exposed
for 26 weeks or in female
mice exposed in utero
from GDI-17. One mouse
High and medium •
confidence studies
Dose-response
relationship •
Limited number of
studies examining
specific outcomes
Changes in body
weight may limit
ability to interpret
these responses
©OO
Slight
Evidence is based on 10
high and medium
confidence studies. There
were no observed effects
on mating or fertility in
the only available two-
generation reproduction
study; however, other
studies observed effects on
length of gestation,
hormone levels, and
estrous cyclicity. Some of
the observed reproductive
effects (e.g., decreased
gestation length in female
rats, decreased prolactin-
family hormones in female
mice) may be secondary
" effects because they
occurred at doses that also
resulted in decreased
gestational body weight.
One study reported modest
increases in testosterone
concentrations in females,
but the response in the
highest dose was affected
Evidence Integration
Summary Judgment
C-41
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
study measured maternal
serum concentrations of
prolactin-family hormones
(i.e., mPL-II, mPLP-Ca,
mPLP-K) during
pregnancy and found
dose-dependent decreases
(1/1).
Estrous cyclicity and
ovarian function
1 High confidence study
3 Medium confidence
studies
No significant effect on
estrous cyclicity were
found in two rat studies
(2/2). However, a high
confidence study in rats
observed a dose-
dependent, but not
significant, increase in
estrous cycle length and
prolonged diestrus (1/1)
compared to controls. No
effects on the number and
distribution of corpora
lutea in the ovaries were
observed in pregnant rats
and rabbits (3/3).
High and medium
confidence studies
Limited number of
studies examining
specific outcomes
Small magnitude of
effect
by decreased body weight.
The increases in
testosterone were
accompanied by dose-
dependent increases in
estrous cycle length,
though this increase was
not statistically significant
and alterations in the
estrous cycle were not
observed in a second study
in female rats.
Female pubertal
No effects on age at •
Medium confidence •
Limited number of
development
vaginal opening were
studies
studies examining
3 Medium confidence
observed in reproduction •
Consistent direction
outcome
studies
and developmental studies
of effects
in rats (3/3).
Organ weights
No effects were observed •
High and medium •
Limited number of
2 High confidence
on absolute or relative
confidence studies
studies examining
studies
weights of the uterus (2/2) •
Consistent direction
outcome
1 Medium confidence
or ovaries (1/1).
of effects
study
C-42
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Evidence Integration
Studies and Summary and Key Factors that Increase Factors that Decrease Evidence Stream Summary Judgment
Interpretation Findings Certainty Certainty Judgment
Histopathology
No exposure-related •
High and medium •
Limited number of
2 High confidence
histopathologic^ findings
confidence studies
studies examining
studies
were reported for the •
Consistent direction
outcome
1 Medium confidence
ovaries (2/2), uterus (3/3),
of effects
study
vagina (2/2), or cervix
(1/1).
Notes: El = estrone; E3 = estriol; E2 = estradiol; FSH = follicle-stimulating hormones; 17-OHP = 17-hydroxyprogesterone; SES = socioeconomic status; DBP = diastolic blood
pressure; GD = gestation day; mPL-II= mouse placental lactogen II; mPLP-Ca = mouse prolactin-like protein-Ca; mPLP-K = mouse prolactin-like protein-K.
C-43
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.2 Endocrine
EPA identified 35 epidemiological and 14 animal studies that investigated the association
between PFOS and endocrine effects. Of the epidemiological studies, 4 were classified as high
confidence, 15 as medium confidence, 9 as low confidence, 4 as mixed (1 high/medium, 1
medium/low, 1 medium/uninformative, and 1 low/uninformative) confidence, and 3 were
considered uninformative (Section C.2.1). Of the animal studies, 1 was classified as high
confidence, 10 as medium confidence, 2 as low confidence, and 1 was mixed {medium/low)
(Section C.2.2). Studies may have multiple judgments depending on the endpoint evaluated.
Though low confidence studies are considered qualitatively in this section, they were not
considered quantitatively for the dose-response assessment (See Main PFOS Document).
C.2.1 Human Evidence Study Quality Evaluation and
Synthesis
C.2.1.1 Introduction
Thyroid disease is more common in females than in males and encompasses conditions such as
hypothyroidism and hyperthyroidism. Hypothyroidism is characterized by elevated thyroid
stimulating hormone (TSH) and concurrently low thyroxine (T4) concentrations, while
subclinical hypothyroidism is characterized by elevated TSH in conjunction with normal T4 and
triiodothyronine (T3) levels. Hyperthyroidism is characterized by elevated T4 and low TSH, and
subclinical hyperthyroidism is characterized by low levels of TSH with normal T4 and T3 levels.
The 2016 Health Advisory {U.S. EPA, 2016, 3982043} andHESD {U.S. EPA, 2016, 3603365}
reports identified evidence of endocrine effects of PFOS for thyroid disease, hypothyroidism,
and hypothyroxinemia. Occupational studies examining the relationship between PFOS exposure
and endocrine outcomes did not find any significant associations. Studies on NHANES
populations {Melzer, 2010, 129081 l;Wen, 2013, 2850943} reported associations between PFOS
exposure (serum PFOS concentrations) and thyroid disease. One study {Melzer, 2010, 1290811}
reported associations with thyroid disease in men, and another study {Wen, 2013, 2850943} saw
associations with subclinical hypothyroidism in men and women. In people without diagnosed
thyroid disease or without biomarkers of thyroid disease, thyroid hormones (i.e., TSH, T3 or T4)
show mixed effects across cohorts. In cross-sectional studies where thyroid hormones were
measured in association with serum PFOS, increased TSH was associated with PFOS exposure
in most cases {Berg, 2015, 2851002;Wang, 2013, 4241230;Webster, 2014, 2850208}. Increasing
PFOS was associated with increased T4 in children aged 1 to 17 years from the C8 cohort
{Lopez-Espinosa, 2012, 1291122}; however, PFOS was not associated with hypothyroidism. A
small South Korean study examining correlations between maternal PFAS during pregnancy and
fetal thyroid hormones in cord blood {Kim, 2011, 1424975} found an association for PFOS and
increased fetal TSH, as well as with decreased fetal T3. TSH was the outcome most frequently
associated with PFOS in studies of pregnant women. In studies of pregnant women, PFOS was
associated with increased TSH levels {Berg, 2015, 2851002;Wang, 2013, 4241230}. Pregnant
women testing positive for the anti-thyroid peroxidase (TPO) biomarker for autoimmune thyroid
disease showed a positive association with PFOS and TSH {Webster, 2014, 2850208}. A case-
control study of hypothyroxinemia (normal TSH and low free T4) in pregnant women {Chan,
2011, 1402500}, did not show associations of hypothyroxinemia with PFOS exposure.
C-44
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
For this updated review, 34 studies (35 publications)4 report on the association between PFOS
exposure and endocrine effects. Seven of the publications were studies in pregnant women
{Aimuzi, 2020, 6512125;Berg, 2017, 3350759;Dreyer, 2020, 6833676;Inoue, 2019,
5918599;Itoh, 2019, 5915990;Reardon, 2019, 5412435;Shah-Kulkarni, 2016, 3859821}, and the
remainder of the publications were on the general population. Different study designs were
utilized, including seven cohort studies {Berg, 2017, 3350759, Blake', 2018\
5080657;Crawford, 2017, 3859813;Kim, 2020, 6833758;Lebeaux, 2020, 6356361;Liu, 2018,
4238396;Reardon, 2019, 5412435}, seven cohort and cross-sectional studies {Dreyer, 2020,
6833676;Itoh, 2019, 5915990;Kato, 2016, 3981723;Preston, 2018, 4241056;Wang, 2014,
2850394;Xiao, 2019, 5918609}, one case-control study {Predieri, 2015, 3889874}, one case-
control and cross-sectional study {Zhang, 2018, 5079665}, and 19 cross-sectional studies
{Abraham, 2020, 6506041;Aimuzi, 2019, 5387078;Aimuzi, 2020, 6512125;Audet-Delage, 2013,
2149477;Byrne, 2018, 5079678;Caron-Beaudoin, 2019, 5097914;Dufour, 2018,
4354164;Heffernan, 2018, 5079713;Inoue, 2019, 5918599;Jain, 2013, 2168068;Jain, 2019,
6315816;Kang, 2018, 4937567;Khalil, 2018, 4238547;Lewis, 2015, 3749030;Li, 2017,
3856460;Seo, 2018, 4238334;Shah-Kulkarni, 2016, 3859821;Tsai, 2017, 3860107;van den
Dungen, 2017, 5080340;Yang, 2016, 3858535}. All observational studies measured PFOS in
blood components (i.e., blood, plasma, or serum). Six studies measured PFOS in cord blood
{Aimuzi, 2019, 5387078;Dufour, 2018, 4354164;Liu, 2020, 6569227;Shah-Kulkarni, 2016,
3859821;Tsai, 2017, 3860107;Yang, 2016, 3858535} and eight studies measured PFOS in
maternal blood or serum during pregnancy {Dreyer, 2020, 6833676;Kato, 2016,
3981723;Lebeaux, 2020, 6356361;Preston, 2018, 4241056;Reardon, 2019, 5412435;Wang,
2014, 2850394;Xiao, 2019, 5918609;Yang, 2016, 3858535}. The studies were conducted in
different study populations including populations from Belgium {Dufour, 2018, 4354164},
Canada {Caron-Beaudoin, 2019, 5097914;Reardon, 2019, 5412435}, China {Aimuzi, 2019,
5387078;Aimuzi, 2020, 6512125;Li, 2017, 3856460, Liu , 2020\ 6569227;Yang, 2016,
3858535;Zhang, 2018, 5079665}, Denmark {Dreyer, 2020, 6833676;Inoue, 2019,
5918599;Xiao, 2019, 5918609}, Germany {Abraham, 2020, 6506041}, Italy {Predieri, 2015,
3889874}, Japan {Itoh, 2019, 5915990;Kato, 2016, 3981723}, Republic of Korea {Kang, 2018,
4937567;Kim, 2020, 6833758;Shah-Kulkarni, 2016, 3859821}, Taiwan {Tsai, 2017,
3860107;Wang, 2014, 2850394}, the United Kingdom {Heffernan, 2018, 5079713}, and the
United States {Blake, 2018, 5080657;Byrne, 2018, 5079678;Crawford, 2017, 3859813;Jain,
2013, 2168068;Jain, 2019, 6315816;Khalil, 2018, 4238547;Lebeaux, 2020, 6356361;Lewis,
2015, 3749030;Liu, 2018, 4238396;Preston, 2018, 4241056}. Two studies {Itoh, 2019,
5915990;Kato, 2016, 3981723} belonged to the same cohort, the Hokkaido Study on the
Environment and Children's Health. While most studies evaluated the relationship between
exposure to PFOS and thyroid hormone concentrations, other endocrine outcomes were
investigated as well, including: thyroid disease, thyroid antibodies (thyroglobulin antibodies
(TgAb) and thyroid peroxidase antibody (TPOAb)), and thyroid hormone-associated proteins
(e.g., thyroglobulin, thyroxine-binding globulin).
C.2.1.2 Study Quality
Several considerations were specific to evaluating the quality of studies. First, timing of
exposure and hormone concentration measurements was important. Several studies on mother-
child dyads examined relationships between maternal serum PFOS measurements and thyroid
4 Itoh et al. (2019, 5915990) reports thyroid-related hormone levels in a population overlapping with Kato et al. (2016, 3981723).
C-45
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
hormones in both mothers (i.e., a cross-sectional analyses) and in cord blood or children's serum
(i.e., a longitudinal analyses). Longitudinal comparisons between maternal PFOS concentrations
measured during pregnancy and thyroid hormone levels in cord blood or the child's blood
attenuate any concerns for potential reverse causality. Measuring PFOS and thyroid hormone
concentrations concurrently in maternal serum was considered adequate in terms of exposure
assessment timing. Given the long half-life of PFOS (median half-life = 3.4 years) {Li, 2018,
4238434}, current blood concentrations are expected to correlate well with past exposures.
Second, timing of thyroid hormone assessment was a recurring concern due to the diurnal
variation in thyroid hormones. Thyroid hormone outcome misclassification due to timing of
blood collection is non-differential, however, study sensitivity may be impacted in cases where
timing of collection was uncontrolled.
There are 35 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and endocrine effects. Study quality evaluations for these 35 studies
are shown in Figure C-13 and Figure C-14.
Of the 35 studies identified since the 2016 assessment, 4 studies were classified as high
confidence, 15 as medium confidence, 9 as low confidence, 4 as mixed (1 high/medium, 1
medium/low, 1 medium/uninformative, and 1 low/uninformative) confidence, and 3 studies
{Abraham, 2020, 650604l;Predieri, 2015, 3889874} as uninformative.
C-46
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
«sd»0
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
^ Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
Khalil et al., 2018, 4238547-
—"—
+
l
+
— —
I
+
i
+
i
3
Kim et al., 2020, 6833758-
++
++
++
++
+
Lebeaux et al., 2020, 6356361 -
++
++
++
++
+
+
Lewis et al., 2015, 3749030 -
+
+
+
+
-
+
+
-
Li et al., 2017, 3856460-
-
+
-
-
+
-
+
-
Liu et al., 2018, 4238396-
-
+
++
+
+
+
+
+
Predieri et al., 2015, 3889874 -
+
+
-
-
+
+
-
~
Preston et al., 2018, 4241056 -
+
+
+
+
+
+
+
+
Reardon et al., 2019, 5412435-
+
+
-
+
+
+
+
+
Seo et al., 2018, 4238334-
-
+
-
-
-
-
~
Shah-Kulkarni et al., 2016, 3859821 -
+
+
+
+
+
+
+
+
Tsai et al., 2017, 3860107-
-
+
-
+
+
+
+
-
Wang et al., 2014, 2850394 -
+
+
+
+
D
+
+
+
Xiao et al., 2020, 5918609-
++
++
+
+
m
+
++
++
Yang et al., 2016, 3858535-
+
+
+
+
+
+
+
+
Zhang et al., 2018, 5079665-
-
-
-
+
+
+
+
-
van den Dungen et al., 2017, 5080340 -
-
+
+
-
+
+
-
-
Figure C-14. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Endocrine Effects (Continued)
Interactive figure and additional study details available on HAWC.
C-48
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The main concerns with low confidence and uninformative studies included a lack of
consideration for outcome sampling time, small sample sizes, or use of statistical methods that
did not account for confounding. Other studies rated as low or uninformative had issues
regarding the analysis, including a lack of accounting for population sampling methods {Lewis,
2015, 3749030}, or use of statistical methods that did not account for confounding {Abraham,
2020, 6506041}. Case-control studies {Kim, 2016, 3351917; Predieri, 2015, 3889874} were
rated uninformative and presented issues with insufficient detail regarding participant
recruitment and case definitions. However, the largest issues identified in these studies included
use of statistical methods that did not account for potential confounding factors, and the
sensitivity of both case-control studies was impacted by small sample sizes.
C.2.1.3 Findings from Children
One high confidence study {Kim, 2020, 6833758} observed an inverse association between
PFOS concentrations and subclinical hypothyroidism (defined by reference thyroid hormone
levels) at age six which was consistent after additional adjustment for dietary iodine intake. The
association was observed in boys, but not in girls. A positive association was also observed for
PFOS and T3 at six years of age which was significant among boys but not girls, before and after
adjustment for dietary iodine intake.
Thyroid hormone levels were examined in 19 studies {Abraham, 2020, 6506041;Aimuzi, 2019,
5387078;Caron-Beaudoin, 2019, 5097914;Dufour, 2018, 4354164;Itoh, 2019, 5915990;Kang,
2018, 4937567;Kato, 2016, 3981723;Khalil, 2018, 4238547;Kim, 2016, 3351917;Kim, 2020,
6833758;Lebeaux, 2020, 6356361;Predieri, 2015, 3889874;Preston, 2018, 4241056;Shah-
Kulkarni, 2016, 3859821;Tsai, 2017, 3860107;Wang, 2014, 2850394;Xiao, 2019,
5918609;Yang, 2016, 3858535} and five observed significant effects (Appendix D). One high
confidence study {Xiao, 2019, 5918609} on children from the Faroe Islands showed a large,
significant positive association between maternal third trimester PFOS concentrations and cord
serum TSH. The effect size for TSH was similar in both sexes, but was no longer significant in
female infants. Additionally, sex-stratified analyses showed positive associations between
maternal PFOS and the free thyroxine index (FTI) in cord serum for girls. A medium confidence
study {Kato, 2016, 3981723} on infants in Sapporo, Japan from the Hokkaido Study observed
positive associations with infant TSH which were consistent after stratifying by the infant's sex.
Analyses by quartile revealed a significant increasing trend (p for trend = 0.024) for infant TSH
and maternal blood. A related medium confidence study {Itoh, 2019, 5915990} of a separate
Japanese cohort from the same region also found a significant positive association between
maternal serum PFOS and TSH among boys. When stratifying by the mother's thyroid antibody
(TA) status, the effect remained among boys born to TA-negative mothers. No effect was seen in
TA-positive mothers, but the sample size was small (n = 48).
Other medium confidence cross-sectional studies in newborns {Aimuzi, 2019, 5387078} showed
significant inverse associations with TSH in single pollutant models. These associations
remained for girls after stratifying by sex. A significant positive association was observed for
free T3 (FT3) among this study sample, but a sensitivity analysis including only those infants
with detectable free FT3 concentrations was conducted due to the low detection rate.
Associations between PFOS and FT3 were no longer significant after removing participants with
non-detectable levels. A medium confidence study {Preston, 2018, 4241056} in infants did not
show significant associations in continuous analyses; however, a significant inverse association
C-49
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
was found for T4 among all infants in the highest PFOS exposure quartile and among boys in in
exposure quartile.
A study in Taiwan {Tsai, 2017, 3860107} found significant positive associations for TSH and
inverse associations for T4 in cord blood among the entire sample and among boys in continuous
analyses. Analyses by exposure quantiles (< 30th, 30th-59th, 60-89th, and > 90th percentile)
were consistent in the direction of effect, but only reached significance for each effect comparing
the highest PFOS exposure quantile to the reference in the overall population. A significant
effect was also seen among boys in the second quantile (30th-59th) for TSH. However, only
27% of the initially recruited population had available PFOS and thyroid measurements, and
reasons for missing data were not provided. This limited the sample size (n = 118) and raised
concern for potential selection bias, contributing to a low confidence rating.
C.2.1.4 Findings from Pregnant Women
Thyroid hormone levels were examined in six studies {Aimuzi, 2020, 6512125;Berg, 2017,
3350759;Inoue, 2019, 5918599;Itoh, 2019, 5915990;Reardon, 2019, 5412435;Shah-Kulkarni,
2016, 3859821} and five observed significant effects (Appendix D). One high confidence study
{Xiao, 2019, 5918609} observed a positive association between third trimester PFOS
concentrations and maternal TSH in mothers giving birth to girls. This association was not seen
in the analysis of the entire cohort or in mothers of boys only. A medium confidence study
{Reardon, 2019, 5412435} on a Canadian cohort of pregnant women investigated associations
between multiple PFOS isomers and thyroid hormones at several timepoints during and after
pregnancy. Accounting for all timepoints, a significant positive association was observed for
increasing branched PFOS concentrations and TSH. The same association was not observed for
linear PFOS, except at 3 months post-partum. In this study, the authors note linear PFOS
contributed to 69.0% of exposure concentrations while branched PFOS constituted only 31.0%.
Total PFOS exposure was not assessed. A medium confidence cross-sectional study {Preston,
2018, 4241056} observed a significant inverse association for maternal TSH among TPOAb-
positive mothers. One low confidence analysis {Kato, 2016, 3981723} of mothers in Sapporo,
Japan from the Hokkaido Study observed significant decreases for maternal TSH concentrations
with increasing serum PFOS, which were also observed after stratifying by the infant's sex.
Analyses by quartile confirmed this decreasing trend (p < 0.001). No significant effects were
observed in mothers from the other Hokkaido cohort {Itoh, 2019, 5915990}. Another low
confidence study {Berg, 2017, 3350759} from Norway showed positive associations between
maternal PFOS concentrations and TSH levels during the second trimester. Analysis by quartile
showed significant associations for the two highest exposure groups, suggesting a consistent
trend.
One cross-sectional study {Dufour, 2018, 4354164} on mother-child dyads showed evidence of
increased risk of hypothyroidism in mothers. Analysis by quartile showed a consistent effect, but
only reached significance for mothers in the third PFOS exposure quartile. This study contained
a great deal of uncertainty regarding timing of outcome ascertainment and the method of disease
classification which diminish confidence in the findings for maternal hypothyroidism.
One high confidence study examined adrenal hormones among pregnant women in the OCC and
showed a significant decrease in diurnal urinary (dU) -cortisone and increase in dU-
C-50
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
cortisol/cortisone with two-fold increases in serum PFOS concentrations {Dreyer, 2020,
6833676}. However, dU- and serum Cortisol showed non-significant decreases.
C.2.1.5 Findings from the General Adult Population
Thyroid function was examined in 13 studies {Audt-Delage, 2013, 2149477; Blake, 2018,
5080657; Byrne, 2018, 5079678; Christensen, 2016, 3350721; Crawford, 2017, 3859813; Jain,
2013, 2168068; Jain and Ducatman, 2019, 6315816; Lebeaux, 2020, 6356361; Lewis, 2015,
3749030; Li, 2017, 3856460; Liu, 2018, 4238396; Seo, 2018, 4238334; van denDungen, 2017,
5080340; Zhang, 2018, 5079665} and six observed significant effects (Appendix D). A medium
confidence study {Blake, 2018, 5080657} in individuals residing near a uranium processing
facility in an area with PFAS-contaminated drinking water (Fernald Community Cohort (FCC))
reported a positive association for TSH in whole study sample. Stratifying by sex showed a
difference in direction of effect between men and women, however, the interaction term did not
reach significance (sex interaction p-value = 0.12). In men, the association for TSH was
consistent and was accompanied by a significant inverse association with total T4; no significant
associations were observed for women.
Results were mixed in three overlapping NHANES studies {Jain, 2013, 2168068;Jain, 2019,
6315816;Lewis, 2015, 3749030}. One low confidence study {Lewis, 2015, 3749030} showed
several significant and borderline significant results among NHANES (2011-2012) participants.
Significant positive associations were found between TSH in males (12-20 years old) and
females (20-40 years old), but other results were not consistent among the same stratified groups
(by sex and age). There is no evidence that the NHANES complex sampling design was
accounted for in the analysis which contributed to a low confidence rating. Jain (2013, 2168068),
another low confidence study, did not find any significant effects among NHANES (2007-2008)
participants. Kmedium confidence follow-up study {Jain, 2019, 6315816} examined effects on
thyroid hormones stratified by GF stage in a pooled NHANES dataset (2007-2012). A
significant effect was found for total T4 in those individuals with stage 3 A GF, the second most
severe stage. Associations for total T4 among other stages were non-significant and inconsistent
in direction of effect.
One additional cross-sectional study {Byrne, 2018, 5079678} of Alaska Natives found a
significant sex interaction for free T3. Women showed a positive association between serum
PFOS and free T3 while an inverse association was found in men. Borderline significant inverse
associations for TSH and total T3 were also observed among men (p = 0.085 and p = 0.08,
respectively). The sensitivity of the study, however, was limited by the population size (total
n = 85; male n = 38) and resulted in a low confidence rating. Another low confidence study {Li,
2017, 3856460} conducted in China found significant associations for TSH, free T3, and free T4
among a population oversampled for thyroid conditions (70%). Inverse associations were
observed for free T3 and free T4, while a positive association was found for TSH amongst the
whole population. Associations were not significant when stratified by thyroid disease state (i.e.,
normal, hypothyroidism, Hashimoto's disease). The study was found to be low confidence due to
missing information on recruitment and participation, especially considering this was a
convenience sample. Additionally, there were concerns for selective reporting and residual
confounding because individuals (n = 202) varied greatly by age (1 month to 90 years) and
lifestyle factors were not addressed.
C-51
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A case-control study {Zhang, 2018, 5079665} examined women with and without POI and
observed positive associations for TSH among both cases and controls. Additionally, inverse
associations were found among cases for free T3 and free T4. The thyroid hormone
concentrations were within normal ranges in both cases and controls. The study was rated as low
confidence due to insufficient information on control recruitment and potential for reverse
causation from irregular menstruation (a PFOS elimination route) for those women with PCOS.
C.2.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 4 studies from the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} and 10 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOS HESD that investigated the association between PFOS and endocrine effects. Study
quality evaluations for these 14 studies are shown in Figure C-15.
C-52
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
G°^V^ v#"! i
Conleyetal., 2022, 10176381 -
Curran et al., 2008, 757871
Dong et al., 2011, 1424949
Fuentes et al., 2006, 757859 - +
Fuentes et al., 2007, 757865 -
Lau et al., 2003, 757854
Li et al., 2016, 3981495
Luebker et al., 2005, 757857
NTP, 2019, 5400978-
Pereiro et al., 2014, 2230732 -
+
+
NR
+
+
+
+*
Salgado-Freiria et al., 2018, 5079767 -
+
+
NR
+
+
+
++
a
++
+
Seacat et al., 2002, 757853 -
++
+
NR
+
+
+
++
++
++
+
Thomford, 2002, 5432419-
+
-
NR
++ ++
-
++
++
++
-
Zhang et al., 2020, 6315674-
-
+
NR
+
-
-
+
+
-
-
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
S Critically deficient (metric) or Uninformative (overall)
Not reported
* Multiple judgments exist
Figure C-15. Summary of Study Evaluation for Toxicology Studies of PFOS and Endocrine
Effects
Interactive figure and additional study details available on HAWC.
Animal studies suggest that exposure to PFOS can result in adverse effects to the endocrine
system. Overall, studies of varying durations in rodent models and a single study in cynomolgus
monkeys {Seacat, 2002, 757853} have reported reductions in endocrine hormone levels and
C-53
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
changes in endocrine organ weights. There are insufficient data to support non-neoplastic lesions
(histopathology), and potential neoplastic lesions (see PFOS Main Document). Moreover,
reductions were observed in thyroid hormone levels, including total and free thyroxine (TT4 and
FT4) and total and free triiodothyronine (TT3 and FT3) {Luebker, 2005, 757857; NTP, 2019,
5400978; Lau, 2003, 757854}, as well as reductions in adrenocorticotropic hormone (ACTH),
corticosterone, and/or corticotropin releasing hormone (CRH) {Pereiro, 2014, 2230732; Salgado-
Freiria, 2018, 5079767}. Absolute and relative adrenal gland weights were reduced in rats {NTP,
2019, 5400978}, however adrenal glands subject to histopathologic examination appeared
normal {Chang, 2009, 757876; Luebker, 2005, 757857; Pereiro, 2014, 2230732} (see PFOS
Main Document).
C.2.2.1 Thyroid and Thyroid-Related Hormone Levels
Several 28-day studies provide evidence that exposure to PFOS can result in adverse effects on
rat thyroid hormone levels (Table C-3). Male and female rats were fed PFOS at doses of 0, 2, 20,
50, or 100 ppm (equivalent to 0, 0.14, 1.33, 3.21, or 6.34 mg/kg/day in males and 0, 0.15, 1.43,
3.73, or 7.58 mg/kg/day in females) for 28-days {Curran, 2008, 757871}. In both males and
females, serum TT4 levels were significantly reduced at doses of >20 ppm. Serum TT3 was
decreased at the 100 ppm and >50 ppm dose groups in males and females, respectively {Curran,
2008, 757871}. In another study in rats, male and female Sprague-Dawley rats were exposed to
PFOS at doses of 0 mg/kg/day, 0.312 mg/kg/day, 0.625 mg/kg/day, 1.25 mg/kg/day,
2.5 mg/kg/day, or 5 mg/kg/day via oral gavage {NTP, 2019, 5400978}. At study termination,
TT4 and FT4 levels were decreased in all male and female dose groups. In addition, TT3 was
significantly decreased in males and females treated with >0.625 mg/kg/day. No treatment-
related effects were seen on TSH levels {NTP, 2019, 5400978}. Yu et al. (2009, 757872)
exposed male Sprague-Dawley rats to 0 mg/L, 1.7 mg/L, 5.0 mg/L, or 15.0 mg/L PFOS in
drinking water for 91 days (drinking water consumption was not reported). Significant dose-
dependent reductions in TT4 were noted in animals treated at > 1.7 mg/L; however, FT4 was
only decreased in the 5.0 mg/L group. A statistically significant increase in TT3 was observed in
the 1.7 mg/L dose group, though TT3 in the 5 mg/L and 15 mg/L groups returned to control
levels. No treatment-related effects were seen in TSH {Yu, 2009, 757872}.
A number of reproductive/developmental studies investigated the effect of PFOS on thyroid
hormone production in parental and Fi rodents (Table C-3).
Lau et al. (2003, 757854) analyzed thyroid hormones in offspring of pregnant rats exposed by
gavage to PFOS at 0 mg/kg/day, 1 mg/kg/day, 2 mg/kg/day, or 3 mg/kg/day from GD 2-GD 21.
The authors reported statistically significant reductions in TT4 and FT4 on PND 5 in rat pups
treated with 2 mg/kg/day and 3 mg/kg/day during gestation. Signs of recovery in TT4 were noted
at weaning, while reduced FT4 persisted through PND 35. No effects were noted in serum TT3
nor TSH of pups when compared to controls {Lau, 2003, 757854}. In a cross-fostering study
conducted by Yu et al. (2009, 757880), pregnant Wistar rats were fed a diet containing
0 mg/kg/day or 3.2 mg/kg/day PFOS throughout gestation and/or lactation. PFOS-exposed
groups consisted of pups treated with PFOS during gestation only, pups treated with PFOS
during lactation only, and pups treated with PFOS during gestation and lactation. Pups in all
exposure groups had significant decreases in TT4 on PND 21 and PND 35. In contrast, TT3 and
reverse T3 (rT3) were not affected with PFOS exposure in rat pups {Yu, 2009, 757880}.
Another study measured serum TSH in pups and dams (GD 20, PND 4, or PND 21) following
C-54
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
oral gavage exposure of pregnant Sprague-Dawley rats to PFOS (0 mg/kg/day, 0.1 mg/kg/day,
0.3 mg/kg/day, or 1.0 mg/kg/day) from GD 0-PND 20. No statistically significant effects were
observed in dams or offspring at any timepoint assayed {Chang, 2009, 757876}.
Luebker et al. (2005, 757857) exposed pregnant Female Crl:CD®(SD)IGS VAF/Plus rats to
0.4 mg/kg/day, 0.8 mg/kg/day, 1.0 mg/kg/day, 1.2 mg/kg/day, 1.6 mg/kg/day, or 2.0 mg/kg/day
for 42 days prior to mating through LD 4. Exposed dams showed decreased TT4 and TT3 at
doses > 0.4 mg/kg/day and >1.2 mg/kg/day, respectively, although no perturbations were seen in
TSH or FT4 levels. In the pups, no perturbations were noted in TT3, FT4, or TSH, however, TT4
was reduced at doses ranging from 0.4 mg/kg/day-1.6 mg/kg/day (2.0 mg/kg/day group not
assessed due to high pup mortality). The authors noted that the contributions of prenatal vs.
postnatal effects of PFOS on thyroid hormones were not clear {Luebker, 2005, 757857}. The
authors also conducted follow-up analyses due to potential for negative bias from immeasurable
levels of FT3 and FT4 using equilibrium dialysis-radioimmunoassay (ED-RIA) methods and
measurements of TT3 and TT4 with chemiluminometric methods to ensure the validity of their
initial radioimmunoassay (RIA)-based results. While the ED-RIA reference method indicated
potential bias in the results for FT4 in pups, a true comparison could not be made due to
insufficient sample sizes {Luebker, 2005, 757857}. Conley et al. (2022, 10176381) also
determined levels of thyroid hormones in maternal serum following gestational exposure to
PFOS. The authors reported TT3 and TT4 on GD 18 in Sprague Dawley rats exposed to PFOS at
0 mg/kg/day, 0.1 mg/kg/day, 0.3 mg/kg/day, 1 mg/kg/day, 3, 10, or 30 mg/kg/day from GD 14-
GD 18. PFOS significantly reduced TT3 and TT4 at 10 and 30 mg/kg/day. Non-significant
decreases ranging from 7%-34% in TT3 and 3%-24% in TT4 were observed in dams exposed to
doses below 10 mg/kg/day. Fuentes et al. (2006, 757859) examined the effects of PFOS on
thyroid hormones in CD1 mice. The dams were exposed during gestation from GD 6-GD 18 to
0 mg/kg/day, 1.5 mg/kg/day, 3 mg/kg/day, or 6 mg/kg/day. At GD 18, the dams had an overall
percent reduction ranging from 11-57% in TT3, 36%-57% in FT3, and 42%-67% in FT4.
Conversely, increases in TT4 levels ranged from 158%—188%. Nonetheless, the differences
between the exposed and control dams were not statistically significant due to high variability.
Only one study was included that investigated the effects of PFOS exposure on hormone levels
during development in mice. Lau et al. (2003, 757854) exposed pregnant CD-I mice to
0 mg/kg/day, 1 mg/kg/day, 5 mg/kg/day, 10 mg/kg/day, 15 mg/kg/day, or 20 mg/kg/day PFOS
from GD 1-GD 17 and evaluated TT4 in sera of pooled mouse pups of each sex at several
timepoints across postnatal development. Due to mortality in the 15 mg/kg/day and
20 mg/kg/day groups, TT4 was only assessed in the 1 mg/kg/day, 5 mg/kg/day, and
10 mg/kg/day groups. TT4 levels varied across the different time points with different trends
based on treatment group. On PND 7, PND 14, and PND 28 there was a general trend for
decreased TT4 in the 5 mg/kg/day and 10 mg/kg/day exposure groups when compared to control
animals {Lau, 2003, 757854}. However, this was not observed on PND 3 or PND 21. Results
were not significant at any time point but may be limited by small sample size (3-7
determinations per group).
Male and female cynomolgus monkeys (4-6/sex/group) were orally exposed to PFOS at doses of
0 mg/kg/day, 0.03 mg/kg/day, 0.15 mg/kg/day, or 0.75 mg/kg/day for 182 days {Seacat, 2002,
757853}. Recovery animals from the 0 mg/kg/day, 0.15 mg/kg/day, and 0.75 mg/kg/day dose
groups were then monitored for an additional year. On the last day of dosing (day 182), thyroid
C-55
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
hormone levels, including TSH, TT3, and TT4 were evaluated. In males, TT3 was significantly
reduced across all dose groups while TSH and TT4 remained unaffected. In females, significant
reductions in TT3 were noted in animals treated with 0.15 mg/kg/day and 0.75 mg/kg/day.
Significant reductions in TT4 were noted in the mid-dose group only (0.15 mg/kg/day). TSH
remained unaffected in females. Sixty-one days after cessation of treatment there was still a trend
for decreased TT3 in 0.15 mg/kg/day males and 0.75 mg/kg/day males and females. Because
there were only 2 animals per group at this time, statistical analyses were not appropriate. TT4
and TSH results were not reported in the recovery period {Seacat, 2002, 757853}.
C-56
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table C-3. Summary of Results for Thyroid and Thyroid-Related Hormones in Toxicological Studies Following Exposure to
PFOS
Study Name
Species
Study Design
Life Stage Sex
Dose (mg/kg/day)
Value (ju.g/d L)"
Percent Change
Total Thyroxine (TT4)
Seacat et al. (2002, Cynomolgus
Chronic (26 wk)
Adult M
0
4.38 ±0.61
NA
757853)b
Monkey
0.03
4.72 ±0.68
7.8
0.15
3.99 ±0.62
-8.9
0.75
5.34 ± 1.57
21.9
F
0
5.66 ±0.89
NA
0.03
4.33 ± 1.46
-23.5
0.15
3.91 ±0.62*
-30.9
0.75
5.61 ± 1.00
-0.9
Fuentes et al.
CD-I Mice
Developmental
Po Adult (GD 18) F
0
0.50 ±0.13
NA
(2006, 757859)°
(GD 6-18)
1.5
1.29 ±0.59
158
3
1.41 ±0.39
182
6
1.44 ±0.57
188
Conley et al.
Sprague-Dawley
Developmental
Po Adult (GD 18) F
0
3.27 ±0.83
NA
(2022, 10176381)°
(GD 14-18)
0.1
2.49 ±0.43
-24
0.3
2.42 ±0.35
-26
1
3.18 ±0.95
-3
3
2.49 ±0.42
-24
10
1.67 ±0.47
-49
30
1.04 ±0.50
-68
Lau et al. (2003,
CD-I Mice
Developmental
Fi Pups (PND 28) M/F
0
4.2 ±0.9
NA
757854)c,d
(GD 1-17)
1
3.8 ±0.5
-9.5
5
3.6 ±0.5
-14.3
10
3.5 ±0.3
-16.7
Curran et al.
Sprague-Dawley
Short-term (28 d)
Adult M
0
6.27 ±0.92
NA
(2008, 75787l)b
Rats
0.14
5.19 ± 1.14
-17.3
1.33
1.11 ±0.32*
-82.3
3.21
1.00 ±0.21*
-84.1
6.34
1.03 ±0.20*
-83.6
F
0
2.92 ± 1.19
NA
0.15
2.51 ±0.81
-14.1
1.43
1.52 ±0.19*
-48.0
3.73
1.17 ± 0.15*
-60.1
C-57
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Name
Species
Study Design
Life Stage Sex
Dose (mg/kg/day)
Value (ju.g/d L)"
Percent Change
7.58
1.27 ±0.36*
-56.5
NTP (2019,
Sprague-Dawley
Short-term (28 d)
Adult M
0
3.51 ±0.3
NA
5400978)°
Rats
0.312
1.33 ±0.19*
-62.1
0.625
0.53 ±0.09*
-84.9
1.25
0.26 ±0.07*
-92.6
2.5
0.22 ±0.04*
-93.7
5
0.48 ±0.07*
-86.3
F
0
2.21 ±0.24
NA
0.312
1.11 ± 0.12*
-49.8
0.625
0.55 ±0.07*
-75.1
1.25
0.33 ±0.07*
-85.1
2.5
0.35 ±0.09*
-84.2
5
0.38 ±0.05*
-82.8
Yu et al. (2009,
Sprague-Dawley
Subchronic (91 d)
Adult M
0
4.09 ±0.18
NA
757872)°
Rats
0.0017
2.39 ±0.13*
-41.6
0.005
1.64 ±0.54*
-59.9
0.015
0.85 ±0.16*
-79.2
Lau et al. (2003,
Sprague-Dawley
Developmental
Fi Adult (PND 35) M/F
0
4.3 ±0.5
NA
757854)°,d
Rats
(GD 2-21)
1
3 ±0.2
-30.2
2
2.5 ±0.2*
-41.9
3
2 ± 0.1*
-53.5
Luebker et al.
Crl:CD®(SD)IGS
Reproductive
Po Adult (LD 5) F
0.0
1.5 ±0.63
NA
(2005, 757857)b
VAF/Plus® Rats
(80 d (42 d pre-
0.4
0.81 ±0.41*
-46.0
mating, GD 0-21,
0.8
0.6 ±0.44*
-60.0
LD 1-4))
1.0
0.73 ± 0.24*
-51.3
1.2
0.28 ±0.32*
-81.3
1.6
0.27 ±0.17*
-82.0
2.0
0.24 ±0.15*
-84.0
Fi Pups M/F
0.0
0.54 ±0.22
NA
(PND 5)e
0.4
0±0
-100.0
0.8
0±0
-100.0
1.0
0.02 ±0.05
-96.3
1.2
0.01 ±0.02
-98.1
1.6
0.01 ±0.0
-98.1
C-58
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Name
Species
Study Design
Life Stage
Sex
Dose (mg/kg/day)
Value (ju.g/d L)"
Percent Change
2.0
_f
-
Fi Pups
M/F
0.0
2.1 ±0.6
NA
(PND5)g
0.4
0.8
1.0
1.2
1.6
2.0
1.6 ±0.4
1.5 ±0.7
1.5 ±0.5
-23.8
-28.6
-28.6
Yu et al. (2009,
Wistar Rats
Reproductive
Fi Pups (PND 14)
M/F
0
6.78 ±0.35
NA
757880)c,h
(GD 0-PND 35)
3.2
(Gestation Only)
3.2
(Lactation Only)
3.2
(Gestation & Lactation)
6.36 ±0.25
5.97 ±0.39
4.29 ±0.17*
-6.2
-11.9
-36.7
Fi Pups (PND 21)
M/F
0
3.2
(Gestation Only)
3.2
(Lactation Only)
3.2
(Gestation & Lactation)
5.81 ± 0.31
4.63 ± 0.27*
4.15 ±0.26*
4.38 ±0.24*
NA
7.9
-3.3
2.1
Fi Pups (PND 35)
M/F
0
3.2
(Gestation Only)
3.2
(Lactation Only)
3.2
(Gestation & Lactation)
6.75 ±0.35
5.44 ±0.33*
4.33 ±0.30*
4.23 ± 0.22*
NA
-19.4
-35.9
-37.3
Free Thyroxine (FT4)
NTP (2019,
5400978)°
Sprague-Dawley
Rats
Short-term (28 d) Adult
M
0
0.312
0.625
1.25
2.5
0.00253 ± 0.00022 NA
0.00095 ±0.0001* -62.5
0.00047 ± 0.00005* -81.4
0.0004 ± 0.00002* -84.2
0.00036 ± 0.00005* -85.8
C-59
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Name Species Study Design Life Stage Sex Dose (mg/kg/day) Value (ju.g/d L)" Percent Change
5 0.00033 ±0.00001* -87.0
F 0 0.00174 ±0.00023 NA
0.312 0.00107 ±0.00009* -38.5
0.625 0.0007 ±0.00003* -59.8
1.25 0.00064 ± 0.00005* -63.2
2.5 0.00056 ± 0.00005* -67.8
5 0.00048 ±0.00003* -72.4
Yu et al. (2009, Sprague-Dawley Subchronic (91 d) Adult M 0 1.9 ±0.13 NA
757872)° Rats 0.0017 1.67 ±0.14 -12.1
0.005 1.26 ±0.15* -33.7
0.015 1.73 ± 0.11 -8.9
Fuentesetal. CDlMice Developmental Po Adult (GD 18) F 0 0.078 ±0.038 NA
(2006,757859)° (GD 6-18) 1.5 0.045 ±0.007 -42%
3 0.060 ±0.011 -23%
6 0.026 ±0.014 -67%
Lau et al. (2003, Sprague-Dawley Developmental Fi Adult (PND 35) M/F 0 0.02 ±0.002 NA
757854)° d Rats (GD2-21) 1 0.014 ±0.000 -30.0
2 0.009 ±0.001 -55.0
3 0.011 ±0.001 -45.0
Luebker et al. Crl:CD®(SD)IGS Reproductive P0 Adult F 0.0 0.00236 ± 0.00061 NA
(2005,757857)b VAF/Plus® Rats (80 d (42 d pre- (LD 5) 0.4 0.00212 ± 0.00058 -10.2
mating, GD 0-21, 0 8 0.00261 ± 0.00056 10.6
LD !"4)) 1.0
1.2 0.00248 ± 0.00022 5.1
1.6 0.00259 ± 0.00082 9.7
2X) - -
FiPups M/F 0.0 0.0019 ±0.0009 NA
(PND 5)
0.4 0.0013 ±0.0004 -31.6
0.8
1.0
1.2
1.6
2.0
C-60
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Name
Species
Study Design
Life Stage Sex Dose (mg/kg/day)
Value (ju.g/d L)"
Percent Change
Free Triiodothyronine (FT3)
Fuentes et al.
(2006, 757859)°
CD1 Mice
Developmental
(GD 6-18)
Po Adult (GD 18) F
0
1.5
3
6
0.014 ±0.003
0.009 ±0.001
0.006 ±0.001
0.008 ±0.003
NA
-36
-57
-43
Luebker et al.
(2005, 757857)b
Crl:CD®(SD)IGS
VAF/Plus® Rats
Reproductive
(80 d (42 d pre-
mating, GD 0-21,
LD 1-4))
Fi Pups M/F
(LD 5)
0.0
0.4
0.8
1.0
1.2
1.6
2.0
0.00019 ±0.00002
0.0002 ± 0.00003
0.000151
0.00018 ±0.00006
NA
5.3
-21.1
-5.3
Total Triiodothyronine (TT3)
Fuentes et al.
(2006, 757859)°
CD1 Mice
Developmental
(GD 6-18)
Po Adult (GD 18) F
0
1.5
3
6
0.105 ±0.034
0.045 ± 0.002
0.051 ±0.008
0.093 ±0.017
NA
-57
-51
-11
Conley et al.
(2022, 10176381)°
Sprague-Dawley
Developmental
(GD 14-18)
Po Adult (GD 18) F
0
0.1
0.3
1
3
10
30
0.106 ±0.013
0.082 ±0.016
0.070 ±0.001
0.099 ±0.022
0.079 ±0.014
0.069 ±0.015*
0.040 ± 0.006*
NA
-23
-34
-7
-25
-35
-62
Seacat et al. (2002, Cynomolgus
757853)b Monkey
Chronic (26 wk)
Adult M
0
0.03
0.15
0.75
0.16 ±0.007
0.119 ± 0.031*
0.125 ±0.015*
0.066 ± 0.027*
NA
-25.6
-21.9
-58.8
F
0
0.03
0.15
0.75
0.135 ±0.031
0.12 ±0.024
0.097 ± 0.008*
0.085 ±0.012*
NA
-11.1
-28.1
-37.0
Curran et al.,
2008, 757871b
Sprague-Dawley
Rats
Short-term (28 d)
Adult M
0
0.14
1.33
10.39 ±2.14
11.75 ± 1.23
8.83 ± 1.69
NA
13.1
-15.0
C-61
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Name
Species
Study Design
Life Stage Sex
Dose (mg/kg/day)
Value (ju.g/d L)"
Percent Change
3.21
8.38 ±8.38
-19.4
6.34
7.86 ± 1.49*
-24.4
F
0
11.88 ± 1.10
NA
0.15
11.17 ± 0.91
-6.0
1.43
11.36 ± 1.75
-4.4
3.73
9.15 ± 1.43*
-23.0
7.58
8.25 ± 1.30*
-30.6
NTP (2019,
Sprague-Dawley
Short-term (28 d)
Adult M
0
0.08737 ±0.00532
NA
5400978)°
Rats
0.312
0.07781 ±0.00544
-10.9
0.625
0.06063 ± 0.00464*
-30.6
1.25
0.0575 ± 0.00267*
-34.2
2.5
0.05535 ± 0.00275*
-36.6
5
0.051*
-42.8
F
0
0.09305 ±0.00504
NA
0.312
0.0814 ±0.00302
-12.5
0.625
0.07252 ± 0.00427*
-22.1
1.25
0.0692 ±0.00363*
-25.6
2.5
0.06203 ±0.00178*
-33.3
5
0.05157 ±0.00143*
-44.6
Yu et al. (2009,
Sprague-Dawley
Subchronic (91 d)
Adult M
0
0.029 ± 0.004
NA
757872)°
Rats
0.0017
0.048 ± 0.008*
65.5
0.005
0.023 ±0.005
-20.7
0.015
0.023 ±0.003
-20.7
Lau et al. (2003,
Sprague-Dawley
Developmental
Fi Adult (PND 35) M/F
0
0.08 ±0.00
NA
757854)°-d
Rats
(GD 2-21)
1
0.09 ±0.00
12.5
2
0.09 ±0.01
12.5
3
0.11±0.01
37.5
Luebker et al.
Crl:CD®(SD)IGS
Reproductive
Po Adult F
0.0
0.0760 ±0.0185
NA
(2005, 757857)b
VAF/Plus® Rats
(80 d (42 d pre-
(LD 5)
0.4
0.0729 ±0.0135
-4.1
mating, GD 0-21,
0.8
0.0638 ± 0.00668
-16.1
LD 1-4))
1.0
0.0624 ±0.0132
-17.9
1.2
0.0529 ±0.015*
-30.4
1.6
0.0470 ± 0.020*
-38.2
2.0
0.0533 ±0.0173*
-29.9
C-62
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Name
Species
Study Design
Life Stage
Sex
Dose (mg/kg/day)
Value (ju.g/d L)"
Percent Change
Fi Pups
M/F
0.0
0.054 ±0.018
NA
(PND 5)e
0.4
0.8
1.0
1.2
1.6
2.0
0.056 ±0.019
0.049 ±0.018
0.048 ±0.009
0.045 ± 0.022
0.033 ±0.008
0.033 ±0.012
3.7
-9.3
-11.1
-16.7
-38.9
-38.9
Fi Pups
M/F
0.0
0.0424 ± 0.0057
NA
(PND 5)s
0.4
0.8
1.0
1.2
1.6
2.0
0.0362 ± 0.0062
0.031
0.03 ±0*
-14.6
-29.2
-29.2
Yu et al. (2009,
Wistar Rats
Reproductive
Fi Pups
M/F
0
0.057 ±0.004
NA
757880)c,h
(GD 0-PND 35)
(PND 14)
3.2
(Gestation Only)
3.2
(Lactation Only)
3.2
(Gestation & Lactation)
0.052 ±0.004
0.051 ±0.003
0.043 ±0.003
-8.8
-10.5
-24.6
Fi Pups
M/F
0
0.058 ±0.003
NA
(PND 21)
3.2
(Gestation Only)
3.2
(Lactation Only)
3.2
(Gestation & Lactation)
0.065 ± 0.007
0.058 ±0.004
0.059 ±0.003
12.1
0.0
1.7
Fi Pups
M/F
0
0.059 ±0.003
NA
(PND 35)
3.2
(Gestation Only)
3.2
(Lactation Only)
3.2
0.052 ±0.003
0.049 ± 0.004
0.055 ±0.002
-11.9
-16.9
-6.8
(Gestation & Lactation)
C-63
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Name
Species
Study Design
Life Stage Sex
Dose (mg/kg/day)
Value (ju.g/d L)"
Percent Change
Reverse Triiodothyronine (rT3)
Yu et al. (2009,
757880)c,h
Wistar Rats
Reproductive
(GD 0-PND 35)
Fi Pups M/F
(PND 14)
0
3.2
(Gestation Only)
3.2
(Lactation Only)
3.2
(Gestation & Lactation)
-
-
Fi Pups M/F
(PND 21)
0
3.2
(Gestation Only)
3.2
(Lactation Only)
3.2
(Gestation & Lactation)
0.0251
0.025 ±0.003
0.029 ±0.001
0.025 ± 0.002
NA
0.0
16.0
0.0
Fi Pups M/F
(PND 35)
0
3.2
(Gestation Only)
3.2
(Lactation Only)
3.2
(Gestation & Lactation)
0.02 ± 0.002
0.02 ± 0.002
0.015 ±0.000
0.02 ±0.001
NA
0.0
-25.0
0.0
Thyroid Stimulating Hormone (TSH)
Seacat et al. (2002, Cynomolgus
757853)b Monkey
Chronic (26 wk)
Adult M
0
0.03
0.15
0.75
0.43 ± 0.52J
0.34 ± 0.3J
0.74 ± 0.75J
0.93 ± 0.57J
NA
-20.9
72.1
116.3
F
0
0.03
0.15
0.75
0.73 ± 1.12J
0.68 ± 0.82J
1.27 ± 1.52J
0.84 ± 0.79J
NA
-6.8
74.0
15.1
NTP (2019,
5400978)°
Sprague-Dawley
Rats
Short-term (28 d)
Adult M
0
0.312
0.625
1.25
2.039 ±0.14
1.494 ±0.174
1.479 ±0.12
2.333 ±0.294
NA
-26.7
-27.5
14.4
C-64
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Name Species Study Design Life Stage Sex Dose (mg/kg/day) Value (ju.g/d L)" Percent Change
25 2.419 ±0.338 18^6
5 1.890 ±0.239 -7.3
F 0 1.286 ±0.073 NA
0.312 1.476 ±0.088 14.8
0.625 1.276 ±0.085 -0.8
1.25 1.325 ±0.115 3.0
2.5 1.4914 ±0.195 16.0
5 1.536 ±0.073 19.4
Yu et al. (2009, Sprague-Dawley Subchronic (91 d) Adult M 0 0.072 ±0.030 NA
757872)° Rats 0.0017 0.067 ± 0.027 -6.9
0.005 0.112 ±0.034 55.6
0.015 0.162 ±0.067 125.0
Chang et al. (2009, Sprague-Dawley Developmental Po Adult F 0 1.304 ±0.102 NA
757876)c d Rats (GD 0-PND 20) (GD 20) 0.1 1.202 ±0.096 -7.8
0.3 1.061 ±0.058 -18.6
1 1.1 ±0.077 -15.6
Po Adult (PND 4) F 0 1.036 ±0.115 NA
0.1 1.119 ± 0.121 8.0
0.3 0.863 ±0.032 -19.3
1 1.023 ±0.083 -1.3
Po Adult F 0 1.714 ±0.205 NA
(PND 21) 0.1 1.758 ±0.166 2.6
0.3 1.483 ±0.128 -13.5
1 1.95 ±0.198 13.8
FiPups M 0 0.765 ±0.060 NA
(PND 21) 0.1 0.994 ±0.089 29.93
0.3 0.949 ±0.080 24.05
1 0.880 ±0.045 15.03
FiPups F 0 0.880 ±0.06 NA
(PND 21) 0.1 0.889 ±0.074 1.0
0.3 0.865 ±0.07 -1.7
1 0.840 ±0.065 -4.5
FiPups M/F 0 1.212 ±0.134 NA
(GD 20) 0.1 1.053 ±0.08 -13.1
C-65
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Name
Species
Study Design
Life Stage
Sex
Dose (mg/kg/day)
Value (ju.g/d L)"
Percent Change
0.3
1
0.934 ±0.075
0.969 ±0.075
-22.9
-20.0
Fi Pups
(PND 4)
M/F
0
0.1
0.3
1
0.557 ±0.065
0.552 ±0.02
0.477 ± 0.07
0.542 ±0.06
NA
-0.9
-14.4
-2.7
Lau et al. (2003,
757854)c,d
Sprague-Dawley
Rats
Developmental
(GD 2-21)
Fi Adult
(PND 35)
M/F
0
1
2
3
0.62 ±0.08
0.73 ±0.16
0.65 ±0.06
0.29 ±0.02
NA
17.7
4.8
-53.2
Luebker et al.
(2005, 757857)b
Crl:CD®(SD)IGS
VAF/Plus® Rats
Reproductive
(80 d (42 d pre-
mating, GD 0-21,
LD 1-4))
Po Adult
(LD 5)
F
0.0
0.4
0.8
1.0
1.2
1.6
2.0
0.163 ±0.096
0.114 ±0.023
0.144 ±0.092
0.111 ±0.052
0.145 ±0.103
0.167 ±0.077
0.153 ±0.068
NA
-30.1
-11.7
-31.9
-11.0
2.5
-6.1
Fi Pups
(PND 5)
M/F
0.0
0.4
0.8
1.0
1.2
1.6
2.0
0.102 ±0.017
NA
0.2361
0.101 ±0.025
0.145 ±0.034*
0.151
131.4
-1.0
42.2
47.1
NTP (2019,
pg/mL FT4;
ng/mL rT3);
Notes: F = female; Fi = first generation; GD = gestation day; LD = lactation day; M = male; NA = not applicable; Po = parental generation; PND = postnatal day
* Statistically significant atp <0.05.
a Values were converted to ng/dL for Seacat et al. (2002, 757853) (ng/dL TT3, FT3, FT4; uU/mL TSH); Curran et al. (2008, 757871) (nmol/L T4; nmol/L TT3);
5400978) (ng/dL FT4, ng/dL TT3; ng/mL TSH); Yu et al. (2009, 757872) (ng/L TT4; ng/L FT4; ng/L TT3; ng/L TSH); Lau et al. (2003, 757854) (ng/mL TT4;
ng/mL TT3; ng/mL TSH); Luebker et al. (2005, 757857) (ng/dL FT4; pg/mL FT3; ng/dL TT3; ng/mL TSH); Yu et al. (2009, 757880) (ng/mL TT4; ng/mL TT3:
Chang et al. (2009, 757876) (ng/mL TSH); Conley et al. (2022, 10176381) (ng/mL TT3, TT4); Fuentes et al. (2006, 757859) (ng/dL TT3, FT3, FT4).
b Data are presented as mean ± standard deviation.
c Data are presented as mean ± standard error.
d Values were estimated from a figure using a digital ruler.
e Analyzed by analog radioimmunoassay (RIA).
fInsufficient sample for analysis.
g Analyzed by analog chemiluminometric assay (CL).
h Cross-foster study.
C-66
-------
DRAFT FOR PUBLIC COMMENT
1 n = 1.
J Units in (iU/mL.
MARCH 2023
C-67
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.2.2.2 Hypothalamic; Pituitary, and/or Adrenal Hormone Levels
Effects of PFOS exposure on hormones of the hypothalamus, pituitary gland, and adrenals were
available in two rat studies conducted by the same laboratory (Figure C-16). Salgado-Freiria et
al. (2018, 5079767) and Pereiro et al. (2014, 2230732) investigated the effect of PFOS exposure
on hypothalamic CRH, ACTH, and corticosterone of male Sprague-Dawley rats treated at
0 mg/kg/day, 0.5 mg/kg/day, 3.0 mg/kg/day, and 6.0 mg/kg/day for 28 days. Following
exposure, decreases in serum CRH and corticosterone concentrations in all dose groups were
observed, but there was no dose-related trend. However, a dose-dependent decrease in ACTH
was observed. In a reproductive/developmental study, pregnant Sprague-Dawley rats were
administered 0 mg/kg/day, 5 mg/kg/day, and 20 mg/kg/day from GD 12-GD 18 via gavage {Li,
2016, 3981495}. Fetal serum corticosterone levels were significantly increased in animals treated
with 5 mg/kg/day and 20 mg/kg/day.
Three studies in mice have examined the effects of PFOS exposure on serum corticosterone
{Fuentes, 2006, 757859; Fuentes, 2007, 757865; Dong, 2011, 1424949}. Fuentes et al. (2006,
757859) observed 1% and 5% decreases at 1.5 mg/kg and 6 mg/kg respectively; and an 8%
increase at 3 mg/kg indicating there was no dose-related trend in pregnant CD1 mice. Dose-
dependent increases of approximately 20% and 50% were recorded in male CD1 mice following
a 4-week exposure to 3 or 6 mg/kg/day PFOS {Fuentes, 2007, 757865}. In male C57BL/6 mice
exposed to 0 mg/kg/day, 0.008 mg/kg/day, 0.017 mg/kg/day, 0.083 mg/kg/day, 0.417 mg/kg/day,
or 0.833 mg/kg/day over the course of 60 days, serum corticosterone decreased by 2%, 13%, and
17% at 0.008, 0.017, and 0.083 mg/kg/day (low doses) and increased by 2% and 19% at
0.417 mg/kg/day and 0.833 mg/kg (high doses), indicating a biphasic dose response trend
{Dong, 2011, 1424949}. Although the changes in serum corticosterone seem to be related to
exposure, they were not statistically significant, likely due to variability.
PFOS Endocrine Effects-Ad renal Hormone Levels
Endpoint Study Name Study Design Observation Time Animal Description Dose (mgfkg/day) | # Statistically significant 0 Not statistically significant! 195% CI
Adrenocorticotropic Hormone (ACTH) Salgado-Freiria etal., 2018, 5079767 short-term (28d) 29d Rat. Sprague-Dawley (e1, N=8) 0
h-m-i
0.5
3
1
1
1
6
1
Corticosterone Fuentes et al., 20D6, 757859 developmental (GD6-1B) GD18 Mouse, CD-1 (¦}, N=10-11) 0
I + I 1
1.5
| | 1
3
h
1
6
1- •
1
Fuentes etal., 2007,757865 short-term (4wk) 39d Mouse, CD-1 ( C, N=10) 0
i
1 1 1
3
• n
6
• 1
Dong etal., 2011, 1424949 subchronic (60d) 60d Mouse. C57BL/6 N=6) 0
1 • 1
0.008
0.017
0.083
—i
i—
i—•—
1
|
0.417
i
I 1 1
0.833
• 1 1
Salgado-Freiria etal., 2018, 5079767 short-term (28d) 29d Rat, Sprague-Dawley (-. N=8) 0
i—(
I 1
0.5
i—•—i
1
|
1
i • i
1
3
i—•—i
1
6
i • i
1
Corticotropin Releasing Hormone (CRH) Salgado-Freiria ot al., 2018. 5079767 short-term (28d) 29d Rat. Sprague-Dawley N=8) 0
*¦ 1
0.5
a
-
©
©
h«H
1
1
1
1
1
0
0 -60 -40 -20 0 20 40 60 80 100 120 U
Percent control response 1%)
Figure C-16. Percent Change in Adrenal Hormones Relative to Controls in Rodents
Following Exposure to PFOSa'b
C-68
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Interactive figure and additional study details available on HAWC.
ACTH = adenocorticotropic hormone; CRH = corticotropin releasing hormone; CI = confidence interval.
aPereiro et al. (2014,2230732) reported on the same data as Salgado-Freiria et al. (2018, 5079767) and is not shown in the figure.
bThe red dashed lines indicate a 100% increase from the control response.
C.2.2.3 Organ Weights
No adverse effects on male and female thyroid weights (Table C-4) were noted in the previously
mentioned NTP study {NTP, 2019, 5400978}. In a longer-term study conducted by Yu et al.
(2009, 757872), no treatment related effects were observed on absolute and relative thyroid
weights in Sprague-Dawley rats exposed to PFOS in drinking water at doses of 0 mg/L,
1.7 mg/L, 5.0 mg/L, or 15 mg/L for 91 days {Yu, 2009, 757872}. However, in Sprague Dawley
rats exposed to 2 mg/kg/day, 20 mg/kg/day, 50 mg/kg/day, or 100 mg/kg/day in the diet for
28 days, relative thyroid weight was significantly increased in females and males in the highest
dose group. No treatment related effects were observed on absolute thyroid weight or thyroid
weight relative to brain weight {Curran, 2008, 757871}.
PFOS exposure was associated with changes in adrenal gland weighs in rats and non-human
primates (Table C-4). In Sprague Dawley rats, absolute right adrenal gland weights in male rats
were reduced at doses > 1.25 mg/kg/day. No effects were observed in females {NTP, 2019,
5400978}. No effects were observed in relative adrenal weights at any dose for either sex after
28 days of exposure to 0 mg/kg/day-5 mg/kg/day via gavage {NTP, 2019, 5400978}.
Additionally, relative adrenal gland weight was decreased in male rats treated at doses of
> 0.5 mg/kg/day for 28 days {Pereiro, 2014, 2230732}. Curran et al. (2008, 757871) observed
significant trends towards increased adrenal gland weight relative to body weights and increased
adrenal gland weight relative to brain weights in male and female Sprague Dawley rats exposed
to 0 mg/kg/day, 2 mg/kg/day, 20 mg/kg/day, 50 mg/kg/day, or 100 mg/kg/day PFOS for 28 days.
Seacat et al. (2002, 757853) measured absolute and relative adrenal weights in male cynomolgus
monkeys exposed to PFOS at doses of 0 mg/kg/day, 0.03 mg/kg/day, 0.15 mg/kg/day, or
0.75 mg/kg/day for 182 days. The only significant treatment related effect was an increase in left
adrenal-to-body weight percentages in males of the high dose group {Seacat, 2002, 757853}. No
studies were available evaluating the effect of PFOS exposure on mouse organ weights.
Effects on the relative weight of the hypothalamus were observed by Salgado et al. (2015,
3981583) (see PFOS Main Document).
Table C-4. Associations Between PFOS Exposure and Endocrine Organ Weights in
Rodents and Non-human Primates
Endpoint
Study Name
Species
Exposure
Length
Dose (mg/kg/day)
Sex
Change
Adrenal
Weight, Right,
Absolute
NTP (2019, 5400978)
Sprague-
Dawley rat
28 days
0,0.312,0.625, 1.25,
2.5, 5 mg/kg/day
M
4 1.25-
5.0 mg/kg/da
y
F
n.s.
Adrenal
Weight, Right,
Relative
NTP (2019, 5400978)
Sprague-
Dawley rat
28 days
0,0.312,0.625, 1.25,
2.5, 5 mg/kg/day
M
n.s.
F
n.s.
C-69
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Endpoint
Study Name
Species
Exposure
Length
Dose (mg/kg/day)
Sex
Change
Adrenal
Curran et al. (2008,
Sprague
28 days
0,0.14, 1.33,3.21,
M
n.s.
Weight
757871)
Dawley rat
6.34 mg/kg/day
Absolute
0,0.15, 1.43,3.73,
7.58 mg/kg/day
F
t 1.43
mg/kg/day
Adrenal
Curran et al. (2008,
Sprague
28 days
0,0.14, 1.33,3.21,
M
n.s.
Weight,
757871)
Dawley rat
6.34 mg/kg/day
Relative to
Body Weight
0,0.15, 1.43,3.73,
7.58 mg/kg/day
F
t 3.73
mg/kg/day
Adrenal
Curran et al. (2008,
Sprague
28 days
0,0.14, 1.33,3.21,
M
n.s.
Weight,
757871)
Dawley rat
6.34 mg/kg/day
Relative to
Brain Weight
0,0.15, 1.43,3.73,
7.58 mg/kg/day
F
t 1.43
mg/kg/day
Adrenal
Pereiro et al. (2014,
Sprague-
28 days
0,0.5, 1,3,
M
4 0.5-6
Weight,
2230732)
Dawley rat
6 mg/kg/day
mg/kg/day
Relative
Adrenal
Seacat et al. (2002,
Cynomolgus
182 days
0,0.03,0.15,
M
t
Weight, Left,
757853)
monkeys
0.75 mg/kg/day
0.75 mg/kg/
Relative to
day
Body Weight
F
n.s.
Adrenal
Seacat et al. (2002,
Cynomolgus
182 days
0,0.03,0.15,
M
n.s.
Weight, Left,
757853)
monkeys
0.75 mg/kg/day
Relative to
F
Brain Weight
Thyroid
NTP (2019, 5400978)
Sprague-
28 days
0,0.312,0.625, 1.25,
M
n.s.
Weight,
Dawley rat
2.5, 5 mg/kg/day
Absolute
F
n.s.
Curran et al. (2008,
Sprague
28 days
0, 2, 20, 50,
M
n.s
757871)
Dawley rat
100 mg/kg/day
Yu et al. (2009,
Sprague-
91 days
0, 1.7, 5.0, or
M
n.s.
757872)
Dawley rat
15 mg/L
Thyroid
NTP (2019, 5400978)
Sprague-
28 days
0,0.312,0.625, 1.25,
M
n.s.
Weight,
Dawley rat
2.5, 5 mg/kg/day
Relative to
F
n.s.
Body Weight
Curran et al. (2008,
757871)
Sprague
Dawley rat
28 days
0, 2, 20, 50,
100 mg/kg/day
M
n.s.
Yu et al. (2009,
Sprague-
91 days
0, 1.7, 5.0, or
M
n.s.
757872)
Dawley rat
15 mg/L
Seacat et al. (2002,
Cynomolgus
182 days
0,0.03,0.15,
M
n.s.
757853)
monkeys
0.75 mg/kg/day
F
n.s.
Thyroid
Curran et al. (2008,
Sprague
28 days
0, 2, 20, 50,
M
n.s.
weight,
757871)
Dawley rat
100 mg/kg/day
C-70
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Endpoint
Study Name
Species
^Length6 D()sc (mg/kg/day) Sex Change
Relative to
Brain Weight
Notes: F = female; M = male; n.s. = nonsignificant.
C.2.2.4 Histopathology
Few histological and morphometric abnormalities were observed in fetal and neonatal thyroid
glands in Sprague-Dawley rats that were orally administered PFOS at doses of 0 or 1 mg/kg/day
from GD 0-PND 20 {Chang, 2009, 757876}. On GD 20, female fetuses had a significantly
higher number of thyroid follicular epithelial cells compared to controls (2.1-fold increase); the
number of follicular epithelial cells were not statistically different from controls in male fetuses.
No other treatment-related histologic changes in number of follicles present and the distribution
of follicle sizes were observed in fetuses at GD 20 or in neonates at PND 4 or PND 21 {Chang,
2009, 757876}. Luebker et al. (2005, 757857) examined the thyroid gland of one male and
female Crl:CD®(SD)IGS VAF/Plus pup exposed to 2 mg/kg/day (highest dose group) PFOS
through LD4. No microscopic changes were noted {Luebker, 2005, 757857}.
Pereiro et al. (2014, 2230732) examined the effect of oral PFOS exposure on the adrenal cortex
of male Sprague-Dawley rats treated with 0 mg/kg/day, 0.5 mg/kg/day, 1.0, 3.0 and
6.0 mg/kg/day for 28 days. Fasculata zona cells appeared more activated (presenting spongy
cytoplasm due to the presence of liposomes) in animals treated with PFOS when compared with
control animals. However, incidence data of non-neoplastic lesions and statistical analysis were
not reported/conducted {Pereiro, 2014, 2230732}. In contrast, NTP (2019, 5400978) did not
observe histopathological changes in the thyroid, adrenal, or pituitary glands of male or female
rats dosed with up to 5 mg/kg/day PFOS for 28 days.
In male and female cynomolgus monkeys orally exposed to PFOS at doses of 0 mg/kg/day,
0.03 mg/kg/day, 0.15 mg/kg/day, or 0.75 mg/kg/day for 182 days, no treatment related effect on
cell proliferation of the pancreas was observed {Seacat, 2002, 757853}.
C.23 Mechanistic Evidence
Mechanistic evidence linking PFOS exposure to adverse endocrine outcomes is discussed in
Sections 3.2.5, 3.3.2, 3.3.6, and 3.4.1.5 of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365}.
There are 29 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD that investigated the mechanisms of action of PFOS that
lead to endocrine effects. A summary of these studies is shown in Figure C-17. Additional
mechanistic synthesis will not be conducted since evidence suggests but is not sufficient to infer
that PFOS leads to endocrine effects.
C-71
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mechanistic Pathway
Animal
In Vitro
Grand Total
Big Data, Non-Targeted Analysis
2
0
2
Cell Growth, Differentiation, Proliferation, Or Viability
3
12
15
Cell Signaling Or Signal Transduction
2
6
S
Extracellular Matrix Or Molecules
0
1
1
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
1
2
3
Hormone Function
8
13
20
Inflammation And Immune Response
0
1
1
Oxidative Stress
3
0
3
Xenobiotic Metabolism
1
1
2
Other
0
2
2
Not Applicabla'Mot Specified/Review Article
1
0
1
Grand Total
12
18
29
Figure C-17. Summary of Mechanistic Studies of PFOS and Endocrine Effects
Interactive figure and additional study details available on Tableau.
C.2.4 Evidence Integration
There is slight evidence for an association between PFOS exposure and endocrine effects in
humans based on studies reporting positive associations for TSH in children and adults. The
2016 HESD for PFOS {U.S. EPA, 2016, 3603365} included two studies reporting positive
associations with thyroid disease in NHANES participants. In this updated review, further
evidence on the relationship between PFOS and thyroid disease was limited to two studies, one
of which reported an inverse association in children {Kim, 2020, 6833758} and the other was
classified as uninformative. The most consistent effects were for TSH in children. Three medium
confidence studies {Xiao, 2019, 5918609; Kato, 2016, 3981723; Itoh, 2019, 5915990} reported
elevated TSH among infants with increasing PFOS exposure, but other studies found the
opposite effect {Aimuzi, 2019, 5387078}. General population studies in adults also suggested a
positive association between PFOS exposure and TSH, but results were limited to one medium
confidence study, while the rest were low confidence. Interestingly, two general population
studies identified seemingly sexually dimorphic effects for TSH {Blake, 2019, 5080657} and T3
{Byrne, 2019, 5079678}. The 2016 Health Assessment included three studies reporting positive
associations between serum PFOS and TSH in pregnant women. In the recent literature, one high
and one medium confidence study reported positive association, while there was inconsistent
C-72
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
evidence in low confidence studies. Additional uncertainty exists due to the potential for
confounding by other PFAS. One study {Aimuzi, 2019, 5387078} on infants reported
correlations across PFAS (i.e., PFOA, PFNA, PFDA, perfluoroundecanoic acid (PFUnDA),
PFHxS, and PFDoA) and found them to be moderately correlated (r = 0.37-82). Results for
PFOS were not significant, however, the direction and magnitude of effect were similar in
single-pollutant and multi-pollutant models.
The animal evidence for an association between PFOS exposure and effects in the endocrine
system is considered moderate based on effects from 13 high or medium confidence studies.
Decreases in free T4, total T4, and total T3 were observed in rats, mice, and monkeys after PFOS
exposure; however, a compensatory increase in TSH was not reported, nor was there evidence of
thyroid gland histopathology, which is consistent with findings of hypothyroxinemia. Although
evidence of thyroid hormone disruption in humans is inconsistent, EPA concluded that the
sensitive and consistent changes in thyroid hormone levels in multiple animal models indicate
toxicity of relevance to humans.
Reductions in ACTH, corticosterone, and CRH in studies with animal models suggest that
exposure to PFOS may interfere with the hypothalamic-pituitary-adrenal axis. However, changes
in adrenal weights were inconsistent among studies and among species. More data on the
interactions between corticosterone and ACTH are required, as well as potential histological
effects in the adrenal gland, to understand the relevance of an effect of PFOS on adrenocortical
hormone levels.
C.2.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOS exposure has the potential to cause endocrine effects in
humans under relevant exposure circumstances (Table C-5). This conclusion is based primarily
on evidence from animal models showing alterations in circulating thyroid and adrenocortical
hormone levels following exposure to doses as low as 0.03 mg/kg/day PFOS. Although a few
associations between PFOS exposure and TSH were observed in medium confidence
epidemiological studies, there is considerable uncertainty in the results due to inconsistency
across studies and limited number of studies.
C-73
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-5. Evidence Profile Table for PFOS Endocrine Effects
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.2.1)
Thyroid and thryoid-
related hormones and
thyroid disease
3 High confidence
studies
17 Medium confidence
studies
10 Low confidence
studies
In adults, findings
indicated significantly
increased levels of the
thyroid-related hormone
TSH (2/11); however, one
of the studies was of low
confidence. Findings for
thyroid hormones (i.e., T3
and T4) were generally
inconsistent across studies,
and considerable
differences were observed
by sex within studies. TSH
was significantly
increased among children
in three studies (3/19),
including a high
confidence study.
However, other studies
reported inverse
associations for TSH,
including one significant
finding. Findings for free
T4 in children were
mixed, but significant
decreases (2/6) in T4 and
significant increases in T3
(2/6) were reported. Two
studies in pregnant women
(2/3) reported non-
significant positive
associations for free T4
and free T3.
High and medium •
confidence studies •
Low confidence studies
Inconsistent direction
of effect in adults
which may be
influenced by timing of
outcome sampling (i.e.,
diurnal variations)
Lmprecision of findings
©oo
Slight
Evidence for endocrine
effects is based on
increased TSH in adults,
decreased T4 in children,
and increased T3 in
children. Findings from
medium confidence
studies were frequently
inconsistent or imprecise.
There was limited
evidence reporting effects
on thyroid disease.
Uncertainties remain
regarding diurnal variation
of thyroid hormones,
differential effects in
males and females, and
consistency across
outcome timing.
©OO
Evidence Suggests
Primary basis:
Animal evidence
demonstrated alterations in
circulating thyroid and
adrenocortical hormone
levels. Although a few
associations between PFOS
exposure and TSH were
observed in medium
confidence epidemiological
studies, there is
considerable uncertainty in
the results due to
inconsistencies across
studies and the limited
number of studies.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
C-74
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Thyroid hormone
antibodies
2 Medium confidence
studies
Findings for thyroid
hormone antibodies were
generally imprecise,
however, hormone
antibody (i.e., TPOAb-
negative) status was
reported to play a role in
the association between
exposure and TSH levels
in male children.
Medium confidence •
studies
Limited number of
studies examining
outcome
Imprecision of findings
Steroid and adrenal
hormones
1 High confidence study
One study reported
decreases in diurnal
urinary cortisone among
pregnant women, and the
diurnal urinary
cortisol/cortisone ratio
was correspondingly
increased.
High confidence
study
Limited number of
studies examining
outcome
Evidence from In Vivo Animal Studies (Section C.2.2)
Thyroid and thyroid-
related hormones
1 High confidence study
7 Medium confidence
studies
Reductions in total T4,
free T4, and/or total T3
was observed following
short-term and
developmental exposure in
male and female rodents
(5/7) and chronic exposure
in male and female non-
human primates (1/1). No
significant change in TSH
levels was reported in rats,
mice, or non-human
primates (4/4).
High and medium
confidence studies
Coherent changes
across thyroid
hormone levels
Consistent findings
across species, sex,
and study design
Dose-response
relationship
observed for free
T4, total T4, and
total T3
Contributions of
prenatal versus
postnatal exposure to
PFOS on thyroid
hormones unclear
Adrenocortical
hormones
Mixed effects on •
corticosterone levels were
High and medium
confidence studies
No factors noted
0©O
Moderate
Evidence was based on
high and medium
confidence studies that
demonstrated decreased
thyroid hormone levels
(free T4, total T4, total
T3). A compensatory
increase in TSH was not
reported, nor was there
evidence of thyroid gland
" histopathology, which is
C-75
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
5 Medium confidence
studies3
observed in rodent studies
but most reported no
significant changes (3/5).
A dose-dependent
decrease in ACTH and a
non-monatomic decrease
in CRH were reported in
male rats (1/1).
consistent with findings of
hypothyroxinemia.
Organ weights
1 High confidence study
3 Medium confidence
studies
In rodents, absolute (1/2)
and relative (1/3) adrenal
gland weights were
decreased in males while
absolute (1/2) and relative
(1/2) adrenal gland
weights were increased in
females following a 28-
day exposure in rats. One
chronic study in non-
human primates reported
an increase in relative
adrenal weights in males
(1/1). No significant
changes were observed in
absolute or relative thyroid
gland weight (4/4).
High and medium
confidence studies
Inconsistent direction
of effect in organ
weights across studies
Limited number of
studies examining
outcomes
Histopathology
1 High confidence study
2 Medium confidence
studies
No significant effects were
observed in incidence of
non-neoplastic lesions in
the thyroid gland, adrenal
gland, and/or pituitary
gland following exposure
to male and female mice,
rats, and non-human
primates (3/3).
High and medium
confidence studies
Limited number of
studies examining
outcomes
C-76
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Notes: TSH = thyroid stimulating hormone; T3 = triiodothyronine; T4 = thyroxine; TPOAb = thyroid peroxidase antibody; ACTH = adrenocorticotropic hormone; CRH =
corticotropin releasing hormone.
"Pereiro et al. (2014, 2230732) reported on the same data as Salgado-Freiria et al. (2018, 5079767) for adrenocortical hormone measurements.
C-77
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.3 Metabolic/Systemic
EPA identified 69 epidemiological and 29 animal studies that investigated the association
between PFOS and systemic and metabolic effects. Of the epidemiological studies, 10 were
classified as high confidence, 36 as medium confidence, 14 as low confidence, 5 as mixed (4
medium/low and 1 medium/uninformative) confidence, and 4 were considered uninformative
(Section C.3.1). Of the animal studies, 3 were classified as high confidence, 20 as medium
confidence, 5 as low confidence, and 1 was considered mixed ( 6.5%, fasting plasma
glucose >126 mg/dL, a 2-hour plasma glucose > 127 in an oral glucose tolerance test, or a
random plasma glucose > 200 mg/dL (in patients with classic symptoms of hyperglycemia or a
hyperglycemic crisis).
Metabolic syndrome is a combination of medical disorders and risk factors that increase the risk
of developing cardiovascular disease (CVD) and diabetes, including abnormalities in
triglycerides, waist circumference, blood pressure, cholesterol, and fasting glucose. It is highly
prevalent in the general population of the United States. Risk factors for metabolic syndrome
include insulin resistance and being overweight or obese.
The 2016 EPA Health Assessment for PFOS concluded that there is no evidence of an
association with metabolic syndrome. One study observed an association with gestational
diabetes {Zhang, 2015, 2857764}, but no associations were observed with type 1 or type 2
diabetes. Among adults, serum PFOS was significantly associated with increased beta cell
function. Serum PFOS concentration was not associated with metabolic syndrome, glucose
concentration, homeostasis model of insulin resistance (HOMA-IR), or insulin levels in adults or
adolescents {Lin, 2009, 1290820}. Another study reported no association with metabolic
syndrome or glucose homeostasis parameters {Fisher, 2013, 2919156}. Overall, these studies
show a lack of association of PFOS with diabetes, metabolic syndrome, and related outcomes.
For this updated review, 69 new epidemiologic studies examined the association between PFOS
and metabolic outcomes. Of these, 32 were cohort studies, six were case-control studies, 27 were
cross-sectional studies, two were nested case-control studies, and two were controlled trials.
C-78
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Most studies measured exposure to PFOS using biomarkers in blood. Di Nisio et al. (2019,
5080655) measured exposure to PFOS using biomarkers in blood and in semen) Shapiro et al.
(2016, 3201206) measured the exposure to PFOS in urine. Biomarkers in maternal blood were
used in 16 studies and cord blood was used in 2 studies. Most studies identified were conducted
in the United States and China. Other study locations included Canada, Croatia, Denmark
(including the Faroe Islands), France, Italy, Japan, Korea, Norway, Spain, Sweden, Taiwan, the
Netherlands, and the United Kingdom.
Twenty-two studies examined diabetes (1 in children, 9 in pregnant women), and four examined
metabolic syndrome in general adult populations. Other metabolic outcomes examined included
blood glucose levels or glucose tolerance, HbAlc, insulin or insulinogenic index, insulin
resistance, insulin sensitivity, adiponectin, leptin, beta cell function, proinsulin, insulin-like
factor 1, c-peptide, body mass index (BMI) or ponderal index, body weight, gestational weight
gain, body fat, and anthropometric measurements (Appendix D).
C.3.1.2 Study Quality
Several criteria were specific to evaluating the quality of studies on metabolic outcomes. Due to
concerns for potential reverse causality (where the exposure may be affected by disease status),
studies evaluating diabetes were considered critically deficient if exposure and prevalent diabetes
were measured concurrently, since the cross-sectional design would not allow for a reliable
characterization of exposure before the onset of diabetes. Another concern is for the evaluation
of insulin, homeostasis model assessment of beta-cell function (HOMA-B), or HOMA-IR
without consideration of diabetes status, as the treatment of diabetes, particularly in those being
treated with hypoglycemic medications, influences insulin production and secretion.
There are 69 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and metabolic effects. Study quality evaluations for these 69 studies
are shown in Figure C-18, Figure C-19, and Figure C-20.
Based on the considerations mentioned, 10 studies were classified as high confidence, 36 as
medium, 14 as low confidence, and 4 as uninformative for all metabolic outcomes. Five studies
have split ratings and were classified as medium confidence for one outcome and low confidence
for other outcomes). One study {Liu, 2018, 4238396} was considered uninformative for insulin
resistance and medium confidence for other metabolic outcomes. Uninformative studies had
critical deficiencies in at least one domain. These deficiencies included a lack of control for
confounding {Predieri 2015, 3889874; Huang, 2018, 5024212; Jiang, 2014, 2850910}, lack of
fasting measures for glucose measurements {Jiang, 2014, 2850910}, and treating PFOS as an
outcome instead of an exposure, which limits the ability to make causal inference for the purpose
of hazard determination {Predieri, 2015, 3889874; Jain 2020, 6833623}. Other concerns leading
to an uninformative rating included inadequate reporting of population selection {Jiang, 2014,
2850910}, small sample size, and narrow ranges for exposure {Predieri, 2015, 3889874}.
The most common reason for a low confidence rating was potential for residual confounding,
particularly by SES {Christensen, 2016, 3858533; Fassler, 2019, 6315820; Heffernen, 2018,
5079713; Koshy, 2017, 4238478; Lin, 2013, 2850967; Convertino, 2018, 5080342; Khalil, 2018,
4238547}, by adiposity {Lin, 2013, 2850967}, by age {Koshy, 2017, 4238478}, or by diabetes
C-79
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
status {Lind, 2014, 2215376}. Low confidence studies presented concerns with the outcome
measures including potential for outcome misclassification {Christensen, 2016, 3858533; He,
2018, 4238388; Zong, 2016, 3350666}, failing to account for diabetes status {Lind, 2014,
2215376} or use of medications that would impact insulin levels or beta-cell function {He, 2018,
4238388; Fleisch, 2017, 3858513}, analytical methods {Koshy, 2017, 4238478}, and failure to
establish temporality between PFOS exposure and diabetes {Lind, 2014, 2215376}. Other
concerns included selection bias {Fassler, 2019, 6315820; van Den Dungen, 2017, 5080340},
which resulted from self-selection {Christensen, 2016, 3858533}, failure to provide information
on control group selection {Heffernan, 2018, 5079713}, or differential recruitment for cases and
controls {Lin, 2013, 2850967}. Small sample size was also a concern in some studies
{Christensen, 2016, 3858533; Heffernan, 2018, 5079713; Khalil, 2018, 4238547; van den
Dungen, 2017, 5080340}. In the evidence synthesis below, high, and medium confidence studies
were the focus, although low confidence studies were still considered for consistency in the
direction of association.
C-80
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
0<\0^
AS ***%& °°^
9^\iSf9 V"V**V5*^ 0^
Alderete et al.
Ashley-Martin et al.
Ashley-Martin et al.
Blake et al.
Legend
| Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Q Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
Figure C-18. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Metabolic Effects
Interactive figure and additional study details available on HAWC.
C-81
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
5 o>
fit***"
He etal., 2018, 4238388
Huang et al., 2018. 5024212
Jain et al., 2019, 5080621
Jensen et al., 2018, 4354143
Jensen et al., 2020, 6833719
Jiang etal., 2014, 2850910
Kang et al., 2018, 4937567
Karlsen et al., 2017, 3858520
Koshy et al., 2017, 4238478
Lauritzen et al., 2018, 4217244
Lopez-Espinosa et al., 2016, 3859832
Mancini et al., 2018, 5079710 ¦
£
-*
+
-
+
+
~
+
-*
+
+
~
++
++
+
-
+
+
+
-
+
•1
+
++
+
+
+
++
++
D
+
+
+
+
-
+
+
+
+
+
J
-
+
D
+
-
+
J
+
++
+
+
El
+
+
+
+
++
1
'A
-
¦
+
-
-
+
+
+
+
+
+
+
+
+
+*
+
+
++
+
+
+
3
-
+
+
-
•
+
++
D
+
+
+
+
+
+
+*
+
+
+
-
+
++
+
+
++
+
+
+
-
+
+
+
+
+
-
+
+
+*
~
+*
+
+
-
+
+
2
+
+
+
+
£
-
+
1
+
+
-
+
-
+
+
++
+
D
++
++
-
+
+
+
+
+
+
+
+
+
++
+
*
+
+
+
Legend
| Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
^ Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
Figure C-19. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Metabolic Effects (Continued)
Interactive figure and additional study details available on HAWC.
C-82
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Manzano-Salgado et al.,
Marks et al.,
Martinsson et al.,
Matilla-Santander et al.,
Minatoya et al.,
Mitro et al.,
2017, 4238509
2019, 5381534-
2020, 6311645-
2017, 4238432-
2017, 3981691
2020, 6833625
Mora et al.,2017, 3859823-
Predieri et al„ 2015, 3889874
Preston et al., 2020, 6833657
Rahman et al., 2019, 5024206
Ren et al., 2020, 6833646-
Shapiroet al., 2016, 3201206
Starling et al., 2017, 3858473
Su et al., 2016, 3860116-
Sun et al., 2018, 4241053
I " WM I + I M
Tian et al.,
Valvi et al.,
Wang et al.,
Wang et al.,
Xu et al.,
Yang et al.,
Zong et al.,
van den Dungen et al.,
2019, 5080586-
2017, 3983872
2018, 5079666-
2018, 5080352-
2020, 6833677
2018, 4238462
2016, 3350666
2017, 5080340
Legend
I Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
Figure C-20. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Metabolic Effects (Continued)
Interactive figure and additional study details available on HAWC.
C-83
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.3.1.3 Findings from Children and Adolescents
Three medium confidence studies and two low confidence studies evaluated glucose levels in
children, with mixed non-significant results. Two medium confidence studies {Domazet, 2016,
3981435; Kang, 2018, 4937567} observed positive, non-significant associations with fasting
blood glucose. Negative, non-significant associations with fasting blood glucose were observed
in three studies, one of medium confidence {Alderete, 2019, 5080614}, and two of low
confidence {Khalil, 2018, 4238547; Fassler, 2019, 6315820}. Alderete et al. (2019, 5080614)
also reported a positive, non-significant association with 2-hour glucose {Alderete, 2019,
5080614}. (Appendix D).
Seven studies examined insulin measures, and two reported statistically significant associations.
Insulin resistance, as described by the HOMA-IR, was examined in five studies with mixed
results. Fleisch et al. (2017, 3858513) observed a significant negative association with HOMA-
IR in mid-childhood in a study of female children. Five studies (two medium and three low
confidence) reported non-significant negative associations with HOMA-IR {Alderete, 2019,
5080614; Fassler, 2019, 6315820; Koshy, 2017, 4238478; Khalil, 2018, 4238547; Domazet,
2016, 3981435}. In a medium confidence study, a non-significant decrease in HOMA-IR at age
15 and 21 years per increase in PFOS exposure from 9 years and a non-significant increase in
HOMA-IR at 21 per increase in PFOS measured at age 15 {Domazet, 2016, 3981435}.
Three studies examined fasting insulin levels. All three of these studies reported negative, non-
significant associations with fasting insulin {Domazet, 2016, 3981435; Khalil, 2018, 4238547;
Fassler, 2019, 6315820}.
A positive non-significant association was observed with insulin sensitivity, measured through
both the insulin sensitivity index and the Children's Health and Environmental Chemicals in
Korea (CHECK) Index/Quantitative Insulin Sensitivity Check Index (QUICKI) {Fassler, 2019,
6315820}.
One medium confidence study of reported significant negative associations with insulin-like
growth factor 1 (IGF-1) in 6-9-year old children in the C8 Health Project {Lopez-Espinosa,
2016, 3859832}. Significant negative associations for both girls and boys persisted after
stratification by sex, and statistically significant decreasing trends across quartiles were also
observed {Lopez-Espinosa, 2016, 3859832}.
One medium confidence study examined HOMA-B. Negative, non-significant associations were
observed between PFOS levels at age 9 and beta cell function at ages 15 or 21, but a positive
non-significant association was observed between PFOS levels at age 15 and beta cell function at
age 21 {Domazet, 2016, 3981435}.
Two high and two medium confidence studies examined adiponectin and leptin, and one
observed significant association. For adiponectin, all studies observed positive associations. A
high confidence study on the Sapporo Cohort of the Hokkaido Study observed a statistically
significant positive association between maternal PFOS and cord blood adiponectin (p-
value = 0.028) {Minatoya, 2017, 3981691}. Three other studies (one high and two medium
confidence studies) reported positive, non-significant associations with adiponectin {Buck, 2018,
5080288; Domazet, 2020, 6833700; Fleisch, 2017, 3858513}. Buck et al. (2018, 3981371)
C-84
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
observed a positive, non-significant association between maternal PFOS and adiponectin, but a
negative-non-significant association between mid-childhood PFOS and adiponectin.
Two medium and one high confidence study reported negative, non-significant association with
leptin {Domazet, 2020, 6833700; Fleisch, 2017, 3858513; Minatoya, 2017, 3981691}. Minatoya
et al. (2017, 3981691) observed a negative association with leptin among male children and a
positive association among female children; the interaction between child sex and PFOS was
statistically significant. Another study observed a positive, non-significant association with
PFOS; after stratification by sex, a negative non-significant association with leptin was observed
among males, but a positive non-significant association was observed among females {Buck,
2018, 5080288}.
Six studies examined body fat measures, and one reported a significant negative association. A
medium confidence study from the Avon Longitudinal Study of Parents and Children (ALSPAC)
reported a statistically significant negative association between maternal PFOS and trunk fat
percentage in female children {Hartman, 2018, 3859812}. One study observed non-significant
negative associations with body fat percentage {Braun, 2016, 3859836}, and two studies
observed a non-significant negative association with body fat mass {Jeddy, 2018, 5079850;
Domazet, 2020, 6833700}.
A high confidence study of 5-year old children observed positive, non-significant associations
with body fat percentage and fat mass; after stratification by sex, the non-significant positive
associations persisted for boys, but non-significant negative associations with fat mass and body
fat percentage were observed among girls {Chen, 2019, 5080578}. Another study of medium
confidence observed positive, non-significant associations with mid-childhood total fat mass
index, total fat-free mass index, and trunk fat mass index among children from Project Viva
{Mora, 2017, 3859823}.
Eleven studies examined BMI and related measures with mixed results. In the European Youth
Heart Study (EYHS) study, Domazet et al. (2016, 3981435) observed a positive significant
association between PFOS at age 9 and BMI at age 15. Positive, but non-significant associations
were observed between PFOS measured at either age 9 or age 15 and BMI measured at age 21
{Domazet, 2016, 3981435}. Additionally, two medium confidence studies observed significant
positive associations with children's BMI {Lauritzen, 2018, 4217244; Mora, 2017, 3859823}.
Mora et al. (2017, 3859823) reported a positive, significant association between maternal PFOS
and early childhood BMI; the association was positive but not significant for the association with
mid-childhood BMI {Mora, 2017, 3859823}. After stratification by sex, the association with
BMI remained positive (though non-significant) for boys and girls in early childhood and for
girls in mid-childhood but was negative and non-significant for boys in mid-childhood {Mora,
2017, 3859823}.
Significant negative associations were observed between maternal serum PFOS levels and BMI
of girls from the ALSPAC study {Hartman, 2017, 3859812} and between serum PFOS levels
and BMI of girls from the Breast Cancer and Environment Research Program (BCERP) study
{Fassler, 2019, 6315820}. Three studies (one of high confidence and two of low confidence)
reported negative, non-significant associations with BMI {Koshy, 2017, 4238478; Khalil, 2018,
4238547; Chen, 2019, 5080578}. In a sex-stratified analysis, Chen et al. (2019, 5080578)
C-85
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
observed a negative, non-significant association among girls, but a positive non-significant
association among boys.
Di Nisio et al. (2019, 5080655) reported no difference between BMI between Italian male high
school students exposed to PFOS pollution compared to those who were not exposed.
A medium confidence study reported a significant negative association between serum PFOS
levels and ponderal index at birth in infants from the Hokkaido Study on Environment and
Children's Health {Kobayashi, 2017, 3981430}.
Seven studies evaluated BMI z-score, and two observed an association with PFOS. In a medium
confidence study of children from the Faroe Islands, a significant positive association was
observed between maternal PFOS and BMI z-score among 18-month old children {Karlsen,
2017, 3858520}. In children from the POPUP study, Gyllenhammar et al. (2018, 4238300)
observed a positive, significant association with BMI z-score among children 4- and 5-years old;
the association with BMI z-score among 3-year old children was positive, but not significant.
Three other studies (two medium and one high confidence) reported positive, non-significant
associations with BMI z-score {Mora, 2017, 3859823; Manzano-Salgado, 2017, 4238509;
Jensen, 2020, 6833719}. In an age-stratified analysis, Jensen et al. (2020, 6833719) observed a
positive, non-significant association with BMI z-score at birth, but a negative, non-significant
association with BMI z-score at 3-months and 18-months of age.
Two studies reported negative, non-significant associations with BMI z-score {Koshy, 2017,
4238478; Braun, 2016, 3859836}.
Seven studies evaluated the risk of being overweight or obese, and three reported significant
associations. A medium confidence study reported increased odds of being overweight at 4 years
old, with significantly increased odds of being overweight in the 4th quartile of maternal PFOS
exposure {Martinsson, 2020, 6311645}. Another medium confidence study observed
significantly increased odds of being overweight with increasing maternal PFOS among 5-year-
old children {Lauritzen, 2018, 4217244}. A medium confidence study of mother-child pairs in
the Faroe Islands reported a significantly increased risk of being overweight at 18 months
{Karlsen, 2017, 3858520}. Two medium confidence studies observed an increased, non-
significant risk of being overweight {Mora, 2017, 3859823; Manzano-Salgado, 2017, 4238509}.
Manzano-Salgado et al. (2017, 4238509) observed an increased, non-significant risk of being
overweight at age 4, but a non-significant, decreased risk of being overweight at age 7.
Two studies (one medium and one low confidence reported non-significant, decreased risks of
being overweight or obese {Koshy, 2017, 4238478; Braun, 2016, 3859836}. Braun et al. (2016,
3859836) observed a non-significant decreased risk of being overweight or obese in the second
tertile of PFOS exposure, but a non-significant increased risk of being overweight or obese in the
third tertile of PFOS exposure.
Six studies examined waist circumference, and two reported an association. A significant,
positive association was observed between PFOS exposure at age 9 and waist circumference at
age 15 and 21 years old; a positive, non-significant association was reported for PFOS exposure
at age 15 and waist circumference at age 21 {Domazet, 2016, 3981435}. Two studies, one high
confidence and one low confidence observed negative, non-significant associations with waist
C-86
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
circumference {Chen, 2019, 5080578; Mora, 2017, 3859823}. After stratification by sex, Mora
et al. (2017, 3859823) observed negative, non-significant associations with waist circumference
among boys, and positive, non-significant associations with waist circumference among girls.
A medium confidence study of mother-daughter dyads reported a statistically significant negative
association with girls' waist circumference at age 9 {Hartman, 2017, 3859812}. In atertiles
analysis, Braun et al. (2016, 3859836) observed a negative association with waist circumference
in the second tertile of PFOS exposure, but a positive association in the third tertile.
One low confidence study reported no statistical difference in waist circumference among PFOS-
exposed children compared to non-exposed children {Di Nisio, 2019, 5080655}.
Two studies assessed waist circumference z-score among children, and none reported an
association. Both studies observed negative, non-statistical associations with waist circumference
z-score {Jensen, 2020, 6833719; Manzano-Salgado, 2017, 4238509}. Manzano-Salgado et al.
(2017, 4238509) observed a negative, non-significant association with waist circumference z-
score at age 4 and a null association at age 7; after stratification by sex, negative, non-significant
associations were observed for both boys and girls at age 7. In an age-stratified analysis, Jensen
et al. (2020, 6833719) reported a positive association with waist circumference z-score at birth,
but a negative association at 3-months and at 18-months.
Three studies evaluated waist-to-height ratio among children, and one observed a significant
association. A low confidence study reported a significant negative association was observed
with waist-to-height ratio among 6-8 year old girls {Fassler, 2019, 6315820}.
A high confidence study of children from the Shanghai Prenatal Cohort observed negative, non-
significant associations with waist-to-height ratio {Chen, 2019, 5080578}. In a medium
confidence study, a decreased risk of high waist-to-height ratio was observed at age 4, while an
increased risk of waist-to-height ratio was observed at age 7 {Manzano-Salgado, 2017,
4238509}.
Two studies examined waist-to-hip ratio in children, with no significant associations reported. A
medium confidence study observed a positive, non-significant association with waist-to-hip ratio
{Fassler, 2019, 6315820}, while a null association was observed in a medium confidence study
{Mora, 2017, 3859823}. After stratification by sex, Mora et al. (2017, 3859823) observed a
positive, non-significant association among girls, but a negative, non-significant association
among boys.
Three studies examined skinfold thickness metrics, with two studies reporting significant
associations. A study from the EYHS reported significant positive associations between PFOS
measured at age 9 and skinfold thickness at age 15 and age 21; the association between PFOS at
age 15 and waist circumference at age 21 was positive, but not significant {Domazet, 2016,
3981435}. Additionally, a significant positive association was observed with tricep skinfold
thickness z-score, while associations with subscapular skinfold thickness z-score were positive,
but non-significant {Lauritzen, 2018, 4217244}.
Mora et al. (2017, 3859823) observed positive, non-significant associations with subscapular and
tricep skin thickness measures in mid- and early-childhood. Negative, non-significant
associations were observed with the sum of subscapular and tricep skinfold thickness among all
C-87
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
children in mid-childhood, as well as with the subscapular-to-tricep skinfold thickness ratio
among girls in early childhood {Mora, 2017, 3859823}.
C.3.1.4 Findings from Pregnant Women
Ten studies examined diabetes or gestational diabetes and overall results were mixed, with no
significant associations (Appendix D).
Positive, non-significant associations with gestational diabetes were reported in four studies
{Preston, 2020, 6833657; Wang, 2018, 5080352; Liu, 2019, 5881135; Matilla-Santander, 2017,
4238432}. A medium confidences study observed an increased, non-significant risk of
gestational diabetes among women with a family history of type 2 diabetes and women were had
an overweight pre-pregnancy BMI; a decreased, non-significant risk of gestational diabetes was
observed among all women, women without a family history of type 2 diabetes, and with a
normal pre-pregnancy BMI {Rahman, 2019, 5024206}.
Four medium and one low confidence studies reported inverse, non-significant associations with
gestational diabetes {Xu, 2020, 6833677; Wang, 2018, 5079666; Valvi, 2017, 3983872; Zong,
2016, 3350666; Shapiro, 2016, 3201206}. With the exception of the low confidence study
{Zong, 2016, 3350666}, gestational diabetes was determined through standard clinical methods.
The nested case-control study conducted by Xu et al. (2020, 6833677) recruited pregnant women
with no history of diabetes and reported inverse, non-significant odds of gestational diabetes
across quartiles of PFOS exposure and log-transformed PFOS exposure. Similarly, Shapiro et al.
(2016, 3201206) observed inverse, non-significant odds of gestational diabetes or gestational
impaired glucose tolerance, but increased odds of gestational diabetes in the second quartile of
PFOS exposure.
Fasting glucose was examined in six studies, and one reported a positive association. A medium
confidence study observed a significant increase in fasting glucose levels with increasing tertiles
of PFOS, but a negative association between PFOS analyzed continuously and fasting glucose
{Wang, 2018, 5080352}. Two high confidence studies and one medium confidence study
reported negative, non-significant associations with fasting glucose {Starling, 2017, 3858473;
Jensen, 2018, 4354143; Liu, 2019, 5881135}. In contrast, two medium confidence studies
reported positive, non-significant associations with fasting glucose among pregnant women
{Ren, 2020, 6833646; Wang, 2018, 5079666}.
Results from oral glucose tolerance tests were assessed in five studies, two of which reported an
association. A high confidence study from Project Viva observed non-significant positive
associations with 1-hour glucose; a significant association with 1-hour glucose was observed in
the fourth quartile of PFOS exposure {Preston., 2020, 6833657}. Additionally, a medium
confidence study reported a significant association with 1-hour glucose levels among pregnant
women in the Shanghai-Minhang Birth Cohort {Ren, 2020, 6833646}. Three studies observed
positive, non-significant associations with oral glucose tolerance test results {Wang, 2018,
5080352; Jensen, 2018, 4354143; Liu, 2019, 5881135}.
Three studies examined impaired glucose tolerance among pregnant women. One low confidence
study reported positive, statistically significant effect estimates between plasma PFOS levels and
impaired glucose tolerance among pregnant women from the INMA birth cohort in Spain
{Matilla-Santander, 2017, 4238432}. A high confidence study and a medium confidence study
C-88
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
both reported positive, non-significant associations with impaired glucose tolerance in the second
and third quartiles of PFOS exposure, and a negative, non-significant association with impaired
glucose tolerance in the fourth quartile of PFOS exposure {Preston, 2020, 6833657; Shapiro,
2016, 3201206}.
Two high confidence studies evaluated associations between plasma PFOS levels and
hyperglycemia or HbAlc among members of Project Viva. Preston et al. (2020, 6833657)
reported a positive, non-significant association with hyperglycemia. Conversely, Mitro et al.
(2020, 6833625) observed a negative, non-significant association with HbAlc; negative non-
significant associations persisted after stratification by maternal age.
Two studies, one of high and one of medium confidence observed positive, non-significant
associations with both fasting insulin and HOMA-IR in pregnant women {Jensen, 2018,
4354143; Wang, 2018, 5079666}. These studies evaluated members of the OCC in Denmark
with high risk of gestational diabetes {Jensen, 2018, 4354143} and women in China in early
pregnancy {Wang, 2018, 5079666}. Jensen et al. (2018, 4354143) reported a negative, non-
significant association with insulin sensitivity as reported by the Matsuda index.
One high confidence study of members of the OCC examined HOMA-B and levels of fasting c-
peptide among pregnant women with high risk of gestational diabetes and reported positive, non-
significant associations with both HOMA-B and fasting c-peptide {Jensen, 2018, 4354143}.
Two high confidence studies compared levels of PFOS and adiponectin or leptin among pregnant
women. One medium confidence study observed a negative, non-significant association with
adiponectin {Mitro, 2020, 6833625} while another medium confidence study reported a positive,
non-significant association with adiponectin {Ashley-Martin, 2017, 3981371}. After
stratification by age during pregnancy, Mitro et al. (2020, 6833625) reported a negative
association with adiponectin among women aged 35 and older, and a positive, non-significant
association among women under 35.
Among the two medium confidence studies examining leptin, one reported a positive, non-
significant association {Mitro, 2020, 6833625}, while the other reported a negative, non-
significant association {Ashley-Martin, 2017, 3981371}.
Three medium confidence studies examined gestational weight gain, with mixed results.
Jaacks et al. (2016, 3981711) observed a positive, non-significant association with gestational
weight gain among all mothers, and mothers with a BMI < 25, and a negative non-significant
association in mothers with a BMI > 25. Increased odds of excessive gestational weight gain and
decreased odds of inadequate weight gain were observed and were non-significant {Jaacks,
2016, 3981711}.
Ashley-Martin et al. (2016, 3859831) used data from mother-infant pairs from the MIREC to
estimate the odds of having high cord blood PFOS (> 0.39 ng/mL) per increase in gestational
weight gain. ORs were significant for both 1kg increase in gestational weight gain and IQR
increase in gestational weight gain {Ashley-Martin, 2016, 3859831}.
C-89
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Marks et al. (2019, 5381534) observed a negative, non-significant association with gestational
weight gain. However, a significant interaction was observed between PFOS and pre-pregnancy
BMI {Marks, 2019, 5381534}.
One high confidence study reported a significant positive association with skinfold thickness, as
well as a non-significant positive association with waist circumference among pregnant women
from Project Viva {Mitro, 2020, 6833625}.
In a high confidence study, a positive non-significant association was observed between plasma
PFOS levels and BMI in pregnant women from the Project Viva study {Mitro, 2020, 6833625}.
C.3.1.5 Findings from the General Adult Population
Eleven studies evaluated diabetes in the general population and four reported significant
associations with diabetes. A medium confidence study of Taiwanese adults aged 20-60 reported
a significant positive association with type 2 diabetes {Su, 2016, 3860116}. In a quartile
analysis, odds of type 2 diabetes significantly increased with increasing quartiles of PFOS {Su ,
2016, 3860116}. Another medium confidence study reported significantly increased odds of type
2 diabetes in the second and third tertile of PFOS exposure among female nurses in the Nurses'
Health Study (NHS) II {Sun, 2018, 4241053}. A medium confidence study from the E3N cohort
reported a non-significant increased risk of type 2 diabetes in the 2nd-4th, 6th, 8th-9th deciles of
PFOS exposure, and a non-significant decreased risk of type 2 diabetes was observed in the 5th
and 10th deciles of PFOS exposure {Mancini, 2018, 5079710} (Appendix D).
Three low confidence studies reported non-significant positive associations with diabetes {Lind,
2014, 2215376; Christensen, 2016, 3858533; He, 2018, 4238388} and prediabetes {Christensen,
2016, 3858533}.
Significant decreased odds of type 1 and type 2 diabetes were observed among 6889 participants
in the C8 Health Project {Conway, 2016, 3859824}. The decrease in odds of uncategorized
diabetes was not significant. After stratifying by age, significant decreased odds of type 1
diabetes were observed among adults and children {Conway, 2016, 3859824}. One high
confidence cohort study from the Diabetes Prevention Program followed adults at increased risk
of type 2 diabetes and observed a decreased non-significant risk of diabetes {Cardenas, 2017,
4167229}. After stratification by sex, a significant decreased risk of type 2 diabetes was
observed among males, and the decreased risk among females was not significant {Cardenas,
2017, 4167229}. Two other medium confidence study reported non-significant negative
associations with type 2 diabetes {Donat-Vargas, 2019, 598342; Cardenas, 2019, 5381549}.
Four studies (three medium confidence and one low confidence) evaluated metabolic syndrome
(MetS) and one study reported an association. In an adult population of the island of Hvar
(Croatia) Chen et al. (2019, 5387400) observed a positive non-significant association with risk of
Metabolic syndrome as defined by the Adult Treatment Panel III (ATP III) criteria (OR: 2.19;
95% CI: 0.88, 5.44). Two medium confidence studies using overlapping data from NHANES
reported non-significant negative associations with metabolic syndrome. Liu et al., 2018
observed adults aged 20 and older from the 2013-2014 NHANES cycle and Christensen et al.
(2019, 5080398) observed adults aged 18 and older from 2007-2014 NHANES. In a model
simultaneously adjusted for PFDE, PFOA, PFHxS, N-methyl-PFOSA (MPAH), PFNA and
PFUnDA, Christensen et al. (2019, 5080398) reported non-significant increased odds of
C-90
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
metabolic syndrome in the third and fourth quartiles of PFOS exposure; the decreased odds
observed in the second quartile of PFOS were not significant.
A low confidence study observed lower non-significant odds of metabolic syndrome for
participants with serum PFOS > 1.90 ng/mL compared to those with serum PFOS < 1.90 ng/mL
{Yang, 2018, 4238462}. However, concerns for selection bias, outcome misclassification, and
residual confounding by SES diminish confidence in the study results.
There were nine studies examining glucose. Three studies reported associations with fasting
blood glucose, one reported an association with 2-hour glucose, one reported an association with
glucose area under the curve (AUC).
A medium confidence study of adults aged 19-87 years from China reported a significant
positive association with fasting blood glucose {Duan, 2020, 5918597}. Additionally, a study
using NHANES 1999-2014 data observed a significant positive correlation between fasting
glucose and serum PFOS {Huang, 2018, 5024212}. Su et al. (2017, 3860116) reported a non-
significant positive association with fasting glucose; in a quartiles analysis, mean fasting blood
glucose significantly increased with increasing quartiles of PFOS. Liu et al. (2018, 4238514)
reported a negative statistically significant association with fasting blood glucose, but non-
significant increased odds of fasting glucose levels > 100 mg/dL.
A low confidence study observed a positive, non-significant association with fasting blood
glucose {Heffernan, 2018, 5079713}, while another reported lower non-significant odds of blood
glucose >1.6 mmol/L for participants with serum n-PFOS > 3 ng/mL compared with those with
serum n-PFOS < 3 ng/mL {Yang, 2018, 4238462}.
Two studies (one high confidence and one medium confidence) observed non-significant positive
associations with 2-hour glucose {Cardenas, 2017, 4167229; Su, 2016, 3860116} and 30-minute
glucose {Cardenas, 2017, 4167229}. Another medium confidence study reported a negative, non-
significant association with 2-hour glucose {Liu, 2018, 4238514}.
One medium confidence study observed a significant decrease in glucose AUC with increasing
quartiles of PFOS and a non-significant negative association between PFOS (measured
continuously) and glucose AUC {Su, 2016, 3860116}. In the POUNDS-Lost clinical trial, a
positive, non-significant correlation was observed between PFOS and glucose levels {Liu, 2018,
4238514}.
Blood glucose levels were examined in a medium confidence study from NHANES (2007-
2014), which reported increased odds of high blood glucose in the second and third quartiles of
PFOS, and decreased odds in the fourth quartile of PFOS exposure {Christensen, 2019,
5080398}. A low confidence study reported a negative association with blood glucose levels
{van den Dungen, 2017, 5080340}. None of the associations for these two studies reached
statistical significance.
Significant associations were reported between resting metabolic rate and PFOS. The association
with resting metabolic rate was assessed in the POUNDS-Lost trial, a clinical trial of overweight
and obese adults aged 30-70. A non-significant negative correlation between PFOS and resting
metabolic rate was observed {Liu, 2018, 4238396}. In the first 6 months of the trial, resting
metabolic rate decreased non-significantly with increasing tertiles of PFOS exposure for the
C-91
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
entire study population, men, and women. The interaction between PFOS and sex were
significant {Liu, 2018, 4238396}. In months 6-24 of the trial, a significant positive association
was observed with mean resting metabolic rate in all tertiles of PFOS exposure, and average
resting metabolic rate significantly decreased with increasing tertiles of PFOS {Liu, 2018,
4238396}. In a sex-stratified analysis, average resting metabolic rate significantly decreased with
increasing tertiles of PFOS among men and women {Liu, 2018, 4238396}.
Twelve studies examined insulin resistance measures and one observed significant association
with fasting insulin, insulin resistance, fasting plasma insulin, 30-minute insulin, fasting
proinsulin, and insulin (corrected response), and one reporting associations with the ratio of
proinsulin to insulin.
Four studies measured fasting insulin. One high confidence study used a subset of data on 954
adults at high risk of type 2 diabetes from the Diabetes Prevention Program and observed a
positive significant association between PFOS and fasting insulin {Cardenas, 2017, 4167229}.
Two low confidence reported non-significant positive associations with fasting insulin {Chen,
2019, 5387400; Sun, 2018, 4241053}, and one reported a non-significant negative association
(He et al., 2018, 4238388). One medium confidence study reported a positive, non-significant
association with insulin levels {Liu, 2018, 4238514}.
Nine studies examined insulin resistance (measured as HOMA-IR), and one reported a
significant association. A high confidence study of 956 adults at high risk for type 2 diabetes in
the Diabetes Prevention Program reported a significant, positive association with HOMA-IR
{Cardenas, 2017, 4167229}. A medium confidence study of 1871 adults in NHANES observed a
non-significant positive association with HOMA-IR {Liu, 2018, 4238514}. However, Donat-
Vargas et al. (2019, 5083542) reported a non-significant negative association with HOMA-IR in
both continuous and tertile analyses. In a sensitivity analysis, a non-significant negative
association was observed between HOMA-IR and the third tertile of baseline PFOS, and between
HOMA-IR and PFOS measured at the end of follow-up for both the second and third tertile of
PFOS exposure. A non-significant positive association with HOMA-IR was reported in the
second tertile of baseline PFOS exposure {Donat-Vargas, 2019, 5083542}.
Four low confidence studies investigated the association between PFOS and insulin resistance.
Of these studies, two reported a positive, non-significant association with insulin resistance
{Lind, 2014, 2215376; Chen, 2019, 5387400; Lin, 2013, 2850967}. In a sex-stratified tertile
analysis, a non-significant negative association was observed between PFOS and insulin
resistance in both males and females; among females, a significant negative association with
insulin resistance was observed in the third quartile of PFOS exposure {He, 2018, 4238388}.
These studies were of low confidence due to concerns with the statistical analysis (not
accounting for design of NHANES) {He, 2018, 4238388}, failure to account for diabetes status
{Lind, 2014, 2215376} or medications that could affect insulin levels {Chen, 2019, 5387400},
and concerns for residual confounding and selection bias {Lin, 2013, 2850967}.
The association between plasma PFOS and insulinogenic index 1 was investigated in a high
confidence study from the Diabetes Prevention Program. A non-significant positive association
was observed with insulinogenic index among 945 adults at high risk for type 2 diabetes
{Cardenas, 2017,4167229}.
C-92
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
In a high confidence study, Cardenas et al. (2017, 4167229) reported significant positive
associations between PFOS and fasting plasma insulin, 30-minute insulin, and fasting proinsulin.
A non-significant positive association was observed with insulin (corrected response) {Cardenas,
2017, 4167229}.
In a low confidence study, a non-significant positive association was reported for the ratio of
proinsulin to insulin and PFOS {Lind, 2014, 2215376}. This study was given a low confidence
rating due to failure to adjust for diabetes status in statistical analyses.
Four studies measured the association between PFOS and beta cell function and two reported a
significant association. Cardenas et al. (2017, 4167229) reported a significant positive
association with beta cell function (measured as HOMA-B) in adults at high risk for type 2
diabetes from the Diabetes Prevention Program. Positive non-significant associations with
HOMA-B were reported in adults from NHANES {Liu, 2018, 4238514} and {Chen, 2019,
5387400}. A medium confidence studies reported negative, non-significant associations with
HOMA-B {Donat-Vargas, 2019, 5083542}.
Four studies examined adiponectin, and none reported significant associations. Two high
confidence studies reported non-significant positive associations with adiponectin {Buck, 2018,
5080288; Ashley-Martin, 2017, 3981371}. In contrast, a non-significant negative association
with adiponectin was observed among 945 adults in the Diabetes Prevention Program {Cardenas,
2017, 4167229}. A medium confidence study reported a negative non-significant correlation
between PFOS and plasma adiponectin {Sun, 2018, 4241053}.
Three studies examined associations with leptin. One study reported a significant association.
Two high quality studies measured associations with leptin; one reported a non-significant
positive association {Buck, 2018, 5080288}, and the other reported a non-significant negative
association {Ashley-Martin, 2017, 3981371}. A medium confidence study reported a positive,
non-significant correlation between plasma PFOS and leptin concentrations, and a non-
significant, positive correlation with soluble leptin receptors {Liu, 2018, 4238396}.
Nine studies examined HbAlc, and three reported associations. A high confidence study on
participants in the Diabetes Prevention Program reported a significant positive association with
HbAlc {Cardenas, 2017, 4167229}. A significant positive association with HbAlc was also
reported among adults under age 55 in a medium confidence study of adults living in China; the
association with HbAlc among adults aged 55 and older was also positive, but not significant
{Duan, 2020, 5918597}. Two medium confidence studies observed positive correlations with
HbAlc; one was non-significant {Sun, 2018, 4241053} and the other was significant {Huang,
2018, 5024212}. Another medium confidence cross-sectional study assessed the association
between plasma PFOS and HbAlc in adults aged 20-60 {Su, 2016, 3860116}. A positive, non-
significant association between HbAlc and continuous PFOS was reported, and a significant
increase in average HbAlc was observed with increasing quartiles of PFOS {Su, 2016,
3860116}.
In the POUNDS-Lost trial, a negative, non-significant correlation was observed between PFOS
and HbAlc {Liu, 2018, 4238396}. Additionally, a medium confidence study of 1871 adults from
NHANES reported a non-significant negative association with HbAlc {Liu, 2018, 4238514}.
C-93
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
One low confidence study reported a non-significant negative association with HbAlc
{Heffernan, 2018, 5079713}. Another low confidence study observed a non-significant positive
association between PFOS and HbAlc {Chen, 2019, 5387400}. Concerns with measurement of
confounders and inclusion of medications that could affect insulin levels {Chen, 2019,
5387400}, as well as concerns with case selection and residual confounding {Heffernan, 2018,
5079713} resulted in low confidence ratings.
There were four studies evaluating body weight measures. Associations were observed in one
study of body weight, and two studies reported associations with being overweight or obese.
One study, from the POUNDS-Lost clinical trial, evaluated body weight and observed a
negative, non-significant association with weight loss in the first 6 months of the trial, and a
positive, significant association with weight loss in months 6-24 of the trial {Liu, 2018,
4238396}. A significant increase in average weight gain during months 6-24 of the trial was
observed with increasing tertiles of PFOS {Liu, 2018, 4238396}.
Two studies evaluated being overweight, one of which reported an association. A medium
confidence study reported significantly greater serum PFOS among obese adults compared to
non-obese adults {Jain, 2019, 5080621}. One medium confidence study evaluated maternal
PFOS and risk of being overweight or obese in their children; this study reported increased, non-
significant odds of being overweight at age 4 in the second and third quartiles of PFOS exposure,
and significant increased odds of being overweight at age 4 in the fourth quartile {Martinsson,
2020, 6311645}.
One low confidence study observed significant increased odds of being overweight or obese
{Tian, 2019, 5080586}. Another low confidence study reported non-significant negative
associations with being overweight and obese {Yang, 2018, 4238462}.
Five studies evaluated body fat measures, and one reported an association Four studies of
medium confidence evaluated body fat. A significant negative association was observed between
maternal plasma PFOS and trunk fat in young girls ALSPAC. After stratification by age at
menarche, the association remained negative but was not significant in either age group
{Hartman, 2017, 3859812}. A negative, non-significant association was observed between
maternal plasma PFOS and body fat percentage {Hartman, 2017, 3859812}.
Three medium confidence studies reported positive, non-significant associations with body fat
measures {Mora, 2017, 3859823; Braun, 2016, 3859836; Liu, 2019, 5881135}.
Two medium confidence studies evaluated fat mass; one reported a non-significant negative
association with fat mass among children {Jeddy, 2018, 5079850} and a non-significant positive
association with fat mass among overweight and obese adults {Liu, 2019, 5881135}.
11 studies assessed BMI; one significant association was reported for BMI, and one significant
association was reported for BMI z-score.
In the Health Outcome Measures of the Environment (HOME) study, a cohort study of 285
mother-child pairs, PFOS exposure was measured during pregnancy and BMI was recorded at
age 8 {Braun, 2016, 3859836}. Negative, non-significant associations with BMI z-score were
C-94
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
observed in the second and third tertile of maternal PFOS exposure {Braun, 2016, 3859836}. Liu
et al. (2018, 4238396) reported a non-significant negative correlation between PFOS and BMI.
One high confidence study and two medium confidence studies observed positive, non-
significant associations with BMI {Cardenas, 2017, 4167229; Chen, 2019, 5387400; Blake,
2017, 5080657}.
In a medium confidence cohort study from the ALSPAC, a significant negative association with
children's BMI was observed among 312 mother-child pairs {Hartman, 2017, 3859812}.
Another medium confidence study reported non-significant positive association with BMI; in a
sex-stratified analysis, a non-significant percent decrease was observed for males, and a non-
significant percent increase was observed among females {Blake, 2018, 5080657}. In the single
low confidence study, Tian et al. (2019, 5080586) reported a non-significant association with
BMI. In a sex-stratified analysis, a non-significant negative association was observed among
men and a positive, non-significant association was reported for women. {Tian, 2019, 5080586}.
This study was given a low confidence designation due to concerns for PFOS to be potentially
related to BMI.
A high confidence study measured PFOS in maternal serum and BMI z-score in children. Non-
significant negative associations with BMI z-score were observed in children at 3- and 18-
months, and a non-significant positive association with BMI z-score was observed at birth.
{Jensen, 2020, 6833719} A medium confidence study of 412 mother-child pairs observed a
positive, significant association between maternal serum PFOS and 5-year old child's BMI z-
score {Lauritzen, 2018, 4217244}.
Five studies examined waist circumference. Two single medium confidence studies observed a
negative, non-significant association with waist circumference {Liu, 2018, 4238396; Liu, 2018,
4238514}. One low confidence study reported a non-significant positive association with waist
circumference {Tian, 2019, 5080586}. Non-significant decreased odds of increased waist
circumference were observed among men, and non-significant increased odds were observed for
women; the interaction between PFOS and sex was significant but was not significant in
continuous analyses {Tian, 2019, 5080586}. In another low confidence study, non-significant
increased odds of increased waist circumference were observed with increasing quartiles for
PFOS; these estimates were adjusted for multiple PFAS {Christensen, 2019, 5080398}.
C.3.1.6 Findings from Occupational Studies
No occupational studies examined metabolic outcomes and PFOS.
C3.2 Animal Evidence Study Quality Evaluation and
Synthesis
C.3.2.1 Metabolic Homeostasis
There are 3 studies from the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} and 4 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOS HESD that investigated the association between PFOS and metabolic effects. Study
quality evaluations for these 4 studies are shown in Figure C-21.
C-95
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Conley et al., 2022, 10176381 -
+
+
NR
++
+
+
~
++
+
Curran et al., 2008, 757871 -
++
NR
NR
++
+
+
++
++
+
Lai etal., 2018, 5080641 -
+
+
NR
++
-
+
+
+
+
+
Luebker et al., 2005, 757857 -
+
+
NR
+
-
-
++
+
+
+
NTP, 2019, 5400978-
Seacat etal., 2003, 1290852-
Thomford, 2002, 5432419-
Legend
3
Good (metric) or High confidence (overall)
+
Adequate (metric) or Medium confidence (overall)
-
Deficient (metric) or Low confidence (overall)
D
Critically deficient (metric) or Uninformative (overall)
NR
Not reported
Figure C-21. Summary of Study Evaluation for Toxicology Studies of PFOS and Metabolic
Effects
Interactive figure and additional study details available on HAWC.
PFOS has been observed to cause perturbations in glucose homeostasis in rodents. Several
studies in adult and perinatal rats and mice investigate glucose homeostasis, including serum
glucose levels, glucose tolerance, and gluconeogenesis, among other measures. Alterations in
these metabolic endpoints were observed, but the data is inconclusive as there are inconsistencies
within the literature with too few studies to assess possible difference across life stages, sexes,
and species.
NTP (2019, 5400978) reported no statistical differences in serum glucose in adult male and
female Sprague Dawley rats exposed to PFOS doses up to 5 mg/kg/day for 28 days. In contrast,
Seacat et al. (2003, 1290852) observed a significant decrease in serum glucose in adult male
Sprague Dawley rats compared to controls following 1.51 mg/kg/day PFOS exposure in the diet
for 4 weeks. No statistically significant change was seen in females at the 4-week interim
timepoint. After 14 weeks, serum glucose concentrations were no longer statistically different in
males from any treatment group. In females at 14 weeks, serum glucose was significantly lower
in the 0.40 mg/kg/day group, but not in the high dose group (1.56 mg/kg/day).
C-96
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
In a rat reproductive toxicity study, Luebker et al. (2005, 757857) noted significantly higher
serum glucose levels on lactational day (LD) 5 in dams treated with 2 mg/kg/day PFOS for
42 days prior to mating until LD 4. This change was not seen in dams sacrificed at GD 21.
Serum glucose levels were not significantly altered in fetuses at GD 21 or in pups at LD 5. In a
glucose tolerance test, Lv et al. (2013, 2850947) observed a dose-related increase in serum
glucose 10 weeks postweaning in rats perinatally exposed to PFOS from GD 0-PND 20 with
significance in the high dose exposure group of 1.5 mg/kg/day. At 15 weeks postweaning, only
the low dose (0.5 mg/kg/day) group had significantly elevated serum glucose during the glucose
tolerance test. Elevated serum glucose in this test indicates decreased glucose clearance or
tolerance. In addition, at 18 weeks postweaning, rats in the high dose group had elevated serum
insulin, higher insulin resistance indices, increased leptin levels, and decreased adiponectin
levels, all of which indicate dysregulation of glucose homeostasis and insulin resistance,
potential signs of prediabetes {Lv, 2013, 2850947}.
Wan et al. (2014, 2850405) exposed CD-I mouse dams to 0 mg/kg/day, 0.3 mg/kg/day, or
3 mg/kg/day PFOS from GD 3-PND 21. Offspring were then fed either a standard or high-fat
diet from PND 21-PND 63. At PND 21, no statistical difference was detected in the fasting
serum glucose or insulin levels in dams. However, the Homeostatic Model Assessment for
Insulin Resistance (HOMA-IR) index was significantly increased in both the 0.3 and
3 mg/kg/day dose groups. Increases in this metric indicate increased risk of insulin resistance,
hypertension, and type 2 diabetes {Wan, 2014, 2850405}. There was no significant difference in
fasting serum glucose or the HOMA-IR index in male or female pups at PND 21, though males
from both the 0.3 mg/kg/day and 3 mg/kg/day groups had significantly increased fasting serum
insulin levels. No difference was found in fasting serum insulin levels in female pups at PND 21.
In pups fed a standard diet, at PND 63, fasting serum glucose levels were significantly higher for
males and females at both PFOS doses. Serum insulin and HOMA-IR were significantly
increased only at the high dose of 3 mg/kg/day PFOS in both sexes. No significant differences
between treatment groups in glucose tolerance were observed in either sex. In the high-fat diet
group, fasting serum insulin was increased at PND 63 in the 3 mg/kg/day PFOS group of both
sexes. Fasting serum glucose was significantly higher in females dosed with both 0.3 and
3 mg/kg/day, but only for the 3 mg/kg/day males. In the glucose tolerance test, serum glucose
was significantly higher only in the high dose group in both sexes, indicating decreased glucose
tolerance in these animals. The HOMA-IR index in each sex was elevated in the high dose
groups compared to the high-fat diet control group. However, the HOMA-IR indices were
significantly higher for the high-fat diet groups compared to the standard diet groups within a
specific PFOS treatment group and sex. In contrast, Ngo et al. (2014, 2850267) did not observe
significant changes in blood glucose at PNW 6, PNW 11, or PNW 20 in wild-type or
tumorigenic transgenic C57BL/6J-M// mice offspring gestationally exposed to 0 mg/kg/day,
0.01 mg/kg/day, 0.1 mg/kg/day, or 3 mg/kg/day PFOS from GD 1-GD 18, though it should be
noted that the animals were not fasted prior to serum sample collection.
Lai et al. (2018, 5080641) exposed CD-I female mice to 0, 0.3, or 3 mg/kg/day for 7 weeks with
conflicting results. The authors conducted an oral glucose tolerance test and an intraperitoneal
insulin tolerance test. In both tests, blood glucose levels were significantly lower in the
3 mg/kg/day dose group compared to controls, potentially indicating increased glucose tolerance
and reduced insulin resistance, respectively. Pyruvate tolerance was also significantly decreased
C-97
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
in both the 0.3 mg/kg/day and 3 mg/kg/day dose groups which could indicate reduced
gluconeogenesis.
C3.2.2 Survival, Clinical Observations, Body Weight, and Food
Consumption
There are 6 studies from the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} and 21 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOS HESD that investigated the association between PFOS and systemic effects. Study quality
evaluations for these 27 studies are shown in Figure C-22 and Figure C-24.
C-98
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Alametal., 2021, 9959508
Butenhoff et al., 2009, 757873
Curran et al„ 2008, 757871
Dong et at., 2011, 1424949
Han et al., 2018, 4238554
Han et al., 2018, 4355066-
Kawamoto et al., 2011, 2919266
Lefebvre et al., 2008, 1276155-
Li et al., 2021, 7643501
Luebker et al.. 2005, 1276160 - +
Luebker et al., 2005, 757857
Lv et al., 2015, 3981558
NTP, 2019, 5400978
Qiu et al., 2016, 3981408
WA
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
B Critically deficient (metric) or Uninformative (overall)
Not reported
* Multiple judgments exist
Figure C-22. Summary of Study Evaluation for Toxicology Studies of PFOS and Systemic
Effects3
Interactive figure and additional study details available on HAWC.
a Lefebvre et al. (2008, 1276155) reported on the same animals as Curran et al. (2008, 757871).
C-99
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Qiu et al., 2020, 7276729-
i.
NR
i
NR
++
I
+
I
+
++
++
+
Qu et al., 2016, 3981454-
+
+
NR
++
+
+
++
++
+
Salgado et al., 2015, 3981583 -
+
+
NR
+
+
-
+
+*
-
-
Salgado et al., 2016, 3179088-
+
+
NR
++
+
+
++
n
D
+
Seacat et al., 2002, 757853 -
++
NR
+
+
+
++
++
++
+
Seacat et al., 2003, 1290852 -
++
NR
+
+
+
++
D
B
+
Thomford, 2002, 5432419-
+
-
NR
++
++
-
++
++
++
-
Wan et al., 2016, 3981504-
+
+
NR
++
-
+
B
D
++
+
Xing et al., 2016, 3981506-
++
+
NR
++
++
++
++
+
+
Yan etal., 2014, 2850901 -
++
+
NR
++
J
B
++
++
Yang etal., 2021, 7643494 -
++
+
NR
++
++
++
++
-
-
Zhang et al., 2019, 5918673 -
-
+
NR
+
+
-
+
-
-
Zhong etal., 2016, 3748828-
-
NR
NR
+
+
+
+
++
+
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
B Critically deficient (metric) or Uninformative (overall)
Not reported
* Multiple judgments exist
Figure C-23. Summary of Study Evaluation for Toxicology Studies of PFOS and Systemic
Effects (Continued)3
Interactive figure and additional study details available on HAWC.
a Lefebvre et al. (2008, 1276155) reported on the same animals as Curran et al. (2008, 757871).
C-100
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A number of subchronic, chronic, and developmental studies suggest that PFOS exposure can
induce whole-body toxicity, which can manifest as decreased body weight, partly due to a
reduction in food consumption. These changes were more prominent following high exposures to
PFOS. Although one study in non-human primates suggests PFOS-related mortality, PFOS-
induced mortality and clinical observations were not supported by rodent studies.
C3.2.2.1 Mortality and Clinical Observations
PFOS-related mortality was observed in 2 of 6 male cynomolgus monkeys administered
0.75 mg/kg/day PFOS for 26 weeks. Pulmonary inflammation was identified as the probable
cause of death of one monkey that died on day 155 of dosing, and hyperkalemia was suggested
for the other monkey that died on day 179 {Seacat, 2002, 757853}. Mortality was not affected in
female monkeys administered 0.75 mg/kg/day PFOS or male or female monkeys receiving
0.03 mg/kg/day or 0.15 mg/kg/day PFOS {Seacat, 2002, 757853}.
Rodent studies did not observe mortality with doses up to 10 mg/kg/day and durations up to
60 days. No mortality was observed in C57 male mice exposed to 0.5 mg/kg/day or
10 mg/kg/day PFOS for 5 weeks, but the study did not report if there were any overt clinical
observations {Qu, 2016, 3981454}. NTP (2019, 5400978) exposed male and female Sprague-
Dawley rats to 0.312-5 mg/kg/day PFOS for 28 days. All rats survived to the end of the study,
except for one female Sprague-Dawley rat administered 5 mg/kg/day {NTP, 2019, 5400978}.
There were no treatment-related clinical observations reported in male or female rats {NTP,
2019, 5400978}. Similarly, Alam et al. (2021, 9959508) reported that there was no mortality in
male Wistar rats over the course of a 60-day study exposure to 0, 0.015, or 0.15 mg/kg/day
PFOS. Xing et al. (2016, 3981506) did not observe an effect on mortality in C57BL/6J male
mice exposed to PFOS at 2.5 mg/kg/day, 5 mg/kg/day, or 10 mg/kg/day for 30 days. Clinical
observations such as rough hair, slow movement, and constipation were reported, although
neither the exposure group associated with these effects nor incidence were specified {Xing,
2016, 3981506}. Study authors indicated that there were no treatment-related clinical signs or
mortality in Po male Crl:Cd(Sd)lgs rats following 6 weeks of pre-mating exposure to
1.6 mg/kg/day, 2.0 mg/kg/day, or 3.2 mg/kg/day {Luebker, 2005, 1276160}. No mortality was
observed in the Po females, but timing of the clinical observations (i.e., localized areas of partial
alopecia) were not specified when they occurred {Luebker, 2005, 1276160; Luebker, 2005,
757857} (see PFOS Main Document).
C.3.2.2.2 Body Weight in Adults
Many studies with rodent models report reductions in body weight following short term to
subchronic PFOS exposure (Figure C-24). A dose-dependent reduction in body weight change
was observed in C57BL/6J male mice exposed to PFOS at 2.5 mg/kg/day, 5 mg/kg/day, or
10 mg/kg/day via gavage for 30 days {Xing, 2016, 3981506}. All dose groups had a significant
difference in body weight gain when compared to the control with the 10 mg/kg/day group
having a 31% reduction in body weight over the study period compared to a 27.75% weight gain
in the controls. This reduction may be attributed to reduced food consumption reported across all
doses, but the correlation between body weight and food intake was not significant in the
treatment groups suggesting that this may not be the only explanation {Xing, 2016, 3981506}.
C57 male mice exposed to 0 mg/kg/day, 0.5 mg/kg/day, or 10 mg/kg/day by oral gavage for
5 weeks also showed decreased body weight, but only in the 10 mg/kg/day group, which
weighed 83% of controls {Qu, 2016, 3981454}. In a separate study, although reductions in body
C-101
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
weight were observed in male BALB/c mice after 1 week of exposure to 10 mg/kg/day PFOS via
gavage, this effect was attenuated at the end of the exposure period at 3 weeks {Lv, 2015,
3981558}. Additionally, a significant increase in body weight was observed in 2.5 mg/kg/day
exposure group at the end of the 3-week exposure period {Lv, 2015, 3981558}. Food
consumption was not reported in these studies {Qu, 2016, 3981454; Lv, 2015, 3981558}. No
change in body weights were observed across 8 timepoints in male ICR mice exposed to
0.5 mg/kg/day, 5 mg/kg/day, or 10 mg/kg/day by oral gavage for 28 days {Qiu, 2016 3981408}.
Three studies using Sprague-Dawley rats reported decreased body weights following PFOS
exposure via oral gavage for 28 days, which usually occurred at the highest dose tested. Of these,
Han et al. (2018, 4355066) and Wan et al. (2016, 3981504) exposed males to 1 mg/kg/day or
10 mg/kg/day and observed an approximate 10% reduction in body weight following
10 mg/kg/day. NTP (2019, 5400978) reported decreased body weights in male and female
Sprague-Dawley rats exposed to 5 mg/kg/day PFOS. However, body weights of all dose male
and female groups were within 10% of control groups. The decrease in body weights was not
associated with reduced food consumption in Han et al. (2018, 4355066), and food consumption
was not reported in the other studies {Wan, 2016, 3981504; NTP, 2019, 5400978}. Two studies
by Salgado et al. (2015, 3981583; 2016, 3179088) using the same animals reported no change in
body weight variation or food consumption in male Sprague-Dawley rats administered
3 mg/kg/day or 6 mg/kg/day PFOS by oral gavage for 28 days, but data were not provided.
A reduction in body weight was also observed following 6 weeks of PFOS exposure via gavage
in male and female Crl:Cd(Sd)lgs Br Vaf rats exposed to 3.2 mg/kg/day (weighing 93 and 88%
of control, respectively), which was associated with decreased food consumption {Luebker,
2005, 1276160}. Although a 6-week exposure to 2 mg/kg/day did not reduce body weights in
female Crl:CD(SD)Igs Vaf/Plus rats, this dose did reduce mean female body weight gain and
food consumption {Luebker, 2005, 757857}. In a study assessing the dietary PFOS exposure in
the same rat strain, no change was observed in body weights or food consumption in male and
female Crl:CD(SD)IGS BR rats exposed to PFOS in the diet at concentrations of 0 ppm,
0.5 ppm, 2 ppm, 5 ppm, or 20 ppm (equivalent to 0 mg/kg, 0.05 mg/kg, 0.18 mg/kg, 0.37 mg/kg,
or 1.51 mg/kg in males and 0 mg/kg, 0.05 mg/kg, 0.22 mg/kg, 0.47 mg/kg, or 1.77 mg/kg in
females) for 4 weeks {Seacat, 2003, 1290852}.
Chronic PFOS exposure studies also suggest an effect of PFOS on body weight. Male and female
Cynomolgus monkeys exposed to 0 mg/kg/day, 0.03 mg/kg/day, 0.15 mg/kg/day, or
0.75 mg/kg/day PFOS (equivalent to cumulative doses of 0 mg/kg, 4.6 mg/kg, 22.9 mg/kg, or
114.7 mg/kg) via intragastric intubation for 26 weeks (182 days) showed a reduction in body
weight change in the highest dose group (8% reduction in males and 4% reduction in females),
although no change in absolute body weight was observed {Seacat, 2002, 757853}. This is in
contrast to the 14% and 5% body weight increases in control males and females, respectively.
However, chronic (14 weeks) exposure to PFOS in the diet at 0 ppm, 0.5 ppm, 2 ppm, 5 ppm,
and 20 ppm (equivalent to 0 mg/kg, 0.05 mg/kg, 0.18 mg/kg, 0.37 mg/kg, and 1.51 mg/kg in
males and 0 mg/kg, 0.05 mg/kg, 0.22 mg/kg, 0.47 mg/kg, and 1.77 mg/kg in females) showed
had no effect on Crl:CD(SD)IGS BR male or female rats. For 20 ppm dose-group males,
terminal body weights appeared to be reduced in a dose-dependent manner, however this
difference was not statistically significant {Seacat, 2003, 1290852}. In line with reduced body
weights, food consumption was significantly decreased in the 20 ppm exposure group, but these
C-102
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
data were not shown and the sex of the animals affected was not specified {Seacat, 2003,
1290852}.
Kndpoint
Study Name
Study Design
Observation Tin
le Animal Description
Body Weight Change
Xing eta!.. 2016, 3981506
subchronic (30d)
3 Id
Mouse, C57BL/6J (cf, N=2)
Luebker etal., 2005, 757857
reproductive (76-8 Id)
l-42d
PO Rat. Crl:Cd(Sd)lgs Vaf/Plus (9. N=20-28)
Body Weight, Absolute
Seacat et al., 2002,757853
chronic (26wk)
184d
Monkey, Cynomolgus (cf, N=4-6)
Monkey, Cynomolgus (9, N=4-6)
Qiuctal., 2016. 3981408
short-term (4wk)
4wk
Mouse, ICR (cf. N=10)
Qu et al., 2016. 3981454
subchronic (35d)
35d
Mouse, C57 (Cf, N=10)
Zhong et al., 2016, 3748828
developmental (GD1-17)
PNW8
Fl Mouse, C57BL/6 (cf, N=12)
F1 Mouse. C57BL/6(9,N=12)
Han et al.,2018,4355066
short-term (28d)
28d
Rat, Sprague-Dawley (cf, N=6)
Salgadoel al., 2015, 3981583
short-term (28d)
28d
Rat. Sprague-Dawley (cf, N=7)
Wan etal., 2016, 3981504
short-term (28d)
28d
Rat, Sprague-Dawley (cf, N=5)
Seacat et al., 2003, 1290852
short-term (4wk)
4wk
Rat, Crl:CD(SD)IGS BR (cf, N=5)
Rat, Crl:CD(SD)IC.S BR (9, N=5)
chronic (14wk)
I4wk
Rat, Crl:CD(SD)IGS BR (cf, N=5)
Rat, Crl:CD(SD)IGS BR (9- N=5)
NTP, 2019, 5400978
short-term (28d)
29d
Rat, Sprague-Dawley (cf, N=10)
Rat, Sprague-Dawley (9, N=9-10)
Salgado et al., 2016, 3179088
short-term (28d)
29d
Rat, Sprague-Dawley (cf, N=6)
Luebker et al.. 2005, 1276160
reproductive (56d)
42d
P0 Rat, Crl:Cd (Sd)lgs Br Vaf (Cf. N=35)
reproductive (42d prior mating-LD20)
42d
P0 Rat, Cri:Cd (Sd)Igs Br Vaf (9. N=35)
reproductive (GD0-PND112)
PND85-97
Fl Rat, Crl:Cd (Sd)Igs Br Vaf (cf, N=22-25)
Fl Rat, Crl:Cd (Sd)Igs Br Vaf (9, N=22-25)
Butenhoff ct al., 2009,757873
developmental (GD0-PND20)
PND72
Fl Rat, CrhCD(SD) (Cf, N=20)
I'FUS Whole Body Eflects - Body Weights
£ No significant change Significant in
Significant decieas
V V
-¥
-W
-V
-¥
log Concentration (nig/kg/day)
Figure C-24. Effects on Body Weight in Rodents and Non-Human Primates Following
Exposure to PFOS (logarithmic scale)
PFOS concentration is presented in logarithmic scale to optimize the spatial presentation of data.
Interactive figure and additional study details available on HAWC.
GD = gestation day; PNW = postnatal week; PND = postnatal day; LD = lactation day; d = day; wk = week.
C.3.2.2.3 Body Weight in Adults Following Developmental Exposure
Offspring body weights during developmental periods have been reported and described (See
PFOS Main Document). However, the effects on body weight may not persist into adulthood. No
change was observed in adult body weight (PND 85-PND 97) compared to control in male and
female Crl:CD(SD)Igs Br Vaf rats exposed perinatally through adulthood to 0.1 mg/kg/day and
0.4 mg/kg/day PFOS {Luebker, 2005, 1276160}. Developmental (GD 1-GD 17) PFOS exposure
in C57BL/6 mice at 0.1 mg/kg/day, 1 mg/kg/day, or 5 mg/kg/day was not observed to affect
male or female body weight at PNW4 or PNW8 {Zhong, 2016, 3748828}. Similarly, body
weights from birth to PND 70 were not statistically different from controls in the offspring of
female Sprague-Dawley rats exposed to 0 mg/kg/day, 0.1 mg/kg/day, 0.3 mg/kg/day, or
1 mg/kg/day PFOS from GD 0-PND 20 {Butenhoff, 2009, 757873}.
C.3.2.2.4 Food Consumption
Although there is some evidence that short-term and subchronic exposure of rodents to PFOS
can lead to reductions in food consumption, this effect is not consistently observed across all
exposures and strains tested. Food consumption was decreased in C57BL/6J male mice exposed
to 2.5, 5, or 10 mg/kg/day PFOS by oral gavage for 30 days at all three doses {Xing, 2016,
3981506}. Decreased food consumption was also observed in female and male Crl:Cd(Sd)lgs Br
Vaf rats following a 6 week exposure via gavage to 1.6 mg/kg/day or 3.2 mg/kg/day {Luebker,
C-103
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
2005, 1276160}, and in female Crl:Cd(Sd)lgs Vaf/Plus rats following a 6 week exposure to
2.0 mg/kg/day {Luebker, 2005, 757857} (see PFOS Main Document).
Food and water consumption was not observed to be affected in Sprague Dawley rats exposed to
PFOS via gavage at doses of 1 mg/kg/day or 10 mg/kg/day {Han, 2003, 4355066}, 3 or
6 mg/kg/day {Salgado, 2015, 3981583}, nor 0.5 mg/kg/day, 1 mg/kg/day, 3 mg/kg/day, or
6 mg/kg/day {Salgado, 2016, 3179088} for 28 days. Seacat et al. (2003, 1290852) fed
Crl:CD(SD)IGS Br male or female rats 0, 0.5, 2, 5, and 20 ppm PFOS for 4 or 14 weeks
(equivalent to 0, 0.05, 0.18, 0.37, and 1.51 mg/kg in males and 0, 0.05, 0.22, 0.47, and
1.77 mg/kg in females). The authors noted that food consumption was slightly reduced in the
20 ppm female dose group during the first 4 weeks of dosing, but these data were not provided
{Seacat, 2003, 1290852}. By 14 weeks, food consumption was noted to be significantly
decreased in the 20 ppm dose group, but these data were not provided and the sex of the animals
affected was not specified.
C.3.3 Mechanistic Evidence
Mechanistic evidence linking PFOS exposure to adverse metabolic outcomes is discussed in
Sections 3.2.2, 3.3.2, and 3.3.4 of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365}. There are
32 and 36 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD that investigated the mechanisms of action of PFOS that
lead to metabolic and systemic effects, respectively. A summary of these metabolic and systemic
studies is shown in Figure C-25 and Figure C-26, respectively. Additional mechanistic synthesis
will not be conducted since evidence suggests but is not sufficient to infer that PFOS leads to
metabolic and systemic effects.
Mechanistic Pathway Animal Human In Vitro Grand Total
Big Data, Non-Targeted Analysis
0
2
1
3
Cell Growth, Differentiation, Proliferation, Or Viability
1
0
11
12
Cell Signaling Or Signal Transduction
3
1
8
11
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
7
1
8
15
Hormone Function
1
4
3
8
Oxidative Stress
2
1
2
5
Xenobiotic Metabolism
0
0
2
2
Other
2
0
0
2
Not Applicable/Not Specified/Review Article
1
0
0
1
Grand Total
11
7
16
32
Figure C-25. Summary of Mechanistic Studies of PFOS and Metabolic Effects
Interactive figure and additional study details available on Tableau.
C-104
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mechanistic Pathway Animal Human In Vitro Grand Total
Atherogenesis And Clot Formation
0
0
1
1
Big Data, Non-Targeted Analysis
3
0
2
4
Cell Growth, Differentiation, Proliferation, Or Viability
4
1
11
16
Cell Signaling Or Signal Transduction
2
1
7
10
Fatty Acid Synthesis, Metabolism, Storage. Transport, Binding, B-Oxidation
4
1
9
13
Hormone Function
1
0
0
1
Inflammation And Immune Response
0
0
2
2
Oxidative Stress
5
1
7
13
Xenobiotic Metabolism
3
1
1
4
Other
0
0
4
4
Not Applicable/Not Specified/Review Article
2
0
0
2
Grand Total
10
3
25
36
Figure C-26. Summary of Mechanistic Studies of PFOS and Systemic Effects
Interactive figure and additional study details available on Tableau.
C.3.4 Evidence Integration
There is slight evidence of an association between PFOS exposure and metabolic effects in
humans based on observed effects for diabetes, gestational weight gain, HOMA-IR, HOMA-B,
leptin, and adiponectin in high and medium confidence studies. Five studies observed non-
significant positive associations with gestational diabetes. In the general population, six studies
reported positive associations with type 2 diabetes. Three epidemiological studies observed
positive associations with gestational weight gain. Seven studies reported non-significant
positive associations with HOMA-IR in pregnant women and in general populations, or in adults
at high risk for type 2 diabetes. Of the six studies on HOMA-IR in children, only one reported a
positive association with HOMA-IR. Four studies reported positive associations with HOMA-B,
but an inverse association was observed in children (one study). There is limited evidence
suggests a potential association between PFOS exposure and adiponectin in children, but not
adults. Findings for an association between PFOS exposure and metabolic syndrome were mixed
in four general population epidemiological studies identified since 2016: two reported negative
associations with metabolic syndrome, and two reported positive associations.
The animal evidence for an association between PFOS and systemic or metabolic effects is
indeterminate. Although some alterations related to glucose homeostasis were reported in the
available animal toxicity literature, the results from 6 high or medium confidence studies are
inconclusive as there are too few studies to assess possible difference across life stages, sexes,
and species. In addition, the effects on body weight, clinical observations, and mortality from 20
high or medium confidence studies indicate that the systemic effects occur only at the high doses
tested. NTP (2019, 5400978) and Seacat et al. (2003, 1290852) reported differing observations
on the impact of PFOS on serum glucose in male rats at 4 weeks, which may be explained by
differing methods of exposure (gavage and dietary, respectively). Additionally, the statistically
C-105
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
significant observations reported by Seacat et al. (2003, 1290852) and Curran et al. (2008,
757871) differ between males and females, are not consistent across timepoints, and sometimes
did not follow a linear dose-response relationship. Given the differences noted in timing of
measurement, duration of exposure, and differences across sex, the biological significance of the
increase or decrease in metabolic endpoints such as serum glucose in these animal models is
unclear, especially considering the sensitivity of these parameters to increases in animal stress.
There were also inconsistencies in results reported in developmental studies. Lv et al. (2013,
2850947) reported dose-dependent increases in serum glucose during a glucose tolerance test at
PNW10 in rat offspring. This trend did not continue through PNW 15 in this study. In addition,
Wan et al. (2014, 2850405) did not report significantly altered results of the glucose tolerance
test at PND 63 in mouse offspring gestationally exposed to PFOS and fed standard diets.
Although multiple studies indicate potential effects of PFOS on glucose homeostasis, the
responses were inconsistent and/or transient for specific endpoints across studies and the
biological significance of the observed effects is uncertain.
Though the observed metabolic effects were inconsistent, evidence from animal studies suggests
that PFOS exposure may induce whole-body toxicity, but only at the higher doses tested.
Decreased body weight and food consumption were observed in a number of subchronic and
chronic studies using rodents and non-human primates. While signs of decreased body weights
can be indicative of poor health in animals and a relevant endpoint demonstrating whole body
toxicity, the effects reported in these studies were generally minimal and only surpassed a >10%
change in body weight at the highest doses tested.
C.3.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOS exposure has the potential to cause systemic and metabolic
effects in humans under relevant exposure circumstances (Table C-6). This conclusion is based
primarily on diabetes, gestational weight gain, HOMA-IR, HOMA-B, leptin, and adiponectin
effects observed in high and medium confidence studies in humans exposed to median PFOS
levels between 5.4 and 35.7 ng/mL. Although there is some evidence of negative effects of PFOS
exposure on metabolic syndrome, there is considerable uncertainty in the results due to
inconsistency across studies and limited number of studies.
C-106
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-6. Evidence Profile Table for PFOS Systemic and Metabolic Effects
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.3.1)
Glucose metabolism
4 High confidence
studies
13 Medium confidence
studies
7 Low confidence Studies
Findings for FBG in adults
were primarily positive
(7/12), but only a few
reached significance.
OGTT results were
examined only in studies
finding significant
increases in FBG and were
congruent with FBG
findings. In children,
decreases in FBG were
observed (3/5), but none
were significant. Findings
for FBG in pregnant
women were similarly
non-significantly inverse
(3/4), however, the three
high and medium
confidence studies
conducting OGTT
observed increases in 1-
hour glucose levels, two of
which were significant.
• High and medium
confidence studies
• Consistent direction
of effect for FBG in
adults
• Low confidence studies
• Lmprecision of findings
• Potential for selection
bias and residual
confounding by SES
Diabetes (and
gestational diabetes)
3 High confidence
studies
16 Medium confidence
studies
5 Low confidence studies
Findings in adults were
mixed. Among the high
and medium confidence
studies (8/11), two
reported significant
positive associations (2/8),
1 reported a significant
inverse association (1/8),
and 5 reported imprecise
associations (5/8). The 3
• High and medium
confidence studies
• Low confidence studies
• Inconsistent direction of
effect
• Lmprecision of findings
• Potential for outcome
misclassification, self-
selection, residual
confounding by SES,
©OO
Slight
Evidence for metabolic
effects is based on
increases in FBG,
increased odds of diabetes,
and increases in measures
of adiposity in adults.
Positive associations were
reported for heightened
glucose levels, effects on
insulin regulation,
diabetes, and adiposity,
but many medium and
high confidence studies
presented non-statistically
significant results and
several studies presented
conflicting associations.
Uncertainties remain due
to mixed results,
contrasting findings, and
potential for residual
confounding in the
analysis of outcomes such
as glucose metabolism,
diabetes, and insulin
levels.
©OO
Evidence Suggests
Primary basis:
Human evidence indicted
effects on diabetes,
gestational weight gain,
HOMA-IR, HOMA-B,
leptin, and adiponectin and
there was limited animal
evidence. Although there is
some evidence of negative
effects of PFOS exposure
on metabolic syndrome,
there is considerable
uncertainty in the results
due to inconsistency across
studies and limited number
of studies.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
C-107
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
low confidence studies all
reported non-significant
positive associations and
typically relied on self-
reported data. Findings for
HbAlc were less
consistent. In pregnant
women, findings for
gestational diabetes were
mixed. The only study
examining diabetes in
children was considered
uninformative.
and failure to establish
temporality
Insulin levels
2 High confidence
studies
7 Medium confidence
studies
10 Low confidence
studies
Findings from a high
confidence study in adults
reported significant
increases in fasting
insulin, HOMA-IR,
HOMA-B, and insulin
responses during an
OGTT, however, this
population was at high
risk for type 2 diabetes.
Findings for adults among
medium and low
confidence studies were
generally mixed, but there
were multiple contrasting
findings for HOMA-IR,
indicating an inverse
association (5/9). Studies
in children reported mixed
and generally imprecise
findings for measures of
insulin resistance.
»High and medium
confidence studies
• Low confidence studies
• Lnconsistent direction of
effects
• Lmprecision of findings
• Potential for residual
confounding by diabetes
status or use of
medications that would
impact insulin levels in
some studies
C-108
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Similarly, findings in
studies among pregnant
women were imprecise.
Adiponectin and leptin
5 High confidence
studies
3 Medium confidence
studies
1 Low confidence study
Inverse associations with
adiponectin were reported
in two studies of adults
(2/2), while one study
(1/1) reported increases in
leptin. None reached
significance. Findings for
adiponectin in children
were positive (5/6), but
only one reached
significance. Findings for
leptin were mixed among
children. Only one study
reported findings from
pregnant women,
observing non-significant
increases in both
adiponectin and leptin.
• High and medium
confidence studies
• Consistent direction
of effect for
adiponectin in
children
• Low confidence study
• Lmprecision of findings
Adiposity
4 High confidence
studies
17 Medium confidence
studies
4 Low confidence studies
In adults, findings for BMI
were primarily positive
(4/6), indicating increased
BMI. Increases in the odds
of being overweight or
obese were also reported,
which was significant for
women in one study.
Results were mixed for
WC, but one study
observed differences in
direction of effect between
men and women. Findings
for BMI in children were
• High and medium
confidence studies
• Low confidence studies
• Lnconsistent direction of
effects
• Lmprecision of findings
C-109
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
mixed, with studies of
medium confidence
reporting significant
positive and significant
inverse associations with
measures of BMI. In
pregnant women, positive
associations were reported
for gestational weight
gain, but results were
inconsistent between
studies after stratification
of weight status (i.e.,
under-, normal-, or over-
weight^
Metabolic syndrome
4 Medium confidence
studies
1 Low confidence study
In adults, findings for
metabolic syndrome were
mixed, and none reached
significance (0/4).
Significant reduction in
the resting metabolic rate
were observed in a single
study of adults. MetS was
not evaluated in children
or pregnant women.
• Medium confidence
studies
• Low confidence study
• Lnconsistent direction of
effects in medium
confidence studies
• Concern for selection
bias, outcome
misclassification, and
residual confounding by
SES in low confidence
study
Evidence from In Vivo Animal Studies (Section C.3.2.1 and Section C.3.2.2)
Glucose homeostasis
1 High confidence study
5 Medium confidence
studies
Mixed results were
reported on glucose levels
in rodent studies (6). Of
these, 2 reported non-
significant effects, and 4
reported significant effects
with inconsistent
directionality. Reduced
glucose levels were
»High and medium
confidence studies
• Lnconsistent direction
and magnitude of effects
across study designs- and
sex
• Limited number of
studies examining
outcomes
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
ooo
Lndeterminate
Alterations related to
glucose homeostasis were
reported in 6 high or
medium confidence
studies were inconclusive
C-110
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
reported in female rodents
(3/4) at the highest PFOS
exposure group tested. No
significant effects on
glucose levels were
observed in males (3/3)
and dams (1/2). One study
in female mice reported
decreased insulin
resistance (1/1) and
pyruvate tolerance (1/1).
as there are too few
studies to assess possible
difference across life
stages, sexes, and species
and results from the
existing studies are
inconsistent or transient.
Systemic effects (e.g.,
body weight, clinical
observations, survival,
food consumption, and
Body weight
3 High confidence
studies
17 Medium confidence
studies
Statistically significant
reductions in body weights
(9/20) and body weight
changes (2/2) were
reported in various
studies, including studies
in rats (11), mice (9), and
monkeys (2).
• High and medium
confidence studies
• Consistent direction
of effects
• Confounding
variables such as
food consumption
were considered in
most studies
•
• Effects do not follow a
linear dose-responsive
relationship
water consumption) from
20 high or medium
confidence studies
indicate that biologically
significant effects (e.g.,
body weight change
exceeding 10% of control)
tend to occur only at the
highest doses tested.
Survival and mortality
1 High confidence
studies
6 Medium confidence
studies
No effects on survival and
mortality were reported in
rodent studies (6/6). One
study in non-human
primates observed
increased mortality at the
highest dose tested (1/1).
• High and medium
confidence studies
• Consistent direction
of effects across sex,
species, and duration
of exposure
• Limited number of
studies examining
outcomes
Evidence Integration
Summary Judgment
Clinical observations
1 High confidence study
3 Medium confidence
studies
Clinical observations were
observed in most rodent
studies (3/4). Findings
found across these studies
included: hyperkalemia,
rough hair, slow
movement, constipation,
»High and medium
confidence studies
• Limited number of
studies examining
outcomes
• Qualitative and
subjective data reporting
C-lll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Summary and Key
Factors that Increase
Factors that Decrease
Evidence Stream
Interpretation
Findings
Certainty
Certainty
Judgment
and localized areas of
partial alopecia.
Evidence Integration
Summary Judgment
Food and water
consumption
9 Medium confidence
studies
Reduced food
consumption (6/9) was
reported in the higher dose
groups in male and female
rodents. No significant
effects were reported on
water consumption in
male rats following short-
term exposure (2/2).
• Medium confidence
studies
• Consistent direction
of effects on water
consumption
• Limited number of
studies examining
outcomes
Notes: FBG = fasting blood glucose; OGTT = oral glucose tolerance testing; HbAlc = hemoglobin Ale; SES = social economic status; HOMA-IR = homeostatic model
assessment for insulin resistance; HOMA-B = homeostasis model assessment of P-cell function; BMI= body mass index; WC = waist circumference; MetS = metabolic syndrome.
C-112
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.4 Nervous
EPA identified 36 epidemiological and 16 animal studies that investigated the association
between PFOS and nervous effects. Of the epidemiological studies, 3 were classified as high
confidence, 28 as medium confidence, and 5 were considered low confidence (Section C.4.1). Of
the animal studies, 1 was classified as high confidence, 8 as medium confidence, 4 as low
confidence, 2 as mixed (2 medium/low) confidence, and 1 was considered uninformative (Section
C.4.2). Studies may have multiple judgments depending on the endpoint evaluated. Though low
confidence studies are considered qualitatively in this section, they were not considered
quantitatively for the dose-response assessment (See Main PFOS Document).
C.4.1 Human Evidence Study Quality Evaluation and
Synthesis
C.4.1.1 Introduction
The 2016 Health Assessment {U.S. EPA, 2016, 3603365} reviewed studies examining
associations between PFOS exposure and neurodevelopmental disorders in children, including
attention deficit hyperactivity disorder (ADHD) and learning disabilities and concluded there
was limited evidence to suggest an effect. A significant increase in risk of development of
cerebral palsy in males was observed in a case-control study of maternal PFOS levels of
participants within the DNBC {Liew, 2014, 2852208)}. One study observed a significant
positive association of child PFOS levels with parent reported ADHD in children aged 12-15 in
the general population {Hoffman, 2010, 1291112}. No association between maternal plasma
PFOS concentrations and Apgar score or between maternal plasma PFOS concentrations and
mother reported assessments of fine motor skills, gross motor skills or cognitive skills in children
at 6 and 18 months of age were observed in one study of pregnant women and their children
{Fei, 2008, 1290822}. No association between parent reported behavioral or coordination
problems in children 7 years of age and prenatal PFOS levels was reported in another study {Fei,
2011, 758428}. No associations were observed between prenatal PFOS and parent reported
motor development scores in children ages 7 to 9; however, the highest PFOS tertile was
associated with a 0.5-point higher hyperactivity score for participants within one country with
higher exposures, but not for participants within other countries {Hoyer, 2015, 2851038}. Data
interpretations within these studies were limited in some cases by use of a cross-sectional study
design {Fei, 2008, 1290822; Hoffman, 2010, 1291112}, potential random misclassification error
resulting from using current PFOS levels as proxy measures of etiologically relevant exposures
{Hoffman, 2010, 1291112}, outcomes defined by parental report {Fei, 2008, 1290822; Fei,
2011, 758428; Hoyer, 2015, 2851038; Hoffman, 2010, 1291112}, and limited sample sizes in
some countries {Hoyer, 2015, 2851038}.
For this updated review, 35 studies (35 publications) investigated the association between PFOS
and neurological outcomes that have been identified since the 2016 document. One was
conducted in a high-exposure community {Spratlen, 2020, 6364693}. One publication {Vuong,
2020, 356876} was conducted in pregnant women. The remainder were conducted in the general
population. Study designs included 3 case-control {Ode, 2014, 2851245; Long, 2019, 5080602;
Shin, 2020, 6507470}, 2 nested case-control {Liew, 2015, 2851010; Lyall, 2018, 4239287}, 26
cohort, and 5 cross-sectional studies (Appendix D). The studies measured PFOS in different
C-113
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
matrices including blood, serum, plasma, cord blood, breast milk {Forns, 2015, 3228833;
Lenters, 2019, 5080366}, maternal serum, maternal plasma, and amniotic fluid {Long, 2019,
5080602}. Several studies {Braun, 2014, 2345999; Vuong, 2016, 3352166; Vuong, 2018,
5079675; Vuong, 2018, 5079693; Vuong, 2019, 5080218; Vuong, 2020, 6356876; Vuong, 2020,
6833684; Zhang, 2018, 4238294} were conducted on subsets of data from the HOME study.
Two studies {Forns, 2015, 3228833; Lenters, 2019, 5080366} utilized data from the Norwegian
Human Milk Study (HUMIS). Two studies {Liew, 2015, 2851010; Liew, 2018, 5079744}
utilized the DNBC data. The studies were conducted in multiple locations including populations
from China, Denmark, the Faroe Islands, Great Britain, Japan, the Netherlands, Norway,
Sweden, Taiwan, and the United States (Appendix D). Neurological effects were determined for
numerous clinical conditions and by assessing performance on neuropsychological tests
assessing various neurological domains, including developmental, general intelligence (i.e.,
intelligence quotient (IQ)), social-emotional, executive function, ADHD and attention, autism
spectrum disorder (ASD) and intellectual disability (ID), and visuospatial performance.
C.4.1.2 Study Quality
There are 34 studies (36 publications)5 from recent systematic literature search and review
efforts conducted after publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that
investigated the association between PFOS and nervous effects. Study quality evaluations for
these 36 studies are shown in Figure C-27 and Figure C-28.
Of the 36 studies identified since the 2016 assessment, three {Niu, 2019, 5381527; Oulhote,
2016, 3789517; Harris, 2018, 4442261} were classified as having high confidence, 28 studies
were classified as medium confidence, and five were low confidence. Studies rated as low
confidence had deficiencies including potential residual confounding, exposure misclassification,
selection bias, and small sample size. One low confidence NHANES study {Berk, 2014,
2713574} had a high likelihood of residual confounding due to the use of an insensitive marker
of SES, and the analysis did not account for the population's complex sampling design.
Differences in laboratory extraction methods, collection timing, and missing details on storage
raised concerns for exposure misclassification in a study on children from the HUMIS cohort
{Forns, 2015, 3228833}. Additionally, children were only evaluated on some, but not all, test
instrument (Ages and Stages Questionnaire (ASQ)) domains, and rationale for domain selection
was not provided. Concerns for Lien et al. (2016, 3860112) included a high loss to follow-up,
lack of detail on completion rates of ADHD questionnaires and low detection rate for PFOS.
Small sample size, temporality and reporting issues were cited as limitations in Weng, 2020,
6718530. Finally, limitations in Ode et al. (2014, 2851245) included sensitivity concerns due to
the limited number of ADHD cases and potential for residual confounding due to the lack of data
on other exposures potentially related to ADHD. In the evidence synthesis below, high, and
medium confidence studies were the focus, although low confidence studies were still considered
for consistency in the direction of association.
5 Vuong et al. (2018, 5079675) reports score trajectories for the same population and test as Vuong et al. (2016, 3352166). Vuong
et al. (2020, 6833684) reports on an overlapping population with the same test as Zhang et al. (2018,4238294).
C-114
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Berk et al., 2014, 2713574-
i
+
J
+
++
i,
i
i
,i.
Braun et al., 2014, 2345999-
++
+
+
+
+
+
+
+
Chen et a!., 2013, 2850933-
+
+
++
+
+
+
Ding and Park, 2020, 6711603-
+
+
+
+
+
+
+
Forns et al., 2015, 3228833 -
+
-
+
+
++
+
+
-
Gallo et al., 2013, 2272847-
++
+
-
+
+
+
+
+
Ghassabian et al., 2018, 5080189 -
+
+
+
+
++
+
+
+
Goudarzi et al., 2016, 3981536 -
+
+
++
+
+
+
+
+
Harris et al., 2018, 4442261 -
+
+
++
++
++
+
+
++
Jeddy et al., 2017, 3859807-
+
++
+
+
+
+
+
+
Lenters et al., 2019, 5080366 -
+
+
+
++ ++
+
+
+
Li, 2020, 6833686-
+
+
+
+
+
+
+
+
Lien etal.,2016, 3860112-
-
-
+
+
+
+
-
-
Liew et al., 2015, 2851010-
+
+
+
++
+
+
+
Liew et al.,2018, 5079744-
+
++
+
++
+
+
+
Long et al., 2019, 5080602-
+
+
+
+
+
+
-
+
Lyall et al.,2018, 4239287-
+
+
+
++
+
+
+
Niu et al., 2019, 5381527-
+
+
+
++
+
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
Figure C-27. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Neurological Effects
Interactive figure and additional study details available on HAWC.
C-115
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
,0d»o
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.4.1.3 Findings from Children and Adolescents
Six cohort studies {Goudarzi, 2016, 3981536; Chen, 2013, 2850933; Jeddy, 2017, 3859807;
Forns, 2015, 3228833; Niu, 2019, 5381527; Shrestha, 2017, 3981382}, and one high-exposure
community study {Spratlen, 2020, 63646931 examined developmental outcomes in children. In a
high confidence study {Niu, 2019, 5381527} from the Shanghai-Minhang Birth Cohort Study
(SMBCS), maternal PFOS concentrations (median = 10.8 ng/mL) during pregnancy were
inversely associated with neuropsychological development (especially for personal-social skills)
assessed by the ASQ in 4-year old children. A medium confidence study of data from the Taiwan
Birth Panel Study {Chen, 2013, 2850933} observed associations between in utero PFOS
(mean = 7.4 ng/mL) and decreases in Comprehensive Developmental Inventory (CDI)
developmental quotients in the highest exposure group compared with the lowest exposure group
for the whole test as well as for gross motor, fine motor, and self-help domains. Effect sizes were
generally greater with increasing PFOS levels. A medium confidence study {Jeddy, 2017,
3859807} utilizing data from the ALSPAC observed significant associations between maternal
PFOS (median = 19.8 ng/mL) and verbal comprehension scores as assessed by the adapted
MacArthur Communicative Development Inventories for Infants (MCDI) in children at
15 months of age, but not for vocabulary comprehension and production, nonverbal
communication, or social development. Significant inverse associations were also observed
between maternal PFOS and language and intelligibility scores in children at 38 months of age.
Results for this study varied by maternal age at delivery. A statistically significant inverse
association was reported for vocabulary comprehension and production scores in 15-month
infants with mothers < 25 years of age. A significant inverse association was observed for
intelligibility scores in children 38 months of age with mothers >30 years of age, and a
significant positive association was observed for intelligibility scores in children 38 months of
age with mothers < 25 years of age. Results from a medium confidence study {Goudarzi, 2016,
3981536} reported no significant associations between prenatal PFOS levels
(median = 5.7 ng/mL at 6 months; median = 5.8 at 18 months) and Mental (MDI) and
Psychomotor (PDI) Development Indices in infants at 6 and 18 months. Similarly, no significant
adverse associations or apparent trends between delivery or cord blood PFOS concentrations
(median = 6.0 ng/mL) and age 1 mental or psychomotor developmental indices were reported in
a high-exposure community study of children prenatally exposed to the World Trade Center
(WTC) Disaster, however a significant interaction by sex with MDI at ages 2 and 3, with
stronger positive associations for females compared with males was observed {Spratlen, 2020,
6364693}.
Ten studies evaluated cognitive function and IQ measures among children, with most conducted
within the general population {Vuong, 2020, 6833684; Zhang, 2018, 4238294; Strom, 2014,
2922190; Harris, 2018, 4442261; Oulhote, 2019, 6316905; Liew, 2018, 5079744; Vuong, 2019,
5080218; Wang, 2015, 3860120; Skogheim, 2019, 5918847}, and one within a high-exposure
community {Spratlen, 2020, 6364693}. In a high confidence analysis of participants within
Project Viva, children born to women with top quartile PFOS (34.9-168.0 ng/mL)
concentrations had higher non-verbal IQ scores, although dose-response patterns appeared non-
linear {Harris, 2018, 4442261}. Positive associations were observed between prenatal PFOS
(median = 12.7 ng/mL) and reading skills at age eight years in a medium confidence study
{Vuong, 2020, 6833684} which utilized data from the HOME study. Childhood serum PFOS
concentrations at ages three and eight years were positively associated with higher children's
C-117
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
reading scores at ages five and eight years, respectively in an additional medium confidence
study of data within the HOME study {Zhang, 2018, 4238294}. No significant associations were
reported between maternal prenatal PFOS (median = 21.4 ng/mL) and offspring scholastic
achievement in a medium confidence prebirth cohort study of participants within the Danish
Fetal Origins 1988 (DaF088) cohort {Stram, 2014, 2922190}. Maternal prenatal PFOS
(median = 27.7 ng/mL) concentrations were associated with lower cognitive function as assessed
by the Boston Naming Test in a medium confidence study of children aged seven years
{Oulhote, 2019, 6316905}.
In a medium confidence study in a highly exposed community, sex-specific trends between
PFOS exposures and some cognitive outcomes (verbal and full-scale IQ only) at 4 and 6 years
were observed, suggesting stronger positive associations for females compared to males
{Spratlen, 2020, 6364693}. Another medium confidence study investigated associations between
prenatal exposure to PFOS and IQ at age five in a sample of children from the DNBC with no
consistent associations observed {Liew, 2018, 5079744}. Consistent adverse associations with
age eight cognitive development as assessed by IQ were not observed in an additional medium
confidence study {Vuong, 2019, 5080218}. Similarly, utilizing data from participants within the
Taiwan Maternal and Infant Cohort Study, a medium confidence prospective cohort study by
Wang {Wang, 2015, 3860120} reported no significant associations between maternal serum
PFOS (median = 13.3 ng/mL) and IQ measurements in children five or eight years of age.
Evidence was inconsistent, with significant decreases in non-verbal working memory only in the
highest quintile and no significant associations with verbal working memory, for the evaluation
of the association between prenatal exposure to PFOS (median =11.5 ng/mL) and cognitive
dysfunction in preschool children in a medium confidence study from The Norwegian Mother,
Father, and Child Cohort Study (MoBa) {Skogheim, 2019, 5918847}.
Six studies assessed the relationship between PFOS and behavioral development problems and
behavioral regulation problems {Quaak, 2016, 3981464; Vuong, 2018, 5079693; Oulhote, 2019,
6316905; Ghassabian, 2018, 5080189; Oulhote, 2016, 3789517; Weng, 2020, 6718530}. No
significant associations between prenatal PFOS (1,650 ng/L) and externalizing problems at age
18 months assessed using the Child Behavior Checklist 1.5-5 (CBCL 1.5-5) were reported in a
high confidence study utilizing data from the Dutch cohort LINC (Linking Maternal Nutrition to
Child Health) {Quaak, 2016, 3781464}. No consistent associations in total Strengths and
Difficulties Questionnaire (SDQ) behavior scores with serum PFOS (median = 16.8 ng/mL) at
age five was observed, but a two-fold increase in serum PFOS (median = 15.3 |ig/L) in children
aged seven years was associated with higher SDQ total behavioral difficulties scores in girls, and
lower scores in boys (gender interaction p < 0.05) in a high confidence study {Oulhote, 2016,
3789517}. Maternal prenatal PFOS concentrations (median = 27.7 ng/mL) were positively
associated with total scores on the SDQ, indicating more behavioral problems, in a medium
confidence study of children seven years of age {Oulhote, 2019, 6316905}. Higher newborn
PFOS levels (median =1.7 ng/mL) in dried blood spots were associated with increased odds of
having behavioral difficulties, driven mostly by problems in conduct and emotional symptoms,
as assessed by the maternal completed SDQ at age 7 in another medium confidence birth cohort
study {Ghassabian, 2018, 5080189}. Child sex modified the associations between prenatal PFOS
and attention, with males having better performance than females, but not enough evidence was
observed to support an overall association between prenatal PFOS (median = 12.9 ng/mL) and
inattention and impulsivity as assessed by the Connors' Continuous Performance Test-II in a
C-118
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
medium confidence study {Vuong, 2018, 5079693}. A low confidence study on adolescents
reported a significant, inverse correlation between prenatal PFOS levels (mean = 14.85 ng/mL)
and in the right putamen brain region associated with impulsive behavior as assessed by MRI in
teenage offspring {Weng, 2020, 6718530}.
One medium confidence study {Strom, 2014, 2922190} from the DaF088 cohort examined the
association between prenatal PFOS exposure and depression among offspring with 20 years of
follow-up. No significant association was observed between clinical depression and maternal
PFOS (median = 21.4 ng/mL) levels.
Three medium confidence studies {Vuong, 2016, 3352166; Vuong, 2018, 5079675; Shrestha,
2017, 3981382} examined the relationship between PFOS concentrations and executive function
in children with mixed results. Executive function was assessed with the parent-rated Behavior
Rating Inventory of Executive Function (BRIEF) in two studies {Vuong, 2016, 3352166; Vuong,
2018, 5079675} among HOME study participants at five and eight years of age. Higher BRIEF
scores indicate executive function impairments. Maternal serum PFOS concentrations were
significantly associated with poorer behavior regulation, metacognition, and global executive
functioning, with approximately a 3-point increase in all summary measures with a 1 ln-unit
increase in PFOS concentrations {Vuong, 2016, 3352166}. Vuong et al. (2018, 5079675) again
utilized data from the HOME study in a medium confidence cross-sectional analysis to examine
associations of child PFOS levels measured in children aged eight years with executive function
and reported no significant associations between PFOS and executive function.
Five medium confidence studies assessed relationships between PFOS exposures and ADHD
{Stram, 2014, 2922190; Liew, 2015, 2851010; Quaak, 2016, 3981464; Skogheim, 2019,
5918847; Lenters, 2019, 5080366}. One medium confidence study {Lenters, 2019, 5080366}
examined early life high PFOS exposures in breast milk in relation to ADHD among children
(range: 7.2-14.1 years old) from the HUMIS and reported significant associations with PFOS
concentrations (median = 117.7 ng/L) and increased odds of ADHD (OR = 1.75, 95% CI: 1.11,
2.76) with significant sex-specific effects. Stram et al. (2014, 2922190) investigated the
association between maternal prenatal PFOS and ADHD among offspring (follow-up to age 20)
of participants within the DaF088 cohort. No significant association between maternal PFOS
(median = 21.4 ng/mL) and offspring ADHD was reported in this medium confidence study. A
medium confidence nested case-control study {Liew, 2015, 2851010} within the framework of
the DNBC examined prenatal PFOS exposures and ADHD in children. No consistent evidence
was observed to suggest that prenatal PFOS exposures (ADHD cases median = 26.8 ng/mL;
controls median = 27.4 ng/mL) increase the risk of ADHD. Quaak et al. (2016, 3981464)
explored the relationship between prenatal PFOS exposures and parent-reported ADHD using
the CBCL 1.5-5. This medium confidence study utilized data from the Dutch cohort, LINC. No
significant associations were reported between cord blood PFOS (median = 1,600 ng/L)
exposures and ADHD scores in the whole population or in the sex-stratified analyses.
Two low confidence studies {Ode, 2014, 2851245; Lien, 2016, 3860112} examined PFOS
exposures in relation to ADHD. Ode et al. (2014, 2851245) investigated the association in a
case-control study between cord blood PFOS (median = 6.9 ng/mL for cases, 6.8 ng/mL for
controls) exposures and ADHD diagnosis in childhood (age range 5-17 years), but no
associations between PFOS and ADHD were observed. Lien, 2016, 3860112 evaluated the
association between cord blood PFOS (mean = 4.8 ng/mL) exposures and neurobehavioral
C-119
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
symptoms related to ADHD among 7-year old participants from the Taiwan Birth Panel Study
and the Taiwan Early-Life Cohort, but no effects were observed.
One high {Oulhote, 2016, 3789517}and five medium confidence studies since the 2016
assessment evaluated PFOS exposures in relation to autism, autistic behaviors, and ID {Braun,
2014, 2345999; Liew, 2015, 2851010; Long, 2019, 5080602; Lyall, 2018, 4239287; Shin, 2020,
6507470}. A two-fold increase in serum PFOS (median = 15.26 |ig/L) at age seven was
associated with significantly higher SDQ autism screening scores at age seven, with higher
autism scores in females than in males, in a high confidence study {Oulhote, 2016, 3789517}. In
a medium confidence prospective birth cohort study from the HOME study, increasing maternal
serum PFOS concentrations (median =13 |ig/L) were associated with increased autistic
behaviors in children 4 to 5 years of age as assessed by maternal completed Social
Responsiveness Scale (SRS) scores, although not significantly so, and PFOS levels were
positively associated with SRS scores in boys, but not girls {Braun, 2014, 2345999}. No
consistent evidence of an association between maternal plasma PFOS (median = 25.4 ng/mL for
cases; 27.4 ng/mL for controls) and diagnosed childhood autism identified by linkage to the
Danish National Hospital Registry was observed in a medium confidence nested case-control
study of mother-child pairs with an average of ten years of follow-up within the DNBC {Liew,
2015, 2851010}. Autism cases had significantly lower PFOS levels in a medium confidence
case-control study of amniotic fluid PFOS (median = 0.6 ng/mL for cases; 1.4 ng/mL for
controls) and diagnosed ASD, with cases identified as born 1982-1999 within the Danish
Psychiatric Central Registry {Long, 2019, 5080602}. Prenatal maternal serum PFOS
(median = 17.5 ng/mL for ASD cases; 15.9 ng/mL for ID cases; 17.9 ng/mL for controls) was
inversely associated with ASD and ID in a medium confidence study of children aged 4.5-
9 years with diagnosed ASD and ID {Lyall, 2018, 4239287}. An association was reported in a
medium confidence study of modeled prenatal maternal PFOS and clinically confirmed ASD
from mother-child pairs in the Childhood Autism Risk from Genetics and Environment
(CHARGE) study of children ages two to five years, with modeled prenatal maternal PFOS
(median = 3.1 ng/mL for cases; 3.3 ng/mL for controls) associated with increased odds of child
diagnosis of ASD and among boys when stratified by sex {Shin, 2020, 6507470}.
The effects on visuospatial performance were evaluated in one high confidence study of
participants of Project Viva {Harris, 2018, 4442261}. Visual-motor test scores (Wide Range
Assessment of Visual Motor Abilities) were consistently lower with increasing prenatal or
childhood PFOS exposures. Children in the upper quartile of prenatal PFOS (Q4 = 34.9-
168.0 ng/mL) had lower mid-childhood visual-motor scores, and participants in the third quartile
of childhood PFOS (Q3 = 6.3-9.7 ng/mL) had significantly decreased visual-motor scores.
Participants from the HOME study were assessed using the Virtual Morris Water Maze
(VMWM), but no significant effects were observed {Vuong, 2018, 5079693}.
C.4.1.4 Findings from Pregnant Women
No evidence was observed to support an adverse relationship between serum PFOS during
pregnancy and maternal depressive symptoms assessed by the Beck Depression Inventory (BDI)
from pregnancy to eight years postpartum in a medium confidence study based on women from
the HOME study {Vuong, 2020, 6356876}.
C-120
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.4.1.5 Findings from the General Adult Population
The effects of PFOS on general intelligence and IQ test outcomes were examined in a medium
confidence study {Shrestha, 2017, 3981382} of adults (ages 55-74 years) in New York State.
Findings indicated higher PFOS was significantly associated with improved performance in tests
of delayed recall.
Findings of a medium confidence study {Shrestha, 2017, 3981382}, described above, indicated
no significant associations between serum PFOS in adults and tests of executive function.
Two medium confidence studies investigated a possible association between PFOS and
depression {Shrestha, 2017, 3981382; Vuong, 2020, 6356876}. No significant associations were
observed in a medium confidence study of depression assessed by the BDI and serum PFOS
(median = 33.7 ng/mL) in a cross-sectional study of adults aged 55 to 74 years {Shrestha, 2017,
3981382}. Additionally, no evidence was observed to support a relationship in adults between
serum PFOS during pregnancy and maternal depressive symptoms assessed by the BDI from
pregnancy to 8 years postpartum in a medium confidence study based on women from the
HOME study {Vuong, 2020, 6356876}. One low confidence study (Berk, 2014, 2713574) of
data from adults participating in NHANES reported no adverse associations between PFOS
levels and depression as assessed by the nine-item depression module of the Patient Health
Questionnaire (PHQ-9).
The effects on visuospatial performance were evaluated in one medium confidence cross-
sectional study of older adults {Shrestha, 2017, 3981382}. A significant association between
serum PFOS and improved tests of visual and spatial function results was reported.
Two medium confidence studies explored the relationships between PFOS and memory loss.
{Gallo, 2013, 2272847; Shrestha, 2017, 3981382}. Statistically significant inverse associations
between PFOS and memory impairment were reported in a medium confidence study of adults in
the C8 Health Project {Gallo 2013, 2272847}. No adverse effects of PFOS on memory
impairment were again reported in a separate medium confidence study of older adults {Shrestha,
2017, 3981382}.
Two medium confidence cross-sectional studies investigated PFOS and hearing impairment in
adult NHANES participants. Li, 2020, 6833686 reported positive correlations between PFOS and
hearing impairment, while Ding and Park (2020, 6711603) observed no significant associations.
C.4.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 3 studies from the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} and 13 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOS HESD that investigated the association between PFOS and nervous effects. Study quality
evaluations for these 16 studies are shown in Figure C-29.
C-121
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Butenhoff et al., 2009, 757873 -
Curran et al., 2008, 757871 -
Era et al., 2009, 2919358-
Fuentes et al., 2007, 757865 -
Kawamoto et al., 2011, 2919266-
Li et al., 2021, 7643501 -
Lopez-Doval et al., 2015, 2848266-
Luebkeret al., 2005, 1276160-
Mehri et al„ 2016, 8776814
Mshaty et al., 2020, 6833692 -
NTP, 2019, 5400978
++ ++ ++3|6
B
++
8
++
++
++
¦
B
Pereiro et al., 2014, 2230732 -
+
+
NR
++
+
+
+
+
~
Salgado et al., 2015, 3981583 -
+
+
NR
+
+
-
+
0
0
Salgado et al., 2016, 3179088 -
+
+
NR
++
+
++
B
B
Thomford, 2002, 5432419-
+
-
NR
++
-
++
++ ++
Zhang etal.,2019, 5080461 -
+
+
NR
+
+
+
++
+
+
+*
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
B Critically deficient (metric) or Uninformative (overall)
Not reported
* Multiple judgments exist
Figure C-29. Summary of Study Evaluation for Toxicology Studies of PFOS and Nervous
Effects
C-122
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Interactive figure and additional study details available on HAWC.
There are few studies evaluating neurotoxicity, including neurodevelopmental toxicity,
associated with short-term, subchronic, or gestational exposure to PFOS in experimental models
(Table C-7). No study indicates morphological changes or damage attributed to PFOS. However,
there is some evidence suggesting that PFOS exposure may be associated with neurobehavioral
and physiological effects (e.g., impairments in spatial learning and memory, increases in
locomotor activity, and changes in neuronal electrophysiology and neurotransmitter levels).
Further research may be warranted.
Brain weight was assessed in only one developmental study and two short-term study in rats.
Absolute and relative brain weights were unchanged in the offspring of Crl:CD (SD) rats dosed
with 0.1-1 mg/kg/day PFOS during gestation and lactation {Butenhoff, 2009, 757873}. In male
and female Sprague-Dawley rats exposed to 2 mg/kg-100 mg/kg PFOS in diet for 28 days,
relative brain weights were increased in the 50 mg/kg and 100 mg/kg exposure groups in a
concentration-dependent manner, which may have been secondary to a decrease in body weights
as absolute brain weights were unchanged {Curran, 2008, 757871}. The relative weights of the
amygdala, hippocampus, and prefrontal cortex were unchanged in male Sprague-Dawley rats
dosed with 0.5 mg/kg/day-6 mg/kg/day PFOS for 28 days (data not provided) {Salgado, 2016,
3179088}; the absolute weights of these brain regions were not provided. One developmental
and one short-term study examined the gross pathology or histopathology of the brain, and no
effects were seen in rats exposed to 0.1 mg/kg/day-5 mg/kg/day PFOS {Butenhoff, 2009,
757873; NTP, 2019, 5400978}. The authors of a subchronic study in female BALB/c mice dosed
with 0.1 mg/kg/day and 1 mg/kg/day PFOS for 2 months noted a small amount of neuron
phagocytosis and that neuronal cells were contracted, deeply stained, and lacked clearly defined
cytoplasm and nuclei {Li, 2021, 7643501}.
One developmental {Mshaty, 2020, 6833692}, one short-term {Fuentes, 2007, 757865}, and one
subchronic study {Long, 2013, 2850984} in mice and several reproductive {Luebker, 2005,
1276160} and developmental studies {Butenhoff, 2009, 757873; Wang, 2015, 2851030;
Johansson, 2008, 1276156; Fuentes, 2007, 757863} in rats assessed the neurobehavioral effects
associated with PFOS. Mshaty et al. (2020, 6833692) assessed learning and memory in male
C57BL/6J mice exposed to 0.1-1 mg/kg/day PFOS from PND 1-PND 14 using the object
location test, object recognition test, and pairwise visual discrimination task. The discriminatory
index for the object location and recognition memory tests were decreased in mice exposed to
1 mg/kg/day, as was the learning curve for the 1 mg/kg/day group during the visual
discrimination task. Spatial learning and memory were also reduced in adult male C57BL6 mice
dosed with 2.15 mg/kg/day and 10.75 mg/kg/day but not 0.43 mg/kg/day PFOS for 3 months, as
seen by increases in escape latency and decreases in the time spent in the target quadrant using
the Morris water maze {Long, 2013, 2850984}. Time spent in the target quadrant was also
decreased in male CD1 mice dosed with 3 mg/kg/day but not 6 mg/kg/day PFOS for 4 weeks
{Fuentes, 2007, 757865}. In this study, swimming speed was increased in mice exposed to 3 and
6 mg/kg/day and distance traveled was increased in mice exposed to 6 mg/kg/day, whereas no
effects on motor activity were seen with the open field test or rotarod test. Similar effects on
spatial learning and memory were seen in the offspring of Wistar rat dams exposed to 15 mg/mL
but not 5 mg/mL PFOS in drinking water throughout gestation and/or lactation (drinking water
consumption not reported); swimming speed was not affected by exposure {Wang, 2015,
C-123
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
2851030}. However, two studies reported no changes in learning and memory, as tested with the
Morris water maze or the Biel swimming maze, in male and female rats exposed to
0.1 mg/kg/day-3.2 mg/kg/day PFOS pre- and postnatally {Luebker, 2005, 1276160; Butenhoff,
2009, 757873}. In a two-generation study, Luebker et al. (2005, 1276160) also reported no
effects on learning, memory, and short-term retention, as measured in a passive avoidance
paradigm, and Butenhoff et al. (2009, 757873) reported no effects on the acoustic startle
response. However, increased motor activity (ambulatory and total locomotor activity) and lack
of habituation was seen at PND 17 in males exposed to >0.3 mg/kg/day or 1 mg/kg/day,
respectively, throughout development {Butenhoff, 2009, 757873}. In male NMRI mice given a
single dose of 11.3 mg/kg at PND 10, during a period of development, lack of habituation was
also observed at 2 and 4 months of age {Johansson, 2008, 1276156}; this effect was not
observed with a single dose of 0.75 mg/kg at PND 10. In this study, locomotion, rearing, and
total activity was significantly decreased in both the 0.75 mg/kg and 11.3 mg/kg dose groups at
2 months of age. Another development study exposed CD-I mice to 6 mg/kg/day PFOS from
GD 12-GD 18 and assessed neuromotor maturation with surface righting reflex, open-field test,
and rotarod test {Fuentes, 2007, 757863}. Surface righting reflex was delayed at PND 4 and
PND 8. Significant effects were also observed during the climb test, with PFOS exposure
resulting in diminished resistance to backwards pull and reduced climb ability at PND 10 and
PND 11 but not PND 12. Climbing ability and forelimb grip strength was reduced with PFOS
exposure at PND 11 but not PND 10 or PND 12. The authors state that these transient effects
may support delayed neuromotor maturation due to gestational PFOS exposure. However, no
effects were observed with the open-field or rotarod tests at 3 months of age.
A short-term study reported that male CD-I mice displayed increased anxiety-like behavior in
the open field test, as seen by decreased time in the center of the chamber in the 3 mg/kg/day
PFOS group and decreased vertical activity in the 6 mg/kg/day group {Fuentes, 2007, 757865}.
However, in a developmental study by the same authors, no effects on anxiety-like behavior
were observed in CD-I mice exposed to 6 mg/kg/day PFOS from GD 12-GD 18 {Fuentes, 2007,
757863}. Similarly, no effects on this behavior were observed in a single-dose study in male
NMRI mice dosed with 0.75 mg/kg or 11.3 mg/kg PFOS {Johansson, 2008, 1276156}.
C-124
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table C-7. Associations Between PFOS Exposure and Neurobehavioral Effects in Rodents
Reference
Study Design
Learning and
Memory
Acoustic
Startle
Motor
Anxiety-like Behavior Activity/ Coordinatio
n
Neuromaturation
Mice
Fuentes et al. (2007,
757863)a
Developmental
exposure (GD12-18) to
0 or 6 mg/kg/day
NT
NT
Open field: No effect Open field: No effect
Rotarod: No effect
Surface righting reflex: J,
at 6 mg/kg/day
Grip strength: J, at
6 mg/kg/day
Mshaty et al. (2020,
6833692)b
Developmental
exposure (PND1-14) to
0,0.1, 0.25, or
1 mg/kg/day
Object location and
recognition test, and
pairwise visual
discrimination task:
| at 1 mg/kg/day
NT
NT NT
NT
Johansson et al.
Single dose (PND10) to
Spontaneous
NT
Elevated plus maze:
Spontaneous behavior,
NT
(2008, 1276156)b
0, 0.75, or 11.3 mg/kg
behavior,
habituation: I at
11.3 mg/kg
No effect
total activity: j at
>0.75 mg/kg in first
test block; f at
11.3 mg/kg in final test
block
Fuentes, et al. (2007,
Short-term exposure to
Morris water maze
NT
Open field, time in
Open field: No effect
NT
757865)b
0, 3, or 6 mg/kg/day
(acquisition): no
effect
Morris water maze
(probe): | at
3 mg/kg/day
center: J, at
3 mg/kg/day;
vertical activity: j at
6 mg/kg/day
Rotarod: No effect0
Morris water maze
(probe), swimming
speed: t at
> 3 mg/kg/day;
distance traveled: t at
6 mg/kg/day
Long et al. (2013,
Subchronic exposure
Morris water maze
NT
NT
NT
NT
2850984)d
(3 months) to 0, 0.43,
2.15, or
10.75 mg/kg/day
(acquisition, probe):
i at >2.15 mg/kg/day
C-125
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Reference
Study Design
Learning and
Memory
Acoustic
Startle
Motor
Anxiety-like Behavior Activity/ Coordinatio
n
Neuromaturation
Rats
Wang et al. (2015, Developmental Morris water maze NT NT Morris water maze, NT
2851030)e exposure (gestational (acquisition, probe): swimming speed: No
and/or lactational) to 0, J, at 15 mg/mL effect
5, 15 mg/L (0, 0.8, or
2.4 mg/kg/dayf)
Butenhoff et al. Developmental
Males, habituation: J, No effect NT
Males, motor activity: NT
(2009, 757873 f exposure (GD 0-PND
at 1 mg/kg/day
t at 0.3 mg/kg/day
20) to 0,0.1, 0.3, or
1.0 mg/kg/day
Biel swimming
Females: No effect
maze: No effect
Luebker et al. (2005, Reproductive exposure
Modified M-maze: NT NT
NT NT
1276160)3 (GD 0-PND 112) to
No effect
0.0, 0.1,0.4, 1.6, or
3.2 mg/kg/day
Passive avoidance:
No effect
Notes:GD = gestation day; NT = not tested; PND = postnatal day.
aMales and females analyzed separately.
b Study conducted in males.
c No quantitative data were presented for this endpoint, which was consequently rated as low confidence.
d Sexes combined.
e Sex was not specified.
fDoses in mg/kg/day were derived and presented in the 2016 PFOS HESD {U.S. EPA, 2016, 3603365}.
C-126
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Several short-term studies in mice and rats {Salgado, 2015, 3981583; Salgado, 2016, 3179088;
Lopez-Doval, 2015, 2848266}, one developmental study in mice {Mshaty, 2020, 6833692}, and
one subchronic study in mice {Long, 2013, 2850984} examined the effects of PFOS on
neurotransmitter levels (Table C-8). Glutamine, glycine, and serotonin were each examined in
only one study. Neither glutamine nor glycine were altered in the dorsal hippocampus of male
C57BL/6J mice exposed to 1 mg/kg/day PFOS from PND 1- PND 14 {Mshaty, 2020, 6833692}.
Serotonin was increased in the anterior hypothalamus, mediobasal hypothalamus, and the median
eminence of male Sprague-Dawley rats dosed with 0.5 mg/kg/day-6 mg/kg/day for 28 days
{Lopez-Doval, 2015, 2848266}. The effect of PFOS on dopamine and/or gamma-aminobutyric
acid (GAB A) in various brain regions was examined in three studies {Mshaty, 2020, 6833692;
Long, 2013, 2850984; Salgado, 2015, 3981583}. A subchronic study found no changes in
GABA in the hippocampus of male C57BL6 mice dosed with 0.43-10.75 mg/kg/day PFOS
{Long, 2013, 2850984}. However, GABA was increased in the dorsal hippocampus of male
C57BL/6J mice exposed to 1 mg/kg/day PFOS from PND 1- PND 14 {Mshaty, 2020, 6833692}.
In adult male Sprague-Dawley rats dosed with 3 and 6 mg/kg/day PFOS for 28 days, GABA was
unaltered in the mediobasal hypothalamus and increased in the anterior hypothalamus in both
dose groups {Salgado, 2015, 3981583}. In male Sprague-Dawley rats dosed with
0.5 mg/kg/day-6 mg/kg/day PFOS for 28 days, dopamine was increased in the hippocampus in
the 0.5 mg/kg, 1 mg/kg, and 3 mg/kg groups, but not the 6 mg/kg/day group {Salgado, 2016,
3179088}. Increased dopamine levels were also detected in the prefrontal cortex of the
1 mg/kg/day group only and in the anterior hypothalamus of the 3 mg/kg/day and 6 mg/kg/day
groups {Salgado, 2015, 3981583; Salgado, 2016, 3179088}. No changes in dopamine levels
were seen in the mediobasal hypothalamus {Salgado, 2015, 3981583}. In male C57BL6 mice
dosed with 0.43 mg/kg/day-10.75 mg/kg/day PFOS, dopamine in the caudate putamen was
decreased only at the highest dose {Long, 2013, 2850984}. In this study, glutamate in the
hippocampus was also increased at the highest dose. However, glutamate was increased in the
dorsal hippocampus of male C57BL/6J mice exposed to 1 mg/kg/day PFOS from PND 1-PND
14 {Mshaty, 2020, 6833692}. Greater sensitivity of the developing brain to PFOS exposure
might explain why glutamate increases in the hippocampus were only seen at higher doses in the
Long et al. (2013, 2850984) study compared to increases seen at a lower dose in the Mshaty et
al. (2020, 6833692) study.
C-127
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table C-8. Associations Between PFOS Exposure and Neurotransmitters in Rodents
Glutamine/
Reference
Study Design
Glutamate
Glycine
Serotonin
GABA
Dopamine
Mice
Mshaty et al. (2020,
Developmental exposure
Dorsal hippocampus,
Dorsal
NT
Dorsal
NT
6833692)a
(PND1-14) to 0 or
glutamate: t at
hippocampus:
hippocampus: t at
1 mg/kg/day
1 mg/kg/day
No effect
1 mg/kg/day
Dorsal hippocampus,
glutamine: No effect
Long et al. (2013,
Subchronic exposure
Hippocampus,
NT
NT
Hippocampus: No
Caudate putamen: J,
2850984)b
(3 months) to 0, 0.43,
glutamate: t at
effect
at 10.75 mg/kg/day
2.15, or 10.75 mg/kg/day
10.75 mg/kg/day
Rats
Salgado et al. (2015,
Short-term exposure
NT
NT
NT
Mediobasal
Mediobasal
3981583)3
(28 days) to 0, 3, or
hypothalamus:
hypothalamus:
6 mg/kg/day
No effect
No effect
Anterior
Anterior
hypothalamus: t at
hypothalamus: t at
>3 mg/kg/day
>3 mg/kg/day
Salgado et al. (2016,
3179088)3
Short-term exposure
(28 days) to 0, 0.5, 1, 3,
or 6 mg/kg/day
NT
NT
NT
NT
Amygdala: No
effect
Prefrontal cortex: f
at 1 mg/kg/day but
not at 3 and 6
mg/kg/day
Hippocampus: t at
0.5, 1, and
3 mg/kg/day but not
at 6 mg/kg/day
C-128
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Study Design
Glutamine/
Glutamate
Glycine
Serotonin
GABA
Dopamine
Lopez-Doval et al.
(2015, 2848266)a
Short-term exposure
(28 days) to 0, 0.5, 1, 3,
or 6 mg/kg/day
NT
NT
Mediobasal NT
hypothalamus: t at
>0.5 mg/kg/day
Anterior
hypothalamus: t at
>0.5 mg/kg/day
Median eminence: t
at >0.5 mg/kg/day
NT
Notes: NT = not tested.
a Study conducted in males.
b Sexes combined.
C-129
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Synaptic transmission and plasticity were assessed in one electrophysiology study in SD rats
exposed to 0.35 mg/kg/day-2.17 mg/kg/day PFOS throughout development until PND 90
{Zhang, 2019, 5080461}. Zhang et al. (2019, 5080461) observed moderate inhibition of paired
pulse facilitation (at highest dose) and the input/output curve (at all doses) in the hippocampus.
Long-term potentiation was also decreased in a dose-dependent manner in the 0.72 mg/kg/day
and 2.17 mg/kg/day dose groups.
C.4.3
Mechanistic Evidence
Mechanistic evidence linking PFOS exposure to adverse nervous outcomes is discussed in
Sections 3.2.4, 3.2.5, 3.2.6, 3.3.4, 3.3.6, and 3.4.1.4 of the 2016 PFOS HESD {U.S. EPA, 2016,
3603365}. There are 54 studies from recent systematic literature search and review efforts
conducted after publication of the 2016 PFOS HESD that investigated the mechanisms of action
of PFOS that lead to nervous effects. A summary of these studies is shown in Figure C-30.
Additional mechanistic synthesis will not be conducted since evidence suggests but is not
sufficient to infer that PFOS leads to nervous effects.
Mechanistic Pathway Animal
Big Data, Non-Targeted Analysis 1
Cell Growth, Differentiation, Proliferation, Or Viability
Cell Signaling Or Signal Transduction
Extracellular Matrix Or Molecules 0
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation 2
Hormone Function 7
Inflammation And Immune Response 1
Oxidative Stress 1
Xenobiotic Metabolism 0
Other 2
Not Applicable/Not Specified/Review Article 3
Grand Total 26
Human
0
0
0
In Vitro Grand Total
31
29
6
11
1
3
3
54
Figure C-30. Summary of Mechanistic Studies of PFOS and Nervous Effects
Interactive figure and additional study details available on Tableau.
C.4.4 Evidence Integration
There is slight evidence of an association between PFOS and nervous system effects in humans
based on the mostly mixed results. There were no new neurological studies identified that
evaluated cerebral palsy. Outcomes investigated include depression, memory impairment,
hearing impairment, ASD, and ID.
Epidemiological studies in this current review provide limited indication of adverse effects of
PFOS on neurodevelopment or neuropsychological outcomes {Chen, 2013, 2850933; Jeddy,
C-130
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
2017, 3859807; Niu, 2019, 5381527}, cognitive development {Harris, 2018, 4442261; Oulhote,
2019, 6316905}, and executive function {Vuong, 2016, 3352166} in human populations. No
adverse effects were observed for PFOS and depression or memory impairment, and only one
study indicated effects of PFOS on hearing impairment {Li, 2020, 6833686}, however the
number of studies was limited. Overall, results from studies of neurodevelopmental,
neuropsychological, and cognitive outcomes were mixed.
The recent studies provide limited indication of adverse effects of PFOS on behavioral problems,
ADHD, ASD, and ID. The studies reviewed provide some indication of behavioral problems
associated with PFOS {Oulhote, 2016, 3789517; Oulhote, 2019, 6316905; Ghassabian, 2018,
5080189}, however overall results were mixed. Of the multiple studies examining associations
between PFOS and ADHD, only one {Lenters, 2019, 5080366} reported a significant
relationship between PFOS and ADHD, with results indicating heterogeneity with respect to
gender. No adverse associations of ID with PFOS were reported in the single study reviewed
{Lyall, 2018, 4239287}. There was an indication of a potential relationship between PFOS and
autistic behaviors or ASD diagnosis in some studies {Braun, 2014, 2345999; Oulhote, 2016,
3789517; Shin, 2020, 6507470}. However, many studies have methodological concerns, as
PFOS exposures in cases and controls within the ADHD and ASD studies were often either
similar to or had mean control exposures greater than cases in some studies. A single category
outcome for ASD may also not adequately encompass the heterogeneity in terms of
developmental history, intelligence, comorbidity, and severity that might be important in
accurately revealing associations. The current evidence examining PFOS exposure and
neurodevelopmental disorders in children, including ADHD and learning disabilities, is limited.
The animal evidence for an association between PFOS and neurological effects is moderate.
There are several medium confidence studies available where changes in neurobehavioral effects
were observed. Although the studies varied by design, endpoints measured, and methods of
measurement leading to some inconsistencies across studies, there is evidence of effects on
learning and memory. Of the studies available in animal models, no effects were noted for brain
weight with limited changes observed for histopathology. Some neurobehavioral effects were
observed but these results and the methods used to quantify them were relatively inconsistent.
Alterations in neurotransmitter levels and synaptic transmission and plasticity were also
observed, though it is often unclear what magnitude of change in neurotransmitters levels can be
considered adverse. Notably, Mshaty et al. (2020, 6833692) observed dose-dependent effects of
PFOS in both the object recognition memory test and object location recognition memory test, as
well as dose-dependent effects of PFOS across 9 days of a visual discrimination task. These
behavioral changes in the 1 mg/kg/day dose group were accompanied by significant increases in
hippocampal neurotransmitter concentrations, including glutamate and GAB A. Increased
hippocampal glutamate levels may cause excitotoxicity which could explain the spatial learning
deficits seen by Mshaty et al. (2020, 6833692). Importantly, the exposure period in this study
encompassed a sensitive period of neurodevelopment (i.e., lactation) and the observed effects
occurred at relatively low doses. In addition, the deficits in spatial learning and increased
hippocampal glutamate concentrations observed by Long et al. (2013, 2850984) in PFOS-
exposed adult mice support these results.
C-131
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.4.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOS exposure has the potential to cause nervous system effects
in humans under relevant exposure circumstances (Table C-9). This conclusion is based
primarily on alterations in neurodevelopment, neurobehavior, and neurotransmitter levels in
animals following exposure to doses as low as 0.5 mg/kg/day PFOS. Although there is some
evidence of adverse effects of PFOS exposure on neurodevelopment or neuropsychological
outcomes, cognitive development, executive function and behavioral problems in humans, there
is considerable uncertainty in the results due to inconsistency across studies and limited number
of studies.
C-132
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-9. Evidence Profile Table for PFOS Nervous System Effect
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.4.1)
Neurodevelopment
2 High confidence study
4 Medium confidence
studies
1 Low confidence study
Cognitive function
1 High confidence
studies
9 Medium confidence
studies
Inverse associations were
reported for
neurodevelopmental
outcomes in three studies
of children (3/7). One
high confidence study
observed a significant
inverse association with
social measures among
girls. Of the medium
confidence studies, one
observed significant
inverse associations with
total neurodevelopment
and motor skill measures.
Another study reported
significant inverse
associations with
communication measures
but a positive association
with cognition. The same
study reported
inconsistent effects when
stratified by maternal age.
Results reported in the
remaining studies were
inconsistent.
Reported results were
largely inconsistent
across studies, with both
positive and inverse non-
significant associations
reported. One high
• High and medium
confidence studies
• Low confidence
studies
• Lnconsistent direction
of effects across
studies
1 High and medium
confidence studies
• Lnconsistent direction
of effects across
studies
• Small magnitude of
effect
©OO
Slight
Evidence of nervous
system effects is based on
high confidence studies
reporting significant
associations, which
varied in magnitude and
were inconsistent across
neurological outcomes.
Uncertainties remain due
to inconsistent findings
within studies and mixed
findings across studies.
Studies with mixed
findings were primarily
of medium or low
confidence.
©OO
Evidence Suggests
Primary basis:
Animal evidence indicated
alterations in
neurodevelopment,
neurobehavior, and
neurotransmitter levels.
Although there is some
evidence of adverse
effects of PFOS exposure
on neurodevelopment or
neuropsychological
outcomes, cognitive
development, executive
function and behavioral
problems in humans, there
is considerable uncertainty
in the results due to
inconsistency across
studies and limited
number of studies.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
C-133
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase Factors that Decrease
Certainty Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Social-emotional and
behavioral regulation
1 High confidence study
4 Medium confidence
studies
1 Low confidence study
confidence study
observed non-
significantly increased
non-verbal IQ scores
among the highest
exposure group. Positive
associations with reading
scores were observed in
some medium confidence
studies (2/9).
One high confidence
study found no
significant associations
with behavioral measures
at age 5 but observed a
positive association
among females and a
negative association
among males at age 7. Of
the medium confidence
studies, two observed
positive associations with
behavioral difficulties
(2/4). Another medium
confidence study
observed that the
association with
impulsivity was modified
by sex, with males
performing better than
females (1/4). One low
confidence study of
adolescents observed a
significant inverse
correlation with the
region of the brain
1 High and medium
confidence studies
• Low confidence study
• Lnconsistent direction
of effects across and
within studies
C-134
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Evidence Integration
Studies and Summary and Key Factors that Increase Factors that Decrease Evidence Stream Summary Judgment
Interpretation Findings Certainty Certainty Judgment
associated with impulsive
behavior.
One medium confidence • Medium confidence • Low confidence study
study reported positive studies
but non-significant results
for depression in general
population adults.
Another medium
confidence study
explored depression in
children followed for 20
years, reporting no
association. An additional
study of medium
confidence reported no
association with
depression among
pregnant women. A low
confidence study reported
no association.
Executive function Two medium confidence
3 Medium confidence studies examined
studies executive function
measures, including
behavior regulation and
metacognition, among
children from the HOME
Study (2/3). One of these
studies reported
significantly inversed
associations with
executive function
measures, while the other
reported no significant
associations. A medium
Depression
3 Medium confidence
studies
1 Low confidence study
• Medium confidence
studies
• Lnconsistent direction
of effects across
studies
• Limited number of
studies examining
outcome
C-135
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Attention
5 Medium confidence
studies
2 Low confidence
studies
Autism, autistic
behaviors, and
intellectual disability
1 High confidence study
5 Medium confidence
studies
confidence study of
adults did not observe
significant associations.
Studies examining
measures of attention
reported mixed findings.
One medium confidence
study reported
significantly increased
odds of ADHD. When
stratified by child sex,
significant effects
remained. The remaining
medium confidence
studies (4/5) did not
report significant
associations.
Additionally, the two low
confidence studies
observed no associations
with measures of
attention.
One high confidence
study observed a positive
association with autism
scores when measured at
age 7. When stratified by
sex, higher scores were
observed in females
compared to males.
Findings from the five
medium confidence
studies were mixed. Two
studies observed positive
associations, with one
• Medium confidence
studies
• Low confidence
studies
• Lnconsistent direction
of effects across
studies
1 High and medium
confidence studies
• Lnconsistent direction
of effects across
studies
C-136
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Visuospatial
performance
1 High confidence study
1 Medium confidence
studies
Memory impairment
2 Medium confidence
studies
study reporting
associations for the
overall study population
and the other study
reporting the association
only among males.
Another medium
confidence study reported
inverse associations.
Other reported results
were not significant.
Two studies examined
visuospatial performance
effects among children.
One high confidence
study among children
observed a significant
inverse association with
visual-motor performance
across quartiles of
exposure. The medium
confidence study reported
no association with
visuospatial
performance.
Two studies reported
associations with memory
loss among adult
populations. One medium
confidence study
observed a significant
inverse association with
memory impairment. No
significant effects were
reported from the
• High and medium
confidence studies
• Large magnitude of
effect
• Medium confidence
studies
• Inconsistent direction
of effects across
studies
• Limited number of
studies examining
outcome
• Inconsistent direction
of effects across
studies
• Small magnitude of
effect
• Limited number of
studies examining
outcome
C-137
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Hearing impairment
2 Medium confidence
studies
remaining medium
confidence study.
Two medium confidence
studies examined hearing
impairment among adults
from NHANES. One
study observed positive
correlations with hearing
impairment, while the
other reported no
associations.
• Medium confidence
study
• Large magnitude of
effect
• Inconsistent direction
of effects across
studies
• Limited number of
studies examining
outcome
Evidence from In Vivo Animal Studies (Section C.4.2)
Evidence Integration
Summary Judgment
Neurobehavior
4 Medium confidence
studies
Changes in
neurobehavior endpoints
were altered and
decreases in learning and
memory tasks were
largely consistent among
studies (3/4). Motor
activity was found to be
increased (2/2), with
anxiety-like behavior
being decreased (1/1). A
single study measured
acoustic startle and found
no changes (1/1).
• Medium confidence
studies
• Coherence of
findings in
neurotransmitters
• Limited number of
studies examining
specific outcomes
• Inconsistent direction
of effects
0©O
Moderate
Several medium
confidence studies are
available where changes
in neurobehavioral effects
were observed. Although
the studies varied by
design, endpoints
measured, and methods of
measurement leading to
some inconsistencies
C-138
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Neurotransmitters
3 Medium confidence
studies
Organ weights
3 Medium confidence
studies
Histopathology
1 High confidence study.
1 Medium confidence
study
Changes in
neurotransmitter levels in
short-term studies in male
mice included a dose-
responsive increase in
serotonin (1/1) and
region-specific decreases
of GABA (1/1) and
dopamine (2/2).
No effects were observed
on absolute brain weights
(2/2). One study reported
a significant increase in
relative brain weights;
however, this increase
was confounded by a
reduction in body
weight.
One study found no
effects on brain
histopathology in male
and female rats, whereas
some phagocytosis in the
brains of PFOS-exposed
mice was noted in
another study.
• Medium confidence
studies
• Coherence of
findings in
neurobehavior
endpoints
• Dose-response
relationship
• Medium confidence
studies
• High and medium
confidence studies
• Limited number of
studies examining
outcome
• Biological significance
of the magnitude of
effect is unclear
• Limited number of
studies examining
outcomes
• Confounding variables
such as decreases in
body weights
• Limited number of
studies examining
outcomes
across studies, there is
evidence of effects on
learning and memory. No
effects were noted for
brain weight and limited
changes were observed
for histopathology.
Alterations in
neurotransmitter levels
and synaptic transmission
and plasticity were also
observed, though it is
often unclear what
magnitude of change in
neurotransmitters levels
can be considered
adverse.
Evidence Integration
Summary Judgment
C-139
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Elect rophysiology
1 Medium confidence
study
One developmental study
in male and female rats
found inhibition of paired
pulse facilitation and the
input/output curve in the
hippocampus.
Hippocampal long-term
potentiation was also
decreased (1/1).
• Medium confidence
study
• Dose-response
relationship
• Limited number of
studies examining
outcome
Notes: ADHD = attention deficit/hyperactivity disorder; GABA = gamma-aminobutyric acid; HOME = Health Outcomes and Measures of the Environment; IQ = intelligence
quotient; NHANES = National Health and Nutrition Examination Survey.
C-140
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.5 Renal
EPA identified 19 epidemiological and 12 animal studies that investigated the association
between PFOS and renal effects. Of the epidemiological studies, 17 were classified as low
confidence and 2 were considered uninformative (Section C.5.1). Of the animal studies, 2 were
classified as high confidence, 8 as medium confidence, and 2 were considered low confidence
(Section C.5.2). Studies may have multiple judgments depending on the endpoint evaluated.
Though low confidence studies are considered qualitatively in this section, they were not
considered quantitatively for the dose-response assessment (See Main PFOS Document).
C.5.1 Human Evidence Study Quality Evaluation and
Synthesis
C.5.1.1 In traduction
PFOS has the potential to affect the kidney's function given the saturable resorption from the
renal tubules {U.S. EPA, 2016, 3603365}. Biomarkers of renal function include blood urea
nitrogen (BUN), estimated glomerular filtration rate (eGFR), serum creatinine, and uric acid.
eGFR is a marker of non-malignant renal disease.
The 2016 HESD for PFOS {U.S. EPA, 2016, 3603365} concluded there was evidence of a
suggestive association between PFOS and chronic kidney disease (CKD; defined as glomerular
filtration rate (GFR) < 60 mL/min/1.73 m2) based on two studies on the general population
{Shankar, 2011, 2919232; Steenland, 2010, 1290810} and two on children {Geiger, 2014,
2851286; Watkins, 2013, 2850974}; however, given the cross-sectional study designs, the
potential for reverse causality could not be ruled out.
For this updated review, 19 studies examined the association between PFOS and renal health
outcomes. Five studies were in children and adolescents {Geiger, 2013, 2919148; Kataria, 2015,
3859835; Khalil, 2018, 4238547; Predieri, 2015, 3889874; Qin, 2016, 3981721}, one study in
pregnant women {Nielsen, 2020, 6833687}, one study in occupational workers {Rotander, 2015,
3859842}, and the remainder of the studies were in the general population. Fifteen of the studies
utilized a cross-sectional study design; the remaining study designs included one case-control
study {Predieri, 2015, 3889874}, and three cohorts {Blake, 2018, 5080657; Conway, 2018,
5080465; Nielsen, 2020, 6833687} (Appendix D). All studies measured PFOS in blood
components (i.e., plasma or serum). Two studies conducted in China investigated the same
population from the Isomers of C8 Health Project {Wang, 2019, 5080583; Zeng, 2019,
5918630}. Among the studies investigating populations in the United States, five studies utilized
data from the NHANES {Geiger, 2013, 2919148; Jain, 2019, 5080378; Jain, 2019, 5381566;
Kataria, 2015, 3859835; Scinicariello, 2020, 6833670}. Outcomes evaluated in these studies
including clinical conditions, such as CKD and gout and biomarkers of renal function, including
uric acid, eGFR, albumin, and creatinine.
C.5.1.2 Study Quality
Several considerations were specific to evaluating the quality of studies examining kidney
function and kidney disease. Since PFOS is removed from the blood by the kidney, cross-
sectional analyses using serum PFOS as the exposure measure are problematic if individuals
C-141
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
with compromised kidney function are included: PFOS concentrations could be increased in
those individuals and an apparent association with GFR would be observed, even if one did not
exist {Dhingra, 2017, 3981432}.
There are 19 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and renal effects. Study quality evaluations for these 19 studies are
shown in Figure C-31.
Of the 19 studies identified since the 2016 assessment, 17 studies were classified as low
confidence and the remaining two as uninformative {Predieri, 2015, 3889874; Seo, 2018,
4238334}. No studies were classified as high or medium confidence. The main concerns with the
low confidence studies included potential for residual confounding, selection bias, and reverse
causality. Another concern included small sample sizes {Khalil, 2018, 4238547; Nielsen, 2020,
6833687}. Additionally, low confidence studies utilizing cross-sectional analyses of kidney
function with serum PFOS were impacted by the potential for reverse causation.
Deficiencies identified in Predieri et al. (2015, 3889874) included a small sample size and
narrow ranges of exposures which contributed to an uninformative rating. Seo et al. (2018,
4238334) presented bivariate correlations between PFOS exposure and renal outcomes, limiting
the ability to interpret the results. Other potential sources of bias were identified, including a lack
of information on participant recruitment and selection, unexplained discrepancies in samples
sizes, and missing details on outcome assessment methods. Neither uninformative study adjusted
for key confounders (e.g., age and SES), resulting in a high potential for residual confounding.
C-142
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
>0e
Arrebola et al., 2019, 5080503 -
J
+
'
+
i
+
++
-
i
+
i
+
Blake et al., 2018, 5080657 -
+
-
++
+
++
+
+
-
Chen et al., 2019, 5387400-
+
+*
++
+
+
+
+
-
Conway et al., 2018, 5080465 -
+
-
++
+
+
+
+
¦
Geiger et al., 2013, 2919148-
++
-
+
+
++
+
+
-
Jain and Ducatman, 2019, 5080378 -
+
-
+
+
-
+
+
-
Jain and Ducatman, 2019, 5381566 -
+
+
-
+
-
+
+
-
Kataria et al., 2015, 3859835 -
+
¦
+
+
++
+
+
-
Khalil et al., 2018, 4238547-
-
+
+
-
+
+
-
-
Lin et al., 2013, 2850967-
-
+
-
-
+
+
+
-
Liu et al., 2018, 4238514-
+
+
+
+
+
+
+
-
Nielsen et al., 2020, 6833687 -
+
+
++
-
-
+
+
•
Predieri et al., 2015, 3889874 -
+
+
-
-
+
+
~
Qin et al., 2016, 3981721 -
+
-
+
+
+
+
+
-
Rotander et al., 2015, 3859842 -
-
+
-
+
+
+
-
¦
Scinicariello et al., 2020, 6833670 -
+
-
+*
+
+
+
+
-
Seo etal., 2018, 4238334-
-
+
-
H
-
-
~
Wang et al., 2019, 5080583 -
+
-
+
+
+
+
+
~
Zeng et al., 2019, 5918630-
+
-
++
+
++
+
+
~
Legend
Q Good (metric) or High confidence (overall)
+ Adequate (metric) or Medium confidence (overall)
- Deficient (metric) or Low confidence (overall)
9 Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
Figure C-31. Summary of Study Evaluation for Epidemiology Studies of PFOS and Renal
Effects
Interactive figure and additional study details available on HAWC.
C-143
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.5.1.3 Findings from Children and Adolescents
Three low confidence studies reported on uric acid among children and adolescents {Geiger,
2013, 2919148; Qin, 2016, 3981721; Kataria, 2015, 3859835} with two also reporting on
hyperuricemia {Geiger, 2013, 2919148; Qin, 2016, 3981721}, defined as serum uric acid levels
> 6 mg/dL. The three studies reported mixed results. Among adolescents aged 12 to 18 years
from NHANES (1999-2008), Geiger et al. (2013, 2919148) observed statistically significant
positive associations between increasing quartiles of PFOS and hyperuricemia (p-
trend = 0.0221), and serum uric acid (p-trend = 0.0575). An overlapping NHANES (2003-2010)
study {Kataria, 2015, 3859835} also reported a positive association with uric acid levels among
adolescents, where the highest PFOS quartile (> 19.4 ng/mL) was associated with a 0.19 mg/dL
(95% CI: 0.032, 0.34 mg/dL, p < 0.05) increase in uric acid levels compared to the lowest PFOS
quartile (< 7.9 ng/mL). Qin et al. (2016, 3981721) did not observe significant associations for
hyperuricemia or uric acid in children aged 12 to 15 years from the GBCA in Taiwan.
One low confidence study {Kataria, 2015, 3859835} reported on GF in children aged 12 to
19 years from NHANES (2003-2010). Significant negative associations were observed for eGFR
in the second, third, and fourth quartiles of PFOS exposure compared to the lowest quartile.
Two low confidence studies and one uninformative study investigated serum creatinine among
children and adolescents {Kataria, 2015, 3859835; Khalil, 2018, 4238547; Predieri, 2015,
3889874}. One low confidence study {Kataria, 2015, 3859835} on NHANES (2003-2010)
adolescents (12-19 years old) reported a significant positive association with serum creatinine in
the third and fourth quartiles of PFOS exposure. One low confidence study {Khalil, 2018,
4238547} examined serum creatinine levels among obese children aged 8 to 12 years, but no
significant effect was observed.
C.5.1.4 Findings from the General Adult Population
Two low confidence studies examined CKD in the general population {Conway, 2018, 5080465;
Wang, 2019, 5080583} and both observed positive associations. CKD was defined as an eGFR
of < 60 mL/min/1.73 m2 In C8 Health Project participants, Conway, 2019, 5080465 observed
significantly elevated odds of CKD among non-diabetic participants; a negative association was
observed among participants with diabetes. The prevalence of CKD in the diabetic population
was higher (22%) than the non-diabetic population (7%). Wang et al. (2019, 5080583) observed
non-significantly elevated odds of CKD in participants from the Isomers of C8 Health Project in
China. However, a concern for reverse causality makes interpretation of the results difficult in
both studies.
Gout was examined in one low confidence study {Scinicariello, 2020, 6833670} in adults from
NHANES (2009-2014). Positive associations were observed between serum PFOS and self-
reported gout, however, none were significant.
Six low confidence general population studies {Arrebola, 2019, 5080503; Chen, 2019, 5387400;
Jain, 2019, 5080378; Lin, 2013, 2850967; Scinicariello, 2020, 6833670; Zeng, 2019, 5918630}
and one low confidence occupational study {Rotander, 2015, 3859842} examined PFOS and uric
acid levels, and three of those studies evaluated uric acids as they pertained to hyperuricemia
{Arrebola, 2019, 5080503; Scinicariello, 2020, 6833670; Zeng, 2019, 5918630}.
C-144
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A low confidence NHANES (2009-2014) study {Scinicariello, 2020, 6833670} observed
significantly elevated serum uric acid across increasing PFOS exposure quartiles, and the trend
was significant (p-trend = 0.003). Higher odds of hyperuricemia among participants in the
highest exposure quartile (> 11.90 ng/mL) compared to the lowest (< 4.43 ng/mL) was also
observed, but the trend was not significant (p-trend = 0.15). Results were similar when restricted
to participants without CKD. Another low confidence study {Zeng, 2019, 5918630} on
participants from the Isomers of C8 Health Project reported significantly elevated uric acid levels
with increasing PFOS exposure, and a marginally significant association (OR: 1.17, 95% CI:
0.99, 1.39, p = 0.074) for hyperuricemia., Jain and Ducatman (2019, 5080378) examined uric
acid by glomerulation stage among NHANES (2007-2014) participants. For males, positive
associations with uric acid were observed for stages GF-1 (p < 0.01) and GF-2 (p = 0.05), but the
effect was negative for stages GF-3A (p = 0.66) and GF-3B/4 (p < 0.01). For females, all
associations were positive across stages of GF with significant associations (p < 0.05) for GF-1
and GF-3A. Two low confidence studies did not observe associations with plasma uric acid in
Croatian adults aged 44-56 years {Chen, 2019, 5387400}, or in adolescents and young adults
aged 12-30 years in the Young Taiwanese Cohort Study {Lin, 2013, 2850967}. Another low
confidence study {Arrebola, 2019, 5080503} using pooled cohort data (the BIOAMBIENT.ES
study) observed a non-s ignificant increase in serum uric acid with increasing PFOS.
One low confidence occupational study examined serum uric acid levels among firefighters with
past exposure to aqueous film forming foam (AFFF) {Rotander, 2015, 3859842}. No significant
association was observed for serum uric acid and increasing PFOS exposure.
Two general population studies evaluated PFOS and eGFR {Blake, 2018, 5080657; Wang, 2019,
5080583}. A low confidence study {Blake, 2018, 5080657} assessed participants of the FCC
with high exposure to PFAS from their household water supplies. A significant inverse
association with eGFR was observed in the latent effects mixed effect model (LME), but not in
the repeated measures LME. These results were consistent with the low confidence study {Wang,
2019, 5080583} which assessed participants of the Isomers of C8 Health Project and observed
negative association between total PFOS serum concentrations and eGFR.
The evidence of association between PFOS and renal effects among pregnant women was
limited. Only one low confidence study reported on pregnant women {Nielsen, 2020, 6833687}
using a small sample of women (n = 73) from the Pregnancy Obesity Nutrition and Child Health
study (PONCH) study. No significant Spearman rank correlations were reported between PFOS
and kidney function parameters.
Two studies examined albumin and creatinine as biomarkers for renal function {Chen, 2019,
5387400; Jain, 2019, 5381566}. The two low confidence studies provided differing conclusions.
Jain, 2019, 5381566 utilized NHANES (2005-2014) data and reported statistically significant
positive associations with serum and urine creatinine, and serum albumin. Statistically
significant negative associations were also reported with urine albumin and urine albumin-
creatinine ratios. Stratification by stages of GF was noted as better representing more severe
stages of renal failure. For PFOS, stratification by stages of GF had inconsistent effects. One low
confidence study {Chen, 2019, 5387400} did not observe significant associations with plasma
creatinine in Croatian adults ages 44-56 years.
C-145
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
One low confidence study, Liu et al. (2018, 4238514) examined serum proteins among NHANES
(2013-2014) participants, and positive associations (p < 0.01) were observed for serum protein
with increasing PFOS exposure. The effect was consistent when stratified by linear and branched
PFOS.
C.5.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 4 studies from the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} and 8 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOS HESD that investigated the association between PFOS and renal effects. Study quality
evaluations for these 12 studies are shown in Figure C-32.
C-146
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
,Ge
Butenhoff et al., 2012, 1276144-
++
++
NR
++
++
++
++
++
++
Curran et al., 2008, 757871 -
++
~
NR
++
D
D
++
++
++
+
Dong et al., 2011, 1424949-
++
+
NR
++
++
++
++
++
++
+
~
¦
Fuentes et al., 2006, 757859 -
+
+
NR
D
¦
++
++
+
Kawamoto et al., 2011, 2919266 -
-
+
NR
¦
++
++
D
++
-
Li et al., 2021, 7643501 -
+
+
NR
-
+
+
-
+
NTP, 2019, 5400978-
++
++
5$
++
++
++
++
++
++
Seacat et al., 2002, 757853 -
++
D
D
D
Seacat et al., 2003, 1290852 -
++ ++
NR
+
+
+
Thomford, 2002, 5432419-
+
-
NR
++
++
-
Xing et al.f 2016, 3981506-
++
+
NR
++
++
++
Zhong et al., 2016, 3748828 -
-
NR
NR
+
+
+
++ ++ ++
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
NR Not reported
* Multiple judgments exist
~ Bias away from null
Figure C-32. Summary of Study Evaluation for Toxicology Studies of PFOS and Renal
Effects
Interactive figure and additional study details available on HAWC.
Few renal effects were observed across multiple studies assessing PFOS toxicity in animal
models. Most studies did not observe significant effects of PFOS exposure on kidney weight or
histopathology {Seacat, 2002, 757853; Seacat, 2003, 1290852; Fuentes, 2006, 757859; Peden-
Adams, 2008, 1424797; Yahia, 2008, 2919381; Dong, 2011, 1424949; Zhong, 2016, 3748828;
Li, 2021, 7643501}. Flowever, two subchronic studies in male mice reported significant
decreases in relative kidney weight with PFOS treatment for 30 days at the highest dose tested of
C-147
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
10 mg/kg/day (approximately 10% decrease) {Xing, 2016, 3981506} and treatment for 60 days
at doses of 0.83 mg/kg/day or 2.083 mg/kg/day (approximately 18% and 16% decreases,
respectively) {Dong, 2009, 1424951}. Neither of these studies reported absolute kidney weight
and, in both studies, PFOS treatment resulted in decreased body weight at these doses which
precludes evaluation of the significance of relative weights. One developmental study in mice
reported no significant changes in maternal relative or absolute kidney weight {Fuentes, 2006,
757859}.
In contrast to the mouse studies, four short-term/sub chronic studies in male rats reported
significant increases in relative kidney weight at doses as low as 1.25 mg/kg/day {NTP, 2019,
5400978}, 5 mg/kg/day {Cui, 2009, 757868}, 6 mg/kg/day {Goldenthal, 1978, 1291068}, and
6.34 mg/kg/day {Curran, 2008, 757871}. NTP (2019, 5400978) observed an approximately 14%
increase in relative kidney weight at the highest dose tested (5 mg/kg/day) that occurred along
with significantly decreased body weight. Small but significant increases (approximately 8%) in
relative kidney weight were also observed at 1.25 and 2.5 mg/kg/day; however, no significant
changes were observed in absolute kidney weight at any dose level. While Cui et al. (2009,
757868) did not provide absolute kidney weight data, no significant difference was observed in
body weight in the 5 mg/kg/day dose group; the study authors indicate that the increased relative
kidney weight may be due to renal hypertrophy. Body weight was affected in all other dose
groups showing changes in relative kidney weight in Goldenthal et al. (1978, 1291068), Cui et
al. (2009, 757868), and Curran et al. (2008, 757871). Curran et al. (2008, 757871) also reported
that absolute kidney weight and kidney weight relative to brain weight were both significantly
decreased in male rats exposed to 6.34 mg/kg/day, which also indicates that the increase in
relative kidney weight in that dose group was driven by decreased body weight.
NTP (2019, 5400978) also observed small but significant increases (approximately 9%) in
relative kidney weight in female rats at doses as low as 0.625 mg/kg/day, but the increase was
not significant at the highest dose tested (5 mg/kg/day). Curran et al. (2008, 757871) observed a
significant increase in relative kidney weight for female rats at doses as low as 3.73 mg/kg/day,
but body weights were significantly decreased in the same dose groups and there were no
significant changes in absolute kidney weight or kidney weight relative to brain weight.
Similarly, a chronic study in female rats reported significant increases in kidney weight relative
to body weight with the highest dose tested (1.25 mg/kg/day) but reported no change in kidney
weight relative to brain weight at the same dose, indicating these effects were also driven by the
significant decreases in body weight seen at this dose {Butenhoff, 2012, 1276144}.
Cui et al. (2009, 757868) observed altered kidney histopathology in male rats, including
turbidness and tumefaction in the epithelia of the proximal convoluted tubule, congestion in the
renal cortex and medulla, and enhanced cytoplasmic acidophilia, though only in the highest dose
group (20 mg/kg/day). Besides Cui et al. (2009, 757868), all other studies reported no treatment-
related changes in kidney histopathology {Seacat, 2003, 1290852; Curran, 2008, 757871; Yahia,
2008, 2919381; Butenhoff, 2012, 1276144; Xing, 2016, 3981506; NTP, 2019, 5400978; Li,
2021, 7643501}.
Several studies also analyzed clinical chemistry endpoints relevant to renal toxicity. At the
highest dose tested in each study (1.3 mg/kg/day-5 mg/kg/day), Seacat et al. (2003, 1290852)
and NTP (2019, 5400978) (males only) both reported significant increases in BUN in rats after
14-week and 28-day exposures, respectively. Similarly, Curran et al. (2008, 757871) observed a
C-148
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
significant trend toward increased serum urea in male rats exposed to doses up to 6.34 mg/kg/day
for 28 days, although no significant differences were detected between exposure groups. In an
extension of the Seacat et al. (2003, 1290852) study, Butenhoff et al. (2012, 1276144) reported
increased BUN in both males and females of the high dose group (approximately 0.98 mg/kg/day
and 1.25 mg/kg/day, respectively) at 27 weeks and significantly increased BUN in doses
>0.1 mg/kg/day in males and >0.3 mg/kg/day in females at 53 weeks. However, the studies that
reported increased BUN did not see concurrent increases in serum creatinine concentrations at
the same dose levels and time points {Seacat, 2003, 1290852; Curran, 2008, 757871; Butenhoff,
2012, 1276144; NTP, 2019, 5400978}; NTP (2019, 5400978) and Butenhoff et al. (2012,
1276144) consider mild increases in BUN without increases in creatinine to be more consistent
with decreased water intake and mild dehydration rather than a direct toxicological effect of
chemical exposure, though these studies did not quantify water intake in exposed animals.
Additionally, increases in BUN were not seen in male mice treated with up to 10 mg/kg/day
PFOS for 30 days {Xing, 2016, 3981506} or in male or female monkeys treated with up to
0.75 mg/kg/day PFOS for 26 weeks {Seacat, 2002, 757853}. Other clinical chemistry endpoints,
including creatine kinase {Seacat, 2002, 757853; Curran, 2008, 757871; NTP, 2019, 5400978},
uric acid {Curran, 2008, 757871}, urinary N-acetyl-b-glucosaminidase (NAG) {Xing, 2016,
3981506}, and urinalysis parameters including urine pH {Seacat, 2002, 757853; Seacat, 2003,
1290852; Curran, 2008, 757871; Butenhoff, 2012, 1276144}, were not widely assessed across
multiple studies and either showed no significant changes or inconsistent responses between
studies.
C.5.3 Mechanistic Evidence
There was no mechanistic evidence linking PFOS exposure to adverse renal outcomes in the
2016 PFOS HESD {U.S. EPA, 2016, 3603365}. There are 3 studies from recent systematic
literature search and review efforts conducted after publication of the 2016 PFOS HESD that investigated
the mechanisms of action of PFOS that lead to renal effects. A summary of these studies is shown in
Figure C-33. Additional mechanistic synthesis will not be conducted since evidence suggests but
is not sufficient to infer that PFOS leads to renal effects.
C-149
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mechanistic Pathway Animal In Vitro Grand Total
Big Data, Non-Targeted Analysis
0
1
1
Cell Growth, Differentiation, Proliferation, Or Viability
2
2
2
Cell Signaling Or Signal Transduction
1
2
2
Extracellular Matrix Or Molecules
1
1
1
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
0
1
1
Inflammation And Immune Response
1
1
2
Oxidative Stress
0
1
1
Renal Dysfunction
1
1
2
Xenobiotic Metabolism
0
1
1
Grand Total
3
2
3
Figure C-33. Summary of Mechanistic Studies of PFOS and Renal Effects
Interactive figure and additional study details available on Tableau.
C.5.4 Evidence Integration
There is slight evidence for an association between PFOS exposure and renal effects in humans
based on observed effects on measures of renal function and kidney disease in 17 low confidence
studies. The 2016 HESD for PFOS {U.S. EPA, 2016, 3603365} concluded there was evidence of
a suggestive association between PFOS and CKD. The epidemiological evidence in this review
observed positive associations between serum PFOS concentrations and CKD only in low
confidence studies {Conway, 2018, 5080465; Wang, 2019, 5080583}. There is suggestive
evidence of associations with decreased kidney function, although reverse causality (i.e.,
increases in serum perfluoroalkyl levels could be due to a decrease in GF and shared renal
transporters for perfluoroalkyls and uric acid) cannot be ruled out. There were mixed results
across the measures of renal function. Results were more consistent for eGFR, in which inverse
associations were reported by two low confidence studies {Blake, 2018, 5080657; Wang, 2019,
5080583}. Regarding hyperuricemia and uric acid levels, results varied across glomerular
function and sex. Among children, there were mixed results for associations with creatinine and
uric acid. One low confidence study reported a statistically significant decrease in eGFR in
adolescents across PFOS quartiles {Kataria, 2015, 3859835}. Additionally, given the limited
evidence, conclusions cannot be drawn between PFOS and renal effects among pregnant women
and occupational workers.
The animal evidence for an association between PFOS exposure and effects on renal toxicity is
considered indeterminate based on 10 high or medium confidence animal studies. The renal
system does not appear to be sensitive to PFOS toxicity. Effects on kidney weight were
inconsistent between species and mainly consisted of changes in relative kidney weights
occurring at relatively high doses where body weights were also decreased. These changes in
relative kidney weight are considered a reflection of changes in body weight rather than adverse
effect on the kidney. Additionally, changes in clinical chemistry parameters such as increased
C-150
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
BUN without further evidence of kidney dysfunction (e.g., increased serum creatinine) are not
generally considered adverse and may be more reflective of changes in water consumption than
effects on the kidney.
C.5.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOS exposure has the potential to cause renal effects in humans
under relevant exposure circumstances (Table C-10). This conclusion is based primarily on
effects on measures of kidney function observed in studies in humans exposed to median PFOS
ranging from 3.5 ng/mL to 11.9 ng/mL. Although there is some evidence of negative effects of
PFOS exposure on CKD, there is considerable uncertainty in the results due to inconsistency
across studies, mixed findings, limited number of studies, and potential for reverse causation.
C-151
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-10. Evidence Profile Table for PFOS Renal Effects
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.6.1)
Uric acid
10 Low confidence
studies
Increases in uric acid were
observed in both children
(3/3) and adults (4/7).
Significant increases in
uric acid were observed in
adults (2/7). Results were
consistently stratified by
CKD status, but the
direction of effect was less
consistent when stratified
by eGFR. Increases in uric
acid led to increased odds
of hyperuricemia in all
studies that assessed
hyperuricemia (5/5).
• Consistent direction
of effects among
children and adults
• Low confidence studies
Serum and urinary Significant increases in
biomarkers serum albumin were
5 Low confidence studies observed in adults (2/2),
while albumin was not
analyzed in children.
Creatinine was
significantly increased in
children (2/3), but two
studies in adults reported
inconsistent directions of
effect. A study in adults
from NHANES observed
significant positive
associations of serum
proteins with PFOS and
when linear and branched
PFOS were analyzed
separately.
• No factors noted
• Low confidence studies
©OO
Slight
All studies were of low
confidence, which found
evidence of decreased
kidney function in adults
and children, including
increased uric acid,
hyperuricemia, and
decreased eGFR. In adults,
studies found evidence of
increased albumin and
total serum proteins, and
children studies reported
evidence of decreased
creatinine. Overall,
inconsistent findings in
direction of effect and
imprecision were observed
for most outcomes. The
limitation of only low
confidence studies, mixed
results, and risk of high
bias leaves uncertainty
regarding renal outcomes
and PFOS exposures.
©OO
Evidence Suggests
Primary basis:
No evidence in animals and
human evidence indicted
effects on kidney function.
Although there is some
evidence of negative
effects of PFOS exposure
on CKD, there is
considerable uncertainty in
the results due to
inconsistency across
studies, mixed findings,
limited number of studies
and potential for reverse
causation.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
C-152
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Chronic kidney disease
2 Low confidence studies
Two studies examined
CKD in adults. Odds of
CKD was increased
among general population
adults (2/2), with one
reporting a significant
increase. The direction of
effect was not consistent
after stratification by
diabetes status.
• No factors noted
• Low confidence studies
• Limited number of
studies examining
outcome
Glomerular filtration
rate
4 Low confidence studies
One study in children
reported significantly
decreased eGFR in all
exposure groups (1/1). In
adults, significant
decreases in eGFR were
observed (2/2), but results
were less consistent after
stratification by sex.
Results in pregnant
women (1/1) were not
significant.
• Consistent direction
of effects
• Low confidence studies
Gout
1 Low confidence study
No significant associations
were observed in the
overall study population,
or in analyses stratified by
CKD status.
• No factors noted
• Low confidence study
• Limited number of
studies examining
outcome
• Potential outcome
misclassification due to
self-reported outcome
Evidence from In Vivo Animal Studies (Section C.6.2)
Kidney weight
2 High confidence
studies
7 Medium confidence
studies
Relative kidney weight
was increased in rats (3/4),
mainly occurring at
relatively high dose levels
that also resulted in
»High and medium
confidence studies
• Lnconsistency of findings OOO
across species Indeterminate
• Changes in body weight
may limit ability to Evidence was based on 10
interpret these responses high and medium
C-153
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
decreased body weight.
Relative kidney weight
was decreased in mice
(1/4) and absolute kidney
weight was decreased in
rats (1/4), both at dose
levels that also resulted in
decreased body weight.
One study in monkeys
reported no effects on
Histopathology
None of the studies that • High and medium • No factors noted ;
2 High confidence
examined kidney confidence studies 1
studies
histopathology (0/6) found • Consistent effects 1
4 Medium confidence
evidence of morphological across study design, i
studies
damage or exposure- sex, and species <
related lesions following 1
short-term, subchronic, or <
chronic exposure to ;
PFOS. J
Serum biomarkers
2 High confidence
studies
4 Medium confidence
studies
Serum BUN was increased
(3/6) mainly at the highest
dose tested and only in rats
(1 study each in monkeys,
rats, or mice found no
effects on BUN). One high
confidence study with
chronic exposure observed
increased BUN in male
and female rats at several
timepoints throughout the
study with a dose response
evident in female rats after
53 weeks of exposure. No
significant changes in
serum creatinine were
1 High and medium
confidence studies
»Incoherence of findings
in serum biomarkers of
renal function
Evidence Integration
Summary Judgment
confidence studies. The
renal system does not
appear to be sensitive to
PFOS toxicity. Effects on
kidney weight were
inconsistent between
species and mainly
consisted of changes in
relative kidney weights
occurring at relatively high
doses with body weights
also decreased. There were
no apparent exposure-
related changes observed
in kidney histopathology
or urinalysis endpoints.
Changes in clinical
chemistry parameters such
as increased BUN without
further evidence of kidney
dysfunction (e.g.,
increased serum creatinine)
are not generally
considered adverse and
may be more reflective of
changes in water
consumption than effects
on the kidney.
C-154
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
observed (5/5), including
all studies that observed
increased BUN. No
exposure-related changes
were observed for serum
uric acid (1/1) or creatine
kinase (2/2).
Urinalysis
1 High confidence study
4 Medium confidence
studies
No exposure-related
changes were observed for
urinalysis endpoints (5/5).
Urine pH was increased or
decreased (2/5), but the
changes were not exposure
related. One subchronic
study in mice found no
changes in urinary N-
acetyl-b-
glucosaminidase.
»High and medium
confidence study
• No factors noted
Notes: BUN = blood urea nitrogen; CKD = chronic kidney disease; eGFR = estimated glomerular filtration rate; NHANES = National Health and Nutrition Examination Survey.
C-155
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.6 Hematological
EPA identified 8 epidemiological and 5 animal studies that investigated the association between
PFOS and hematological effects. Of the epidemiological studies, 3 were classified as medium
confidence, 2 as low confidence, and 3 were considered uninformative (Section C.6.1). Of the
animal studies, 1 was classified as high confidence, 3 as medium confidence, 1 was considered
low confidence (Section C.6.2). Studies may have multiple judgments depending on the endpoint
evaluated. Though low confidence studies are considered qualitatively in this section, they were
not considered quantitatively for the dose-response assessment (See Main PFOS Document).
C 6.1 Human E vide nee Study Quality E valuation and
Synthesis
C.6.1.1 In traduction
The mechanisms for PFOS effects on hematological parameters might include immune
suppression, shifts in nutrients absorbed from the diet, or the influences related to other health
outcomes such as cardiometabolic or kidney dysfunction {Abraham, 2020, 6506041; Chen,
2019, 5387400; Jain, 2020, 6333438}. PFOS has been implicated in endocrine disruption, which
may affect vitamin D homeostasis {Etzel, 2019, 5043582}. It could also alter epigenetics via
DNA methylation {van den Dungen, 2017, 5080340}. The effects of PFOS on hematological
outcomes may differ by characteristics such as age, gender, race, and genetics.
Hematological health outcomes in humans were previously reviewed in the 2016 HESD for
PFOS {U.S. EPA, 2016, 3603279}. Six occupational studies and one general population study,
published prior to 2010, provided hematology data. No statistically significant associations
between PFOS exposure and hematology parameters were identified. The HESD did not
specifically discuss or draw conclusions about these parameters independent of other health
outcomes.
For this updated review, eight studies examined the association between PFOS hematological
health outcomes (Appendix D). The specific hematological parameters investigated included
hematology tests (calcium, erythrocytes, ferritin, fibrinogen, hematocrit, hemoglobin, iron),
blood coagulation tests, Vitamin D levels and deficiency and anemia.
All studies assessed exposure to PFOS using biomarkers in blood. Samples were taken from
participating pregnant women, children, adolescents, or adults. All included studies were cross-
sectional designs. Four were from the United States, three from Europe, and one from Asia.
Three studies used overlapping data from a large, ongoing survey in the United States, NHANES
{Etzel, 2019, 5043582; Jain, 2020, 6333438; Jain, 2020, 6833623}. Etzel et al. (2019, 5043582)
used 2003-2010 NHANES data for adolescents and adults 12 years and older, and Jain (2020,
6333438) and Jain (2020, 6833623), used 2003-2016 NHANES data for adults 20 years and
older. Also in the United States, Khalil et al.(2018, 4238547) included 48 obese children 8-
12 years old from a hospital lipid clinic in Dayton, Ohio. Abraham et al.(2020, 6506041)
included 101 healthy one-year-old German children in the Berlin area, including 27 children
living near a former copper smelting site. Jiang et al.(2014, 2850910) recruited 141 pregnant
women in Tianjin, China. Chen et al.(2019, 5387400) conducted a pilot study with 1,430 male
and female adults from the island of Hvar, off the coast of Croatia. A study conducted by van
C-156
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
den Dungen et al.(2017, 5080340) included 80 men aged 40-70 years in the Netherlands who
regularly consumed eel.
C.6.1.2 Study Quality
Several considerations were specific to evaluating the quality of studies on hematological
parameters. Important considerations included the influence of diet, supplement or medication
use, adiposity (due to lipid binding), disease status, and SES on both PFOS exposure and
hematology. In particular, the duration of breastfeeding is expected to be associated with both
PFOS exposure and nutrition intake {Abraham, 2020, 6506041}. The blood matrix (whole blood
vs. plasma or serum) could also affect the interpretation of results. Measuring PFOS and serum
lipids concurrently was considered adequate in terms of exposure assessment timing. Given the
long half-life of PFOS (median half-life = 3.5 years) {Li, 2018, 4238434}, current blood
concentrations are expected to correlate well with past exposures.
There are 8 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and hematological effects. Study quality evaluations for these 8
studies are shown in Figure C-34.
Based on the considerations mentioned, three studies were classified as medium confidence, two
as low confidence and three as uninformative. Two low confidence studies had deficiencies in
participant selection, confounding, or sample size. Khalil et al. (2018, 4238547) was affected by
a small sample size, the cross-sectional design, and potential residual confounding attributable to
differences in participants' SES. van den Dungen et al. (2017, 5080340) was affected by a small
sample size, concerns about selection bias, and a lack of information on key confounders such as
SES.
Three studies were rated as uninformative for hematological outcomes. For Jain (2020,
6833623), the use of PFOS as the dependent variable and health outcomes as the independent
(predictive) variable rendered the study uninformative for hazard assessment {Jain, 2020,
6833623}. Abraham et al. (2020, 6506041) and Jiang et al. (2014, 2850910) only performed
unadjusted correlation analyses and therefore did not consider the influence of potential
confounding factors.
C-157
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
¦<*<*
V
Abraham et al., 2020, 6506041 -
Chen et al., 2019, 5387400
Etzel et al., 2019, 5043582^
Jain, 2020, 6333438-
Jain, 2020, 6833623
Jiang et al., 2014, 2850910
Khalil et al., 2018, 4238547-
van den Dungen et al., 2017, 5080340 -
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
+
+
++
++
-
+
+
~
+
+
+
+
+
+
++
+
+
+
++
+
+
+
+
+
++
+
-
+
+
+
+
+
+
+
+
+
+
-
-
++
+
¦
-
+
-
--
-
+
+
+
+
-
-
-
+
+
-
+
+
-
-
Figure C-34. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Hematological Effects
Interactive figure and additional study details available on HAWC.
C.6.1.3 Findings
Two studies examined levels of 25-hydroxy vitamin D and vitamin D deficiency and a
significant association was observed in one study {Etzel, 2019, 5043582}. In adolescents and
adults from NHANES (2003-2010), Etzel et al.(2019, 5043582) observed a statistically
significant decrease in total serum 25-hydroxy vitamin D per a 2-fold increase in PFOS and
comparing the top quintile of PFOS exposure (25.9 ng/mL-435.0 ng/mL) to the lowest quintile.
Statistically significant decrease in total serum 25-hydroxy vitamin D were also observed in
participants 60 and older. A positive non-significant association with prevalence ORs for vitamin
D deficiency was also observed. In 8-12-year old U.S. children, Khalil et al. (2018, 4238547)
also observed a decrease in 25-hydroxy vitamin D levels, but it did not reach significance.
C-158
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
In adults from NHANES (2003-2016), Jain(2020, 6333438) observed small statistically
significant increases in whole blood hemoglobin levels (WBHGB) with increased PFOS
exposure among adult males or females > 20 years (Appendix D). This was true for subgroups
with or without anemia, although the magnitude of the effect was larger among those defined as
anemic. Anemia was defined as WBHGB concentrations < 12 g/dL for females or < 13 g/dL for
males. Jain (2020, 6333438) also evaluated impact of deteriorating kidney function, by
stratifying results by stages of GF. For anemic males, association between WBHGB and PFOS
concentrations were uniformly positive across worsening stages of renal failure. For anemic
females, association between WBHGB and PFOS concentrations were positive except at GF-1
(eGFR > 60 mL/min/1.73 m2). Overall, the association between WBHGB and PFOS followed U-
shaped distributions. Hemoglobin levels were also examined in pregnant women {Jiang, 2014,
2850910}. Small significant positive correlations were observed between total PFOS and
hemoglobin levels (r = 0.280, p < 0.01) as well as total PFOS and red blood cell count (RBC)
(r = 0.206, p < 0.01), although these results did not consider the influence of confounding factors
and should be interpreted with caution. In high-exposed population {van den Dungen, 2017,
5080340}, observed non-significant decreases in hemoglobin and hematocrit levels, and non-
significant increases in retinol.
Chen et al.(2019, 5387400) found that serum calcium levels among Croatian adults were
statistically significantly decreased in association with an increase in the natural log of PFOS
exposure.
C.6.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 3 studies from the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} and 2 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOS HESD that investigated the association between PFOS and hematological effects. Study
quality evaluations for these 5 studies are shown in Figure C-35
C-159
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Curran et al., 2008, 757871 -
++
NR
++
B
++
++
++
B
NTP, 2019, 5400978-
++ ++
NR
++
++
++
++
++
++
Seacat et al., 2002, 757853 -
++
B
NR
B
B
++
++
++
B
Seacat et al., 2003, 1290852 -
++ ++
NR
B
B
++
B
B
Thomford, 2002, 5432419-
+
-
NR
++
++
++
++
++
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
NR Not reported
Figure C-35. Summary of Study Evaluation for Toxicology Studies of PFOS and
Hematological Effects
Interactive figure and additional study details available on HAWC.
Hematological measures, along with other biomarkers or histopathological findings, may be
informative for assessment of the health and function of blood-forming tissues such as the spleen
and bone marrow. The focus of this section is clinical hematological endpoints including
alterations in hemoglobin and hematocrit levels and changes in red blood cell production and
structure. Four oral studies in rodents or monkeys with short-term to chronic exposure durations
evaluated the effects of PFOS on the hematological system (see PFOS Main Document).
Significantly decreased reticulocyte counts were observed in male and female Sprague Dawley
rats following 28-day oral gavage exposure to 2.5 mg/kg/day or 5 mg/kg/day {NTP, 2019,
5400978}. The percent decrease from control was 42% and 49% in the 5 mg/kg/day dose group
for males and females, respectively, indicating potential deficiencies in red blood cell
maturation. Increased incidences of decreased splenic hematopoiesis, as well as increased bone
marrow hypocellularity characterized by minimal increases in the number of adipocytes and
reductions in hematopoietic cells, were observed in both males and females at these doses (see
PFOS Main Document). NTP (2019, 5400978) suggests that a combination of these findings
may indicate a suppression in erythropoiesis.
No other effects on hematocrit, hemoglobin, mean cell volume, platelet count, and red blood
cells were reported in male or female Sprague Dawley rats in the NTP (2019, 5400978) report or
in male or female Sprague Dawley rats administered up to 20 ppm PFOS (equivalent to
C-160
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1.51 mg/kg/day or 1.77 mg/kg/day in females and males, respectively) in feed for 28 days
{Seacat, 2003, 1290852}. In a third 28-day study, female Sprague Dawley rats exposed to
100 mg/kg of PFOS in diet (highest dose tested, equivalent to 7.58 mg/kg/day), displayed
significantly reduced red blood cell numbers, hemoglobin levels, hematocrit, and mean cell
hemoglobin concentrations, though these effects were generally within 10% of control levels
{Curran, 2008, 757871}. In male rats, there was a trend toward reduced red blood cell
distribution widths (i.e., decreased range in the volume and size of erythrocytes) with increasing
PFOS dose. Circulating blood platelet numbers were unaffected, but mean platelet volume was
significantly reduced in male rats at 6.34 mg/kg/day (100 mg/kg of PFOS in the diet) and in
female rats at 3.73 mg/kg/day (50 mg/kg of PFOS in the diet). In both males and females
exposed to 100 mg/kg PFOS in the diet, equivalent to 6.34 mg/kg/day and 7.34 mg/kg/day,
respectively, the red blood cell deformability index was significantly reduced over a range of
shear stress levels. Effects on blood electrolyte levels were also noted in these rats. Notably, the
sodium/potassium ratio was increased in males and females at 100 mg/kg PFOS in the diet (7.34
mg/kg/day) while inorganic phosphate was decreased in females only at this same dose {Curran,
2008, 757871}.
Other reported hematologic effects following subchronic or chronic exposure to PFOS appear to
be minimal in the low dose range. For example, male and female Sprague Dawley rats exposed
to 0.5-20 ppm PFOS in feed (equivalent to 0.03 mg/kg/day-1.33 mg/kg/day and
0.04 mg/kg/day-1.56 mg/kg/day in males and females, respectively) for 14 weeks showed no
effects on hematocrit, hemoglobin, mean cell volume, platelet count, and red blood cells {Seacat,
2003, 1290852}. Hemoglobin levels were decreased in male Cynomolgus monkeys following a
chronic 182-day exposure to 0.75 mg/kg/day, although no changes were observed in female
monkeys. While the hemoglobin levels in males reported by Seacat et al. (2002, 757853) are
statistically significant, they are within 10% of control and no other hematologic changes were
reported in the study.
C.6.3 Mechanistic Evidence
Mechanistic evidence linking PFOS exposure to adverse hematological outcomes is discussed in
Section 3.1.1.1 of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365}. There are 2 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOS HESD that investigated the mechanisms of action of PFOS that lead to hematological
effects. A summary of these studies is shown in Figure C-36. Additional mechanistic synthesis
will not be conducted since evidence is inadequate to infer that PFOS leads to hematological
effects.
Mechanistic Pathway Animal Human Grand Total
Atherogenesis And Clot Formation
0
1
Other
0
1
Grand Total
1
1
2
Figure C-36. Summary of Mechanistic Studies of PFOS and Hematological Effects
C-161
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Interactive figure and additional study details available on Tableau.
C.6.4 Eviden ce in tegra tion
The evidence evaluating an association between PFOS exposure and hematological effects in
humans is indeterminate. The limited number of studies reporting on hematological effects of
PFOS in humans is limited and relevant outcomes were not studied in more than in one study,
hence coherence is impossible to evaluate. There is evidence for an association between
increased PFOS and slightly increased WBHGB levels {Jain, 2020, 6333438}, particularly
among anemic adults in a large NHANES study. Increases in hemoglobin and RBC may also
affect pregnant women {Jiang, 2014, 2850910}. However, it is unclear whether the observed
changes are clinically adverse. The two studies that examined 25-hydroxy vitamin D levels
reported mixed non-significant effects; three studies examined hemoglobin and also reported
mixed effects.
There is indeterminate animal evidence of an association between PFOS exposure and
hematological effects. Although the available 28-day studies in rats observed some
hematological effects, the alterations were generally within 10% of control, except for reduced
reticulocyte counts observed by NTP (2019, 5400978). These reductions in reticulocyte counts
support histopathological changes in the spleen (splenic extramedullary hematopoiesis) that have
been identified as notable immune endpoints (see PFOS Main Document). Reticulocyte counts
do not appear to be as sensitive as the corresponding histopathological findings in the spleen;
decreases in reticulocytes were observed at doses > 2.5 mg/kg/day whereas histopathological
alterations were observed at a slightly lower dose of 1.25 mg/kg/day and higher. Further, the
available subchronic and chronic studies measured hematology at various timepoints did not
observe any consistent effect of treatment on red blood cells.
C.6.4.1 Evidence Integration Judgment
Overall, there is inadequate evidence to assess whether PFOS exposure can cause hematological
effects in humans under relevant exposure circumstances (Table C-l 1).
C-162
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-ll. Evidence Profile Table for PFOS Hematological Effects
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.6.1)
25-hydroxy vitamin D
1 Medium confidence
study
1 Low confidence study
Two studies observed
decreases in 25-hydroxy
vitamin D. One of the
studies observed a
significant decrease
among the whole study
population. Results were
similar in all stratifications
and study authors reported
increased vitamin D
deficiency.
• Medium confidence
study
• Consistent direction
of effects
• Low confidence study
• Limited number of
studies examining
outcome
Anemia and whole
blood hemoglobin
(WBHGB)
1 Medium confidence
study
1 Low confidence study
One study (1/2) observed
significantly increased
WBGHB, and one study
(1/2) observed non-
significant decreases in
hemoglobin.
• Medium confidence
study
• Low confidence study
• Lnconsistent direction of
effects
OOO
Lndeterminate
Evidence for
hematological effects is
based on two studies
reporting decreased 25-
hydroxy vitamin D and
one study reporting
increased WBGHB.
Considerable uncertainty
due to limited number of
studies and unexplained
inconsistency across
studies and endpoints.
OOO
Inadequate Evidence
Primary basis:
Evidence in humans and
animals were limited and
largely non-significant.
Human relevance, cross-
stream coherence, and
other inferences'.
No specific factors are
noted.
Serum electrolytes One study observed • Medium confidence • Limited number of
1 Medium confidence significantly decreased study studies examining
study serum calcium among outcome
adults.
Evidence from In Vivo Animal Studies (Section C.6.2)
Complete blood count
1 High confidence study
3 Medium confidence
study
One short-term study in
rats found evidence of
decreased reticulocyte
counts in male and female
following PFOS exposure
(1/1). Hematocrit levels
were decreased in female
rats at the highest dose
tested following short-
• High and medium
confidence studies
• Limited number of
studies examining
outcome
• Lnconsistent direction of
effect for reticulocyte,
hematocrit, hemoglobin,
and RBC levels
OOO
Lndeterminate
Evidence was limited,
inconsistent with direction
of effect, and largely non-
significant for
C-163
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Evidence Integration
Studies and Summary and Key Factors that Increase Factors that Decrease Evidence Stream Summary Judgment
Interpretation Findings Certainty Certainty Judgment
term exposure (1/4). hematological endpoints
Decreased hemoglobin in animal models.
(2/4) was observed in
male monkeys following
chronic exposure (1/4) and
in female rats following
short-term exposure (1/4).
No effects on hemoglobin
were found after short-
term and chronic exposure
in rats (2/4). RBC was
decreased (1/4) in females
at the highest dose tested
and only in rats (2
additional studies in rats
and 1 study in monkeys
found no effects on RBC).
No significant changes in
mean cell volume (2/2)
and red cell distribution
width (1/1) were observed.
An increase in the RBC
deformity index associated
with increased PFOS dose
and log shear stress in
both male and female rats
in a short-term study (1/1).
Decreased mean platelet
volume (1/1) was
observed in male and
female rats following
short-term exposure to
PFOS. No significant
exposure-related changes
were observed in platelet
count (3/3).
C-164
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Serum electrolytes
inorganic phosphate,
chloride, and Na/K
ratio
2 Medium confidence
studies
Inorganic phosphate levels
were decreased (1/2) in
female rats chronically
exposed to the highest
dose tested (1/1) but no
significant findings were
observed in a short-term
study for male or female
monkeys (1/1). In a
chronic rat study,
increased Na/K ratio (1/1)
was observed in males and
females and no exposure-
related changes were
observed in chloride levels
(1/1).
• Medium confidence
studies
• Limited number of
studies examining
outcome
Notes: WBHGB = whole blood hemoglobin; RBC = red blood count; Na/K = sodium/potassium ratio.
C-165
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.7 Respiratory
EPA identified 5 epidemiological and 5 animal studies that investigated the association between
PFOS and respiratory effects. All 5 of the epidemiological studies were classified as medium
confidence (Section C.7.1). Of the animal studies, 1 was classified as high confidence, 3 as
medium confidence, and 1 was considered low confidence (Section C.7.2). Studies may have
multiple judgments depending on the endpoint evaluated. Though low confidence studies are
considered qualitatively in this section, they were not considered quantitatively for the dose-
response assessment (See Main PFOS Document).
C 7.1 Human E vide nee Study Quality E valuation and
Synthesis
C.7.1.1 In traduction
Respiratory health can be ascertained by several measurements. The most informative are
measurements of pulmonary function (e.g., lung volume and air flow measures determined by
spirometry, as well as respiratory sounds, sputum analysis, and blood gas tension) or pulmonary
structure (e.g., lung weight, histopathology, and chest radiography), while respiratory symptoms
(shortness of breath, cough/presence of sputum, chest tightness), history of respiratory illnesses,
and respiratory mortality have low specificity and sensitivity.
The 2016 Health Assessment for PFOS {U.S. EPA, 2016, 3603365} did not examine any
epidemiological evidence of association between exposure to this chemical and respiratory
health effects.
For this updated review, five epidemiological studies investigated the association between PFOS
and respiratory outcomes. All studies measured PFOS using biomarkers in blood. Three studies
were mother-child cohort studies conducted in Europe {Agier, 2019, 5043613; Impinen, 2018,
4238440; Manzano-Salgado, 2019, 5412076}, one was a cross-sectional case-control study
(cross-sectional analyses were performed in asthmatic cases and non-asthmatic controls)
conducted in Taiwan {Qin, 2017, 3869265}; and one was a cross-sectional study of adolescents
and young adults residing near the WTC {Gaylord, 2019, 5080201}. The five available studies
examined lung function measures in children and young adults, including forced expiratory
volume in one second (FEVi), forced vital capacity (FVC), FEVi/FVC ratio, forced expiratory
flow at 25%-75% (FEF25-75%), peak expiratory flow rate (PEF), lung volume, resistance at
oscillation frequencies of 5 Hz or 20Hz, lung function at birth, and severity of obstructive
airways disease (Appendix D).
Studies that examined respiratory illnesses or symptoms reflecting immune system responses
(e.g., asthma and allergies) and respiratory tract infections (e.g., cough) are analyzed in the
immune system section.
C. 7.1.2 Study Quality
There are 5 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and respiratory effects. Study quality evaluations for these 5 studies
are shown in Figure C-37. The five general population studies identified since the last
C-166
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
assessment were all classified medium confidence. These studies had minor deficiencies,
including concerns that co-exposures in the WTC disaster could confound the results {Gaylord,
2019, 5080201}, reduced sensitivity because of low exposure levels and narrow ranges
{Impinen, 2018, 4238440}, or concerns with potential bias in selection of non-asthmatic controls
{Qin, 2017, 3869265}.
09°^o^0o^ "o-je'
¦®xe ^
,sie
,c®
Agier et al., 2019, 5043613-
I
+
I
+
++
i
+
i
+
i
i
+
+
Gaylord et al., 2019, 5080201 -
+
+
+
-
+
+
+
+
Impinen et al., 2018, 4238440 -
+
B
++
+
+
+
-
+
¦Salgado et al., 2019, 5412076-
+
++
++
+
++
+
+
+
Qin et al., 2017, 3869265-
-
+
+
+
+
+
+
+
Legend
| Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
Figure C-37. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Respiratory Effects
Interactive figure and additional study details available on HAWC.
C. 7.1.3 Findings in Children and Adolescents
Four studies examined respiratory health effects in children up to 15 years old {Agier, 2019,
5043613; Impinen, 2018, 4238440; Manzano-Salgado, 2019, 5412076; Qin, 2017, 3869265} and
one examined adolescents and young adults ages 13-22 years {Gaylord, 2019, 5080201}
(Appendix D).
Of the four studies examining FEV1, three reported negative associations (i.e., decrease in FEV1
with higher PFOS levels), while one reported a positive association. Qin et al. (2017, 3869265)
observed significant inverse associations for children ages 10-15 years old with asthma
(beta = -0.061, 95% CI: -0.101, -0.021), and in boys with asthma, but not in girls with asthma.
There was also a significantly decreasing trend by quartiles of PFOS in children with asthma (p-
trend = 0.003). No effects were observed in children without asthma. Results from other studies
C-167
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
examining FEV1 were inconsistent and non-significant, with two studies {Gaylord, 2019,
5080201; Manzano-Salgado, 2019, 5412076} observing inverse associations and one study
{Agier, 2019, 504613} reporting a positive association.
For other lung function measures examined there was also limited evidence of associations. Qin
et al. (2017, 3869265) reported a statistically significant association with FVC (beta = -0.055,
95% CI: -0.1, -0.01) but a non-significant decreasing trend by quartiles of PFOS (p-
trend = 0.186). Non-significant associations were observed for FEF25"., 75".,or PEF or for any lung
function measures in children without asthma. Impinen et al. (2018, 4238440) reported a
statistically significant association with severe obstructive airways disease at age 2 measured by
the Oslo Severity Score (OSS), but only for the lowest severity category (OSS 1-5) (OR per log2
increase PFOS = 1.71, 95% CI: 1.16, 2.53). The study also reported a non-significant decrease in
odds of reduced lung function at birth, as measured by tidal flow volume. Clear patterns were not
observed for other lung function measures (i.e., FVC, FVC/FEVi, lung resistance, total lung
capacity, functional residual capacity, and residual volume) in the remaining studies {Gaylord,
2019, 5080201; Manzano-Salgado, 2019, 5412076}.
C 7.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 5 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and respiratory effects. Study quality evaluations for these 5 studies
are shown in Figure C-38.
C-168
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Argus, 2000, 5080012-
++
++
NR
Li et al., 2021, 7643501 -
+
+
NR
NTP, 2019, 5400978-
++
0
Thomford, 2002, 5432419-
B
Yang etal., 2021, 7643494-
++
+
r
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
B Critically deficient (metric) or Uninformative (overall)
Not reported
* Multiple judgments exist
Figure C-38. Summary of Study Evaluation for Toxicology Studies of PFOS and
Respiratory Effects
Interactive figure and additional study details available on HAWC.
Several studies have reported adverse pulmonary effects resulting from oral PFOS exposure. The
available literature primarily focuses on fetal and neonatal outcomes as several groups
hypothesized that the interactions of PFOS with pulmonary surfactants and subsequent
reductions in lung function or maturity may play a role in the increased perinatal mortality
resulting from gestational PFOS exposure {Argus Research Laboratories, 2000, 5080012;
Grasty, 2003, 1332670; Grasty, 2005, 2951495; Yahia, 2008, 2919381; Chen, 2012, 1276152;
Ye, 2012, 2919212; U.S. EPA, 2016, 3603365}. There are also several available studies that
reported pulmonary effects in adult mammalian models {Goldenthal, 1979, 9573133; Cui, 2009,
757868; NTP, 2019, 5400978; Li, 2021, 7643501; Yang, 2021, 7643494}.
Yahia et al. (2008, 2919381) exposed mouse dams to 0, 1, 10, or 20 mg/kg/day PFOS from GD
0-GD 17 and assessed neonatal and maternal lung histopathology. Initially, a single surviving
pup from each dam (n = 5/treatment group) was analyzed at PND 0; all 5 pups in the
20 mg/kg/day group showed lung atelectasis (i.e., complete or partial lung collapse) which was
characterized by alterations in the alveolar epithelium, congestion of alveolar capillary vessels,
and reduced alveolar space. Focal or severe atelectasis was also present in some of the pups from
the 10 mg/kg/day group (incidence not provided) but not in pups from the control or 1 mg/kg/day
groups. No observed histological effects of PFOS exposure were observed on the maternal lung.
Yahia et al. (2008, 2919381) dosed additional dams with 20 mg/kg/day PFOS from GD 0-GD 17
C-169
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
or 10 mg/kg/day PFOS from GD 0-GD 18 to further examine pulmonary effects in fetuses and
pups, respectively. Immediately at birth, 27% (4/15) of pups (n = 3 pups/dam) from 3/5 dams
dosed with 10 mg/kg/day PFOS showed at least mild lung atelectasis. In contrast, all fetuses in
the 20 mg/kg/day group showed normal lung histopathology at GD 18. The authors suggested an
increase in the incidence of moderate to severe intracranial blood vessel dilation in fetuses at GD
18 as a cause of the pulmonary effects that were not seen until birth {Yahia, 2008, 2919381}.
Chen et al. (2012, 1276152) similarly assessed rat pup lung histopathology at PND 0 and PND
21 after gestational exposure to 0 mg/kg/day, 0.1 mg/kg/day, or 2 mg/kg/day PFOS from GD 1-
GD 21. With PFOS exposure of 2 mg/kg/day, pups showed marked alveolar hemorrhaging,
thickened interalveolar septum, and focal lung consolidation at PND 0 (incidence data not
provided). These effects lasted through PND 21, when pups from the 2 mg/kg/day treatment
group also showed alveolar hemorrhaging, thickened interalveolar septum, and inflammatory cell
infiltration. The 2 mg/kg/day group PND 0 and PND 21 pups also had higher percentages of
pulmonary apoptotic cells. There were no pulmonary abnormalities observed in pups from the
control or 0.1 mg/kg/day groups.
Zhang et al. (2021, 6988534) reported that Sprague-Dawley rat pups exposed to 1 or
5 mg/kg/day from GD 12 to GD 18 had higher lung injury scores and that pups in the
5 mg/kg/day group had lower radial alveolar counts on PND 1,3,7, and 14 compared to
controls.
In an attempt to identify the prenatal window of susceptibility to PFOS in neonatal rats, Grasty et
al. (2003, 1332670) dosed dams with 0 mg/kg/day, 25 mg/kg/day, or 50 mg/kg/day PFOS during
several 4-day gestational timepoints, including GD 17-GD 20, a period of development they
identified in this study as a particularly sensitive window for neonatal mortality. As the last
few days of fetal development involve central nervous system and pulmonary maturation, the
authors conducted a second exposure of 0 mg/kg/day, 25 mg/kg/day, or 50 mg/kg/day PFOS
from GD 19-GD 21 and sacrificed fetuses at GD 21 or pups at PND 0 to examine lung histology
{Grasty, 2003, 1332670}. No histological differences between lung samples of control and
treated fetuses sacrificed at GD 21 were observed, though it appeared that PFOS reduced lung
expansion and slowed or compromised lung maturation of pups by PND 0; epithelial thickness of
lungs of PFOS-treated pups at PND 0 was similar to that of lungs from fetal control animals at
GD 21 (incidence data not provided). Grasty et al. (2005, 2951495) conducted a follow-up study
with the same GD 19-GD 21 exposure paradigm to further explore mechanisms of
developmental pulmonary dysfunction and potential methods of therapeutic rescue of delayed
lung maturation and effects on pulmonary surfactants seen after gestational PFOS exposure.
Grasty et al. (2005, 2951495) found several morphometric changes in pup lung tissue after
25 mg/kg/day or 50 mg/kg/day PFOS exposure, including increases in the proportion of lung
occupied by solid tissue, decreases in the proportion of lung occupied by small airways, and
increases in the ratio of solid tissue to small airway space. The authors also note that some lung
samples from the 50 mg/kg/day group did not appear to fill fully upon perfusion, potentially
indicating a failure of inflation upon birth or atelectasis. Similar to the results of Grasty et al.
(2003, 1332670), the lungs of some PFOS-exposed pups at PND 0 resembled the lungs of
control fetuses at GD 21 (incidence of 17% and 50% of pups from the 25 mg/kg/day and
50 mg/kg/day groups, respectively). Co-treatment with the therapeutic agents dexamethasone or
retinyl palmitate did not increase neonatal survival, indicating the pulmonary effects of PFOS do
C-170
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
not drive neonatal mortality, though the authors did not report histological analyses showing
improved pulmonary outcomes in co-treated animals. Ye et al. (2012, 2919212) did not observe
effects on rat fetal lung histopathology following gestational exposure to 5 or 20 mg/kg/day,
though the exposure period lasted from GD 12-GD 18 and may have missed the sensitive period
of lung development in rats {Grasty, 2003, 1332670; Grasty, 2005, 2951495}.
In a rabbit teratology study, Argus (2000, 5080012) reported a significant increase in the number
of fetuses with absent intermediate lung lobes after exposure to 0.1 mg/kg/day PFOS from GD
7-GD 20 (7/172 fetuses compared to 2/175 in controls). However, this increase was not
statistically significant when analyzed by litter (4/19 litters compared to 2/20 in controls) and no
increase was observed in the higher dose groups of 1 mg/kg/day, 2.5 mg/kg/day, or
3.75 mg/kg/day. Argus (2000, 5080012) noted that this fetal malformation was likely not related
to the test article as varied lung development is frequently observed in New Zealand White
rabbits.
Pulmonary effects were observed in adult animals after short-term and subchronic exposures to
PFOS. Cui et al. (2009, 757868) reported dose-related increases in pulmonary congestion and
focal or diffuse thickening of epithelial walls in the lungs of male rats gavaged with 5 mg/kg/day
or 20 mg/kg/day PFOS for 28 days (incidence data not provided). Focal or diffuse neutrophil,
acidophilia, and lymphocyte cellular infiltration and vasodilatation due to leakage of erythrocytes
was also especially apparent in the 20 mg/kg/day dose group (incidence data not provided). In a
study with limited sample size (n = 2/sex/treatment), Goldenthal et al. (1979, 9573133) reported
increased moderate diffuse atrophy of the serous alveolar cells in 3/4 rhesus monkeys from the
highest dose group (4.5 mg/kg/day) treated with PFOS for 90 days. NTP (2019, 5400978) did not
report nasal, olfactory, or pulmonary histopathological effects in adult male or female rats dosed
with up to 5 mg/kg/day PFOS for 28 days. However, female rats dosed with 1.25 mg/kg/day,
2.5 mg/kg/day, or 5 mg/kg/day had significantly increased relative lung weight. The biological
significance of this increase is unclear as absolute lung weight was only significantly increased
in the 1.25 mg/kg/day group and there were no accompanying histopathological alterations in the
lung. Yang et al. (2021, 7643494) examined the impacts of PFOS exposure on male C57BL/6
mice pulmonary system in a 28-day oral gavage study. Relative lung weights displayed an 1%
and 6% increase in 0.25 mg/kg/day and 2.5 mg/kg/day groups, respectively, compared to the
control group. The toxicological significance of the increase is unclear due to both low sample
size with 6 animals per group and lack of report on body weight or absolute lung weight. Li et al.
(2021, 7643501) examined the histopathological effects of PFOS exposure on female BALC/c
mice pulmonary system in a 60-day oral gavage study. Authors reported zero incidence of
lesions following respiratory histopathological examination among the female mice gavaged
with 0.1 mg/kg/day and 1 mg/kg/day.
Immunological responses in lungs were investigated in Yang et al. (2021, 7643494). No
significant differences in bronchoalveolar lavage fluid (BALF) macrophages, eosinophils,
neutrophils, and total cell counts were observed among control or dosed groups. Cytokine IL-4 in
BALF displayed significant increases in both the 0.25 mg/kg/day and 2.5 mg/kg/day dose
groups. IL-13 in BALF showed a significant increase in the 2.5 mg/kg/day dose group whereas
IFN-y in BALF did not display a significant difference. In the same study, PFOS was found to
likely exacerbate asthmatic responses. In the BALF, total cell count and eosinophil numbers
were higher in ovalbumin (OVA)-induced mice exposed to 0.25 mg/kg/day or 2.5 mg/kg/day
C-171
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOS than to OVA-induced alone. 2.5 mg/kg/day PFOS-treated OVA-induced mice showed a
33% increase in the eosinophil infiltration and 67% increase in mucus production compared to
OVA-induced alone mice.
C. 7.3 Mechanistic Evidence
Mechanistic evidence linking PFOS exposure to adverse respiratory outcomes is discussed in
Sections 3.2.5 and 3.4.1.2 of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365}. There are 3
studies from recent systematic literature search and review efforts conducted after publication of
the 2016 PFOS HESD that investigated the mechanisms of action of PFOS that lead to
respiratory effects. A summary of these studies is shown in Figure C-39. Additional mechanistic
synthesis will not be conducted since evidence suggests but is not sufficient to infer that PFOS
leads to respiratory effects.
Mechanistic Pathway In Vitro Grand Total
Cell Growth, Differentiation, Proliferation, Or Viability
3
3
Inflammation And Immune Response
1
1
Oxidative Stress
1
1
Grand Total
3
3
Figure C-39. Summary of Mechanistic Studies of PFOS and Respiratory Effects
Interactive figure and additional study details available on Tableau.
C.7.4 Evideri ce In tegra tion
The evidence evaluating associations between PFOS exposure and respiratory effects in humans
is slight, with an indication of decreased lung function in infants, children, and adolescents.
However, the results across studies are inconsistent, and there are a lack of studies examining
respiratory effects in both children and adults. Specifically, no studies were available that
assessed respiratory health effects in older adults. While there is some evidence of detrimental
respiratory health effects, particularly in children with asthma, the available epidemiological
evidence examining PFOS exposure and respiratory health is limited.
The animal evidence for an association between PFOS exposure and respiratory effects is slight,
with an indication that the developing lung may be affected in animal models, but at high doses.
Evidence in adults is less consistent with no lesions observed in medium or high confidence
studies {NTP, 2019 5400978; Li, 2021, 7643501}, but an exacerbated immune response appears
to occur in the lung based on a medium confidence study {Yang, 2021, 7643494}. Several
studies in animal models indicate that PFOS may influence fetal and neonatal lung development
which may be consistent with epidemiological assessments of reduced lung function in children,
though none of the animal studies provide quantifiable incidence data. Additionally, effects on
the pulmonary systems of fetuses and neonates generally occurred at doses above those that
result in other adverse developmental effects (see PFOS Main Document), indicating that
respiratory toxicity is not likely a highly sensitive health outcome for PFOS exposure.
C-172
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C. 7.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOS exposure has the potential to cause respiratory effects in
humans under relevant exposure circumstances (Table C-12). The conclusion is based on limited
evidence of an association between PFOS and detrimental respiratory health effects, particularly
in children with asthma, in a small number of epidemiologic studies with median exposure levels
from 5.2 ng/mL-31.5 ng/mL, and on evidence from animal models showing changes in pup lung
tissue following exposure to doses as low as 2 mg/kg/day PFOS. However, limited number of
studies and issues with inconsistency across studies raise considerable uncertainty.
C-173
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-12. Evidence Profile Table for PFOS Respiratory Effects
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.7.1)
Lung function
measures
4 Medium confidence
studies
Two studies (2/4)
observed decreases in
forced expiratory volume
in Is (FEV1) and forced
vital capacity (FVC) in
children, with one study
reporting significant
decreases among
asthmatic children. Other
studies observed small
increases in FEV1/FVC
and FEF25-75% at age 4,
but the associations were
imprecise at age 7.
• Medium confidence
studies
• Imprecision of study
findings in children
©oo
Slight
©OO
Evidence Suggests
Obstructive disease
1 Medium confidence
study
One study in infants
under 2 years old
observed significantly
increased odds of low
severity obstructive
airway disease.
• Medium confidence
study
• Limited number of studies
examining outcome
Primary basis'.
Several studies of Human evidence indicted
medium confidence f0Unddetrimental respiratoiy
evidence for decreases in health effects' Particularly
lung function measures in children with asthma
among infants, children, wlule arumal evldence
and adolescents, though indicated changes in pup
other medium confidence 'un^ l'SSLIC following
exposure. However,
limited number of studies
and issues with
imprecision across studies
raise considerable
studies did not observe
significant effects. Few
studies examined
obstructive disease
effects. Uncertainty
remains about respiratory uncertainty
outcomes among adults
in occupational settings
and in the general
population.
Evidence from In Vivo Animal Studies (Section C.7.2)
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
Histopathology
1 High confidence study
3 Medium confidence
studies
One teratology study in
rabbits (1/1) reported a
significant increase in the
number of fetuses with
absent intermediate lung
lobes after gestational
exposure to the lowest
dose of PFOS. This
increase was not
significant when analyzed
by litter and no increase
was observed following
1 High and medium
confidence studies
> Inconsistency of findings
across species and life stage
©OO
Slight
Evidence indicates that
the developing lung may
be affected. Evidence in
adults is less convincing
as limited findings were
observed in adult mice
and rats.
C-174
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
exposure to higher doses.
Three short-term and
subchronic studies in
adult male and female
mice and rats reported no
histopathological effects
in the respiratory system
after exposure (3/3).
Organ weight One short-term study • High confidence study • Limited number of studies
1 High confidence study reported female rats had examining outcome
significantly increased
relative lung weight while
the absolute weight only
increased in one dose
group. No change in lung
weight was reported in
male rats.
Notes: FEF25-75% = forced expiratory flow at 25-75%. ; FEV1 = forced expiratory volume; FVC = forced vital capacity.
C-175
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.8 Musculoskeletal
EPA identified 6 epidemiological and 1 animal studies that investigated the association between
PFOS and musculoskeletal effects. Of the epidemiological studies, 6 were classified as medium
confidence and 2 as low confidence (Section C.8.1). The animal study was classified as low
confidence (Section C.8.2). Studies may have multiple judgments depending on the endpoint
evaluated. Though low confidence studies are considered qualitatively in this section, they were
not considered quantitatively for the dose-response assessment (See Main PFOS Document).
C.8.1 Human Evidence Study Quality Evaluation and
Synthesis
C.8.1.1 Introduction
Musculoskeletal health outcomes include bone mineral density, risk of bone fractures, and risk of
osteoarthritis. Osteoporosis (characterized by weak, brittle bone) and osteoarthritis
disproportionately affect women, older individuals, and certain racial/ethnic groups {Uhl, 2013,
1937226; Khalil, 2016, 3229485}.
The 2016 HESD for PFOS {U.S. EPA, 2016, 3603365} did not previously evaluate
musculoskeletal health outcomes in humans.
For this updated review, eight studies (eight publications) examined the association between
PFOS exposure and musculoskeletal health outcomes. All studies were in the general population.
Different study designs were used, including cross-sectional, prospective cohort, and one clinical
trial {Hu, 2019, 6315798}. All studies measured PFOS in blood components (i.e., blood, plasma,
or serum), and one study {Di Nisio, 2019, 5080655} measured PFOS in semen. Three studies
{Khalil, 2016, 3229485; Lin, 2014, 5079772; Uhl, 2013, 1937226} used data from participants
in the NHANES, but the study years and outcomes examined in these studies did not overlap.
Other studies used data from various cohorts for cross-sectional analyses, including Project Viva
{Cluett, 2019, 5412438}, the POUNDS-Lost clinical trial {Hu, 2019, 6315798}, and the
ALSPAC {Jeddy, 2018, 5079850}. The studies were conducted in different populations,
including participants from England, Italy, and the United States. The specific outcomes
investigated were osteoporosis; osteoarthritis; bone area, mineral content, mineral density,
thickness (e.g., endosteal and periosteal thickness), or circumference; bone stiffness; ultrasound
attenuation and speed of sound; lean body mass; height; arm span; bone fracture; and plasma
concentrations of P-C-telopeptides of type I collagen (CTX), a marker for bone turnover.
C.8.1.2 Study Quality
Considerations specific to evaluating the quality of studies on the musculoskeletal system relate
to the causal pathways for PFOS to alter musculoskeletal development. Expectations for
musculoskeletal condition should be interpreted relative to participants' age, pubertal and/or
menopause status, thyroid hormone levels, and adiposity (BMI), which could likewise be
influenced by PFOS exposure {Cluett, 2019, 5412438; Jeddy, 2018, 5079850; Khalil, 2016,
3229485; Khalil, 2018, 4238547}. Ideally, studies would characterize these factors, adjust
models for confounding where appropriate, and capture a range of human life stages with
prospective measurement of PFOS exposure relative to health outcomes. The outcomes should
C-176
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
be well-defined and validated by biometric testing, a physician diagnosis, or medical records
where possible. An exception may be acute traumatic injuries such as fractures, which are less
likely to be subject to recall bias.
There are 8 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and musculoskeletal effects. Study quality evaluations for these 8
studies are shown in Figure C-40.
Based on the considerations mentioned, six studies were classified as medium confidence and
two as low confidence. The two cross-sectional studies {Di Nisio, 2019, 5080655; Khalil, 2018,
4238547} classified as low confidence had deficiencies in participant selection, confounding, and
study sensitivity. Participant selection was considered a deficiency mainly due to underreporting
about participation rates and participant characteristics. Other deficiencies included potential for
residual confounding by SES, small sample sizes and limited ranges of participant exposure to
PFOS.
C-177
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Cluett et al., 2019, 5412438-
++
+
++
+
++
i .
+
i
+
+
Di Nisio et al., 2019, 5080655 -
-
+
+
-
+
+
-
-
Hu et al., 2019, 6315798-
+
++
-
++
+
+
+
Jeddy et al., 2018, 5079850-
+
++
-
+
+
+
+
Khalil et al., 2016, 3229485-
++
+
+
+
+
+
+
Khalil et al., 2018, 4238547-
-
+
++
-
+
+
-
-
Lin et al., 2014, 5079772-
+
+
79
+
+
+
+
+
Uhl et al., 2013, 1937226-
+
+
+
+
++
+
+
+
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
Figure C-40. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Musculoskeletal Effects
Interactive figure and additional study details available on HAWC.
C.8.1.3 Findings from Children and Adolescents
Three studies {Cluett, 2019, 5412438; Jeddy, 2018, 5079850; Khalil, 2018, 4238547} examined
musculoskeletal outcomes in children and adolescents, and two observed effects. While the
medium confidence studies observed few statistically significant associations between PFOS and
musculoskeletal health outcomes, the associations consistently supported a harmful, rather than
beneficial, direction of effect (Appendix D). Cluett et al. (2019, 5412438) observed a statistically
significant inverse association with areal bone mineral density (aBMD) z-score (a standardized
measure of bone mineral amount relative to bone area) in children aged 6-10 years. The sex-
stratified results were not statistically significant. Inverse non-significant associations were also
C-178
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
observed with a BMD in boys and in girls with bone mineral content (BMC) z score. Jeddy et al.
(2018, 5079850) identified a statistically significant inverse association between prenatal PFOS
exposure and total lean body mass and height in 17-year old girls. The same study initially
showed inverse associations between PFOS exposure and bone mineral content or bone area, but
these were not statistically significant after adjusting for participant height.
A low confidence study in 8-12-year old children from a hospital lipids clinic in Dayton, Ohio,
{Khalil, 2018, 4238547} observed non-significant inverse associations with bone stiffness index,
broadband ultrasound attenuation, or speed of sound.
None of the studies identified in this updated review examined musculoskeletal outcomes in
pregnant women and infants.
C.8.1.4 Findings from the General Adult Population
Five studies {Khalil, 2016, 322948; Uhl, 2013, 1937226; Lin, 2014, 5079772; Hu, 2019,
6315798; Di Nisio, 2019, 5080655} examined musculoskeletal outcomes in adults in the general
population and three observed effects (Appendix D).
The four medium confidence studies observed a small number of statistically significant
associations but a consistently harmful direction of effect. The same outcomes were not
examined by multiple studies. Uhl et al. (2013, 1937226) observed higher odds of osteoarthritis
with increased PFOS exposure only in women aged 20-84 from NHANES (2003-2008), who
may have differing susceptibility to endocrine disruption. Significant associations were observed
only by younger women aged 20-49. In an overlapping NHANES study {Lin, 2014, 5079772},
observed decreased total lumbar spine bone mineral density only among younger women not in
menopause; no statistically significant association with a history of bone fractures were observed
in women aged 20 or older. Khalil et al.(2016, 3229485) observed a statistically significant
inverse association with bone mineral density of the total femur or femoral neck in women aged
12-80 years from NHANES (2009-2010). The same was true for the femoral neck only in males
aged 12-80 years. In adults aged 30-70 years from the POUNDS-Lost study, Hu et al.(2019,
6315798) observed small but statistically significant inverse associations with bone mineral
density (or two-year change in bone mineral density) in three of the six sites examined: the spine,
total hip, and hip intertrochanteric area.
A low confidence study in young men (18-24 years) from the Padova area of northeastern Italy
{Di Nisio, 2019, 5080655} did not find evidence of associations between PFOS exposure and
arm span.
C.8.2 Animal Evidence Study Quality Evaluation and
Synthesis
There is 1 study from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and musculoskeletal effects. Study quality evaluation for this 1 study
is shown in Figure C-41.
C-179
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Thomford, 2002, 5432419
NR
Legend
| Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
Not reported
' 0S>° o^°° 9^
Figure C-41. Summary of Study Evaluation for Toxicology Studies of PFOS and
Musculoskeletal Effects
Interactive figure and additional study details available on HAWC.
Limited data are available on the effect of PFOS on the musculoskeletal system other than
developmental skeletal defects resulting from gestational exposure (see PFOS Main Document).
EPA did not identify any publications that reported musculoskeletal effects outside of those
associated with developmental toxicity from the 2016 PFOS HESD {U.S, EPA, 2016, 3603365}
or the recent literature searches that were PECO relevant and determined to be medium or high
confidence rating during study quality evaluation.
C.8.3 Mechanistic Evidence
There was no mechanistic evidence linking PFOS exposure to adverse musculoskeletal outcomes
in the 2016 PFOS HESD {U.S. EPA, 2016, 3603365}. There are 6 studies from recent
systematic literature search and review efforts conducted after publication of the 2016 PFOS
HESD that investigated the mechanisms of action of PFOS that lead to musculoskeletal effects.
A summary of these studies is shown in Figure C-42. Additional mechanistic synthesis will not
be conducted since evidence suggests but is not sufficient to infer that PFOS leads to
musculoskeletal effects.
C-180
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mechanistic Pathway Animal In Vitro Grand Total
Big Data, Non-Targeted Analysis
0
2
2
Cell Growth, Differentiation, Proliferation, Or Viability
0
4
Cell Signaling Or Signal Transduction
1
4
Extracellular Matrix Or Molecules
0
1
1
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
0
1
1
Hormone Function
1
0
1
Other
1
0
1
Grand Total
2
4
6
Figure C-42. Summary of Mechanistic Studies of PFOS and Musculoskeletal Effects
Interactive figure and additional study details available on Tableau.
C.8.4 Evidence Integration
There is slight evidence of an association between PFOS exposure and musculoskeletal effects in
humans based on observed effects on bone mineral density and bone health in a limited number
of medium confidence studies. Limited evidence from individual studies supported possible
negative effects of PFOS on skeletal size (height), lean body mass, and connective tissue
disorders (osteoarthritis). No musculoskeletal health outcome epidemiologic studies were
previously reviewed in the 2016 HESD for PFOS {U.S. EPA, 2016, 3603365}.
Although relatively few studies have investigated musculoskeletal health outcomes related to
PFOS exposure, some shared conclusions can be drawn. This review observed evidence of
statistically significant associations in about 13% of all tests conducted. The observed
associations were primarily between increased PFOS exposure and decreased bone mineral
density (inconsistently among various skeletal sites), height and lean body mass in adolescence,
and osteoarthritis. These issues with bone density may correspond with the reports of reduced
ossification and skeletal deformities in developmental animal models with gestational PFOS
exposure (see PFOS Main Document). Arm span measures in adolescents were not associated
with PFOS exposure. More severe clinical outcomes, such as fracture, were not observed to be
associated with PFOS exposure. No evidence supported beneficial musculoskeletal effects of
PFOS exposure. In general, links to musculoskeletal disease were more commonly observed
among older women. Some outcomes, such as osteoporosis and osteoarthritis, may be more
relevant to examine in females, due to greater prevalence and potentially greater susceptibility to
endocrine-disrupting chemicals. Study limitations have somewhat reduced the confidence of
most studies; common issues included cross-sectional design or potential for residual
confounding.
The animal evidence for an association between PFOS and effects in the musculoskeletal system
is considered indeterminate based on lack of information in animal models.
C-181
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.8.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOS exposure has the potential to cause musculoskeletal effects
in humans under relevant exposure circumstances (Table C-13). This conclusion is based
primarily on effects on bone mineral density and bone health observed in studies in humans
exposed to median PFOS ranging from 6.4 ng/mL to 32.2 ng/mL. Although there is some
evidence of negative effects of PFOS exposure on skeletal size (height and arm span) and
connective tissue disorders (osteoarthritis, especially in older women), there is considerable
uncertainty in the results due to inconsistency across studies and limited number of studies.
C-182
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-13. Evidence Profile Table for PFOS Musculoskeletal Effects
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.8.1)
Bone parameters
5 Medium confidence
studies
1 Low confidence study
Decreases in bone mineral
content (BMC) were
observed in two studies
(2/6), with significant
decreases observed among
female children.
Reductions in bone
mineral density (BMD)
were also observed in
children and adults (4/6),
including site specific
BMD measures.
Significant decreases in
BMD were also observed
in analyses stratified by
sex. Decreases in other
measures of bone health,
such as the stiffness index,
bone area, and broadband
ultrasound attenuation,
were observed in children.
• Medium confidence
studies
• Consistency of BMD
reduction findings
across three medium
studies
• Imprecision of findings
across exposure groups
and studies
• Low confidence study
Fractures
1 Medium confidence
study
Findings regarding
incidence of fractures in
adults ages 20 years or
older were largely
imprecise.
• Medium confidence
study
Size measures
1 Medium confidence
study
1 Low confidence study
One study reported
significantly decreased
height in girls at age 17
(1/2). Findings for arm
span were largely
• Medium confidence
study
©OO
Slight
Evidence for
musculoskeletal effects is
based on studies reporting
reductions in bone health,
bone density, lean body
mass, and increased odds
of osteoarthritis.
Uncertainties remain due
to inconsistent or
imprecise results, and
limited evidence for
fractures, size measures,
and odds of osteoarthritis
or osteoporosis.
• Imprecision of findings
• Limited number of
studies examining
outcome
• Imprecision of findings
• Limited number of
studies examining
outcome
• Low confidence study
©OO
Evidence Suggests
Primary basis:
No animal evidence and
human evidence indicated
effects on bone mineral
density and bone health.
Although there is some
evidence of negative effects
of PFOS exposure on
skeletal size (height and
arm span) and connective
tissue disorders
(osteoarthritis, especially in
older women), there is
considerable uncertainty in
the results due to
inconsistency across studies
and limited number of
studies.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
C-183
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
imprecise in a study on
male high school students.
Lean body mass
1 Medium confidence
study
One study found a
significant reduction of
total lean body mass in
girls at age 17.
• Medium confidence
study
• Limited number of
studies examining
outcome
Osteoarthritis
1 Medium confidence
study
Odds of osteoarthritis
among adults aged 20-84
and among females aged
20-49 were significantly
increased.
• Medium confidence
study
• Limited number of
studies examining
outcome
Osteoporosis
1 Medium confidence
study
Findings for osteoporosis
in women aged 12-80
were largely imprecise.
• Medium confidence
study
• Lmprecision of findings
• Limited number of
studies examining
outcome
Notes: BMC = bone mineral content; BMD = bone mineral density.
C-184
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.9 Gastrointestinal
EPA identified 4 epidemiological and 2 animal studies that investigated the association between
PFOS and gastrointestinal effects. Of the epidemiological studies, 3 were classified as medium
confidence and 1 as low confidence (Section C.9.1). Of the animal studies, 1 was classified as
high confidence, and 1 was considered low confidence (Section C.9.2). Studies may have
multiple judgments depending on the endpoint evaluated. Though low confidence studies are
considered qualitatively in this section, they were not considered quantitatively for the dose-
response assessment (See Main PFOS Document).
C 9.1 Human E vide nee Study Quality E valuation and
Synthesis
C.9.1.1 In traduction
GI health outcomes were not previously evaluated in the 2016 HESD for PFOS, although
gastroenteritis frequency was considered as a marker of immune system function. Causation of
gastroenteritis cases may be difficult to disentangle, as underlying susceptibility varies, and the
infectious agent or irritant is rarely confirmed. Granum et al. (2013, 1937228) did not observe a
statistically significant association between prenatal PFOS exposure and the frequency of
gastroenteritis episodes in a child's first three years of life, as they did for PFOA {Granum, 2013,
1937228}.
PFOS exposure may affect GI health by altering molecular processes (such as those involved in
inflammation), gut mucosa integrity (by acting as surfactants) and intestinal permeability, gut
microbiota, and/or systemic susceptibility to infection {Steenland, 2018, 5079806; Xu, 2020,
6315709}. GI outcomes only assessed in the context of immune system health, including
ulcerative colitis and Crohn's disease, are discussed (see PFOS Main Document). However, some
research suggests an overall immunosuppressive effect of PFOS could reduce the efficiency of
routine childhood immunizations {Dalsager, 2016, 3858505} which might include that for
rotavirus, a common childhood cause of diarrhea and vomiting. In addition, inflammatory bowel
disease (IBD), or the chronic inflammation of the GI tract in response to environmental triggers,
can be considered an immune dysregulation response occurring in genetically susceptible
individuals {Hammer, 2019, 8776815}.
For this updated review, four studies examined the association between PFOS and GI health
outcomes. The specific outcomes investigated were diarrhea, vomiting, IBD, and IBD
biomarkers (zonulin and calprotectin). PFOS was measured in serum or blood
Dalsager et al. (2016, 3858505) used data from the ongoing, prospective OCC, a group of
pregnant women recruited 2010-2012 and their children living in the Odense area of Denmark.
Hammer et al. (2019, 8776815) examined participants in the Children's Health and the
Environment in the Faroes (CHEF) cohort, which enrolled mother-child pairs, the children's
fathers and grandparents, and young men from the Faroe Islands hospital system between 1986
and 2009. Xu et al. (2020, 6315709) examined child and adult participants from the Ronneby,
Sweden exposed to PFAS in drinking water), and unexposed individuals from a nearby town.
Timmermann et al. (2020, 6833710) examined a subset of 4-18-month-old children from a
C-185
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
randomized controlled trial of early measles vaccination, conducted in Guinea-Bissau in West
Africa from 2012 to 2015.
C.9.1.2 Study Quality
Several considerations were specific to evaluating the quality of the studies of GI symptoms. For
example, fever or a stool test might help to confirm that diarrhea and vomiting are attributable to
infection, as opposed to a chronic underlying condition or other chemical or dietary irritant.
Medical diagnoses are preferred to self-reported symptoms, although knowledge of GI disorders
has developed substantially over recent decades and diagnostic indicators continue to rapidly
evolve. Causal factors in developing GI conditions have likewise shifted over time, such as
changes in emerging contaminants, hygiene, the gut microbiome, activity and stress levels, and
dietary trends. These underlying trends may affect cohort studies with extended recruitment or
follow-up periods. Reverse causation is possible if GI conditions lead to increased intake of
PFOS from food packaging or preparation methods, increased PFOS absorption through the GI
tract, or reduced fecal excretion {Xu, 2020, 6315709}. Measuring PFOS and GI outcomes
concurrently was considered adequate in terms of exposure assessment timing. Given the long
half-life of PFOS (median half-life = 3.5 years) {Li, 2018, 4238434}, current blood
concentrations are expected to correlate well with past exposures.
There are 4 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and gastrointestinal effects. Study quality evaluations for these 4
studies are shown in Figure C-43.
Based on the considerations mentioned, one study was considered medium confidence
{Timmermann, 2020, 6833710} and three as low confidence {Dalsager, 2016, 3858505;
Hammer, 2019, 8776815; Xu, 2020, 6315709}. The medium confidence study {Timmermann,
2020, 6833710} relied on retrospective reporting of GI outcomes, which is subject to recall bias,
and did not detail the interview question used. Study sensitivity was also limited by small case
numbers and relatively low PFOS exposure levels. However, the concerns were considered
relatively minor and likely to minimally impact interpretation of the results.
Concerns in the low confidence studies included potential for selection bias, including using
unclear recruitment methods and, a convenience sample {Xu, 2020, 6315709}. Another concern
was potential for outcome misclassification or underreporting due to inconsistent participation
and adherence to the parent reporting mechanism {Dalsager, 2016, 3858505}. Another common
reason for low confidence was a serious risk for residual confounding by SES {Hammer, 2019,
8776815}. Exposure misclassification was also a concern in Xu et al. (2020, 6315709), due to
use of residential history as a proxy. Deficiencies in multiple domains contributed to an overall
low confidence rating.
C-186
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
rv\0^
,0®
Dalsager et al., 2016, 3858505-
Hammeret al., 2019, 8776815-
Timmermann et al., 2020, 6833710 -
Xu et al., 2020, 6315709
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
| Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
-
++
-
+
+
+
-
-
+
-
-
+
-
-
-
+
+
-
+
++
+
-
+
&
-*
+
-
+
+
-*
-
Figure C-43. Summary of Study Evaluation for Epidemiology Studies of PFOS and
Gastrointestinal Effects
Interactive figure and additional study details available on HAWC.
C.9.1.3 Findings
Both studies examining diarrhea observed non-significant increased association with PFOS
(Appendix D). Timmermann et al. (2020, 6833710) observed increased odds of diarrhea in very
young children (up to 9 months old) in Guinea-Bissau. Dalsager et al. (2016, 3858505) observed
non-significant increased incidence and inconsistent odds of diarrhea; similar inconsistent
associations were observed for vomiting when comparing exposure tertiles to the referent one in
1-4-year old children in Denmark.
Both studies examining IBD observed no associations with PFOS. Hammer et al. (2019,
8776815) observed a non-significant decrease in incidence of IBD in Faroese children and
adults. Xu et al. (2020, 6315709) observed non-significant decreases in levels of IBD biomarkers
calprotectin or zonulin in children and adults from Sweden.
C.9.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 2 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
C-187
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
association between PFOS and gastrointestinal effects. Study quality evaluations for these 2
studies are shown in Figure C-44.
NTP, 2019, 5400978-
Thomford, 2002, 5432419-
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
NR Not reported
Figure C-44. Summary of Study Evaluation for Toxicology Studies of PFOS and
Gastrointestinal Effects
Interactive figure and additional study details available on HAWC.
Studies on the GI effects of PFOS exposure are limited. In a study conducted by NTP (2019,
5400978), male and female Sprague-Dawley rats were orally administered 0 mg/kg/day,
0.312 mg/kg/day, 0.625 mg/kg/day, 1.25 mg/kg/day, 2.5 mg/kg/day, or 5 mg/kg/day PFOS for
28 days. Animals treated at 0 or 5 mg/kg/day showed no effects in the forestomach, glandular
stomach, intestines, pancreas, or salivary gland during histopathological examination {NTP,
2019, 5400978}.
The 2016 HESD identified an acute study in which male and female CD rats were gavaged with
a single dose of 0 mg/kg, 100 mg/kg, 215 mg/kg, 464 mg/kg, or 1,000 mg/kg of PFOS suspended
in a 20% acetone/80% corn oil mixture. Rats were observed for abnormal signs for 4 hours after
exposure and then daily for up to 14 days. All rats died in the 464 mg/kg and 1,000 mg/kg
groups, and 3/10 rats died in the 215 mg/kg group. Necropsy results indicated stomach distension
and irritation of the glandular mucosa. Based on the findings, the acute oral LD50 was 233 mg/kg
in males, 271 mg/kg in females, and 251 mg/kg combined {Dean, 1978, 9579905}.
The 2016 HESD also identified a sub-acute study in rhesus monkeys in which Goldenthal et al.
(1979, 9573133) exposed 2 rhesus monkeys/sex/dose to 0 mg/kg/day, 0.5 mg/kg/day,
1.5 mg/kg/day, or 4.5 mg/kg/day of PFOS in distilled water by gavage for 90 days. All monkeys
in the 4.5 mg/kg/day group died or were euthanized in extremis by week 7 and exhibited signs of
GI tract toxicity (anorexia, emesis, black stool) {Goldenthal, 1979, 9573133}.
C-188
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.9.3 Mechanistic Evidence
There was no mechanistic evidence linking PFOS exposure to adverse GI outcomes in the 2016
PFOS HESD {U.S. EPA, 2016, 3603365}. There are 10 studies from recent systematic literature
search and review efforts conducted after publication of the 2016 PFOS HESD that investigated
the mechanisms of action of PFOS that lead to GI effects. A summary of these studies is shown
in Figure C-45. Additional mechanistic synthesis will not be conducted since evidence is
inadequate to infer that PFOS leads to GI effects.
Mechanistic Pathway Animal Human In Vitro Grand Total
Big Data, Non-Targeted Analysis
1
0
0
1
Cell Growth, Differentiation, Proliferation, Or Viability
1
0
2
2
Extracellular Matrix Or Molecules
1
0
0
1
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
4
0
0
4
Inflammation And Immune Response
1
0
2
2
Other
5
1
1
7
Grand Total
7
1
3
10
Figure C-45. Summary of Mechanistic Studies of PFOS and Gastrointestinal Effects
Interactive figure and additional study details available on Tableau.
C.9.4 Evidence Integration
The evidence evaluating an association between PFOS exposure and GI effects in humans is
indeterminate due to the limited number of studies available for evaluation and the
methodological shortcomings of those studies. In the 2016 HESD for PFOS, GI outcomes in
humans were only assessed in the context of immune system health. Evidence is limited due to a
paucity of research and the quality of the available studies. The available research has not
demonstrated conclusive effects of PFOS on GI effects including vomiting, or diarrhea.
The animal evidence for an association between PFOS exposure and GI effects is indeterminate
based on the limited data available. The few studies that demonstrated GI effects in animal
models appeared to only observe effects in moribund or deceased individuals.
C.9.4.1 Evidence Integration Judgment
Overall, there is inadequate evidence to assess whether PFOS exposure can cause GI effects in
humans under relevant exposure circumstances (Table C-14).
C-189
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-14. Evidence Profile Table for PFOS Gastrointestinal Effects
Evidence Stream Summary and Interpretation
Evidence Integration
Summary Judgment
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence from Studies of Exposed Humans (Section C.9.1)
Diarrhea and vomiting
1 Medium confidence
study
1 Low confidence study
Two studies observed
modest, non-significant
positive associations for
diarrhea in children under
4 years of age. One study
observed inconsistent
non-significant
associations with
vomiting across exposure
tertiles in children ages
1-4 years. No studies
were conducted in
adults.
• Medium confidence
study
• Low confidence study
• Lnconsistent direction of
effects across exposure
levels and endpoints
• Limited number of
studies examining
outcome
• Lmprecision of findings
• Potential outcome
misclassification or
underreporting due to
inconsistent parental
participation
Inflammatory bowel
disease
2 Low confidence
studies
One study in children and
adults observed a modest,
non-significant negative
association for IBD
incidence. One
community-based study
observed no clear
associations for IBD
biomarkers calprotectin
and zonulin.
• No factors noted
• Low confidence studies
• Limited number of
studies examining
outcome
• Lmprecision of findings
• Potential for residual
confounding by
socioeconomic status
and decreased study
sensitivity
OOO
Lndeterminate
Evidence for
gastrointestinal effects is
based on one study
observing a modest, non-
significant association for
diarrhea and vomiting in
children under 4 years of
age. Considerable
uncertainty due to limited
number of studies and
unexplained
inconsistency across
exposure levels and
endpoints.
OOO
Inadequate Evidence
Primary basis:
Evidence in humans and
animals are largely non-
significant.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
C-190
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Evidence Integration
Summary Judgment
Histopathology
1 High confidence
study
Evidence from In Vivo Animal Studies (Section C.9.2)
No changes in
forestomach, glandular
stomach, intestines,
pancreas, or salivary
gland histopathology in
one 28-day study in male
and female rats.
• High confidence
study
• Limited number of
studies examining
outcome
OOO
Indeterminate
Evidence was limited to
one study reporting no
findings of
gastrointestinal toxicity.
Notes: IBD = inflammatory bowel disease.
C-191
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.10 Dental
EPA identified 2 epidemiological studies that investigated the association between PFOS and
dental effects. No animal studies were identified. The 2 epidemiological studies were both
classified as medium confidence (Section C.10.1). Studies may have multiple judgments
depending on the endpoint evaluated. Though low confidence studies are considered qualitatively
in this section, they were not considered quantitatively for the dose-response assessment (See
Main PFOS Document).
C.10.1 Human Evidence Study Quality Evaluation and
Synthesis
C.10.1.1 In traduction
PFOS exposure could potentially adversely affect both dentin and bone mineralization, skeletal
formation, thyroid hormones that stimulate tooth maturation and enamel sufficiency, and
immune responses to cariogenic bacteria {Puttige Ramesh, 2019, 5080517}. At a molecular
level, PFAS such as PFOS may influence tooth growth and development via activation of
peroxisome proliferator-activated receptor alpha, initiation of oxidative stress, altering gene
expression in the vascular endothelial growth factor signaling pathway for gastric cells,
hemoprotein binding, estrogen disruption, or disruption of carbonic anhydrase (needed for
enamel development) {Wiener, 2019, 5386081}.
For this updated review, two studies examined the association between PFOS exposure and
dental caries {Puttige Ramesh, 2019, 5080517; Wiener, 2019, 5386081}. The dental caries effect
was defined as presence of decay or a restoration on any tooth surface or the loss of a tooth
following tooth decay, excluding third molars {Puttige Ramesh, 2019, 5080517}. Trained
dentists performed visual and tactile exams using appropriate tools, but X-rays were not taken.
No other dental health outcomes were evaluated.
The two cross-sectional studies used data from the NHANES: Puttige Ramesh et al. (2019,
5080517} assessed data from 2,869 12-19-year-old adolescents included in 1999-2012
NHANES and Wiener and Waters (2019, 5386081) examined data from 639 children ages 3-
11 years in the 2013-2014 NHANES cycle. Therefore, no participant overlap is expected
between these studies. Exposure to PFOS was assessed via biomarkers in blood.
C.10.1.2 Study Quality
Important considerations specific to evaluating the quality of studies on dental outcomes relate to
the difficulty of characterizing risk factors for dental caries, such as diet and oral hygiene
practices. Self-reported frequency of brushing, fluoridated product use, and dental visits may be
useful indicators. Fluoride levels in local public drinking water supplies are also thought to
influence development of dental caries and tap water consumption habits differ among
households and individuals {Wiener, 2019, 5386081}. Measuring PFOS and dental outcomes
concurrently was considered adequate in terms of exposure assessment timing. Given the long
half-life of PFOS (median half-life = 3.5 years) {Li, 2018, 4238434}, current blood
concentrations are expected to correlate well with past exposures.
C-192
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
There are 2 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and dental effects. Study quality evaluations for these 2 studies are
shown in Figure C-46.
Based on the considerations mentioned, the two included studies were considered medium
confidence, wherein limitations were not expected to severely affect results interpretation.
Limitations included cross-sectional study design, which introduces some concern about whether
the exposure preceded the outcome or vice-versa {Puttige Ramesh, 2019, 5080517; Wiener,
2019, 5386081}. Puttige Ramesh et al. (2019, 5080517) was primarily limited by participant
selection, since NHANES data necessarily excluded participants who were unable or unwilling
to submit to a dental examination. This could have resulted in selection bias against individuals
with the most severe tooth decay. Dental examinations were performed on all NHANES
participants aged 2+ who did not have orofacial pain, specific medical conditions, physical
limitations, inability to comply, or were uncooperative.
^6\e°^e^e^ ^ &
l l I l l l I l
\C©
Puttige Ramesh et al., 2019, 5080517
Wiener et al., 2019, 5386081 -
I Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
Critically deficient (metric) or Uninformative (overall)
Figure C-46. Summary of Study Evaluation for Epidemiology Studies of PFOS and Dental
Effects
Interactive figure and additional study details available on HAWC.
C. 10.1.3 Findings
The two studies observed mixed effects {Puttige Ramesh, 2019, 5080517; Wiener, 2019,
5386081}. Wiener and Waters (2019, 5386081) observed borderline significant increased odds
of dental caries with increased PFOS exposure in children (OR: 1.41; 95% CI: 0.97, 2.05;
p-value = 0.069). The analysis adjusted for age, sex, race/ethnicity, ratio of family-income-to-
poverty guidelines, tooth brushing frequency, fluoride in water, percentage of sugar in the diet,
and dental visits. Puttige Ramesh et al. (2019, 5080517) observed increased odds of dental caries
only in the third quartile of exposure, but decreased odds in the second and highest quartiles
compared to the lowest, and per doubling of PFOS. Analyses did not account for age, but
C-193
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
considered gender, race, education level of parent/guardian, family-poverty-to-income ratio,
blood lead level, and serum cotinine level (an indicator of exposure to smoking). No studies of
dental health outcomes were available for pregnant women, adults, or occupational workers
(Appendix D).
C.10.2 Animal Evidence Study Quality Evaluation and
Synthesis
In the available literature, there is no reported biological consequence of PFOS exposure on
dental outcomes in animals.
C.10.3 Mechanistic Evidence
There was no mechanistic evidence linking PFOS exposure to adverse dental outcomes in the
2016 PFOS HESD {U.S. EPA, 2016, 3603365}. There are no studies from recent systematic
literature search and review efforts conducted after publication of the 2016 PFOS HESD that
investigated the mechanisms of action of PFOS that lead to dental effects. Additional
mechanistic synthesis will not be conducted since evidence is inadequate to infer that PFOS may
cause dental effects.
C. 10.4 Eviden ce In tegra tion
The evidence evaluating an association between PFOS exposure and dental effects in humans is
indeterminate based on the limited number of available studies. Dental health outcomes were not
previously reviewed in the 2016 HESD for PFOS {U.S. EPA, 2016, 3603365}. The present
review was limited by the availability of only two studies. Only one outcome was examined
(prevalence of dental caries), and while both studies observed increased odds of dental carries,
the associations were non-significant {Puttige Ramesh, 2019, 5080517; Wiener, 2019,
5386081}. These studies have exposure levels typical in the general population, large sample
sizes and low risk of bias.
The animal evidence for an association between PFOS exposure and dental effects is
indeterminate because there are no available studies in animal models examining the effects of
PFOS exposure on dental outcomes.
C. 10.4.1 Evidence Integration Judgment
Overall, there is inadequate evidence to assess whether PFOS exposure can cause dental effects
in humans under relevant exposure circumstances (Table C-15).
C-194
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-15. Evidence Profile Table for PFOS Dental Effects
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.10.1)
Dental caries
2 Medium confidence
studies
Two studies observed
non-significant increases
and decreases in odds of
dental caries. No
significant associations
observed in studies of
children and adolescents
fromNHANES.
• Medium confidence • Inconsistent direction of
studies
effects across studies
and across exposure
levels
• Limited number of
studies examining
outcome
• Imprecision of findings
ooo
Indeterminate
Evidence was limited to
two studies that reported
non-significant positive
associations to dental
caries in children and
adolescents, but results
are imprecise.
Uncertainty remains
regarding adults and
other age groups from the
general population.
OOO
Inadequate Evidence
Primary basis:
No evidence in animals
and evidence in humans is
largely non-significant.
Human relevance, cross-
stream coherence, and
other inferences'.
No specific factors are
noted.
Notes: NHANES = National Health and Nutrition Examination Survey; N/A = not applicable.
C-195
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.ll Ocular
EPA identified 1 epidemiological and 2 animal studies that investigated the association between
PFOS and ocular effects. The epidemiological study was classified as medium confidence
(Section C.ll.l). Of the animal studies, 1 was classified as high confidence, and 1 was
considered low confidence (Section C.11.2). Studies may have multiple judgments depending on
the endpoint evaluated. Though low confidence studies are considered qualitatively in this
section, they were not considered quantitatively for the dose-response assessment (See Main
PFOS Document).
C.ll.l Human Evidence Study Quality Evaluation and
Synthesis
C. 11.1.1 Introduction
For this updated review, there is one epidemiological study that investigated the association
between PFOS and ocular effects {Zeeshan, 2020, 6315698}.
This cross-sectional study conducted in Shenyang, China as part of the Isomers of C8 Health
Project in China focused on a high-exposed population, including adults aged 20 years and older,
who were randomly selected using multistage, stratified cluster sampling. Median total PFOS
serum concentrations among the 1,202 study participants were 24.07 ng/mL. Participants were
subject to a complete ophthalmic examination which included ocular history, visual acuity, and
anterior and posterior segment examinations. Several ocular conditions, reflecting effects on
different segments of the eyes, were assessed, including visual impairment (VI), vitreous
disorder, synechia, macular disorder, corneal pannus, anterior chamber depth (ACD)-shallow,
retinal disorder, lens opacity, and conjunctival disorder.
C.ll.1.2 Study Quality
There is 1 study from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and ocular effects. Study quality evaluation for this 1 study is shown
in Figure C-47.
Zeeshan et al. (2020, 6315698) was classified as medium confidence. The main limitation of this
study is the cross-sectional design, which does not allow for establishing temporality.
Participants' serum samples were collected at study enrollment only and the utilization of a
single exposure measurement may not adequately represent exposure variability; additionally, it
is unclear if exposure occurred at an etiologically relevant time period to reflect changes in
ocular function.
C-196
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure C-47. Summary of Study Evaluation for Epidemiology Studies of PFOS and Ocular
Effects
Interactive figure and additional study details available on HAWC.
C. 11.1.3 Findings
Zeeshan et al. (2020, 6315698) examined the effects of exposure to PFOS in adults aged 22-
96 years, who had lived for at least 5 years in in Shenyang, China (Appendix D). Ocular
outcomes examined include VI, vitreous disorder, synechia, macular disorder, corneal pannus,
and ACD, and combined eye disease (aggregating all ocular conditions examined). A positive
statistically significant association between VI and total serum PFOS was observed (OR: 3.11;
95% CI: 2.30, 4.20). When stratified by age, the association between combined eye disease and
total serum PFOS was statistically significant for participants aged < 65 years (OR: 1.52; 95%,
1.21, 1.91), but not for the older participants (OR: 0.91; 95% CI: 0.55, 1.51). No other
associations were observed.
C.11.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 2 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and ocular effects. Study quality evaluations for these 2 studies are
shown in Figure C-48.
C-197
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
rjj- o''
I I I
NTP, 2019, 5400978-
Thomford, 2002, 5432419-
Legend
p
Good (metric) or High confidence (overall)
+
Adequate (metric) or Medium confidence (overall)
-
Deficient (metric) or Low confidence (overall)
D
Critically deficient (metric) or Uninformative (overall)
NR
Not reported
Figure C-48. Summary of Study Evaluation for Toxicology Studies of PFOS and Ocular
Effects
Interactive figure and additional study details available on HAWC.
An eye irritation study in rabbits suggests that PFOS acts as an ocular irritant {Biesemeier, 1974,
4467668}; however, in a 28-day oral toxicity study conducted by NTP, no histological
abnormalities were noted in male or female Sprague-Dawley rats exposed to 5 mg/kg/day PFOS
{NTP, 2019, 5400978}.
C.11.3 Mechanistic Evidence
There was no mechanistic evidence linking PFOS exposure to adverse ocular outcomes in the
2016 PFOS HESD {U.S. EPA, 2016, 3603365}. There is 1 study from recent systematic
literature search and review efforts conducted after publication of the 2016 PFOS HESD that
investigated the mechanisms of action of PFOS that lead to ocular effects. A summary of these
studies is shown in Figure C-49. Additional mechanistic synthesis will not be conducted since
evidence is inadequate to infer that PFOS leads to ocular effects.
C-198
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mechanistic Pathway
In Vitro
Grand Total
Atherogenesis And Clot Formation
Cell Signaling Or Signal Transduction
Cell Growth, Differentiation, Proliferation, Or Viability
Inflammation And Immune Response
Grand Total
Figure C-49. Summary of Mechanistic Studies of PFOS and Ocular Effects
Interactive figure and additional study details available on Tableau.
The evidence evaluating an association between PFOS exposure and ocular effects in humans is
indeterminate due to limited evidence available from epidemiological studies. In the 2016 Health
Assessment for PFOS, no epidemiological evidence of an association between PFOS exposure
and ocular health effects was examined. One epidemiological study reported an association
between PFOS and VI and combined eye disease in humans. However, since only one study was
available for review and given its cross-sectional design, existing epidemiological evidence does
not allow for a definitive conclusion regarding potential detrimental ocular health effects due to
exposure to PFOS.
The association between PFOS and ocular effects is indeterminate due to the limited evidence
available in animal models. One available study in an animal model did not report
histopathological ocular abnormalities.
C. 11.4.1 Evidence Integration Judgment
Overall, there is inadequate evidence to assess whether PFOS exposure can cause ocular effects
in humans under relevant exposure circumstances (Table C-16).
C.11.4 Evidence Integration
C-199
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-16. Evidence profile table for PFOS Ocular effects
Evidence Stream Summary and Interpretation
Evidence Integration
Summary Judgment
Studies and
Interpretation
Summary and Key Factors that Increase Factors that Decrease
Findings Certainty Certainty
Evidence Stream
Judgment
Evidence from Studies of Exposed Humans (Section C.ll.l)
OOO
Eye disease
1 Medium confidence
study
The only study
examining eye disease
was a cross-sectional
study that observed
significant positive
associations between
visual impairment and
serum PFOS. The
association was also
significant for combined
eye disease, but only in
participants aged <65
years.
• Medium confidence
study
• Limited number of
studies examining
outcome
Indeterminate
ooo
Evidence was limited to
one study reporting
increases in visual
impairment in all ages
and increases in
combined eye disease in
participants aged <65
years.
Evidence from In Vivo Animal Studies (Section C.11.2)
Histopathology
1 High confidence
study
No changes in ocular
histopathology were
reported in one 28-day
study in male and female
rats.
1 High confidence
study
• Limited number of
studies examining
outcome
Indeterminate
OOO
Evidence was limited to
one study reporting no
findings of ocular
toxicity.
Primary basis:
Evidence in humans is
limited and evidence in
animals is largely non-
significant.
Human relevance, cross-
stream coherence, and
other inferences'.
No specific factors are
noted.
C-200
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.12 Dermal
EPA identified 1 epidemiological and 2 animal studies that investigated the association between
PFOS and dermal effects. The epidemiological study was classified as medium confidence
(Section C.12.1). Of the animal studies, 1 was classified as high confidence, and 1 was
considered low confidence (Section C.12.2). Studies may have multiple judgments depending on
the endpoint evaluated. Though low confidence studies are considered qualitatively in this
section, they were not considered quantitatively for the dose-response assessment (See Main
PFOS Document).
C.12.1 Human Evidence Study Quality Evaluation and
Synthesis
C.12.1.1 Introduction
For this updated review, one study examined the association between age at the occurrence of
acne and PFOS exposure. In the Puberty Cohort, a large sub-cohort of the DNBC in Denmark,
Ernst et al. (2019, 5080529) examined the association between prenatal PFOS exposure and
pubertal development. Mother-child pairs were recruited for the DNBC from 1996-2002, and
eligibility for the Puberty Cohort was determined in 2012. PFAS levels in maternal blood,
largely collected during the first trimester of pregnancy, were used to assess prenatal exposure,
and age at the occurrence of acne was self-reported by children via bi-annual questionnaire
starting in 2012 or at 11 years of age.
C.12.1.2 Study Quality
There is 1 study from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and dermal effects. Study quality evaluation for this 1 study is shown
in Figure C-50.
Ernst et al. (2019, 5080529) was considered a medium confidence study, with no major concerns
with the overall quality of the study and any identified concerns were not likely to impact the
results. Self-reporting was used to assess the occurrence of acne, a study limitation that could
introduce minor bias to the outcome assessment. Additionally, some children were sampled for
the Puberty Cohort after the onset of puberty, thus their self-reported outcome information has
increased risk of inaccurate recall. However, this was not expected to substantially impact the
accuracy of the outcome measures.
C-201
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure C-50. Summary of Study Evaluation for Epidemiology Studies of PFOS and Dermal
Effects
Interactive figure and additional study details available on HAWC.
C. 12.1.3 Findings
Ernst et al. (2019, 5080529) observed negative non-significant associations between prenatal
PFOS exposure and age at the occurrence of acne in both boys and girls. Associations remained
negative and non-significant in analyses stratified by tertiles, except for girls in the second tertile
of PFOS exposure compared to the lowest (P: 0.09; 95% CI: -4.69, 4.87) {Ernst, 2019,
5080529}. Associations in boys were negative and non-significant (Appendix D).
C.12.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 2 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOS HESD {U.S. EPA, 2016, 3603365} that investigated the
association between PFOS and dermal effects. Study quality evaluations for these 2 studies are
shown in Figure C-51.
C-202
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
,
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The evidence for potential dermal effects in experimental animals in indeterminate and limited to
a single high confidence study with no dermal lesions observed. In the available literature from
animal models, there is no reported biological consequence of oral PFOS exposure on dermal
tissue.
C. 12.4.1 Evidence Integration Judgment
Overall, there is inadequate evidence to assess whether PFOS exposure can cause dermal effects
in humans under relevant exposure circumstances (Table C-17).
C-204
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-17. Evidence Profile Table for PFOS Dermal Effects
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Summary and Key
Findings
Factors that Increase
Certainty
Factors that Decrease
Certainty
Evidence Stream
Judgment
Evidence Integration
Summary Judgment
Evidence from Studies of Exposed Humans (Section C.12.1)
Acne
1 Medium confidence
study
One study found negative
non-significant
associations with age of
acne onset among
adolescent girls and boys.
Medium confidence
study
Limited number of OOO
studies examining Indeterminate
outcome
Imprecision of findings Evidence was limited to
one study reporting non-
significant associations.
Evidence from In Vivo Animal Studies (Section C.12.2)
Histopathology
1 High confidence study
No changes in skin
histopathology were
reported in one 28-day
study in male and female
rats.
High confidence
study
Limited number of
studies examining
outcome
OOO
Indeterminate
Evidence was limited to
one study reporting no
findings of dermal
toxicity.
OOO
Inadequate Evidence
Primary basis:
Evidence in humans and
animals are largely non-
significant.
Human relevance, cross-
' stream coherence, and
other inferences'.
No specific factors are
noted.
C-205
-------
DRAFT FOR PUBLIC COMMENT
Appendix D. Detailed Information from Epidemiology Studies
D-l
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.l Developmental
D.l.l Forest Plots
Sampling Exposure Study
Period Reference Matrix Design Exposure Levels Sub-population Comparison EE
Effect Estimate
0 2 4 6 8 10 12 14 16
OR for 02(2.37-3.34
Later Souzaet Maternal Case Control Median=3.41ng/mL -- ng/mL)vsQl(<2.37 0.58
pregnancy al., 2020 Blood (25th-75th ng/mL)
percentile: 2.34-5.77
ng/mL)
1
1
1
1
1
1
1
1
1
1
1
1
1
OR for Q3 (3.35-5.73
ng/mL) vsQl(< 2.37 0.91
ng/mL)
1
1
1
1
1
1
1
—^
1
1
1
1
1
1
OR for 04 (> 5.73
ng/mL) vs Q1 (< 2.37 3.67
ng/mL)
1
1
1
1
1
1
1
1 •
1
1
1
1
1
1
Median (25th-75th ,
... . - Odds ratio (per
XuetaL, „ Cross percentiles =4.07
Cord Blood _ -J Q nc -- loglO-unit change in 4.14
2019 Sectional ng/mL (2.86-8.05 " .
ng/mL) '
1
1
1
1
1
1
1
1 *
1
1
1
1
1
1
0 2 4 6 8 10 12 14 16
Figure D-l. Odds of Small-for-gestational-age in Children from Low Confidence Epidemiology Studies Following Exposure to
PFOS
Interactive figure and additional study details available on Tableau.
D-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Small-for-gestational-age defined as birthweight below the 10th percentile for the reference population.
Souza et al. (2020, HERO 6833697) reports the odds of the fetal growth ratio <0.85.
D.1.2 Tables
Table D-l. Associations Between PFOS Exposure and Developmental Effects in Recent Epidemiological Studies
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Ashley-Martin
etal.,2017,
3981371
High
Canada, 2008-
2011
Cohort
Pregnant women Maternal blood BW (z-score): Regression
(enrolled if <14
weeks gestation,
>18 years of age)
and their infants at
recruitment and
from MIREC
N = 1,509
Early pregnancy
4.6 (3.2-6.8)
adequate,
inadequate,
and excess
weight gain
coefficient per
loglO-unit
increase PFOS
BW: 0.05 (-0.18, 0.29)
Females: 94.31 (-76.3, 264.92)
Males: -11.15 (-174.26, 151.95)
Adequate weight gain: -0.03 (-
0.49,0.41)
Excess weight gain: 0.25 (-0.11,
0.62)
Inadequate weight gain: -0.24 (-
0.95,0.45)
MIREC = Maternal-Infant Research on Environmental Chemicals (MIREC)
Outcome: Weight gain adequacy based on Institute of Medicine (IOM) guidelines
Confounding: Maternal age, pre-pregnancy BMI, parity, household income, smoking, each PFAS°
Bachetal.,
2016,3981534
High
Denmark,
2008-2013
Cohort
Pregnant women
Maternal serum
BL (cm), BW
Regression
BL: 0 (-0.1, 0.2)
and their infants
Early pregnancy
(g, z-score),
coefficient or
Q2
-0.3 (-0.7, 0)
from the Aarhus
8.3 (6.0-10.8)
gestational
OR (PTB) per
Q3
-0.1 (-0.4,0.3)
Birth Cohort
length
IQR increase in
Q4
-0.1 (-0.5,0.2)
N = 1,507
(weeks), HC
PFOS and by
(cm), PTB
quartiles
BW (g): -8 (-30, 14)
Q2
-66 (-122,-11)
Q3
-30 (-86, 26)
Q4
-58 (-105, 8)
Females: -32 (-71, 7)
Q2
-44 (-140, 52)
Q3
-55 (-148, 38)
Q4
-71 (-174,31)
Males: 26 (-13, 65)
Q2
-129 (-239, -19)
Q3
9 (-93, 110)
D-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
Q4:-37 (-141, 67)
BW (z-score): -0.02 (-0.07, 0.04)
Q2
Q3
Q4
-0.15 (-0.29,-0.02)
-0.06 (-0.19,0.07)
-0.11 (-0.25,0.02)
Gestational length: 0 (-0.1, 0.1)
Q2: -0.1 (-0.4,0.1)
Q3: 0 (-0.2,0.3)
Q4: 0 (-0.3, 0.2)
HC: 0 (-0.1, 0.1)
Q2
Q3
Q4
-0.2 (-0.5, 0)
-0.1 (-0.4,0.1)
-0.1 (-0.3,0.2)
Results: Lowest quartile used as reference.
Confounding: Maternal age, pre-pregnancy BMI and educational level, GA
PTB: 0.85 (0.6, 1.21)
Q2: 0.96 (0.53, 1.74)
Q3: 0.65 (0.34, 1.26)
Q4: 0.82 (0.44, 1.53)
Bell et al., 2018, United States,
5041287 2008-2010
High
Cross-sectional Singleton and twin Blood
BL (cm), BW Regression
BL
infants born in Later pregnancy (g), GA
coefficient per S: -0.04 (-0.10, 0.1)
from Upstate
KIDS
N = 2,071
singletons; 1,040
twins
Singletons: 1.72 (weeks), HC
(1.14-2.44) (cm),
Twins: 1.64 ponderal
(1.09-2.33) index
log(PFOS+l)
unit increase
T: 0.23 (-0.07, 0.53)
BW
S:-18.32 (-12.41, 5.78)
T: 3.91 (-31.07, 38.89)
GA
S: 0.05 (-0.03, 0.13)
T:-0.02 (-0.15, 0.11)
HC
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
S: 0.03 (-0.19, 0.24)
T: 0.23 (-0.04, 0.49)
Comparison: Logarithm base not specified.
Results: S = Singletons; T = Twins
Confounding: Maternal age, maternal BMI, maternal education, infertility treatment, parity
Ponderal index
S: -0.01 (-0.03,0.01)
T: -0.01 (-0.04, 0.01)
Bjerregaard- Denmark,
Olesenetal., 2011-2013
2019,5083648
High
Cohort Pregnant women
and their children
from FETOTOX
N = 671
Maternal serum BL (cm), BW
Early pregnancy (g), HC (cm)
IQR = 4.12
Regression BL: -0.1 (-0.3, 0.2)
coefficient per Females: -0.4 (-0.8, 0)
IQR increase in Males: 0.2 (-0.1, 0.5), Interaction
serum PFOS p-value = 0.022
BW: -15 (-62, 32)
Females:-81 (-147,-14)
Males: 38 (-28, 105), Interaction p-
value = 0.013
HC: 0 (-0.2, 0.1)
Females: -0.1 (-0.4, 0.1)
Males: 0 (-0.2, 0.2), Interaction p-
value = 0.404
Confounding: Age at delivery, pre-pregnancy BMI, educational level, smoking, alcohol intake, GA at birth
Buck Louis et
United States, Cohort Pregnant women
Maternal blood
Umbilical Regression
Umbilical circumference: 0.04 (-
al., 2018,
2009-2013 (age range 18-40
Early pregnancy
circumference coefficient per
0.09,0.16)
5016992
years) with
5.13 (3.39-7.89) (cm), upper SD increase in
Upper arm length: -0.04 (-0.1,0.1)
High
singleton
arm length log-PFOS
Upper thigh length: -0.03 (-0.1,
pregnancies from
(cm), upper
0.04)
the NICHD Fetal
thigh length
Growth Studies
(cm)
N = 2,106
NICHD = National Institute of Child Health and Human Development
Comparison: Logarithm base not specified.
Confounding: Maternal age, education, pre-pregnancy body mass index, serum cotinine, infant sex, chemical-maternal race/ethnic interaction
D-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population, Ages,
N
Exposure
Matrix, Sample
Timing, Levels3
Outcome Comparison
Resultsb
Chu et al., 2020, China,
6315711 2013
High
Cohort Pregnant women
(aged 18-45 years)
and infants from
Guangzhou Birth
Cohort Study
N = 372
Maternal serum
Later pregnancy
1.538 (0.957-
2.635)
Females: 1.497
(0.920-2.642)
Males: 1.558
(0.988-2.628)
BW (g), GA
(weeks),
LBW, PTB
Regression
coefficient
(BW, GA) or
OR (LBW,
PTB) per ln-unit
increase in
PFOS or by
quartiles
BW:-83.28 (-133.2,-33.36)
Females: -71.91 (-143.86, 0.05)
Males: -71.52 (-142.44, -0.61)
p-value for interaction by sex =
0.678
GA:-0.32 (-0.53,-0.11)
Females: -0.61 (-0.9, -0.32)
Males: 0.004 (-0.31, 0.32)
p-value for interaction by sex =
0.003
LBW: 2.43 (1.08, 5.47)
Q2: 0.83 (0.11,6.47)
Q3: 1.41 (0.23,8.82)
Q4: 3.7 (0.61,22.58)
p-trend <0.001
PTB: 2.03 (1.24, 3.32)
Q2: 2.22 (0.55, 9.05)
Q3: 4.52 (1.21, 16.88)
Q4: 4.99 (.134, 18.56)
p-trend = 0.003
Outcome: LBW defined as BW < 2500 g
Results: Lowest quartile used as reference.
Confounding: Maternal age, maternal occupation, maternal education, family income, parity for all outcomes; GA for BW and LBW; child
sex for B W and GA
Costa et al.,
2019,5388081
High
Spain, 2003-
2008
Cohort
Pregnant women Maternal plasma AC, FL, BPD, Percent change AC
and their children
from INMA study
N = 1,230 (Girls =
597, Boys = 633)
6.05 (4.52-7.82)
estimated
fetal weight
at 12 weeks,
20 weeks,
and 34
weeks
per twofold
increase in
PFOS
12 wk: 1.4 (-2.1, 4.9)
Girls: 2.3 (-2.8, 7.1)
Boys: 0.8 (-3.8, 5.4)
20 wk: 2.2 (-1.3, 5.6)
Girls: 4.0 (-0.9, 8.8)
Boys: 0.5 (-4.1, 5.0)
34 wk: 2.1 (-1.3, 5.5)
Girls: 1.2 (-3.6, 5.8)
D-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
Boys: 2.8 (-1.8,7.2)
FL
12 wk: 1.2 (-2.3, 4.8)
Girls: 0.3 (-4.7, 4.9)
Boys: 2.0 (-2.6, 6.6)
20 wk:-0.6 (-4.1, 2.9)
Girls:-1.7 (-6.5, 3.1)
Boys: 0.0 (-4.6, 4.7)
34 wk: 1.2 (-4.1, 6.5)
Girls: 1.3 (-3.6,6.1)
Boys: 1.7 (-2.9,6.2)
BPD
12 wk: 0.5 (-3.0, 3.9)
Girls: 1.6 (-3.3, 6.4)
Boys: -0.9 (-8.2, 6.3)
20 wk: 1.3 (-2.3,4.8)
Girls: 1.2 (-3.7, 6.0)
Boys: 1.2 (-3.5, 5.9)
34 wk: 0.9 (-2.7, 4.4)
Girls: 0.0 (-4.9, 4.7)
Boys: 1.2 (-3.5, 5.9)
Estimated Fetal Weight
12 wk: 1.9 (-1.7, 5.4)
Girls: 1.3 (-3.5, 6.2)
Boys: 2.5 (-2.3,7.1)
20 wk: 2.6 (-0.9, 6.1)
Girls: 2.4 (-2.4, 7.2)
Boys: 1.0 (-3.7, 5.3)
34 wk: 2.6 (-0.9, 6.1)
Girls: 1.8 (-3.2, 6.5)
Boys: 3.0 (-1.7,7.5)
INMA = INfancia y Medio Ambiente (Environment and Childhood) Project
D-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Darrow et al.,
2013,2850966
High
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample
^ Timing, Levels3
Outcome Comparison
Resultsb
Confounding: Cohort, parity, maternal age, country of birth, smoking at week 12, maternal pre-pregnancy BMI, studies, season of last
menstrual period
United States Cohort Pregnant women
2005-2011 from the C8HP
exposed through
drinking water,
Ages >19
LBW, all births
N = 1,629
LBW, first
prospective birth
N = 783
BW, all births
N = 1,470
BW, first
prospective birth
N = 710
PTB, all births
N = 1,628
PTB, first
prospective birth
N = 783
Maternal serum LBW, BW OR (LBW, LBW
at enrollment (g), PTB PTB) and All births
13.9(9.5-19.7) regression Per ln-unit increase: 1.12 (0.75,
coefficient 1.67
(BW) Per IQR increase: 1.12 (0.87, 1.44)
per ln-unit Q2:l.48 (0.71, 3.08)
increase in Q3: 1.23 (0.57, 2.65)
PFOS, per IQR Q4: 1.31 (0.59, 2.94)
increase in Q5:1.33 (0.60, 2.96)
PFOS, or by p-value for trend = 0.651
quintiles First prospective birth
Per ln-unit increase: 0.97 (0.61,
1.54)
Per IQR increase: 0.93 (0.63, 1.37)
Q2: 1.65 (0.52, 5.20)
Q3: 0.95 (0.30,3.01)
Q4: 1.17 (0.36,3.78)
Q5: 0.82 (0.25,2.70)
p-value for trend = 0.484
BW
All births
Per ln-unit increase: -29 (-66, 7)
Per IQR increase: -23 (-48, 3)
Q2: -25 (-96, 48)
Q3:-37 (-109, 35)
Q4: -83 (-152, -13)
Q5: -54 (-124, 17)
p-value for trend = 0.045
First prospective birth
Per ln-unit increase: -49 (-90, -8)
Per IQR increase: -29 (-58, 0)
Q2: -33 (-140, 74)
Q3: -115 (-216,-14)
D-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
Q4: -149 (-244, -54)
Q5: -105 (-196,-13)
p-value for trend = 0.006
PTB
All births
Per ln-unit increase: 1.02 (0.78,
1.35)
Per IQR increase: 1.03 (0.83, 1.27)
Q2: 1.11 (0.63, 1.94)
Q3: 0.76 (0.42, 1.36)
Q4: 1.00 (0.56, 1.78)
Q5: 1.07 (0.58, 1.95)
p-value for trend = 0.976
First prospective births
Per ln-unit increase: 1.02 (0.72,
1.45)
Per IQR increase: 0.95 (0.73, 1.25)
Q2: 1.07 (0.44,2.59)
Q3: 0.63 (0.25, 1.59)
Q4: 1.08 (0.47,2.46)
Q5: 0.86 (0.36,2.04)
p-value for trend = 0.818
C8HP = C8 Health Project
Outcome: PTB defined as births occurring before 37 weeks gestation. LBW defined as those weighing less than 2,500 g.
Results: Lowest quintile used as reference.
Confounding: Maternal age, educational level, smoking status, parity, BMI, self-reported diabetes, time between conception and serum
management (year strata). Additional confounding for BW: indicator variables for gestational week.
Eick et al.,
2020,7102797
High
United States
2014-2018
Cohort
Second trimester Maternal serum BW (g, z- Regression
BW (g)
pregnant women from the second score), GA coefficient by T2: 1.62 (-105.53, 108.77)
from the CIOB
cohort
BW (g)
N = 461
trimester (weeks),
1.93 (1.18-3.13) PTB
tertile
PTB:
OR by tertile
T3: 14.26 (-101.51, 130.03)
BW (z-score)
T2: -0.01 (-0.24, 0.22)
T3: 0.02 (-0.23,0.27)
D-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
GA, BW (z-score),
PTB
N = 506
GA
T2: -0.19 (-0.64,0.26)
T3: -0.08 (-0.59,0.43)
PTB
T2: 1.21 (0.50,2.91)
T3: 1.87 (0.72,4.88)
CIOB = Chemicals in our Bodies
Outcome: PTB defined as birth at <37 weeks gestation.
Results: Lowest tertile used as reference.
Confounding: Maternal age, maternal race/ethnicity, pre-pregnancy BMI, maternal education, smoking status, parity, food insecurity.
Gardener et al., United States Cohort Pregnant women Maternal serum GA at birth GA at birth and
2021,7021199 Recruitment: in third trimester from primarily (weeks), BW: Mean by
High 2009 (ages 18^19) and third trimester BW (z- quartile
children at birth 3.9(2.6-5.9) score), GA
from the Vanguard <37 weeks GA <37 weeks
Pilot Study of the and B W: OR by
NCS quartile
GA at birth
Mean
Ql: 38.92 (38.58, 39.26)
Q2: 38.53 (38.19, 38.87)
Q3: 38.77 (38.43, 39.09)
Q4: 38.77 (38.42, 39.10)
p-trend = 0.77
GA at birth
N = 433
BW
N = 403
BW
Mean
Ql:-1.15 (-4.63,2.32)
Q2: 0.56 (-2.72, 3.84)
Q3: 1.16 (-2.06, 4.38)
Q4: 1.10 (-2.29, 4.46)
p-trend = 0.35
OR
Q2: 0.93 (0.43, 2.04)
Q3: 1.41 (0.66,3.03)
Q4: 0.81 (0.36, 1.82)
p-trend = 0.40
GA <37 weeks
OR
Q2: 1.94 (0.66, 5.68)
Q3: 1.13 (0.34,3.73)
D-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Q4: 1.41 (0.46,4.33)
p-trend = 0.82
NCS = National Children's Study
Results: Lowest quartile used as reference.
Confounding: Maternal age, education, race/ethnicity, pre-pregnancy BMI, prenatal smoking, parity, GA at serum collection.
Govarts et al.,
2016,3230364
High
Belgium, 2008- Cohort
2009
Mother-newborn Cord blood BW (g) Regression
pairs from FLEHS coefficient per
II 2.63 ^L (1.70- IQR increase in
N = 213 3.90 nL) PFOS
FLEHS II = Flemish Environmental and Health Study II
Confounding: GA, child's sex, smoking of the mother during pregnancy, parity, maternal pre-pregnancy BMI
10.82 (-72.4, 94.05), p-value :
0.798
D-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population, Ages,
N
Exposure
Matrix, Sample
Timing, Levels3
Outcome Comparison
Resultsb
Huo et al., 2020, China, 2013-
6835452 2016
High
Mothers (aged >
20 years) and their
children from the
Shanghai Birth
Cohort
N = 2,849
Cohort Mothers (aged > Maternal blood GA (weeks), Regression GA: 0.02 (-0.08, 0.12)
Later pregnancy PTB coefficient (GA) Tl:-0.27 (-0.62, 0.08)
9.33 (6.54- (indicated, perln-unit T2:0.26 (-0.43,0.96)
13.65) non- increase in T3:0.03 (-0.24,0.29)
spontaneous, PFOS and per OR T2: 0.08 (-0.06, 0.21)
spontaneous, tertile OR T3: 0.06 (-0.08, 0.19)
and overall)
OR (PTB) per PTB, overall: 0.86 (0.63, 1.17)
ln-unit increase T2: 0.61 (0.4, 0.94)
in PFOS and per T3: 0.73 (0.48, 1.1)
tertile Tl (per ln-unit increase): 2.67
(0.85, 8.29)
T2 (per ln-unit increase): 0.63
(0.05, 8.04)
T3 (per ln-unit increase): 0.83
(0.33, 2.08)
Females: 0.74 (0.45, 1.16)
Males: 0.94 (0.62, 1.41)
PTB, indicated: 1.13 (0.64, 2.01)
T2: 0.79 (0.35, 1.78)
T3: 0.99 (0.46,2.12)
PTB, non-spontaneous
Females: 1.35 (0.56, 3.26)
Males: 0.98(0.46,2.09)
PTB, spontaneous: 0.77 (0.53, 1.11)
T2: 0.56 (0.34, 0.94)
T3: 0.65 (0.4, 1.05)
Females: 0.59 (0.33, 1.06)
Males: 0.93 (0.57, 1.5)
Results: Lowest tertile used as reference.
Confounding: Maternal age, pre-pregnancy BMI, parity, parental education levels, pregnancy complicated with chronic disease, infant sex,
GA at blood drawing
D-12
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Lauritzen et al., Norway and Cohort
2017, 3981410 Sweden, 1986-
High 1988
Mother-infant Maternal serum BL (cm), BW Regression
pairs from NICHD Later pregnancy (g), GA coefficient or
SGA Norway: 9.74 (weeks), HC OR (SGA) per
N = 424 (265 from (Range = 0.95- (cm), SGA
Norway, 159 from 59.6) inPFOS
BL: -0.3 (-0.7, 0.1), p-value =
0.139
NO: 0 (-0.4, 0.4), p-value = 0.987
Sweden (78 girls,
81 boys))
Sweden: 16.4
(Range = 2.28-
55.2)
ln-unit increase SE: -1.2 (-2.1, -0.3), p-value =
0.007
BW: -15.1 (-111, 80.7), p-value =
0.757
NO: 74 (-31, 178), p-value = 0.167
SE: -292 (-500, -84), p-value =
0.006
GA: -0.07 (-0.34, 0.2), p-value =
0.601
NO: -0.01 (-0.3, 0.3), p-value -
0.952
SE: -0.4 (-0.9, 0.2), p-value =
0.201
HC: 0.04 (-0.19, 0.27), p-value =
0.748
NO: 0.2 (-0.1, 0.4), p-value = 0.189
SE: -0.4 (-0.9, 0.04), p-value =
0.073
SGA: 0.95 (0.62, 1.48), p-value =
0.833
NO: 0.71 (0.42, 1.2), p-value =
0.201
SE: 2.51 (0.93, 6.77), p-value =
0.068
NICHD SGA = The US National Institute of Child Health and Human Development (NICHD) Scandinavian Successive Small for Gestational
Age Births Study
Outcome: SGA defined as BW below the 10th percentile for GA, sex, and parity.
Results: NO = Norway; SE = Sweden
D-13
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Confounding: Maternal age, height, pre-pregnancy BMI, education, parity, smoking status at conception, interpregnancy interval, offspring
sex
Lindetal., Denmark Cohort Infants prenatally
2017,3858512 2010-2012 exposed to PFAS
High from the Odense
Child Cohort
N = 212 girls, 299
boys
Maternal serum
Early pregnancy
8.1 (6.0-11.1)
BW (g), HC Regression BW
(cm), coefficient per Males
gestational ln-unit Continuous: -17 (-130,97)
length increase in p-trend by quartiles = 0.73
(days) PFOS or by Females
quartiles Continuous: 92 (-15, 199)
p-trend by quartiles = 0.15
HC
Males
Continuous: -0.2 (-0.6, 0.2)
p-trend by quartiles = 0.38
Females
Continuous: 0.3 (-0.1, 0.7)
p-trend by quartiles = 0.12
Gestational length
Males
Continuous: -0.5 (-3.4, 2.3)
p-trend by quartiles = 0.74
Females
Continuous: -1.0 (-4.2, 2.1)
p-trend by quartiles = 0.83
Quartile analysis did not show any
statistically significant associations
Results: Lowest quartile used as reference.
Confounding: Age at examination, weight for age z-score, pre-pregnancy BMI, parity, smoking
Luo et al., 2021, China Cohort Mother-newborn Maternal blood BW (g), BL Regression BW
9959610 2017-2019 pairs and cord blood (cm), coefficient per -93.34 (-157.92,-28.75), p-value
High for BW N = 224 within three ponderal ln-unit increase <0.05
days of delivery index (kg/m3) in PFOS
D-14
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Medium for
birth length and
ponderal index
5.01 (3.32, 7.62)
BL
-0.05 (-0.38, 0.28)
Ponderal index
-0.67 (-1.08, -0.26), p-value <
0.05
Confounding: Maternal age, pre-pregnancy BMI, education, parity, environmental tobacco smoke exposure, alcohol drinking, GA, newborn
sex.
Manzano- Spain, 2003- Cohort Mother (aged >16
Salgado et al., 2008 years)-child pairs
2017, 4238465 from INMA
High N = 1,202
Maternal plasma BL (cm), BW
Early pregnancy
Mean = 6.05
(SD = 2.74)
(g), GA
(weeks), HC
(cm), LBW,
LBW at term,
PTB, SGA
Regression BL: 0.03 (-0.12, 0.17)
coefficient per p-value for sex interaction = 0.98
doubling of
PFOSorby BW: 0.44 (-32.48, 33.36)
quartiles p-value for sex interaction = 0.75
LBW, LBW at GA: -0.06 (-0.19, 0.06)
term, PTB, Q2:-0.09 (-0.33, 0.16)
SGA: OR per Q3: -0.02 (-0.26, 0.23)
log2-unit Q4: -0.31 (-0.55, -0.06); p-value <
increase in 0.05
PFOS p-value for sex interaction = 0.38
HC: 0 (-0.1, 0.1)
p-value for sex interaction = 0.53
LBW: 1.06 (-0.71, 1.58)
Females: 0.73 (0.46, 1.19)
Males: 1.90 (0.98, 3.68)
p-value for sex interaction = 0.01
LBW at term: 0.91 (0.55, 1.50)
p-value for sex interaction = 0.15
PTB: 1.10(0.70, 1.74)
p-value for sex interaction =0.35
SGA: 0.92 (0.70, 1.22)
D-15
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
p-value for sex interaction = 0.57
BL, BW, HC: No statistically
significant associations by quartiles
All outcomes: No statistically
significant associations by sex
INMA = INfancia y Medio Ambiente [Environment and Childhood Project]
Outcome: SGA defined as newborns weighing below the 10th percentile for GA and sex according to national references.
Results: Lowest quartile used as reference.
Confounding: Maternal age, parity, pre-pregnancy BMI, fish intake during pregnancy, type of delivery
Minatoya et al., Japan Cohort Pregnant women Maternal serum
2017,3981691 2002-2005 and their children
High from the Sapporo 5.1(3.7-6.7)
Cohort (Hokkaido Female mean:
Study) 5.04 (SD = 2.33)
N = 168 (90 girls, Male mean: 5.85
78 boys) (SD = 2.63)
BW (g), Regression BW
ponderal coefficient per -29 (-289, 232); p-value = 0.828
index (kg/m2) loglO-unit Females: -251 (-645, 143)
increase in Males: 190 (-162, 543)
PFOS and LSM p-value for sex interaction = 0.201
by tertiles LSM Tl: 3196 (3095, 3298)
LSM T2: 3076 (2976,3176)
LSM T3: 3158 (3057,3258)
p-trend = 0.424
Confounding: Maternal BMI, parity, smoking during pregnancy, blood sampling period, GA
Ponderal index
-2.25 (-4.01, -0.50); p-value =
0.012
Females: -2.11 (-4.86, 0.64)
Males: -2.46 (-4.74, -0.18)
p-value for sex interaction = 0.658
LSMT1: 28.39 (27.71, 29.06)
LSM T2: 26.68 (26.02, 27.34)
LSM T3: 27.23 (26.57, 27.90)
p-trend = 0.003
Rokoff et al., United States
2018,4238310 1999-2002
High
Case-control
Pregnant women
and their children
from Project Viva
N = 1,597
Maternal plasma BW for GA z- Regression
score coefficient per
Mean =29.1 IQR increase in
(SD = 16.5) PFOS
-0.03 (-0.07, 0.02)
D-16
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Confounding: Maternal age, race/ethnicity, education, pre-pregnancy BMI, and parity, black carbon, prenatal smoking
Sagiv et al., United States, Cohort Pregnant women
2018,4238410 1999-2002 and infants from
High Project Viva
N = 1,644
Maternal blood
Early pregnancy
25.7 (IQR =
16.0)
BW-for-GA Regression BW-for-GA
(z-score), coefficient per -0.04 (-0.08, 0.01)
gestational IQR increase in Q2: -0.09 (-0.22, 0.04)
length PFOS and by Q3: -0.09 (-0.22, 0.04)
(weeks), PTB quartiles Q4:-0.13 (-0.26, 0.00)
No statistically significant
PTB: associations or interactions by sex
OR per IQR
increase in Gestational length
PFOS and by -0.08 (-0.17, 0.02)
quartiles Q2: -0.20 (-0.47, 0.06)
Q3: -0.08 (-0.35,0.19)
Q4: -0.36 (-0.64, -0.09)
Females: 0.01 (-0.11,0.14)
Males:-0.19 (-0.33,-0.05)
p-value for sex interaction = 0.09
PTB
1.1 (1.0, 1.3)
Q2: 2.0 (1.1, 3.7)
Q3: 2.0 (1.1, 3.7)
Q4: 2.4 (1.3, 4.4)
Outcome: PTB was defined as <37 weeks
Results: Lowest quartile used as reference.
Confounding: Maternal age at enrollment, race/ethnicity, education, prenatal smoking, parity, history of breastfeeding, pre-pregnancy BMI,
paternal education, household income, child's sex, GA at blood draw
Shoaff et al.,
2018,4619944
High
United States,
2003-2006;
follow-up 4
weeks to 2 years
from
recruitment
Cohort Pregnant women
(aged >18 years)
and their children
at birth, 4 weeks
and 2 years from
the HOME study
N = 345
Maternal blood
Later pregnancy
14 (9.6-18)
BW (z-score),
length-for-age
(z-score),
rapid weight
gain, weight-
for-age (z-
score),
weight-for-
Regression
coefficient by
tertile (per
doubling in
PFOS)
Rapid weight
gain: RR by
tertile
BW z-score
T2: -0.05 (-0.29,0.19)
T3: -0.12 (-0.36,0.13)
p-value for trend = 0.36
Length-for-age z-score
T2: 0.05 (-0.33, 0.44)
T3: -0.24 (-0.64,0.15)
p-value for trend = 0.08
D-17
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
length (z-
score)
Weight-for-age z-score
T2:0.01 (-0.31,0.32)
T3: -0.33 (-0.65,-0.01)
p-value for trend = 0.07
Weight-for-length z-score
T2: -0.16 (-0.41,0.09)
T3: -0.31 (-0.56,-0.05)
p-value for trend = 0.66
Rapid weight gain
T2: 0.79 (0.55, 1.14)
T3: 1.11 (0.81, 1.53)
HOME = Health Outcomes and Measures of the Environment
Outcome: Rapid weight gain defined as increase in weight z-score > 0.67 SDs any time between age 4 weeks and 2 years.
Results: Lowest tertile used as reference
Confounding: Maternal age at delivery, race, marital status, insurance, income, education, parity, serum cotinine, depressive symptoms, mid-
pregnancy BMI, food security, fruit/vegetable and fish consumption during pregnancy, prenatal vitamin use
Starling et al., United States, Cohort Pregnant women
2017,3858473 2009-2014 (aged >16 years)
High and infants from
Healthy Start at
birth
N = 628
Maternal serum Adiposity (% Regression
fat mass), BW coefficient per
2.4(1.5-3.7) (g) ln-unit increase
in PFOS and by
tertiles
Adiposity: 0.08 (-0.33,
T2: 0.26 (-0.46, 0.98)
T3: -0.41 (-1.15,0.33)
0.49)
BW: -13.8 (-102.8, 35.2)
T2: -33.8 (-102.8, 35.2)
T3: -71.1 (-142.6,0.5)
Results: Lowest tertile used as reference.
Confounding: Maternal age, pre-pregnancy BMI, race/ethnicity, education, gestational weight gain, smoking during pregnancy, gravidity, GA
at blood draw, infant sex, and GA at birth
Starling et al.,
2019,5412449
High
United States,
2009-2014
Cohort
Pregnant women Maternal serum Adiposity Regression
(aged >16 years)
and infants from
Healthy Start
assessed up to 5
months
2.2 (1.4-3.4)
(%), weight-
for-age z-
score (WAZ),
weight-for-
length z-score
(WLZ), WAZ
coefficient per
ln-unit increase
in PFOS and by
tertiles
Adiposity at 5 months
-0.13 (-0.83,0.57)
Females:-0.91 (-1.84, 0.02)
Female T3: -2.08 (-3.81, -0.35)
Males: 0.73 (-0.36, 1.81)
Male T2: 1.85 (0.14, 3.47)
p-value for sex interaction = 0.05
D-18
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
N = 415 (202 girls,
213 boys)
and WLZ
growth from
birth to 5
months, rapid
growth in
WAZ or WLZ
Rapid growth:
OR per ln-unit
increase in
PFOS
WAZ at 5 months: -0.10 (-0.23,
0.02)
T3: -0.28 (-0.51,-0.05)
Females: -0.26 (-0.43, -0.10)
Female T3: -0.56 (-0.87, -0.26)
Males: 0.07 (-0.13,0.27)
p-value for sex interaction = 0.10
Valvi et al.,
2017,3983872
High
WLZ at 5 months: -0.08 (-0.23,
0.06)
Females: -0.08 (-0.23, 0.06)
Female T3: -0.52 (-0.88, -0.17)
Males: 0.06 (-0.17,0.28)
p-value for sex interaction = 0.17
WAZ or WLZ growth from birth to
5 months, rapid growth: No
statistically significant associations
Outcome: Rapid growth defined as change in WAZ or WLZ >0.67 between birth and 5 months
Results: Lowest tertile used as reference
Confounding: Maternal age, race/ethnicity, pre-pregnancy BMI, any previous pregnancies, any smoking during pregnancy, education,
gestational weight gain z-score, infant sex, exclusive breastfeeding to follow-up visit, infant age (days) at follow-up
Faroe Islands
1997-2000
Cross-sectional
Pregnant women
and their children
N = 604 (288 girls,
316 boys)
Maternal serum HC (cm), Regression HC
body length coefficient per 0 (-0.28, 0.27)
27.2 (23.1-33.1) (cm), BW doubling of Girls: 0.48 (0.05,0.90)
(g) PFOS Boys: -0.28 (-0.65, 0.09)
p-value for sex interaction = 0.01
Body length
0.05 (-0.33, 0.43)
Girls: 0.32 (-0.24, 0.89)
Boys:-0.18 (-0.60, 0.23)
p-value for sex interaction = 0.17
BW
D-19
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
-81 (-173, 11)
Girls: 5 (-124, 135)
Boys: -150 (-282, -17)
p-value for sex interaction = 0.08
Confounding: Maternal age at delivery, education, parity, pre-pregnancy BMI, smoking during pregnancy, child sex
Whitworth et
al., 2012,
2349577
High
Norway Cohort Pregnant women
2003-2004 and their children
from MoBa
Maternal plasma PTB, B W (z- OR by quartile
PTB, LGA,
N = 901
BW
N = 838
SGA
around 17
weeks of
gestation
13.0(10.3-16.6)
score),
SGA, LGA
BW:
Regression
coefficient per
unit increase in
PFOS, or by
quartile
PTB
Q2: 0.9 (0.3, 2.8)
Q3: 0.9 (0.3, 2.7)
Q4: 0.3 (0.1, 1.0)
p-trend = 0.03
LGA
Q2: 0.8 (0.5, 1.6)
Q3: 1.0(0.5, 1.7)
Q4: 0.7 (0.3, 1.4)
p-trend = 0.33
SGA
Q2: 1.2 (0.5, 3.0)
Q3: 2.2 (1.0, 5.1)
Q4: 1.3 (0.5, 3.4)
p-trend = 0.51
BW
Per increase: -0.01 (-0.02, 0.01)
Q2
Q3
Q4
-0.08 (-0.29,0.13)
-0.17 (-0.39, 0.05)
-0.18(0.41,0.05)
p-trend = 0.12
MoBa = Norwegian Mother and Child Cohort Study
Outcome: PTB defined as GA <37 weeks. SGA defined as gender-and gestation age-specific BW less than the 10th percentile. LGA defined
as gender- and GA-specific BW greater than the 90th percentile.
Confounding: Maternal age, pre-pregnancy BMI, parity. Additional confounding for BW: albumin concentration, maternal education,
interpregnancy interval, quadratic interpregnancy interval, consumption of lean fish.
Wikstrom et al., Sweden
2020,6311677 2007-2010
Cohort Infants exposed Maternal serum BW (g), BW- Regression
prenatally to PFAS Early pregnancy SDS, SGA coefficient
BW
Per increase: -46 (-88, -3)
D-20
-------
DRAFT FOR PUBLIC COMMENT
Reference, Location, . Population, Ages, Exposure
rk"'— Matrix, Samp
Timing, Levels3
„ ... Design Matrix, Sample
Confidence Years N
High from the SELMA 5.38 (3.97-7.60)
study
N = 1533 (732
girls, 801 boys)
D-21
MARCH 2023
Outcome Comparison Resultsb
(BW, BW-SDS) Q2: -27 (-89, 35)
and OR (SGA) Q3: -22 (-84, 41)
per ln-unit Q4: -80 (-144, -16)
increase in Girls
PFOS or by Per increase: -85 (-145, -25)
quartiles Q2:-32 (-115, 52)
Q3: -51 (-137, 34)
Q4: -142 (-231,-54)
Boys
Per increase: -13 (-73, 47)
Q2: -28 (-118, 63)
Q3: 5 (-86, 96)
Q4: -28 (-119, 63)
BW-SDS
Per increase: -0.100 (-0.197,
-0.004)
Q2:-0.045 -0.185,0.096)
Q3:-0.024 (-0.166,0.118)
Q4:-0.172 (-0.317,-0.027)
Girls
Per increase: -0.167 (-0.301,
-0.034)
Q2:-0.044 (-0.232,0.143)
Q3:-0.092 (-0.283,0.100)
Q4: -0.296 (-0.494, -0.098)
Boys
Per increase: -0.027 (-0.166,
0.112)
Q2:-0.055 (-0.263,0.153)
Q3: 0.038 (-0.171,0.246)
Q4:-0.066 (-0.276,0.144)
SGA
Per increase: 1.19(0.87, 1.64)
Q2: 0.69 (0.43, 1.08)
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Q3: 0.79 (0.53, 1.18)
Q4: 1.56(1.09,2.22)
Girls
Per increase: 1.40 (0.83, 2.35)
Q2: 0.89 (0.39,2.03)
Q3: 0.82 (0.36,2.03)
Q4: 2.05 (1.00,4.21)
Boys
Per increase: 1.08 (0.72, 1.63)
Q2: 1.26 (0.67,2.37)
Q3: 0.86 (0.45, 1.67)
Q4: 1.30(0.7,2.4)
SELMA = Swedish Environmental Longitudinal Mother and Child, Asthma and Allergy
Outcomes: SGA defined as BW below the 10th percentile for GA and sex.
Results: Lowest quartile used as reference.
Confounding: Sex, GA, maternal weight, parity, cotinine levels
Wikstrom et al.,
2021,7413606
High
Sweden, 2007-
2010
Nested case-
control
Pregnant women
from the SELMA
study
N = 1,527
Miscarriage
Serum during
first trimester
Case: 6.09
(3.99-8.77)
Control: 5.45
(4.00-7.68)
SELMA = Swedish Environmental Longitudinal Mother and Child, Asthma and Allergy
Confounding: Parity, age, cotinine (tobacco smoke) exposure.
OR per
doubling in
PFOS
Per doubling: 1.13 (0.82, 1.52)
Xiao et al., Denmark Cohort Pregnant women
2019,5918609 1994-1995 and their children
High N = 171
Maternal blood
Later pregnancy
GM = 20.8 (ig/g
(range: 6.9^17.6
l-ig/g)
Z-scores for
BL, BW, and
cranial
circumference
Regression
coefficient per
log2-unit
increase in
PFOS
BL z-score
-0.33 (-0.69, 0.03)
Girls: -0.23 (-0.75, 0.30)
Boys: -0.41 (-0.87, 0.05)
BW z-score
-0.47 (-0.85, -0.09)
Girls: -0.56 (-1.12,0.00)
Boys: -0.40 (-0.89, 0.08)
Cranial circumference z-score
-0.26 (-0.68,0.16)
D-22
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Girls: -0.42 (-1.05,0.21)
Boys:-0.15 (-0.68,0.39)
Confounding: Child sex, parity, maternal BMI, maternal height, maternal education, maternal age, smoking and drinking alcohol during
pregnancy, total PCB, mercury
Yao et al., 2021,
9960202
High
China
2010-2013
Cross-sectional
Parents and their
children from
LWBC
N = 369
BW (g)
Regression
coefficient per
ln-unit increase
inPFOS
B W by maternal exposure
Model A: -32.28 (-116.2, 51.64)
B W by paternal exposure
Model A: 0.19 (-74.26, 74.65)
Maternal and
paternal serum
within three
days of birth
Maternal: 4.55
(Range = 0.55-
29.85)
Paternal: 10.15
(
-------
DRAFT FOR PUBLIC COMMENT
Reference. Location, Population, Ages, Exposure
kcicc c, Design P Matrix, Sample
Confidence Years N t,. . T , a
Timing, Levels3
D-24
MARCH 2023
Outcome Comparison Resultsb
Q4: -0.26 (-0.26, -0.10); p-value <
0.05
S-boys: -0.05 (-0.11,0.01)
Q2: -0.01 (-0.16,0.15)
Q3: -0.11 (-0.27,0.06)
Q4: -0.15 (-0.32,0.02)
T: -0.03 (-0.10, 0.05)
Q2: 0.11 (-0.09,0.32)
Q3: 0.07 (-0.14,0.28)
Q4: 0.0005 (-0.2, 0.2)
Length
S: 0.07 (-0.06,0.19)
S-girls: 0.03 (-0.14,0.20)
S-boys: 0.10 (-0.07,0.27)
T: 0.18 (-0.07, 0.42)
Length z-score
S: 0.03 (-0.03, 0.08)
S-girls: 0.008 (-0.07, 0.08)
S-boys: 0.05 (-0.03,0.12)
T: 0.07 (-0.04, 0.18)
Weight
S: -21.99 (-59.52, 15.55)
S-girls: -51.57 (-102.32, -0.82); p-
value < 0.05
S-boys: 6.15 (-48.31,60.61)
T: 62.47 (-13.97, 138.92)
Weight z-score
S: -0.03 (-0.08,0.01)
S-girls: -0.07 (-0.13, -0.01); p-
value < 0.05
S-boys: -0.001 (-0.06, 0.06)
T: 0.04 (-0.04, 0.12)
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
WFL z-score
S: -0.08 (-0.12, -0.04)
S-girls: -0.10 (-0.16, -0.05); p-
value < 0.05
S-boys: -0.05 (-0.11,0.01)
T: -0.03 (-0.11,0.05)
Rapid weight gain, obesity: not
statistically significant for all
children
Outcome: Rapid weight gain defined as the child's weight gain SD above 0.5 for 4 or 9 months or about 0.67 for 12 months.
Comparison: Logarithm base not specified.
Results: Lowest quartile used as reference.
Confounding: Child's age at measurement, age squared, age cubed, sex-age interactions, maternal age, pre-pregnancy BMI category, maternal
education, maternal race, private insurance, infertility treatment
Andersen et al.,
2010,1429893
Medium
Denmark,
1996-2002
Cohort
Pregnant women Maternal plasma BW (g, z- Regression
and their children
followed up at
birth, 5 months,
and 12 months
from DNBC
N at birth = 1114
(552 boys, 562
girls)
from first and
second trimester
33.4 (6.4, 106.7)
score), BMI
at 5 and 12
months,
height at 5
and 12
months
(cm), weight
at 5 and 12
months (g)
coefficient per
unit increase
inPFOS
BW
z-score: -0.002 (-0.006, 0.002)
g: "I ("3.1,1.0)
Boys
z-score: 0.003 (-0.003, 0.008)
g: 1.3 (-1.6, 4.2)
Girls
z-score: -0.006 (-0.011, -0.001),
p-value <0.05
g: -3.2 (-6.0, -0.3), p-value <0.05
BMI at 5 months
z-score: -0.001 (-0.006, 0.003)
g: -0.002 (-0.10,0.005)
Boys
z-score: -0.004 (-0.011, 0.002)
g: -0.007 (-0.018, 0.003)
Girls
z-score: 0.001 (-0.005, 0.007)
g: 0.002 (-0.008, 0.012)
D-25
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference. Location, Population, Ages, Exposure
„ ... Design Matrix, Sample Outcome Comparison Results
Confidence Years N t,. . T , a
Timing, Levels3
BMI at 12 months
z-score: -0.007 (-0.011, 0.002), p-
value <0.05
g: -0.011 (-0.019,-0.003)
Boys
z-score: -0.01 (-0.017, -0.003), p-
value <0.01
g: -0.017 (-0.028, -0.005), p-value
<0.01
Girls
z-score: -0.005 (-0.011, 0.002)
g: -0.007 (-0.018, 0.003)
Height at 5 months
z-score: 0.002 (-0.002, 0.006)
g: 0.006 (-0.004, 0.017)
Boys
z-score: 0.0004 (-0.006, 0.006)
g: 0.001 (0.014, 0.016)
Girls
z-score: 0.004 (-0.001, 0.010)
g: 0.011 (-0.004,0.026)
Height at 12 months
z-score: 0.003 (-0.001, 0.008)
g: 0.010 (-0.003, 0.023)
Boys
z-score: 0.003 (-0.004, 0.009)
g: 0.008 (-0.011,0.027)
Girls
z-score: 0.004 (-0.002, 0.010)
g: 0.011 (-0.007,0.030)
Weight at 5 months
z-score: -0.001 (-0.005, 0.003)
D-26
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
g: -0.8 (-4.2, 2.6)
Boys
z-score: -0.004 (-0.009, 0.001)
g: -3.7 (-8.7, 1.3)
Girls
z-score: 0.002 (-0.004, 0.007)
g: 1.3 (-3.3, 5.9)
Weight at 12 months
z-score:-0.005 (-0.009, 0.001), p-
value <0.05
g: -5.8 (-10.4, -1.2), p-value <0.05
Boys
z-score: -0.008 (-0.013, -0.002),
p-value <0.05
g: -9 (-15.9, -2.2), p-value <0.05
Girls
z-score: -0.003 (-0.009, 0.003)
g: -3.3 (-9.3,2.7)
DNBC = Danish National Birth Cohort
Results: "Models for weight at 5 or 12 months included BW, models for length at 5 or 12 months included birth length, and models for body
mass index at 5 or 12 months included birth body mass index."; adjusted models were used for all results.
Confounding: Maternal age, parity, pre-pregnancy BMI, smoking, socioeconomic status, GA at blood drawing, breastfeeding. Additional
confounding for BMI and 5 and 12 months: birth BMI. Additional confounding height at 5 and 12 months: birth height. Additional
confounding for weight at 5 and 12 months: BW.
Apelberg et al., United States
2007,1290833 2004-2005
Medium
Cross-sectional Pregnant women Cord blood at
and their newborns birth
from Baltimore 5(3.4-7.9)
THREE Study,
N = 293
BW (g), HC
(cm), BL
(cm),
ponderal
index
(g/cmA3*10
0), GA
(days)
Regression
BW
coefficient per Per ln-unit increase:-69 (-149, 10)
ln-unit
increase in
PFOS,
regression
coefficient per
IQR increase
in PFOS
Per IQR increase: -58 (-125, 9)
HC
Per ln-unit increase: -0.32 (-0.56,
-0.07), p-value <0.05
Per IQR increase: -0.27 (-0.48,
-0.06), p-value <0.05
BL
D-27
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
Per ln-unit increase: 0.13 (-0.26,
0.52)
Per IQR increase: 0.11 (-0.22,
0.44)
Ponderal index
Per ln-unit increase: -0.074
(-0.123, -0.025), p-value <0.05
Per IQR increase: -0.062 (-0.104,
-0.021), p-value <0.05
GA
Per ln-unit increase: 1.9 (-1.3, 5)
Per IQR increase: 1.0 (-0.7, 2.8)
Confounding: GA, maternal age, BMI, race, parity, smoking, baby sex, height, net weight gain, diabetes, hypertension. Additional
confounding for HC: delivery mode.
Arbuckle et al., Canada, 2008- Cohort Pregnant women Maternal blood
2020, 6356900 2011 (age range = 17-
Medium 42 years) and their 4.50 |ig/L
infants from (3.30-6.10
MIREC (ig/L)
N = 205
Anoclitoris
distance
(ACD, mm),
anofourchett
e distance
(AFD, mm),
anopenile
distance
(APD, mm),
anoscrotal
distance
(ASD, mm)
Regression ACD: 0.07 (-1.03, 1.18)
coefficient per Q2: -0.06 (-1.7, 1.58)
ln-unit
increase in
PFOS and by
quartiles
Q3: 0.17 (-1.5, 1.85)
Q4: -0.05 (-1.68, 1.57)
AFD: -0.29 (-1.62, 1.04)
Q2
Q3
Q4
-0.12 (-2.09, 1.85)
0.89 (-1.12, 2.9)
-0.33 (-2.31, 1.65)
APD: 0.13 (-1.13, 1.38)
Q2
Q3
Q4
-0.97 (-2.81, 0.87)
-1.28 (-3.22, 0.66)
0.22 (-1.68, 2.13)
ASD: 1.05 (-0.24, 2.35)
Q2: -0.87 (-2.78, 1.04)
Q3: 0.33 (-1.67,2.33)
Q4: 0.49 (-1.47, 2.46)
D-28
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
No statistically significant trends
MIREC = Maternal-Infant Research on Environmental Chemicals (MIREC)
Results: Lowest quartile used as reference.
Confounding: Household income, education, active smoking status, GA, weight-for-length Z-score, and recruitment site
Chang et al.,
United States Cohort Mother-infant Maternal serum, BW (g), SGA
BW: Regression BW
2022
2014-2018 pairs from the Early
coefficient per Per doubling: -7 (-48, 34)
Medium
Emory University pregnancy,
doubling in Q2: 78 (-98, 196)
9959688
African American 2.19 (1.45-3.24)
PFOS and by Q3: 20 (-98, 138)
Vaginal, Oral, and
quartiles Q4: -16 (-136, 105)
Gut Microbiome
p-trend = 0.48
in Pregnancy
SGA: Odds
Study
ratio per SGA
N = 370
doubling in Per doubling: 1.12(0.88, 1.42)
PFOS and by Q2: 0.92 (0.47, 1.78)
quartiles Q3: 1.32 (0.69, 2.53)
Q4: 1.09 (0.56,2.13)
p-trend = 0.65
Outcome: SGA defined as a BW below the 10th percentile for GA.
Confounding: maternal age, education, BMI, parity, tobacco use, marijuana use, and infant's sex (BW only)
Chen et al.,
2012,1332466
Medium
Taiwan, 2004-
2005
Cross-sectional
Mother-infant
pairs from TBPS
N = 429
Cord blood at
birth
GM (SD) = 5.94
(1.95)
BW (g), BL
(cm), GA
(weeks), HC
(cm), LBW,
ponderal
index
(g/cm3),
PTB, SGA
BW: Regression BW
coefficient per Per ln-unit increase: -110.2 (-176,
umt increase
inPFOS
BW, BL, GA,
HC, ponderal
index:
Regression
-44.5), p-value <0.01
Per unit increase: -11.3 (-17.4,
-5.2)
Q2: 54 (-44, 152)
Q3: 2 (-95, 102)
Q4: -92 (-190, 6)
coefficient per p-trend = 0.045
ln-unit
increase in
PFOS, or by
quartile
PTB, LBW,
SGA: OR per
ln-unit
BL
Per ln-unit increase: -0.17 (-0.42,
0.09)
Q2
Q3
Q4
0.08 (-0.39, 0.55)
0.14 (-0.33, 0.62)
-0.32 (-0.80, 0.15)
increase in p-trend = 0.234
D-29
-------
DRAFT FOR PUBLIC COMMENT
Reference. Location, Population, Ages, Exposure
kcicc c, Design P Matrix, Sample
Confidence Years N t,. . T , a
Timing, Levels3
D-30
MARCH 2023
Outcome Comparison Resultsb
PFOS, or by
quartile GA
Per ln-unit increase: -0.37 (-0.6,
-0.13), p-value <0.01
Q2: 0.13 (-0.30,0.57)
Q3: -0.65 (-1.07,-0.20)
Q4: -0.44 (-0.88, 0.00)
p-trend = 0.004
HC
Per ln-unit increase: -0.25 (-0.46,
-0.05), p-value <0.05
Q2:-0.16 (-0.53, 0.21)
Q3: -0.26 (-0.63,0.12)
Q4: -0.42 (-0.80, -0.05)
p-trend = 0.025
Ponderal index
Per ln-unit increase: -0.01 (-0.05,
0.02)
Q2: 0.03 (-0.03, 0.09)
Q3: -0.02 (-0.08,0.04)
Q4: -0.03 (-0.09, 0.04)
p-trend = 0.232
PTB
Per ln-unit increase: 2.45 (1.47,
4.08), p-value <0.001
Q2: 1.0 (0.2, 5.0)
Q3: 6.5 (2.0, 24)
Q4: 5.5 (1.5,20)
p-trend = 0.0006
LBW
Per ln-unit increase: 2.61 (0.85,
8.03)
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
Q2: 0.5 (0.02, 13)
Q3: 1.0(0.06, 18)
Q4: 4.5 (0.50, 57)
p-trend = 0.062
SGA
Per ln-unit increase: 2.27 (1.25,
4.15), p-value <0.01
Q2: 0.8 (0.2, 2.5)
Q3: 0.5 (0.1, 2.0)
Q4: 1.5 (0.6, 4.5)
p-trend = 0.422
TBPS = Taiwan Birth Panel Study
Outcome: PTB defined as GA <37 weeks. Low BW defined as a BW <2,500 g. SGA defined as a BW below the 10th percentile for GA.
Confounding: Maternal age, pre-pregnancy BMI, education level, ln-transformed cord blood cotinine levels, type of delivery, parity, infant
sex. Additional confounding for BW, BL, HC, ponderal index, low BW, PTB: GA.
Chen et al.,
2017,3981292
Medium
Taiwan, 2004- Cohort
2005
Mother-infant Cord blood BMI (z-score, Regression BMI
pairs from the kg/m2), coefficient per Birth:-0.11 (-0.25,0.02)
Taiwan Birth 5.7(IQR=5.0) height (z- ln-unit 0-6 mo: 0.002 (-0.17,0.18)
Panel Study score, cm), increase in 6-12 mo:-0.12 (-0.31,0.08)
(TBPS) weight (z- PFOS Girls 6-12 mo: -0.33 (-0.59,
N = 429 score, kg) -0.08); p-value < 0.05
12-24 mo: -0.09 (-0.29, 0.11)
Girls 12-24 mo: -0.25 (-0.45,
-0.05); p-value < 0.05
24-60 mo: -0.17 (-0.41, 0.06)
60-108 mo: -0.02 (-0.33, 0.28)
Girls 60-108 mo: 0.34 (0.007,
0.68); p-value < 0.05
Height
Birth: -0.16 (-0.31, -0.02), p-value
<0.05
0-6 mo:-0.04 (-0.23, 0.16)
6-12 mo:-0.02 (-0.23, 0.18)
12-24 mo: 0.04 (-0.17, 0.26)
D-31
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
24-60 mo: 0.09 (-0.12, 0.3)
Boys 24-60 mo: 0.18 (0.03, 0.33);
p-value < 0.05
60-108 mo: 0.06 (-0.19, 0.31)
Boys 60-80 mo: 0.19 (0.01, 0.38);
p-value < 0.05
Weight
Birth: -0.14 (-0.26, -0.01), p-value
<0.05
0-6 mo:-0.008 (-0.17, 0.16)
6-12 mo:-0.13 (-0.32,0.07)
Girls 6-12 mo: -0.25 (-0.47,
-0.04); p-value < 0.05
12-24 mo: -0.0.05 (-0.25, 0.16)
Girls 12-24 mo: -0.24 (-0.41,
-0.06); p-value < 0.01
24-60 mo: -0.07 (-0.3, 0.16)
60-108 mo: 0.02 (-0.27, 0.31)
BMI, height, and weight: no
statistically significant interactions
by sex at any age
Population: Infants were followed up at 4, 6, 13, 24, 60, 84, and 108 months
Confounding: Maternal age, pre-pregnancy BMI, education level, ln-cord blood cotinine, infant sex, PTB, postnatal ETS exposure,
breastfeeding
Chen et al.,
2021,7263985
Medium
China
Recruitment:
2013-2015
Cohort
Mother-child pairs Maternal plasma BW (g), BL Regression
BW
from the SBC,
Ages > 20,
N = 214
95 male children,
119 female
children
from the first (cm), HC coefficient per 2.7 (-84.3, 89.7)
trimester (cm) ln-unit
9.70 (6.75- increase in BL
15.35) PFOS -0.27 (-0.51,-0.02), p-value <0.05
Males
-0.14 (-0.55, 0.26)
Females
-0.4 (-0.74, -0.06), p-value <0.05
D-32
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
HC
-10.6 (-60.7, 39.6)
SBC = Shanghai Birth Cohort
Confounding: Maternal age, BMI, educational level, occupation, income, fetal sex, parity, GA, smoking and alcohol.
Darrowetal., United States, Cohort Pregnant women
2014,2850274 Recruitment: with known PFAS
Medium 2005-2006, exposure (ages
Follow-up: >20 years) from
2008-2011 C8HP
N = 1,438
First pregnancy
N = 1,129
Serum collected
before
pregnancy
15.1 (10.4-21.2)
Primary
analysis
miscarriage,
first
pregnancy
miscarriage
OR per ln-unit
increase in
PFOS, OR by
quintile
Primary analysis: 1.21 (0.94, 1.55)
Q2: 1.34(0.84,2.16)
Q3: 1.40 (0.88,2.25)
Q4: 1.59 (0.99,2.54)
Q5: 1.41 (0.88,2.26)
First pregnancy: 1.34 (1.02, 1.76)
Q2: 1.68 (0.99,2.84)
Q3: 1.93 (1.13,3.31)
Q4: 1.94(1.14,3.31)
Q5: 1.80 (1.06,3.06)
C8HP = C8 Health Project
Outcome: Primary analysis includes more than one pregnancy for some women (304 miscarriages). First pregnancy is restricted to the first
pregnancy conceived per woman after serum measurement (213 miscarriages)
Confounding: Maternal age, educational level, smoking status, BMI, self-reported diabetes, time between conception, serum measurement.
de Cock et al.,
The Netherlands Cohort Mother-child pairs
Cord blood
BMI (kg/m2),
Regression
BMI, HC, height, and weight: no
2014,2713590
Recruitment: N = 89
HC (cm),
coefficient for
statistically significant associations
Medium
2011-2013
1,600.0 ng/L
height (cm),
quartiles of
Follow-up at 1,
(Range = 570-
weight (kg)
PFOS
2, 4, 6, 9, and 11
3,200 ng/L)
months after
birth
Confounding: BW, GA, maternal height
deCocketal., The Cross-sectional Mother-infant Cord blood BW(g) Regression T2:254.8 (-99.47,609.09), p-value
2016,3045435 Netherlands, pairs coefficient by =0.153
Medium 2011-2013 N = 64 1,600 ng/L tertiles T3:438.4 (55.09,821.68), p-value
(Range = 570- = 0.026
3,200 ng/L) Females
T2: 143.3 (-361.63, 648.32), p-
value - 0.566
D-33
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
T3: 301.1 (-124.87, 727.05), p-
value = 0.159
Males
T2: 486.9 (-1.21, 975.03), p-value
= 0.051
T3: 724.4 (193.83, 1,254.97), p-
value = 0.009
Results: Lowest tertile used as reference.
Confounding: GA, maternal BMI, maternal height, maternal age at birth, and parity, paternal BMI, paternal height, education, fish intake
Fei et al., 2008,
1290822
Medium
Denmark
Recruitment:
1996-2002,
Assessment 6-
18 months later
Cohort Pregnant women
and their children
at 6 and 18 months
from the DNBC
Total
N = 1,400
18-month olds
N = 1,380
Maternal plasma Gross motor
during the first milestone,
trimester language
33.3 (26.0-43.2) milestone,
Apgar score
<10
Gross motor
milestone:
Hazard ratio
by quartile
Language
milestone: OR
by quartile
Apgar score:
OR for Q4 vs.
Qi
Gross motor milestone
Q2: 0.93 (0.79, 1.08)
Q3: 0.85 (0.72,0.99)
Q4: 0.86 (0.73, 1.01)
p-trend = 0.041
Language milestone
Q2: 1.39 (0.46,4.25)
Q3: 1.58(0.51,4.91)
Q4: 2.93 (1.00, 8.56)
p-trend = 0.039
Apgar score
Q4: 1.2(0.67,2.14)
DNBC = Danish National Birth Cohort
Outcome: Gross motor milestone defined as sitting without support. Language milestone defined as child not using word-like sounds to tell
what he/she wants.
Results: Lowest quartile used as reference group.
Confounding: Maternal age, maternal occupational and educational status, parity, pre-pregnancy BMI, smoking and alcohol consumption
during pregnancy, gestational weeks at blood drawing, child's sex. Additional confounding for gross motor milestone and language milestone:
parity, out-of-home childcare, home density (rooms/person). Additional confounding for language milestone: child's age at interview.
Fei et al., 2008,
2349574
Medium
Denmark
1996-2002
Cohort
Pregnant women Maternal plasma Placental
and their newborns between 4-14 weight (g),
from the DNBC weeks gestation HC (cm),
33.4 (26.1-43.3) BL (cm),
Placental weight abdominal
N = 1,337
Regression Placental weight
coefficient per Per unit increase: -0.24 (-0.85,
unit increase 0.37)
inPFOS
Q2
Q3
Q4
-6.6 (-28.8, 15.5)
-13.7 (-36.4, 8.9)
-10.8 (-33.4, 11.8)
D-34
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Birth length
N = 1,376
HC
N = 1,347
Abdominal
circumference
N = 1,325
circumferen Mean difference
ce (cm)
by quartile HC
Per unit increase: 0.0 (-0.006,
0.007)
Q2: 0.14 (-0.09, 0.36)
Q3: 0.09 (-0.14,0.32)
Q4: 0.03 (-0.20, 0.27)
BL
Per unit increase: -0.002 (-0.011,
0.006)
Q2: 0.21 (-0.08,0.51)
Q3: 0.06 (-0.24,0.36)
Q4: 0.05 (-0.25, 0.35)
Abdominal circumference
Per unit increase: -0.003 (-0.012,
0.005)
Q2: 0.24 (-0.07, 0.55)
Q3: 0.10 (-0.21, 0.42)
Q4: 0.00 (-0.32, 0.32)
DNBC = Danish National Birth Cohort
Results: Lowest quartile used as reference group
Confounding: GA, quadratic GA, infant sex, maternal age, socio-occupational status, parity, cigarette smoking, pre-pregnancy BMI,
gestational week at blood drawing.
Govarts et al., Belgium, the Cohort Mother-child pairs Cord blood SGA OR per IQR
2018,4567442 Netherlands, from FLEHSI and increase in
Medium Norway, and II, HUMIS, LINC, 1,984 ng/L PFOS
Slovakia
2002-2012
Mother-child pairs Cord blood
from FLEHS I and
II, HUMIS, LINC, 1,984 ng/L
and PCB Cohort (1,200-3,008
N = 657 ng/L)
0.823 (0.742, 0.913)
FLEHS = Flemish Environmental and Health Study; HUMIS = Human Milk Study; LINC = Linking EDCs in Maternal Nutrition to Child
Health
Outcome: SGA defined as newborns weighing below the 10th percentile for the norms defined by GA, country, and infant's sex.
Confounding: Maternal education, maternal age at delivery, maternal height, maternal pre-pregnancy BMI, smoking during pregnancy,
parity, child's sex
D-35
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Gyllenhammar
et al., 2018,
4238300
Medium
Sweden, 1996-
2011 and
follow-up at 5
years of age
Cohort and
cross-sectional
Mother-infant
pairs of singleton
births from
POPUP study
N = 377
Maternal serum
Later pregnancy
13 (7.4-19)
BL (SD
scores), BW
(SD scores),
gestational
length
(days), HC
(SD scores),
length (SD
scores),
weight (SD
scores)
Regression BL: 0.1377(-0.0971, 0.3725)
coefficient per BW: 0.0167 (-0.1878, 0.2225)
IQR increase Gestational length: -2.0342 (-
in maternal 4.1139, 0.0455)
PFOS HC: 0.0703 (-0.1602, 0.2974)
HC, length, and weight: no
statistically significant associations
by sex
POPUP = Persistent Organic Pollutants in Uppsala Primiparas
Confounding: Sampling year, maternal age, pre pregnancy BMI, maternal weight gain during pregnancy, maternal weight loss after delivery,
years of education, smoking during pregnancy, total fish consumption
Hamm et al.,
2010,1290814
Medium
Canada
Recruitment:
2005-2006
Follow-up at
delivery: 2006-
2007
Cohort
Pregnant women Maternal serum BW (g, z-
(>18 years of age)
and their singleton
children delivered
at or after 22
weeks gestation
N = 252
collected at 15-
16 weeks
gestation
GM (SD) = 7.4
(2.0)
score),
SGA, PTB,
length of
gestation
(weeks)
BW: Regression
coefficient per
ln-unit or per
unit increase
in PFOS and
by tertiles
SGA, PTB:
Relative risk
by tertile
BW (g per ln-unit): -31.3 (-43.3,
105.9), p-value = 0.03
T2: -13.51 (-136.57, 109.55)
T3: 71.25 (-54.97, 197.48)
BW(g per unit): 1.5 (-7.6, 10.6)
BW (z-score per ln-unit): 0.06
(-0.11,0.23)
T2: -0.006 (-0.29, 0.27)
T3: 0.16 (-0.13, 0.44)
Length of
gestation:
Regression
coefficient per
ln-unit
increase in
PFOS and by
tertile
SGA:
T2: 0.99 (0.27, 3.61)
T3: 0.26 (0.10,0.70)
PTB:
T2: 1.06 (0.33, 3.45)
T3: 1.11 (0.36,3.38)
Length of gestation:
Per ln-unit: 0.21 (-0.12, 0.53)
T2: 0.13 (-0.42,0.67)
D-36
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
T3: 0.046 (-0.51,0.60)
Outcome: SGA defined as BW <10th percentile for GA and infant gender; PTB defined as delivery at 22-36 weeks
Results: Lowest tertile used as reference
Confounding: Maternal age, maternal race, gravida, maternal weight and height, smoking. Additional confounding for PTB and BW (g):
infant gender. Additional confounding for BW (g): GA at birth.
Hjermitslev et Greenland, Cohort Pregnant women
al., 2020, Recruitment: (>18 years of age)
5880849 2010-2011, and their children
Medium 2013-2015 from ACCEPT
N = 256
Maternal serum
Early
pregnancy, later
pregnancy
8.99 (Range =
1.50-61.3)
BW (g), GA
at birth
(weeks), HC
(cm),
preterm
birth
Regression
coefficient per
ln-unit
increase in
PFOS
Preterm birth:
OR per ln-unit
increase in
PFOS
BW: -5.47 (-12.6, 1.67)
Females: -5.65 (-14.9, 3.55)
Males: -1.9 (-14, 10.2)
GA: 0.001 (-0.02,0.03)
Females: 0.002 (-0.03, 0.03)
Males: -0.006 (-0.05, 0.04)
HC: -0.01 (-0.04, 0.01)
Females: -0.02 (-0.05, 0.01)
Males: 0.005 (-0.04, 0.05)
Preterm birth: 0.95 (0.87, 1.05), p-
value = 0.321
No statistically significant
associations
ACCEPT = Adapting to Climate Change, Environmental Pollution and Dietary Transition
Confounding: Maternal age, plasma cotinine, alcohol consumption during pregnancy, pre-pregnancy BMI, GA at birth
Jensen et al., Denmark, Cohort Pregnant women Maternal serum Ponderal
2020,6833719 2010-2012 and and infants at 3 index
Medium follow-up at 18 and 18 months of 8.04 (3.82- standard
months of age age from Odense 15.45) deviation
Child Cohort score (SDS)
N = 593
Regression -0.004 (-0.03, 0.02)
coefficient per Birth: 0.03 (0.01, 0.05), p-value =
unit increase 0.02
in PFOS 3 months: -0.005 (-0.03, 0.016)
18 months: -0.003 (-0.03, 0.02)
3 and 18 months: no statistically
significant associations
Outcome: Ponderal index (kg/m3) was calculated as weight (kg) divided by the length cubed (m3)
Confounding: Maternal age, parity, pre-pregnancy BMI, pre-pregnancy BMI2, education, smoking, sex, visit, adiposity marker at birth
D-37
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
Kashino et al., Japan, 2003-
2020,6311632 2009
Medium
Cohort Mother-infant
pairs from the
Hokkaido Study
on Environment
and Children's
Health
N = 1,949
Plasma Birth HC
Later pregnancy (cm), BL
3.4(2.6-1.7) (cm), BW
(g)
Regression HC: -0.067 (-0.418, 0.283)
coefficient per Females: 0.001 (-0.531, 0.532)
Males: -0.142 (-0.605, 0.321)
loglO-unit
increase in
PFOS
Length: 0.092 (-0.311, 0.494)
Females: 0.25 (-0.321, 0.821)
Males: -0.019 (-0.589, 0.551)
BW: -35 (-109, 39)
Females: -19.9 (-128, 88.2)
Males: -46.3 (-148.4, 55.8)
HC, BL, and BW: no statistically
significant associations overall or
stratified by sex
Confounding: GA, maternal age, pre-pregnancy BMI, parity, infant sex, maternal educational level, plasma cotinine concentration during
pregnancy
Kishi et al.,
Japan, 2002- Cross-section Pregnant women Maternal blood BW (g)
Regression Females
2015,2850268
2005 (aged 28-34 years)
coefficient by Q2: -70.1 (-242.5, 102.2)
Medium
and infants from Mean = 5.89
quartiles Q3: -39.1 (-216.1, 137.8)
the Hokkaido (SD = 0.20)
Q4: -186.6 (-363.4, -9.8), p-value
Study
<0.05
Females, N = 165
p-trend = 0.031
Males, N= 141
Males
Q2: -56.7 (-255.9, 142.4)
Q3: 95.9 (-116.5, 308.4)
Q4: 30.5 (-169.7, 230.8)
p-trend = 0.187
Results: Lowest quartile used as reference.
Confounding: GA, maternal age, pre-pregnancy BMI, smoking and drinking during pregnancy, parity, annual household income, blood
sampling period
Kobayashi et
Japan, 2002- Cross-sectional Pregnant women Maternal serum BL (cm), BW
Regression Length: 0.32 (-0.19, 0.82)
al., 2017,
2005 at 22-35 weeks (g)
coefficient per
3981430
gestation and 5.3(3.9-7.2)
ln-unit BW: -56 (-162.8, 50.8)
Medium
infants from
D-38
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Kobayashi et
al., 2022,
10176408
Medium
Length and BW: no statistically
significant associations
Hokkaido Study increase in
on Environment PFOS
and Children's
Health
N = 177
Confounding: Maternal age, pre-pregnancy BMI, parity, maternal education, maternal smoking during pregnancy, GA, infant sex, maternal
blood sampling period
Japan Cohort Mother-child pairs
Recruitment: from the Sapporo
2002-2005 Cohort of the
Hokkaido Birth
Cohort
N = 372 (198
female children,
174 male children)
Regression
coefficient
loglO-unit
increase in
PFOS
Maternal blood BW (g), BL
in the third (cm)
trimester
5.2 (3.7-7.2)
Females
5.2 (3.4-7.3)
Males
5.3 (3.9-7.0)
Confounding: Maternal age (continuous), pre-pregnancy BMI (continuous), maternal smoking in the
alcohol consumption during pregnancy (yes/no), parity (primiparous, multiparous), educational level,
section (yes/no), maternal blood sampling period, GA (continuous), infant sex.
BW
per -182.3 (-336.5, -28.2), p-value =
0.021
Females: -292.1 (-504.3, -79.8),
p-value = 0.007
Males: 17.7 (-207, 242.5), p-value
= 0.876
BL
-0.552 (-1.433, 0.328), p-value =
0.218
Females: -1.384 (-2.472, -0.297),
p-value = 0.013
Males: 0.635 (-0.832, 2.102), p-
value = 0.394
third trimester (yes/no), maternal
annual household income, cesarean
Kwon et al., Korea, 2006- Cohort Pregnant women
2016,3858531 2010 and infants from
Medium EBGRC
N = 268
Cord blood
0.64 (0.29-1.09)
BW (g) Regression -49.41 (-95.57, -3.25), p-value :
coefficient per 0.04
log-unit
increase in
PFOS
EBGRC = Ewha Birth & Growth Retrospective Cohort
Comparison: Logarithm base not specified.
Confounding: Mother's age, pre-pregnancy BMI, past history of alcohol consumption and child's GA, gender, parity
D-39
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Lenters et al.,
2016,5617416
Medium
Greenland,
Poland, and
Ukraine
2002-2004
Cohort
Pregnant women
and singleton
infants from
INUENDO
N = 1,250
INUENDO = Biopersistent Organochlorines in Diet and Human Fertility
Confounding: Study population, maternal age, pre-pregnancy BMI, parity
Maternal serum
Later pregnancy
GM= 9.357 (2-
SD ln-PFOS =
1.600)
BWatterm Regression -114.36 (-206.81,-21.91), p-value
(g) coefficient per =0.015
2-SD increase
in ln-PFOS
Liewetal., Denmark, Nested
2016,6387285 1996-2002 case-
Medium control
Females from Plasma
the Danish Control: 23.35
National (18.1,30.30)
Birth Cohort, Cases: 24.55
N = 438 (19.5,32.25)
Miscarriage
OR per doubling of
PFOS or by
quartiles
1.2 (0.9,1.8)
Q2: 1.1 (0.6, 1.9)
Q3: 1.3 (0.8,2.4)
Q4: 1.4 (0.8,2.4)
Results: Lowest quartile used as the reference group.
Confounding: Maternal age, parental socio-occupational status, maternal smoking in the first trimester, maternal alcohol intake in the first
trimester, gestational week of blood sampling, parity
Louis et al.,
2016,3858527
Medium
United States, Cohort Females from Plasma
Pregnancy loss
HR per log-unit 0.81 (0.65, 1.00)
2005-2009
the LIFE
Study,
Ages 18-40,
N = 344
Pregnant: 12.2
(8.3, 17.8)
Infertile: 12.1
(7.1, 17.1)
increase in PFOS or
by tertiles
T2: 0.81 (0.50,
T3: 0.60 (0.35,
1.33)
1.03)
Comparison: Logarithm base not specified.
Confounding: Age, BMI, prior pregnancy loss conditional on previous pregnancy, any alcohol consumption during pregnancy, any cigarette
smoking during pregnancy
Maisonet et al.,
2012,1332465
Medium
Great Britain
Recruitment:
1991-1992,
followed-up
until 20 months
of age
Cohort
Pregnant women Maternal serum BW (g), BL Regression
BW
and their singleton
girls assessed at
birth, 9, and 20
months from
ALSPAC
BW
N = 422
BL
N = 356
during
pregnancy
(median 15
weeks)
19.6 (Range :
3.8-112.0)
(cm), GA coefficient by T2:-111.71 (-208.24,-15.17)
(weeks),
ponderal
index (g/cm3),
weight at 20
months (g)
tertile
T3: -140.01 (-238.14, -41.89)
p-trend = 0.0053
BL
T2: -0.72 (-1.19,-0.25)
T3: -0.63 (-1.11,-0.15)
p-trend = 0.0103
GA
D-40
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
GA
N = 444
Ponderal index
N = 360
Weight at 20
months
N = 320
(106 upper tertile
of BW,
107 middle tertile
of BW,
107 lower tertile
of BW)
T2: -0.02 (-0.39, 0.35)
T3: -0.15 (-0.53,0.23)
p-trend = 0.4352
Ponderal index
T2: 0.00 (-0.07, 0.06)
T3: 0.05 (-0.01,0.12)
p-trend = 0.1120
Weight at 20 months
T2: 310.64 (27.19, 594.08)
T3: 579.82 (301.4, 858.25)
p-trend < 0.0001
Upper tertile of BW
T2: 333.57 (-301.28, 968.42)
T3: 596.22 (-52.98, 1245.42)
p-trend = 0.0714
Middle tertile of BW
T2: -262.83 (-884.25, 358.60)
T3: 165.43 (-439.52, 770.37)
p-trend = 0.5886
Lower tertile of BW
T2: 602.64 (-150.79, 1356.07)
T3: 932.71 (186.90, 1678.52)
p-trend = 0.0148
ALSPAC = Avon Longitudinal Study of Parents and Children
Results: Lowest tertile used as reference
Confounding: BW: maternal smoking during pregnancy, maternal pre-pregnancy BMI, previous live births, and GA; BL additionally adjusted
for maternal education. GA: GA when maternal serum sample was obtained. Ponderal index: maternal pre-pregnancy BMI, previous live
births, and GA when maternal serum sample was obtained. Weight at 20 months (all tertiles): height at 20 months, BW, maternal education,
maternal age at delivery, and previous live birth; intratertile analyses adjusted for maternal education, maternal age at delivery, previous live
birth, and BW.
Manzano- Spain, 2003-
Salgado et al., 2008
2017,4238509
Medium
Cohort Mother (aged >16
years)-child pairs
from INMA
Maternal blood
GM = 5.80
(4.52-7.84)
Weight gain
z-score,rapid
growth
Regression
coefficient or
RR per log2-
Weight gain z-score
-0.02 (-0.11,0.07)
Girls: -0.09 (-0.21,0.04)
Boys:-0.05 (-0.08,0.19)
D-41
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
unit increase in
PFOS
p-value for sex interaction = 0.54
assessed at birth
and 6 months
N = 1,154 (568
girls, 586 boys)
INMA = INfancia y Medio Ambiente [Environment and Childhood Project]
Outcome: Rapid growth defined as a z-score >0.67 standard deviation for weight gain from birth until 6 months.
Confounding: Maternal characteristics (i.e., region of residence, country of birth, previous breastfeeding, age, pre-pregnancy BMI), age and
sex of child
Rapid growth
0.92 (0.80, 1.06)
Mengetal., Denmark, Cohort Pregnant women
2018,4829851 1996-2002 and their infants
Medium from DNBC
N = 3,522 (1,533
girls, 1,969 boys)
Maternal serum BW (g), GA
Early (days), low
pregnancy, Later BW, PTB
pregnancy
30.1 (22.9-39.0)
Regression BW
coefficient -45.2 (-76.8., -13.6)
(BW, GA) or Q2: 24.7 (-24.8, 74.1)
OR (LBW, Q3:-50.1 (-101.1, 0.9)
PTB) per Q4: -48.2 (-99, 2.5)
doubling of Females: -65.3 (-111.7, -18.9)
PFOS and by Males: -24.3 (-67.1, 18.6)
quartiles p-value for sex interaction =0.31
GA
-1.1 (-1.7,-0.4)
Q2: -1.1 (-2.1,-0.1)
Q3:-2 (-3.1,-1)
Q4: -1.5 (-2.6, -0.5)
Females: -1 (-2, 0)
Males: -1.1 (-2.0, -0.3)
p-value for sex interaction = 0.72
LBW
1.3 (0.9, 2.0)
Q2: 1.4 (0.7, 2.8)
Q3: 1.8(0.9,3.6)
Q4: 1.2 (0.6, 2.4)
PTB
1.5(1.1,2.2)
Q2: 2.0 (1.1, 3.6)
Q3: 3.3 (1.8, 5.8)
D-42
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Q4: 1.9(1.0,3.5)
DNBC = Danish National Birth Cohort
Results: Lowest quartile used as reference.
Confounding: Infant sex, infant birth year, gestational week of blood draw, maternal age, parity, socio-occupational status, pre-pregnancy
body mass index, smoking during pregnancy, alcohol intake during pregnancy, study sample
Ou et al., 2021,
7493134
Medium
Pregnant women
and their children
with (cases) and
without (controls)
CHD
N = 316
Septal defects,
conotruncal
defects, total
CHD
China, 2014- Nested case- Pregnant women Maternal blood
2018 control and their children and cord blood
at delivery
Maternal blood
Cases: 5.752
(3.655-8.683)
Controls: 5.742
(4.156-6.850)
Cord blood:
Cases: 1.928
(0.823-3.295)
Controls: 2.237
(1.505-3.072)
CHD = Congenital heart defects
Outcome: Total congenital heart defects included septal defects and conotruncal defects,
with a large number of cases.
Confounding: Maternal age, parity, infant sex.
OR for >75*
percentile vs.
<75lh percentile
PFOS
Maternal PFOS
Septal defects: 1.92 (0.80, 4.60)
Conotruncal defects: 1.65 (0.59,
4.63)
Total CHD: 1.61 (0.91, 2.84), p-
value <0.10
Cord PFOS
Septal defects: 1.15 (0.38, 3.54)
Conotruncal defects: 0.63 (0.16,
2.57)
Total CHD: 1.03 (0.46,2.3)
as well as individual congenital heart defect subtypes
Robledo et al., United States, Cohort Couples and their
2015,2851197 2005-2009 children from the
Medium LIFE study
N = 234
Serum
Early pregnancy
Girls: GM =
BW (g), HC
(cm), BL
(cm),
12.44 (95% CI = ponderal
11.50,13.44) index (g/cm3)
Boys: GM =
21.6 (95% CI =
19.97, 23.39)
Regression Maternal PFOS
coefficient for Girls:
mean change BW: 14.16 (-81.83, 110.15)
per 1-SD HC: -0.04 (-0.46, 0.38)
increase in BL: 0.30 (-0.26, 0.86)
ln(maternal Ponderal Index: -0.03 (-0.10, 0.03)
PFOS) and in Boys:
ln(paternal BW: 37.51 (-73.45, 148.46)
PFOS) HC: 0.07 (-0.45, 0.60)
BL: 0.22 (-0.43, 0.86)
Ponderal Index: 0.00 (-0.07, 0.08)
D-43
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Paternal PFOS
Girls:
BW: 38.58 (-59.29, 136.45)
HC: 0.29 (-0.14,0.71)
BL: -0.05 (-0.62, 0.52)
Ponderal Index: 0.05 (-0.02, 0.11)
Boys:
BW: 36.85 (-73.14, 146.84)
HC: 0.16 (-0.37,0.68)
BL: -0.20 (-0.84, 0.43)
Ponderal Index: 0.06 (-0.02, 0.13)
LIFE = Longitudinal Investigation of Fertility and the Environment
Confounding: Maternal and paternal serum lipids, serum cotinine, BMI, maternal age, difference in paternal age, infant gender, individual
and partner sum of remaining chemical concentrations in each chemical's respective class
Stein etal., United States Cohort Pregnant women
2009, 1290816 2005-2006 and their infants
Medium from the C8HP
Birth defects
N = 3,996
PTB
N = 4,512
Low BW
N = 4,561
Maternal serum
within 5 years
after pregnancy
13.6 (9.0-17.7)
Birth defects,
PTB, LBW
OR per IQR
increase in
PFOS
PTB, LBW:
OR by
percentile
Birth defects
Per IQR increase: 1.1 (0.9, 1.3)
PTB
Per IQR increase: 1.1 (1.0, 1.3)
50lh-75lh percentile: 1.1 (0.9, 1.3)
75th-90th percentile: 1.1 (0.9, 1.3)
>90^ percentile: 1.4 (1.1, 1.7)
LBW
Per IQR increase: 1.3 (1.1, 1.6)
50^-75thpercentile: 1.3 (0.9, 1.8)
75th-90th percentile: 1.6(1.1,2.3)
>90'percentile: 1.8 (1.2, 2.8)
C8HP = C8 Health Project
Population: Includes "women who lived in the same contaminated water district from the approximate start of the pregnancy through the time
of enrollment... to ensure that the PFOA level measured at C8 Health Project enrollment would reflect the level at the time of pregnancy."
Outcome: PTB defined as birth at <37 weeks; low BW defined as <5.5 pounds at birth.
Results: <50lh percentile used as reference group.
Confounding: Maternal age, parity, educational level at interview, smoking status at interview, PFOA in the analysis of PFOS.
D-44
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Tianetal., China Cohort Pregnant women Maternal serum Weight gain Regression
2019,5390052 2012-2014 and their sons at z-score (0-6 coefficient per
Medium birth, 6 months, 10.70(7.61- months or 6- ln-unit increase
and 12 months 15.71) 12 months), inPFOSorby
from the S-MBCS AGDap, quartiles
Birth N = 439 AGDas
6-month N = 322
12-month N= 301
Weight gain z-score
0-6 mo: -0.06
6-12 mo: 0.12; p-value < 0.05
AGDap
Quartile analysis showed no other
statistically significant associations
Vesterholm et
al., 2014,
2850926
Medium
Weight gain z-
score: Pearson
correlation
coefficient
S-MBCS = Shanghai-Minhang Birth Cohort Study; AGDap = anopenile distance; AGDas = anoscrotal distance
Results: Lowest quartile used as reference.
Confounding: Maternal age at delivery, GA, maternal education, parity, pre-pregnancy BMI, infant age at physical examination, infant body
size
Toft et al., 2016,
3102984
Medium
Denmark
1980-1996
Case-control
Pregnant women
and their sons
from the DMBR
N = 270
cryptorchidism
cases, 75
hypospadias cases,
and 300 controls
Amniotic fluid
Second
exposure tertile:
0.8-1.4
Cryptorchidis
m,
hypospadias
OR per ln-unit
increase in
PFOS or by
tertiles
Cryptorchidism
0.99 (0.75, 1.30)
T2: 1.08 (0.71, 1.63)
T3: 1.01 (0.66, 1.53)
Hypospadias
0.87 (0.57, 1.34)
T2: 0.97 (0.51, 1.87)
T3: 0.69 (0.35, 1.38)
DMBR = Danish Medical Birth Registry
Outcome: Cryptorchidism defined as both a diagnosis of undescended testis and a corrective surgical procedure recorded in the Danish
National Patient Registry (DNPR). Hypospadias defined as diagnosis in the DNPR.
Results: Lowest tertile used as reference
Confounding: GA of amniocentesis, maternal age, smoking (cotinine groups), and case or control status
Denmark and Nested case- Boys with (cases) Cord blood
Finland control or without
Recruitment (controls)
1997-2002, cryptorchidism
follow-up 3 N = 215
months after
birth
Outcome: Cryptorchidism defined as by Scorer (1964).
Cryptorchidis
m
9.1 (5th-95th
percentile: 4.8-
16.4)
OR per ln-unit
increase in
PFOS or by
tertiles
Continuous: 0.83 (0.44, 1.58)
T2: 0.70 (0.34, 1.46)
T3: 0.83 (0.39, 1.78)
p-trend = 0.64
D-45
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
Exposure Level: Denmark cases: 2.4 (5th - 95th percentile: 1.4-4.4); controls: 2.70 (5th - 95th percentile: 1.4, 4.0); Finland cases: 1.9 (5th - 95'1
percentile: 1.0-3.9); controls: 2.3 (5th - 95th percentile: 1.2—4.8)
Results: Lowest tertile used as reference.
Confounding: BW, GA, parity
Wangetal., China
2019,5080598 2013
Medium
Cross-sectional Pregnant women Cord blood
and their children Later pregnancy
at birth 0.65 (0.40-1.19)
N = 340 (171 girls,
169 boys)
BL (cm), BW
(g), BW z-
score, HC
(cm),
ponderal
index (g/cm3)
Regression BL, BW, HC, ponderal index: no
coefficient per statistically significant associations
loglO-unit or interactions by sex
increase in BL
PFOS -0.01 (-0.40, 0.39); p-value =
0.982
Girls: -0.01 (-0.60, 0.58); p-value
= 0.968
Boys: -0.17 (-0.71, 0.37); p-value
= 0.535
p-value for interaction by sex =
0.557
BW
54.5 (-149.07, 40.06); p-value =
0.259
Girls: -57.3 (-201.38, 86.78); p-
value = 0.436
Boys: -61.6 (-184.61, 61.42); p-
value = 0.326
p-value for interaction by sex =
0.844
BW z-score
-0.15 (-0.41, 0.11); p-value =
0.258
HC
0.02 (-0.26, 0.29); p-value = 0.915
Girls: -0.01 (-0.42, 0.39); p-value
= 0.947
D-46
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
Boys: -0.04 (-0.41, 0.32); p-value
= 0.821
p-value for interaction by sex =
0.709
Ponderal index
-0.04 (-0.09, 0.001); p-value =
0.054
Girls: -0.04 (-0.11, 0.02); p-value
= 0.198
Boys: -0.02 (-0.08, 0.03); p-value
= 0.427
p-value for interaction by sex =
0.637
Confounding: Pregnant age, family income, maternal education level, maternal career, husband's smoking, energy daily intake, daily physical
activity, GA, parity, pre-pregnant maternal body mass index, gestational diabetes mellitus, infant sex, delivery mode, gestational weight gain
Woods et al.,
United States, Cohort
Pregnant women Maternal serum BW (g)
Regression -8.7 (-52.8, 34.9)
2017,4183148
Recruitment:
and their children Later pregnancy
coefficient per
Medium
2003-2006;
at birth from the 14.4 (10-17.0)
loglO-unit
outcome
HOME study
increase
assessed at birth
N = 272
maternal PFOS
HOME = Health Outcomes and Measures of Environment
Confounding: Maternal race, age at delivery, infant sex, maternal education, tobacco exposure, household annual income, employment,
maternal insurance status, marital status, prenatal vitamin use, maternal BMI, GA
Yangetal., China Nested case-
2022, 10176806 2018-2019 control
Medium
Infants from the
KBCS,
N = 768
(403 males, 365
females)
PTBs
N = 384
(205 males,
females)
Term births
N = 384
179
Cord blood at PTB, GA PTBs: OR per PTB
birth (weeks) IQR increase in 1.44 (1.18, 1.79), p-value <0.01
Term births PFOS PFAS-residuals model: 1.71 (1.26,
0.266(0.144- 2.4), p-value <0.001
0.444) GA: Males
PTBs Regression 1.45 (1.10,2.03)
0.213(0.112- coefficient per Females
0.483) IQR increase in 1.40 (1.10,1.93)
PFOS p-value for interaction by infant's
sex = 0.99
GA
D-47
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
PTBs, total
-1.26 (-2.46, -0.05), p-value =
0.04
PFAS-residuals model: -2.01
(-3.42,-0.61), p-value = 0.01
PTBs, males
-0.41 (-2.2, 1.37)
PTBs, females
-1.06 (-2.87,0.74)
p-value for interaction by infant's
sex = 0.14
Term births
-0.16 (-1.81, 1.48), p-value = 0.85
KBCS = Kashgar Birth Cohort Study
Confounding: Maternal age, maternal ethnicity, maternal BMI, household income, maternal education level, maternal tobacco smoking
during pregnancy, maternal alcohol drinking during pregnancy, parity, living near a factory, periconceptional folic acid intake, gestational
diabetes, gestational hypertension. Additional confounding for analyses with both sexes: infant's sex. Additional confounding for PFAS-
residuals model: residuals regressed from PFDoA with PFOA, PFDA, PFUdA, PFNA, and PFTrDA.
Callanet al.,
2016,3858524
Low
Australia 2008-
2011
Cross-sectional
Mother-infant
pairs enrolled in
AMETS,
Ages 19-44,
N = 98
Maternal blood
1.99 (0.45-8.1)
BW (g), BL
(cm),
Proportion of
optimal BW
(POBW), HC
(cm),
ponderal
index (g/cm3
x 100),
proportion of
optimal birth
length
(POBL),
proportion of
optimal HC
(POHC)
Regression
coefficient per
ln-unit increase
inPFOS
BW
-69 (-231,94)
BL
-0.22 (-1, 0.57)
POBW
0.48 (-4.2, 5.2)
HC
-0.39 (-0.98, 0.2)
Ponderal Index
-0.03 (-0.14,0.08)
POBL
-0.12 (-1.4, 1.7)
D-48
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
POHC
-0.6 (-2.3, 1.1)
AMETS = Australian Maternal Exposure to Toxic Substances
Confounding: For BW, BL, HC, and ponderal index: GA, maternal height, pre-pregnancy BMI, weight gain during pregnancy, sex of infant.
For POHC: Weight gain during pregnancy, annual household income. For POBL: Weight gain during pregnancy, maternal age, annual
household income.
Cao et al., 2018, China, Cohort Infants from
5080197 2013-2015 Zhoukou City,
Low China,
N = 337 (183
males, 154
females)
Postnatal weight,
postnatal length,
postnatal HC
N = 282 (157
males, 125
females)
Cord blood
1.01 (0.60-1
.76)
BW (g), BL
(cm),
ponderal
index
(g/cm3),
postnatal
weight (g),
postnatal
length (cm),
postnatal HC,
birth defects
BW, BL, HC
and ponderal
index:
Regression
coefficient by
tertiles
BW
T2: 103.5 (-17.8, 224.8)
T3: -17.6 (-141.2, 106)
Males
T2: 76.2 (-91.1,243.6)
T3: 9.6 (-165.6, 184.8)
Females
T2: 146.8 (-36.2, 329.9)
T3:-6.7 (-184.8, 171.4)
BL
T2: 0.33 (-0.01, 0.68)
T3: 0.07 (-0.27,0.42)
Males
T2: 0.4 (-0.05, 0.84)
T3: 0.27 (-0.19,0.74)
Females
T2: 0.3 (-0.25, 0.86)
T3: -0.04 (-0.58,0.5)
Ponderal index
T2: 0.02 (-0.07,0.1)
T3: -0.04 (-0.13,0.06)
Males
T2: -0.03 (-0.17,0.12)
T3: -0.06 (-0.21,0.09)
Females
T2: -0.03 (-0.17 0.12)
T3: -0.06 (-0.21,0.09)
D-49
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome
^ Timing, Levels3
Comparison
Resultsb
Postnatal weight
T2:-138.1 (-573.7, 297.6)
T3: -78.3 (-531.6, 374.9)
Males
T2: -427.6 (-959.2, 104)
T3:-321.2 (-894.3, 252)
Females
T2: 239.6 (-519.6, 998.8)
T3: 128 (-620.3,876.3)
Postnatal length
T2: 0.08 (-1.78, 1.95)
T3: -0.1 (-2.04, 1.84)
Males
T2: -1.05 (-3.4, 1.29)
T3: 0.17 (-2.36, 2.7)
Females
T2: 1.07 (-2,4.13)
T3: -0.72 (-3.74,2.31)
Postnatal HC
T2: 0.17 (-0.76, 1.09)
T3: -0.23 (-1.19,0.73)
Males
T2: 0.27 (-0.92, 1.45)
T3: 0.28 (-1, 1.56)
Females
T2: -0.19 (-1.69, 1.31)
T3: -1.22 (-2.7,0.25)
Comparison: Tertiles were defined as follows: T2 = 0.74-1.52 vs. <0.74. T3 = >1.52 vs. <0.74.
Results: Lowest tertile used as reference.
Birth defects
T2 OR: 0.84 (0.37, 1.91)
T3 OR: 1.27 (0.59, 2.73)
D-50
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
Confounding: Maternal age, household income, parity, infant's gender. Additional confounding for BW, birth defects, ponderal index:
smoking of father, drinking of father. Additional confounding for BW, birth defects, ponderal index, postnatal weight, postnatal length,
POHC: maternal education. Additional confounding for postnatal weight, postnatal length, and POHC: infant's age.
Espindola Brazil Cross-sectional Mother-child pairs Cord blood from BW (z-score),
Santos et al., Recruitment: of women enrolled newborns BL (z-score),
2021,8442216 2017 inthePIPA weight for
Low project 2.06(1.06-5.21) length (z-
score), HC (z-
BW: N = 72 score)
BL: N = 65
Weight for length:
N = 64
HC: N = 62
Regression
coefficient per
loglO-unit
increase in
PFOS
BW
0.06 (-0.42, 0.54)
BL
-0.02 (-0.54,
0.50)
Weight for length
0.38 (-0.28, 1.04)
HC
0.18 (-0.46, 0.82)
PIPA = Rio Birth Cohort Study
Population: Mothers were recruited between 29th-32nd weeks of gestation and were over 16 years of age.
Exposure: Year of assessment not reported.
Confounding: Education, income, race, pre-gestational BMI, smoking active and passive, alcohol consumption, GA, primiparity, age
(continuous), and fish consumption.
Gennings et al., Sweden, Cohort Mothers and their
2020, 7643497 Recruitment: children (age 7)
Low 2007-2010, from the SELMA
Follow-up at 7 study
years N= 1,312
Prenatal serum BW (g) Regression BW
Mean(SE)= coefficient per -70.39 (SE = 16.31), p-value
0.82(0.19) loglO-unit <0.001
loglO-ng/mL increase in
PFOS
SELMA = Swedish Environmental Longitudinal, Mother and child, Asthma and allergy
Confounding: My Nutrition Index (MNI, z-score), sex, maternal smoking status, maternal weight (z-score), premature birth status, maternal
education, total energy intake (z-score)
Gross et al.,
2020,7014743
Low
United States
2012-2014
Nested Case-
control
Healthy and
overweight 18-
month-old
Hispanic children
from StEP,
N = 98
Newborn blood
Mean (SD) =
0.440 (0.364)
BW (z-score),
overweight
Regression
coefficient
(BW) and OR
(overweight) for
PFOS >mean
level i'.v. PFOS
< mean level
BW
-0.62 (-0.96, -0.29), p-value
<0.00714
Overweight
0.43 (0.17, 1.09)
D-51
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix, Sample Outcome Comparison
^ Timing, Levels3
Resultsb
StEP = Starting Early Program
Outcome: Overweight defined as 18-month weight for length z-score > 85th percentile
Confounding: Maternal age, maternal education, maternal depressive symptoms, pre-pregnancy BMI, GA, parity, intervention status.
Notes: BL = Birth Length; BMI = Body Mass Index; BW = Birth Weight; GA = gestational age; HC = head circumference; AC = Abdominal Circumference; FL = Femur Length;
BPD = Biparietal Diameter; SGA = Small-for-Gestational-Age; LSM = least squares mean; GM = Geometric Mean; SD = Standard Deviation; SE = Standard Error; OR = Odds
Ratio; PTB = Preterm Birth; RR = relative risk ratio; T2 = Tertile 2; T3 = Tertile 3
a Exposure reported as median percentile) in ng/mL unless otherwise specified.
b Results reported as effect estimate (95% confidence interval) unless otherwise specified.
c Confounding indicates factors the models presented adjusted for.
D.2 Reproductive
D.2.1 Male
Table D-2. Associations Between PFOS Exposure and Male Reproductive Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location, _ .
Design
Years
Population,
Ages,
N
Exposure „ ,
Matrix, Levels3 U come
Comparison
Resultsb
Children and Adolescents
Jensen et al. Denmark Cohort
(2020,6311643) 2010-2012
High
Infants from Maternal serum Levels of FSH
Odense Child 8.33 (IU/L), testosterone
Cohort (nmol/L), LH (IU/L),
N = 208 boys testosterone /LH
ratio, DHEAS
(nmol/L), DHEA
(nmol/L),
Androstenedione
(nmol/L), 17-OHP
(nmol/L)
Confounding: Age of the child at examination time, maternal parity0
Regression
coefficient
(testosterone), or
percent change per
doubling of PFOS
No statistically significant
associations
Lind et al. Denmark
(2017,3858512) 2010-2012
High
Cohort
Infants from
Odense child
cohort
Maternal serum Penile width (mm),
Total cohort: 8.1 anogenital distance
Regression
coefficient per ln-
unit increase in
AGDas
Continuous: 1.2 (-0.4, 2.7)
Q2: 0.9 (-0.9, 2.8)
D-52
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels"
Outcome
Comparison
Resultsb
N = 649 (scrotal, as; penile, PFOS, or by
Q3: 0.9 (-0.8, 2.7)
(296 boys) ap) (mm) quartiles
Q4: 1.9 (0.04, 3.7)
p-trend by quartiles = 0.06
AGDap, penile width: no
statistically significant
associations
AGDap: p-trend by quartiles =
0.55
Penile width: p-trend by
quartiles = 0.67
Results: Lowest quartile used as reference.
Confounding: Age at examination, weight for age z-score, pre-pregnancy BMI, parity, smoking
Itoh et al. (2016,
3981465)
Medium
Japan
2002-2005
Cohort
Infants from
Sapporo Cohort
of the Hokkaido
study
N = 83 boys
Maternal serum In cord blood, loglO- Regression
5.40 transformed levels of coefficient per
E2 (ng/mL), FSH
(mlU/mL), Inhibin B
(pg/mL), insulin-like
3 (ng/mL), LH
(mlU/mL),
progesterone
(ng/mL), prolactin
(ng/mL), SHBG (not
log 10-transformed,
nmol/L), testosterone
(pg/mL)
Testosterone/E2
ratio,
testosterone/SHBG
ratio
loglO-unit increase
in PFOS, least
squares mean (LSM)
by quartiles
E2
0.372 (0.057, 0.687)
p-value = 0.021
Ql: 4.34 (3.07, 6.15)
Q2: 5.84 (4.34, 8.01)
Q3: 8.74 (6.33, 12.05)
Q4: 6.39 (4.52, 8.98)
p-trend = 0.027
Inhibin B
-0.439 (-0.620, 0.257)
p-value < 0.001
Ql: 53.4 (42.4, 65.6)
Q2: 50.1 (41.2, 60.5)
Q3: 39.1 (31.8, 47.6)
Q4: 33.3 (26.6, 40.0)
p-trend <0.001
Progesterone
-0.344 (-0.678, 0.01)
p-value = 0.043
D-53
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels"
Outcome
Comparison
Resultsb
Confounding: Age, parity, body mass index before pregnancy, annual income, smoking during pregnancy.
pregnancy, gestational weeks of blood sampling for PFOS/PFOA measurement, gestational age at birth
Ql: 238.5 (161.5, 354.9)
Q2: 267.6 (192, 375.3)
Q3: 241.5 (168.7, 346.2)
Q4: 184.7 (126.5, 267.6)
p-trend = 0.231
Testosterone/E2
-0.399 (-0.643,-0.156)
p-value = 0.002
Ql: 20.3 (15.2, 26.8)
Q2: 19.5 (15.2, 24.8)
Q3: 14.5 (10.7, 18.6)
Q4: 14.5 (10.8, 18.8)
p-trend = 0.015
FSH, insulin-like 3, LH, prolactin,
SHBG, testosterone,
testosterone/SHBG: No
statistically significant
associations or trends
, caffeine consumption during
Lopez-Espinosa
etal. (2016,
3859832)
Medium
United
States
2005-2006
Cross-
sectional
Male children
ages 6-9 years
N = 1,169
Serum
22.4
Total testosterone
(ln-ng/dL)
Percent difference
between 75 th and
25th percentile of ln-
unit PFOS or by
quartiles
Total testosterone:
-5.8 (-9.4, -2.0)
Q2
Q3
Q4
-4.2 (-11.4, 3.6)
-9.2 (-16.1,-1.6)
-11.8 (-18.6,-4.3)
p-trend = 0.002
Results: Results by quartile used lowest quartile as reference.
Confounding: Age, month and time of sampling
Goudarzi et al.
(2017, 3981462)
Medium
Japan
2002-2005
Cohort
Children from
the Hokkaido
Study
N = 185 (81
males)
Serum
Total cohort:
5.20
Levels (log 10 ng-
mL) of DHEA,
androstenedione
Regression
coefficient per
loglO-unit increase
in PFOS or by
quartiles
Among males
DHEA: 0.308 (0.099, 0.755);
p-value = 0.011
D-54
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels"
Outcome
Comparison
Resultsb
Androstenedione: -0.011 (-0.312,
0.284); p-value = 0.926
Confounding: Gestational age, maternal age, parity, smoking and caffeine intake during pregnancy, maternal educational level, and blood
sampling period
Ernst et al.
(2019, 5080529)
Medium
Denmark Cohort
1999-2017
Children from
the Puberty
Cohort of the
Danish National
Birth Cohort
N = 565 boys
Maternal blood
Sample 1: 31.9
Sample 2: 27.2
No statistically significant
associations
Confounding: Highest social class of parents, maternal age at
daily number of cigarettes smoked in first trimester
Age (months) at
axillary hair
attainment, voice
break, first
ejaculation, Tanner
stages 2-5 for
genital development
or pubic hair growth;
combined sex-
specific puberty
indicator
menarche, maternal age at delivery, parity, prepregnancy body mass index, and
Regression
coefficient per log2-
unit increase in first
trimester maternal
serum PFOS
Puberty indicator:
mean difference in
age at puberty by
tertiles
Tian et al. China
(2019, 5390052) 2012-2013
Medium
Cohort
Male infants at
birth, 6 months,
and 12 months
N = 500
Maternal plasma
10.70
Anopenile distance
(AGDap) (mm),
anoscrotal distance
(AGDas) (mm)
Regression
coefficient per ln-
unit increase in
maternal PFOS or by
quartiles
AGDap
GEE (Birth, 6 mo, and 12 mo):
-0.34 (-1.38, 0.69); p-value =
0.516
Birth: -0.04 (-0.78, 0.69);
p-value = 0.925
6mo.:-1.20 (-3.29,0.88);
p-value = 0.262
12 mo.: 0.69 (-1.83,3.22);
p-value = 0.589
Q2: 1.57 (-1.95, 5.09)
Q3: 5.17(1.53,8.81);
p-value < 0.05
Q4: -0.49 (-4.04, 3.07)
AGDas
GEE (Birth, 6 mo, and 12 mo):
-0.83 (-1.71, 0.06); p-value =
0.067
D-55
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels"
Outcome
Comparison
Resultsb
Birth: -0.65 (-1.27, -0.02);
p-value = 0.0429
Q2: 0.17 (-0.79, 1.13)
Q3:-0.10 (-1.10, 0.90)
Q4: -1.46 (-2.44, -0.49);
p-value < 0.05
p-value for trend < 0.05
6 mo.:-2.21 (-4.28,-0.14);
p-value = 0.0372
12 mo.: 0.47 (-1.63,2.58);
p-value = 0.6587
Results: Lowest quartile used as reference.
Confounding: Maternal age at delivery, gestational age, maternal education, parity, pre-pregnancy BMI, infant age at physical examination,
and infant body size (birth weight at birth; WLZ at 6 and 12 months of age)
Wang et al.
(2019, 5080598)
Medium
China Cross- Pregnant women
2013 sectional and their
children
N = 340
(169 boys)
Confounding: Pregnant age, family income,
physical activity, gestational age, parity, pre
gestational weight gain
Cord blood Levels (log 10-
Total cohort: ng/mL) of estrone
0.65 (0.40-1.19) (El), E2, estriol (E3)
Regression
coefficient per
loglO-unit increase
inPFOS
El: 0.071 (-0.05, 0.18);
p-value = 0.247
E2: 0.02 (-0.10,0.14);
p-value = 0.761
E3: 0.36 (0.16, 0.55);
p-value < 0.001
maternal education level, maternal career, husband's smoking, energy daily intake, daily
¦pregnant maternal body mass index, gestational diabetes mellitus, infant sex, delivery mode,
Arbuckle et al.
(2020, 6356900)
Medium
Canada Cohort
2008-2011
Newborns from
the MIREC
cohort
N = 205 boys
Maternal plasma Anopenile distance Regression
AGDap
4.4
(AGDap) (mm),
anoscrotal distance
(AGDas) (mm)
coefficient per ln-
unit increase in Q2
maternal PFOS or by Q3
quartiles Q4
Per In increase: 0.13 (-1.13, 1.38)
-0.97 (-2.81,0.87)
-1.28 (-3.22,0.66)
0.22 (-1.68,2.13)
p-value for trend = 0.908
AGDas
Per In increase: 1.05 (-0.24, 2.35)
Q2: -0.87 (-2.78, 1.04)
Q3: 0.33 (-1.67,2.33)
Q4: 0.49 (-1.47, 2.46)
p-value for trend = 0.3936
D-56
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels"
Outcome
Comparison
Resultsb
Results: Lowest quartile used as reference.
Confounding: AGDap: recruitment site, education, active smoking status, gestational age; AGDas: household income, active smoking status,
gestational age
Zhou et al. Taiwan Cross-
(2016, 3856472) 2009-2010 sectional
Low
Adolescents
ages 13-15
N = 225 (102
boys)
Serum
Total: 28.9
Boys: 29.9
Levels (In-
transformed) of E2
(pmol/L),
testosterone
(nmol/L)
Regression
coefficient per unit
increase in PFOS
Testosterone, boys: -0.0029
(-0.0055, -0.0003)
p-value for interaction by
sex = 0.060
E2: No statistically significant
associations or interactions
Confounding: Age, sex, BMI, environmental tobacco smoke exposure, parental education, regular exercise, month of survey
Zhou et al. Taiwan Case-
(2017, 3858488) 2009-2010 control
Low
Regression
coefficient per unit
increase in PFOS
Testosterone
Cases: -0.004 (-0.005, -0.003)
Controls: -0.002 (-0.008, 0.003)
Children ages Serum Levels of
10-15 with Cases: 33.94 testosterone (In-
cases) or Controls: 28.91 nmol/L)
without
(control) asthma
N = 231 cases,
225 controls
Confounding: Age, sex, BMI, parental education, environmental tobacco smoke exposure, physical activity, month of survey
Di Nisio et al. Italy Cross-
(2019,5080655) 2017-2018 sectional
Low
Male high Serum Anogenital distance Mann-Whitney test
school students Unexposed (cm), crown-to-pubis (Exposed vs.
N = 100 (50 controls: 0.82 distance (cm), pubis- Controls)
unexposed Exposed: 1.11 to-floor distance
controls, 50 (cm), crown-to-
exposed) Semen pubis/pubis to floor
Unexposed ratio, penis
controls: 0.11 circumference (cm),
Exposed: 0.11 penis length (cm),
testicular volume
(mL), normal
morphology (%),
semen pH, immotile
sperm (%),
nonprogressive
motility (%),
AGD
Controls: 4.50 (4.0, 5.2)
Exposed: 4.00 (3.5, 5.0)
Adjusted p-value for comparison
of medians = 0.114
Pubis-to-floor distance
Controls: 97.0 (93.0, 101.1)
Exposed: 95.0 (90.3, 99.0)
Adjusted p-value for comparison
of medians = 0.320
Penis circumference
Controls: 10.10(9.9, 11.0)
Exposed: 9.50 (9.0, 10.0)
D-57
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Population,
Reference, Location, . A Exposure „ . „ „ b
„ Design Ages, „„ ^ „ Outcome Comparison Results
Confidence Years Matrix, Levels
N
progressive motility
Adjusted p-value for comparison
(%), total sperm
of medians < 0.001
count (106), semen
volume (mL), sperm
Penis length
concentration
Controls: 10.0 (9.0, 11.0)
(106/mL), viability
Exposed: 9.00 (8.0, 10.0)
(%), FSH (U/L),
Adjusted p-value for comparison
testosterone
of medians < 0.001
(nmol/L)
Testicular volume
Controls: 16.13 (14.8, 19.0)
Exposed: 14.00 (12.6, 16.0)
Adjusted p-value for comparison
of medians < 0.001
Normal morphology
Controls: 7.0 (4.0, 12.0)
Exposed: 4.0 (2.0, 6.0)
Adjusted p-value for comparison
of medians < 0.001
Semen pH
Controls: 7.60 (7.5, 7.7)
Exposed: 7.70 (7.6, 7.7)
Adjusted p-value for comparison
of medians = 0.042
Testosterone
Controls: 18.98 (12.9, 17.9)
Exposed: 18.98(16.3,21.8)
Adjusted p-value for comparison
of medians < 0.001
Crown-to-pubis, Crown-to-
pubis/pubis-to-floor, sperm
motility, sperm count, semen
D-58
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels"
Outcome
Comparison
Resultsb
volume, sperm concentration,
viability, FSH: No statistically
significant associations after
adjusting for comparison of
medians
Results: Values for each outcome are reported as median (25th-75th percentile).
Confounding: Age
General Population
Cui et al. (2020, China Cross- Adult men
6833614) 2015-2016 sectional N = 651
Medium
Serum
9.94
Semen
0.15
Serum levels (In-
transformed) of E2
(pmol/L), FSH
(IU/L), LH (IU/L),
SHBG (nmol/L),
free testosterone,
total testosterone
(nmol/L); free
androgen index, total
testosterone/LH ratio
Percent change per
ln-unit increase in
serum or semen
PFOS, or by
quartiles
SHBG
Serum PFOS:-4.94 (-8.71,
-1.02); p-value = 0.014
p-trend by quartiles = 0.004
Ages < 30: -3.11 (-6.58, 0.48);
p-value = 0.069
Semen PFOS: -5.29 (-8.94,
-1.49); p-value = 0.007
p-trend by quartiles = 0.026
Ages < 30: -3.13 (-6.25, -0.10);
p-value = 0.009
Total testosterone
Serum PFOS: -3.36 (-6.40,
-0.22); p-value = 0.036
p-trend by quartiles = 0.022
Ages < 30: -4.25 (-7.77, -0.59);
p-value = 0.023
Semen PFOS: -4.20 (-7.13,
-1.18); p-value = 0.007
p-trend by quartiles = 0.019
Ages < 30: -4.82 (-7.96, -1.58);
p-value = 0.004
Total testosterone/LH,
Serum PFOS: -4.53 (-8.99, 0.15);
p-value = 0.058
D-59
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels"
Outcome
Comparison
Resultsb
p-trend by quartiles = 0.044
Semen PFOS: -5.00 (-9.32,
-0.48); p-value = 0.031
p-trend by quartiles = 0.042
No statistically significant
associations by age groups
E2, FSH, free androgen, LH, free
testosterone: No statistically
significant associations or trends
Confounding: Age, BMI, smoking status, blood sampling time, fasting status
Petersen et al. Denmark Cross-
(2018, 5080277) 2007-2009 sectional
Medium
Faroese men
born between
1981 and 1984
N = 263
Serum
19.5
Levels (log-
transformed) of E2
(nmol/L), FSH
(IU/L), free
testosterone
(pmol/L), inhibin B
(pg/mL), LH, (IU/L),
SHBG (nmol/L),
testosterone
(nmol/L)
Ratios of free
testosterone/E2, free
testosterone/LH,
Inhibin B/FSH,
testosterone/E2,
testosterone/LH
Regression
coefficient per log-
unit increase in
PFOS
LH: 0.35 (0.02, 0.68);
p-value = 0.04
SHBG: 0.31 (0.02,0.60);
p-value = 0.04
No other statistically significant
associations
Normal morphology
(%), motile sperm
(logit-%), total
sperm count
((106)i/3) semen
volume (mL1/3),
D-60
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Population,
Design Ages, Mf.|,0™re,, Outcome
Years Matrix, Levels
N
Comparison
Resultsb
sperm concentration
((106/mL)1/3)
Comparison: Logarithm base not specified.
Confounding: Age, BMI groups, current smoking, time of sampling
Kvist et al. Greenland, Cross-
(2012, 2919170) Poland, and sectional
Medium Ukraine
2002-2004
Male partners of Serum
pregnant women Mean
fromlNUENDO Greenland:
N = 359 51.65
Poland: 12.12
Ukraine: 8.20
Confounding: Age, abstinence time, alcohol intake and CB-153
Y :X chromosome
ratio of sperm
Linear regression
adjusted r2
0.016; p-value = 0.026
Leter et al. Greenland, Cross-
(2014, 2967406) Poland, and sectional
Medium Ukraine
2002-2004
Male partners of Serum
pregnant women Mean = 27.2
from INUENDO
N = 262
Sperm DNA Regression
methylation level (% coefficient per ln-
5-mC) at LINE-1, unit increase in
Alu, or Sat-alpha;
global DNA
methylation level
(FCM DGML
channel no.)
PFOS
Sat-alpha
Total: 1.1 (-3.1, 5.3)
Greenland: -1.8 (-8.6, 5.1)
Poland: -7.2 (-16, 1.6)
Ukraine: 8.2 (0.6, 15.8)
Global
Total: -21 (-63.2, 21.3)
Greenland: -32.1 (-105.6, 41.3)
Poland: -108.4 (-191.5, -25.2)
Ukraine: 27.2 (-43.1, 97.6)
Confounding: Site, age (ln-transformed), smoking status
LINE-1, Alu: No statistically
significant associations
Pan etal. (2019, China Cross- Adult men in
6315783) 2015-2016 sectional Nanjing
Medium N = 664
Serum
8.378
Semen
0.097
Sperm normal
morphology (%),
count ((106)1/3),
concentration
((106/mL)1/3),
progressive motility
(%), curvilinear
velocity (VCL)
(|im/s): straight-line
Regression
coefficient per ln-
unit increase in
serum or serum
PFOS, or by
quartiles
No statistically significant
associations by serum PFOS
levels; following results are by
semen PFOS
Progressive motility: -1.700
(-2.867, -0.532); p-value = 0.03
Q2: -2.30 (-5.27, 0.68)
Q3: -1.53 (-4.61, 1.56)
D-61
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Population,
Reference, Location, . A Exposure „ . „ „ b
„ Design Ages, „„ ^ „ Outcome Comparison Results
Confidence Years Matrix, Levels
N
velocity (VSL)
Q4
-5.54 (-8.72, -2.36)
(|im/s). DNA
p-trend = 0.01
fragmentation index
(DFI) (ln-%), high
VCL: -0.767 (-1.447, -0.087);
DNA stainability
p-value = 0.1
(HDS) (ln-%);
Q2
-1.60 (-1.50,2.01)
semen volume (ln-
Q3
-2.78 (-2.40, 1.10)d
mL)
Q4
-4.8 (-2.97, -0.72)
p-trend = 0.1
VSL: -0.773 (-1.337, -0.209);
p-value = 0.04
Q2
-1.00 (-2.44,0.45)
Q3
-1.40 (-2.89,0.09)
Q4
-2.06 (-3.60, -0.52)
p-trend = 0.1
DFI: 0.087 (0.033,0.142);
p-value = 0.02
Q2: 0.03 (-0.11,0.17)
Q3: 0.08 (-0.07,0.22)
Q4: 0.25 (0.10,0.40)
p-trend = 0.01
Normal morphology, sperm count,
sperm concentration, sperm HDS,
semen volume: No statistically
significant associations
Results: Lowest quartile used as reference.
Confounding: Age, BMI, BMI2, smoking, alcohol intake, abstinence time
iVofes:17-OHP = 17-hydroxyprogesterone; AGD = anogenital distance; AGDap = anopenile distance; AGDas = anoscrotal distance; BMI = body mass index;
DHEA = dehydroepiandrosterone; DFI = DNA fragmentation index; DNA = deoxyribonucleic acid; El = estrone; E2 = estradiol; E3 = estriol; FSH = follicle stimulating hormone;
GEE = generalized estimating equation; HDS = high DNA stainability; LH = luteinizing hormone; LSM = least squares mean; MIREC = Maternal-Infant Research on
Environmental Chemicals; PFOA = perfluorooctanoic acid; PFOS = perfluorooctane sulfonic acid; SHBG = sex hormone-binding globulin; VCL = curvilinear velocity;
VSL = straight-line velocity.
a Exposure levels reported as median in ng/mL unless otherwise specified.
D-62
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
b Results reported as effect estimate (95% confidence interval) unless otherwise specified.
c Confounding indicates factors the models presented adjusted for.
d Values are reproduced as reported in publication.
D.2.2 Female
Table D-3. Associations between PFOS Exposure and Female Reproductive Effects in Female Children and Adolescents
Reference,
Confidence
Location,
Year(s)
Study
Design
Population, Exposure
Ages, Matrix,
N Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Jensen et al.
(2020,
6311643)
High
Denmark, Cohort
2010-2012
Female
infants from
the Odense
Child Cohort,
Age
4 months,
N = 165
Maternal serum
8.07 (5th-95th
percentile = 4.21,
15.50)
Levels of 17-OHP
(nM), androstenedione
(nM), DHEA (nM),
DHEAS (nM), FSH
(IU/L), LH (IU/L)
Percent change per
doubling in PFOS
17-OHP
2.1 (-11.9, 18.2)
Androstenedione
0.6 (-14.3, 18.2)
DHEA
-9.4 (-22.5, 5.9)
DHEAS
-10.4 (-28.4, 12.2)
FSH
0.2 (-12.5, 14.7)
LH
9.5 (-12.8, 37.6)
Confounding: Age of the child at examination time, maternal parity0
Lind et al.
(2017,
3858512)
High
Denmark
2010-2012
Cohort
Infants from
Odense child
cohort
N = 649 (353
girls)
Maternal serum
Total cohort: 8.1
Anogenital distance
(AGD) (mm); clitoral
(AGDac), fourchette
(AGDaf)
Regression
coefficient per ln-unit
increase in PFOS, or
by quartiles
AGDac
Continuous: -2.3 (-3.8, -0.7)
Q2
Q3
Q4
-1.0 (-2.6, 0.6)
-1.7 (-3.5, 0)
-2.8 (-4.5,-1.1)
p-trend by quartiles < 0.01
Results: Lowest quartile used as reference.
Confounding: Age at examination, weight for age z-score, pre-pregnancy BMI, parity, smoking.
AGDaf
Continuous: -0.4 (-1.6, 0.8)
No statistically significant
associations by quartiles, p-trend
by quartiles = 0.31
D-63
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population, Exposure
Ages, Matrix,
N Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Yao et al.
(2019,
5187556)
High
Regression
coefficient per loglO-
unit increase in PFOS
China, Cross- Pregnant Cord blood Testosterone (loglO-
2010-2013 sectional women 1.39 (0.92,2.01) ng/mL),
(aged > 18 ye Estradiol (loglO-
ars) and pg/mL),
female infants Testosterone to
N = 171 estradiol ratio (loglO-
transformed)
Confounding: Maternal age, pre-pregnancy BMI, parity, mode of delivery, passive smoking during pregnancy, gestational age, household
income level among male and female infants separately
Testosterone
0.15 (0.01, 0.29), p-value < 0.05
Estradiol
0.01 (-0.05,0.07)
Testosterone to estradiol ratio
0.14 (0.01, 0.27), p-value < 0.05
Donley et al.
United Nested
Mothers and Maternal serum AMH (loglO-ng/mL)
Regression
Complete AMH data:
(2019,
Kingdom, case-
their 19.8(15.1,24.9)
coefficient per unit
0.01 (0.00, 0.02)
5381537)
Recruitment control
daughters
increase in PFOS
Multiple imputation model:
Medium
1991-1992,
from the
0.01 (0.00, 0.015)
outcome
ALSPAC,
assessed at
N = 446
adolescence
Confounding: Maternal age at delivery, pre-pregnancy BMI, maternal education
Ernst et al.
(2019,
5080529)
Medium
Denmark,
Recruitment
1996-2002,
outcome
assessed
2012-2017
Female Maternal blood
adolescents Sample 1
from the (N= 366):
Danish 32.3 (10th-90th
National Birth percentiles = 19.3
Cohort, , 50.8)
N = 555
Combined puberty
indicator:
Mean difference by
tertiles of PFOS
Cohort Female Maternal blood Breast development, Combined puberty Combined puberty indicator
T2: -3.73 (-6.59, -0.87)
T3:-0.17 (-2.83, 2.49)
Breast development
-3.01 (-7.96, 1.95), p-value = 0.03
Pubic hair development
1.81 (-2.42,6.04)
Axillary hair
0.50 (-2.79, 3.79), p-value = 0.02
Menarche
-0.68 (-3.13, 1.77)
Exposure Levels: [Sample 2] Median = 27.9 ng/mL (10th-90th percentiles = 16.5, 42.2 ng/mL). Samples 1 and 2 combined for analysis.
Outcome: Age in months at Tanner stage 5 used to measure breast development and pubic hair development. For combined puberty indicator,
lowest tertile was used as the reference group.
Confounding: Highest social class of parents, maternal age at menarche, maternal age at delivery, parity, pre-pregnancy BMI, daily number of
cigarettes smoked in first trimester
Breast development,
pubic hair
development, age at
attainment of axillary
hair (months), age at
menarche, age at
attainment of
combined puberty
indicator
All other outcomes:
Regression
coefficient per log2-
unit increase in PFOS
D-64
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population, Exposure
Ages, Matrix,
N Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Maternal serum Levels of Regression
5.20(1.50,16.20) androstenedione coefficient per loglO-
(loglO-ng/mL), DHEA unit increase in PFOS
(loglO-ng/mL)
Androstenedione
0.004 (-0.29, 0.30),
p-value = 0.059
DHEA
0.24 (-0.02, 0.80)
Goudarzi et al. Japan, Cohort Pregnant
(2017, 2002-2005 women and
3981462) their infants
Medium from the
Hokkaido
Study on the
Environment
and
Children's
Health,
N = 104
Confounding: Gestational age, maternal age, parity, smoking and caffeine intake during pregnancy, maternal educational level, blood
sampling period
Itoh et al.
Japan, Cohort
Female
Maternal serum
Cord blood levels of
Regression
Estradiol
(2016,
2002-2005
infants from
5.15 (3.45,7.00)
estradiol (loglO-
coefficient per loglO-
0.08 (-0.15,0.31)
3981465)
the Sapporo
ng/mL), testosterone
unit increase in PFOS
Testosterone
Medium
Cohort of the
(loglO-pg/mL),
0.07 (-0.26, 0.40)
Hokkaido
prolactin (loglO-
Prolactin
Study,
ng/mL), progesterone
-0.49 (-0.76, -0.22),
N = 106
(loglO-ng/mL), SHBG
p-value = 0.001
(nmol/L); testosterone
Progesterone
to SHBG ratio,
-0.55 (-0.89, -0.21),
testosterone to
p-value = 0.002
estradiol ratio
SHBG
-0.18 (-0.42, 0.06)
Testosterone/SHBG ratio
0.25 (-0.16,0.66)
Testosterone/estradiol ratio
-0.01 (-0.03 0.26)
Confounding: Maternal age, parity, BMI before pregnancy, annual income, smoking during pregnancy, caffeine consumption during
pregnancy, gestational weeks of blood sampling for PFOS/PFOA measurement, gestational age at birth
Liu et al.
(2020,
6569227)
Medium
China,
2013-2014
Cross- Female
sectional neonates,
N = 191
Cord blood
4.15(2.81,6.18)
Levels of 17-OHP
(ng/mL), progesterone
(ng/mL)
Percent change per 17-OHP
IQR increase in PFOS -1.27 (-7.52, 5.39)
Progesterone
-1.68 (-6.93, 3.88)
D-65
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population, Exposure
Ages, Matrix,
N Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Confounding: Maternal age at delivery, pre-pregnancy BMI, maternal education status, passive smoking during smoking, parity, gestational
weeks, sample-collection time
Lopez-
Espinosa et al.
(2016,
3859832)
Medium
United
States,
2005-2006
Cohort
Females from
the C8 Health
Project,
Ages 6-9,
N = 1,123
Serum
20.9(15.3,
29.4)
Levels of estradiol (ln-
pg/mL), total
testosterone (ln-ng/dL)
Percent difference for
75th vs. 25th
percentiles, or by
quartiles
Estradiol
75th vs. 25th percentiles
-0.3 (-4.6, 4.2), p-value = 0.048
Q2: 5.2 (-3.7, 14.9)
Q3: 3.7 (-5.2, 13.4)
Q4: -1.3 (-9.9, 8.2)
Testosterone
75th vs. 25th percentiles
-6.6 (-10.1,-2.8)
Q2
Q3
Q4
-1.1 (-8.6,7.1)
-7.8 (-15.0,-0.1)
-11.1 (-18.2,-3.5)
Results: Lowest quartile used as the reference group.
Confounding: Age, month, time of sampling
Maisonet et al.
(2015,
3859841)
Medium
United
Kingdom,
1991-1992
Maternal serum
19.2(15.1,25.0)
Levels of serum total
testosterone (nmol/L),
SHBG (nmol/L)
Regression
coefficient by tertiles
ofPFOS
Cohort Female
adolescents
from
ALSPAC,
Age 15,
N = 72
Results: Lowest tertile used as the reference group.
Confounding: Maternal education, maternal age at delivery, maternal pre-pregnancy BMI, maternal smoking during pregnancy, time of day
daughter's blood sample was obtained, daughter's age at menarche, daughter's BMI at 15 years. SHBG concentration included in testosterone
model.
Testosterone
T2: 0.1 (-0.07, 0.28)
T3: 0.18 (0.01,0.35)
SHBG
T2: -2.86 (-18.8, 13.09)
T3: 3.46 (-12.06, 18.98)
Tsai et al.
(2015,
2850160)
Medium
Taiwan,
2006-2008
Cross-
sectional
Female
adolescents,
Ages 12-17,
N = 95
Serum, Levels of serum FSH
8.65 (5.37, 13.29) (ln-mlU/mL), serum
SHBG (ln-nmol/L)
Means by quartiles of FSH
PFOS Ql: 1.56 (SE = 0.23)
Q2: 1.67 (SE = 0.23)
Q3: 1.36 (SE = 0.19)
Q4: 1.23 (SE = 0.35)
SHBG
Ql: 3.58 (SE = 0.29)
Q2: 3.36 (SE = 0.29)
Q3: 3.49 (SE = 0.24)
D-66
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Population,
Location, Study .
Year(s) Design ^S'
Exposure
Matrix,
Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Confounding: Age, BMI, high fat diet
Q4: 3.41 (SE = 0.44)
Wang et al.
(2019,
5080598)
Medium
China.
2013
Cross-
sectional
Pregnant
women and
their children,
N = 171
Cord blood
0.65 (0.40, 1.19)
Levels of estrone
(loglO-ng/mL), (3-
estradiol (loglO-
ng/mL), estriol (loglO-
ng/mL)
Regression
coefficient per ln-unit
increase in PFOS
Estrone
0.15 (0.04, 0.26), p-value = 0.007
(^-estradiol
-0.17 (-0.31,-0.02),
p-value = 0.023
Estriol
0.48 (0.27, 0.70), p-value < 0.001
Confounding: Pregnant age, family income, maternal education level, maternal career, husband's smoking, energy daily intake, daily physical
activity, gestational age, parity, pre-pregnant maternal BMI, gestational diabetes mellitus, infant sex, delivery mode, gestational weight gain
Notes: 17-OHP = 17-hydroxyprogesterone; AMH = anti-Mullerian hormone; BMI = body mass index; DHEA = dehydroepiandrosterone; DHEAS = dehydroepiandrosterone-
sulfate; FSH = follicle stimulating hormone; LH = luteinizing hormone; SHBG = sex hormone-binding globulin; T1 = tertile 1; T2 = tertile 2; T3 = tertile 3; Q1 = quartile 1;
Q2 = quartile 2; Q3 = quartile 3; Q4 = quartile 4; ALSPAC = Avon Longitudinal Study of Parents and Children.
a Exposure levels reported as median (25th-75th percentile) unless otherwise specified.
b Results reported as effect estimate (95% confidence interval) unless otherwise specified.
c Confounding indicates factors the models presented adjusted for.
Table D-4. Associations between PFOS Exposure and Female Reproductive Health Effects in Pregnant Women
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure
Matrix,
Levels" (ng/mL)
Outcome
Comparison
Resultsb
Huo et al.
(2020,
6505752)
High
Mitro et al.
(2020,
6833625)
High
OR per ln-unit
increase in PFOS
China, Cohort Females from Plasma Gestational
2013-2016 the Shanghai 9.36 (6.57, 13.69) hypertension,
Birth Cohort Preeclampsia/Eclampsia
Study,
Ages > 20,
N= 3,220
Confounding: Maternal age, pre-pregnancy BMI, parity, parental educational levels, gestational age of blood drawn, fetal sex
Gestational hypertension
0.91 (0.57, 1.43)
Preeclampsia/Eclampsia:
1.24 (0.82, 1.90)
United States,
Recruitment
1999-2002,
outcome
assessed 3-
Cohort Females from Plasma
Project Viva, 24.7 (18.1, 33.9)
N = 812
Sex hormone binding
globulin (nmol/L)
Percent difference
per log2-unit
increase in PFOS
Sex hormone binding globulin:
-0.6 (-7.6, 6.9)
Ages <35:-0.8 (-11.9, 11.7)
Ages >35:-1.5 (-10.0,7.8)
D-67
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Population, Exposure
Location, Studv .
„ . Ages, Matrix, Outcome Comparison
Year(s) Design T .a/ / t \
N Levels3 (ng/mL)
Resultsb
years
postpartum
Confounding: Age, pre-pregnancy BMI, marital status, race/ethnicity, education, income, smoking, parity
Borghese et al.
Canada, Cohort Females from Plasma DBP (mmHg), SBP
Regression
DBP
(2020,
2008-2011 the MIREC GM = 4.56 (95% (mmHg), preeclampsia,
coefficient (DBP,
Trimester 1 to delivery:
6833656)
study, CI: 4.44,4.69) gestational hypertension
SBP), OR
0.47 (0.10, 0.85)
Medium
Ages >18,
(preeclampsia,
Trimester 1: 0.46 (0.01, 0.90)
N = 1,739
gestational
Trimester 2: 0.33 (-0.10, 0.76)
hypertension)
Trimester 3: 0.66 (0.18, 1.14)
per log2-unit
SBP
increase in PFOS or
Delivery: 1.19(0.28,2.1)
by tertiles
Preeclampsia
1.25 (0.84, 1.82)
T2: 1.72 (0.77, 3.82)
T3: 1.55 (0.68, 3.49)
Gestational hypertension
1.15 (0.91, 1.45)
T2: 1.43 (0.90, 2.29)
T3: 1.38 (0.84,2.23)
Results: Lowest tertile used as the reference group.
Confounding: Maternal age, education, smoking status, pre-pregnancy BMI, parity
Huang et al. China, Cross- Females from Plasma Gestational
(2019, 2011-2012 sectional mother-infant 2.38 (1.81,3.23) hypertension,
5083564) pairs, preeclampsia
Medium N = 687
Comparison: Standardized PFOS calculated by subtracting PFOS concentration from
Confounding: Age, pre-pregnancy BMI, parity, education level
OR per increase in
standardized PFOS
Gestational hypertension
0.87 (0.57, 1.34)
Preeclampsia
0.83 (0.52, 1.32)
mean PFOS concentration and dividing by the SD.
Lyugso et al.
(2014,
2850920)
Medium
Greenland,
2002-2004
Cross- Pregnant
sectional women from
the
INUENDO
cohort,
N = 1,623
Serum,
8.0 (10th-90th
percentile = 3.6,
25.6)
Menstrual cycle length
(long), irregularity
OR per log-unit
increase in PFOS or
by tertiles
Length
1.1 (0.8, 1.6)
T2: 1.3 (0.8,2.2)
T3: 1.2(0.6,2.5)
Irregularity
1.2 (0.9, 1.8)
T2: 1.1 (0.6,2.1)
T3: 1.7(0.8,3.5)
D-68
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure
Matrix,
Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Results: Lowest tertile used as the reference group.
Comparison: Logarithm base not specified.
Confounding: Age at menarche, age at pregnancy, parity, pre-pregnancy BMI, smoking, country
Romano et al.
United States, Cohort Females from Serum
Breastfeeding RR by quartiles of
Termination at 3 months
(2016,
2003-2006 the HOME 13.9(9.6,18.2)
termination (by 3 months PFOS
Q2: 1.08 (0.79, 1.46)
3981728)
study,
postpartum),
Q3: 1.39 (1.04, 1.88)
Medium
Ages >18,
Breastfeeding
Q4: 1.32 (0.97, 1.79)
N = 336
termination (by 6 months
Termination at 6 months
postpartum)
Q2: 1.17 (0.93, 1.48)
Q3: 1.16(0.91, 1.48)
Q4: 1.25 (0.98, 1.58)
Results: Lowest quartile used as the reference group.
Confounding: Maternal age at delivery, household income, total weeks of prior breastfeeding, gestational week at blood draw, marital status,
race, parity, maternal serum cotinine during pregnancy, alcohol use during pregnancy
Rylander et al.
Sweden, 1989 Case- Females with Serum
Preeclampsia OR by quartiles of
Q2: 0.81 (0.5, 1.32)
(2020,
control or without Primiparous
PFOS
Q3: 1.23 (0.78, 1.93)
6833607)
pre- cases:
Q4: 0.96 (0.60, 1.53)
Medium
eclampsia, 12.9 (Minimum,
Ages 15-44, maximum = 2.15,
N = 876 50.0)
Exposure Levels: [Multiparous cases] Median = 10.9 ng/mL (Minimum, maximum = 1.49, 66.6 ng/mL); [Primiparous controls]
Median = 12.4 ng/mL (Minimum, maximum = 0.52, 54.5 ng/mL); [Multiparous controls] Median = 9.36 ng/mL (Minimum, maximum = 1.13,
47.0 ng/mL)
Results: Lowest quartile used as the reference group.
Confounding: Maternal age, BMI in early pregnancy, maternal smoking in early pregnancy, parity
Timmermann
etal. (2017,
3981439)
Medium
Denmark,
1997-2000
2007-2009
Cohort
Pregnant and Serum Total breastfeeding
postpartum 19.47(8.67, duration (months),
females, 28.22) Exclusive breastfeeding
N = 987 duration (months)
Confounding: Cohort, maternal age, pre-pregnancy BMI, pregnancy alcohol intake,
Total breastfeeding duration
-1.4 (-2.1,-0.6)
Exclusive breastfeeding duration
-0.3 (-0.6,-0.1)
pregnancy smoking, education, employment, parity
Regression
coefficient per
doubling of PFOS
Toft et al.
(2016,
3102984)
Medium
Denmark
1980-1996
Case-
control
Pregnant
females and
their male
infants,
N = 545
Amniotic fluid
Tertile 2: (Range:
0.8, 1.4)
Amniotic fluid levels of
17-OHP (ln-nmol/L),
androstenedione (ln-
nmol/L), DHEAS (ln-
nmol/L), progesterone
Percent difference
in median level per
1% increase in
PFOS or by tertiles
17- OHP
0.15(0.11,0.20)
T2: 7 (-1, 13)
T3: 18(11,26)
p-value for trend <0.001
D-69
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure
Matrix,
Levels3 (ng/mL)
Outcome
Comparison
Resultsb
(ln-nmol/L), testosterone
(ln-nmol/L)
Androstenedione
0.15(0.10, 0.21)
T2: 8 (0, 17)
T3: 17(8, 25)
p-value for trend = 0.001
DHEAS
0.07 (-0.03,0.16)
T2: 5 (-10, 20)
T3: 2 (-14, 17)
p-value for trend = 0.93
Progesterone
0.21 (0.14, 0.29)
T2: 11 (0, 23)
T3: 22 (11, 34)
p-value for trend = 0.001
Testosterone
0.16(0.09, 0.23)
T2: 9 (-2, 20)
T3: 18(7, 29)
p-value for trend = 0.002
Results: Lowest tertile used as the reference group.
Confounding: Gestational age of amniocentesis, maternal age, smoking (cotinine groups), case or control status.
Wikstrom et al.
Sweden, Cohort Females from Serum Preeclampsia
OR per log2 1.53 (1.07,2.20)
(2019,
2007-2010 the SELMA 5.39 (3.95, 7.61)
increase inPFOS or Q4: 2.68 (1.17, 6.12)
5387145)
study,
by quartiles
Medium
Ages 28-35,
N = 1,773
Results: Lowest quartile used as the reference group
Confounding: Parity, women's age, body weight, smoke exposure
iVofes:17-OHP = 17-hydroxyprogesterone; BMI = body mass index; DBP = diastolic blood pressure; DHEAS= dehydroepiandrosterone sulfate; GM = geometric mean;
HOME = Health Outcomes and Measures of the Environment; HR = hazard ratio; INUENDO = Biopersistent Organochlorines in Diet and Human Fertility; LIFE = Longitudinal
Investigation of Fertility and the Environment Study; MIREC = Maternal Infant Research on Environmental Chemicals; OR = odds ratio; RR = relative risk ratio; SBP = systolic
blood pressure; SELMA = Swedish Environmental Longitudinal, Mother and child, Asthma and allergy study; T1 = tertile 1; T2 = tertile 2; T3 = tertile 3.
a Exposure levels reported as median (25th-75th percentile) unless otherwise specified.
b Results reported as effect estimate (95% confidence interval) unless otherwise specified.
c Confounding indicates factors the models presented adjusted for.
D-70
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table D-5. Associations between PFOS Exposure and Female Reproductive Health Effects in Non-Pregnant Adult Women
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure
Matrix,
Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Ding et al.
United States, Cohort Pre- Serum Natural menopause
HR per doubling
Sm-PFOS:
(2020,
1999-2017 menopausal Sm-PFOS: 7.2
increase in PFOS or
1.08 (0.99, 1.19)
6833612)
women from (4.6, 10.8)
by tertiles
T2: 1.11 (0.90, 1.37)
High
the Study of n-PFOS: 17.1
T3: 1.27 (1.01, 1.59)
Women's (12.2,24.5)
p-value for trend = 0.03
Health Across
the Nation,
n-PFOS:
Ages 42-52,
1.11 (0.99, 1.23)
N = 1,120
T2: 1.06 (0.86, 1.31)
T3: 1.26 (1.02, 1.57)
p-value for trend = 0.03
Results: Lowest tertile used as the reference group.
Confounding: Age at baseline, race/ethnicity, study site, education, parity, BMI at baseline, physical activity, smoking status, prior hormone
use at baseline0
Crawford et al.
United States, Cohort Females from Serum Cycle-specific time to
Time to pregnancy
Cycle-specific time to pregnancy
(2017,
2008-2009 the Time to 9.29 (8.31, 10.38) pregnancy, day-
outcomes:
0.89 (0.49, 1.60)
3859813)
Conceive specific time to
Fecundability ratio
Day-specific time to pregnancy
Medium
Study, pregnancy, AMH (ln-
per ln-unit increase in
0.99 (0.28, 2.32)
Ages 30-44, ng/mL)
PFOS
AMH
N = 99
0.07
AMH:
Regression
coefficient per ln-unit
increase in PFOS
Confounding: Age, mean cycle length (added for cycle-specific time to pregnancy model)
Kim et al.
Australia, Cross- Females Follicular fluid Fertilization rate
Regression
2.28 (-0.56,5.11)
(2020,
2006-2011 sectional undergoing Mean =4.8
coefficient per unit
6833596)
fertility (Minimum,
increase in PFOS
Medium
treatment, Maximum = 0.7,
Ages 23-42, 22.4)
N = 97
Confounding: Age
D-71
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure
Matrix,
Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Lum et al.
United States, Cohort Females from
Serum
Day-specific
Regression
All women:
(2017,
2005-2009 the LIFE
Women
probability of
coefficient by tertiles
T2: 1.0 (0.7, 1.5)
3858516)
study, Ages
with < 24-day
pregnancy
ofPFOS
T3: 0.9 (0.6, 1.3)
Medium
18-40,
cycle:
N = 483
12.3 (9.7, 17.0)
Women with 25
to 31-day cycle:
12.6 (8.2, 17.6)
Women
with> 32-day
cycle: 11.5 (7.3,
16.9)
Results: Lowest tertile used as the reference group
Confounding: Couple intercourse pattern, female menstrual cycle length, age, BMI, active smoking at enrollment
Tsai et al.
Taiwan, Cross- Females,
Serum,
Levels of FSH in
Means by quartiles of FSH
(2015,
2006-2008 sectional Ages 18-30,
8.65 (5.37, 13.29) serum (ln-mlU/mL),
PFOS
Ql: 1.71 (SE = 0.25)
2850160)
N = 265
SHBG in serum (ln-
Q2: 1.66 (SE = 0.23)
Medium
nmol/L)
Q3: 1.71 (SE = 0.25)
Q4: 1.69 (SE = 0.25)
SHBG
Ql: 3.90 (SE = 0.21)
Q2: 3.82 (SE = 0.20)
Q3: 3.89 (SE = 0.22)
Q4: 3.80 (SE = 0.21)
Confounding: Age, BMI, high fat diet
Wang et al.
China, Case- Females of
Plasma,
Endometriosis-related
OR by tertiles of
T2: 1.11 (0.61, 1.99)
(2017,
2014-2015 control reproductive
Cases:
infertility
PFOS
T3: 0.66 (0.36, 1.21)
3856459)
age,
6.40 (4.02, 11.42)
Medium
N = 335
Controls:
6.60 (3.92, 13.54)
Confounding: Age, BMI, household income, and education
Notes: AMH = anti-Mullerian hormone; BMI = body mass index; T1 = tertile 1; T2 = tertile 2; T3 = tertile 3.
a Exposure levels reported as median (25th-75th percentile) unless otherwise specified.
D-72
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
b Results reported as effect estimate (95% confidence interval) unless otherwise specified.
c Confounding indicates factors the models presented adjusted for.
D.3 Hepatic
D.3.1.1 Forest Plots
Confidence
Rating
Exposure Matrix Study Design Exposure Levels
Sub-population Comparison
Effect Estimate
-0 045 -0.040 -0.035 -0.030 -0.025 -0.020 -0.015 -0.010 -0.005
Low
confidence
Mean change in serum
PFOS from baseline to
project end for 3M
employees: -101.3
ng/mL; for contractors
1 ng/mL
Regression
coefficient (per n n.
1-ng/mL increase
in PFOS)
-0.045 no4 on i -oo:o -oo;5 -on;o -oois -0.010 -0005
Figure D-l. Overall ALT Levels from Epidemiology Studies Following Exposure to PFOS
D-73
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Interactive figure and additional study details available on Tableau.
D3.1.2 Tables
Table D-6. Associations Between PFOS Exposure and Hepatic Effects in Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure Matrix,
Levels (ng/mL)a
Outcome
Comparison
Select Resultsb
Adults
Omoike et al.
(2020,
6988477)
Medium
United States
2005-2012
Cross-
sectional
Adults from
NHANES,
Age > 20,
N = 6,652
Serum
11.40 (20th-80th
percentile = 5.80-
23.18)
Levels of iron in
serum, bilirubin,
and albumin
Percent change per Iron concentration in serum
one percent 0.05 (0.03, 0.07), p-value < 0.05
increase in PFOS
Bilirubin
0.03 (0.02, 0.05), p-value < 0.05
Confounding: Age, sex, race, education, poverty income ratio, serum cotinine, BMI
Albumin
0.02 (0.02, 0.03), p-value < 0.05
Jain (2019,
5381541)
Medium
United States
2003-2014
Cross-
sectional
Adults from
NHANES,
Ages > 20,
N = 108-3,562
Serum
Levels of ALT
(loglO-IU/L),
AST (loglO-
IU/L)
Regression ALT,
coefficient per Non-obese,
loglO-unit increase GF-1:-0.008
inPFOS GF-2:0.011
GF-3A: -0.013
GF-3B/4: -0.088, p-value < 0.01
Obese,
GF-1: 0.048, p-value<0.01
GF-2: 0.005
GF-3A: 0.038
GF-3B/4: 0.0696, p-value < 0.01
AST
Non-obese,
GF-1: -0.013
GF-2: 0.007
GF-3A : -0.015
GF-3B/4: -0.004
Obese,
GF-1: 0.011
GF-2: -0.013
D-74
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location, _ . Population, Exposure Matrix, „ ^ ^ .
„ ~, _7 Design x , / i x \a Outcome Comparison Select Results
Commence Years Ages, N Levels (ng/mL)a
GF-3A : 0.041, p-value = 0.01
GF-3B/4: 0.023
Confounding: Gender, race/ethnicity, smoking status, age, loglO(BMI), diabetes status, hypertension status, fasting time, poverty income ratio,
survey year, alcohol consumption0
Liu et al.
United States, Controlled Overweight and Plasma Hepatic fat mass
Partial Spearman Hepatic fat mass: 0.11
(2018,
2004-2007 trial Obese patients Males
correlation
4238396)
from the 27.2(19.9-45.2)
coefficient among
Medium
POUNDS-Lost, Females
baseline PFOS
Age 30-70 22.3 (14.3-34.9)
(ng/ml) and
study,
hepatic fat mass
N = 150
Confounding: age, sex, race, education, smoking status, alcohol consumption, physical activity, menopausal status (women only), hormone
replacement therapy (women only), and dietary intervention groups
Liu et al.
United States, Cross- Adults from Serum Levels of
Regression Albumin
(2018,
2013-2014 sectional NHANES, GM=5.28 albumin (g/dL)
coefficient per In- 0.04, SE = 0.01, p-value < 0.005
4238514)
Age >18, (SE = 1.02)
unit increase in
N = 1871
PFOS
Confounding: age, gender, ethnicity, smoking status, alcohol intake, household income, waist circumference, and medications (anti-
hypertensive, anti-hyperglycemic, and anti-hyperlipidemic agents)
Salihovic et al. Sweden
Cohort
Elderly adults in
Plasma
Levels of ALT
Regression 0.03 (0.02, 0.04), p-value < 0.0016
(2018, 2001-2014
Sweden,
Age 70
(|ikat/L)
coefficient per ln-
5083555)
Ages 70
13.2 (9.95, 17.8)
unit increase in
Medium
N = 1002
Age 75
PFOS
Ages 75
12.6 (7.97, 19.2)
N = 817
Age 80
Age 80
0.57 (5.36, 11.5)
N = 603
Confounding: Sex, LDL and HDL cholesterol, serum triglycerides, BMI, fasting glucose levels, statin use, smoking
D-75
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure Matrix,
Levels (ng/mL)a
Outcome
Comparison
Select Resultsb
Nian et al. China Cross- Adults in high Serum Levels of ALT
(2019, 2015-2016 sectional exposure area in 24.22(14.62-37.19) (ln-U/L), AST
5080307) China, (ln-U/L)
Medium Ages 22-96,
N = 1,605
Percent change per ALT
ln-unit increase in 4.1 (0.6, 7.7), p-value < 0.05
PFOS
AST
2.0 (-0.3, 4.3)
Confounding: Age, sex, career, income, education, drink, smoke, giblet, seafood consumption, exercise, BMI
Yamaguchi et
al. (2013,
2850970)
Medium
Japan
2008-2010
Cross-
sectional
Participants Blood
Levels of GGT
Spearman rank
GGT
from the 5.8(3.7-8.8)
(IU/L), AST
correlation
0.06, p-value = 0.120
"Survey on the
(IU/L), ALT
Accumulation of
(IU/L)
AST
Dioxins and
0.11, p-value = 0.010
Other Chemical
Compounds"
ALT
project from
0.12, p-value = 0.004
urban,
agricultural and
fishing areas,
Ages 15-76,
N = 590
Confounding: Age, sex, BMI, regional block, smoking habits, frequency of alcohol intake
Levels of ALT ALT, GGT, direct
(ln-IU/L), GGT bilirubin:
(ln-IU/L), Direct Regression
bilirubin (In- coefficient per ln-
mg/dL), ALT unit increase in
(IU/L, elevated) PFOS
Elevated ALT:
OR per ln-unit
increase in PFOS,
or by deciles
Gallo et al.
(2012,
1276142)
Medium
United States
2005-2006
Cross-
sectional
Adults from the
C8 Health
Project,
Ages >18 years,
N = 46, 452
Serum
20.3 (13.7-29.4)
ALT
0.02 (0.014,0.026), p-
value < 0.001
Direct bilirubin
0.029 (0.024, 0.034), p-
value < 0.001
ALT, elevated
Decile 2
Decile 3
Decile 4
Decile 5
Decile 6
Decile 7
Decile 8
1.01
1.06
1.11
1.19
1.19
1.20
1.24
(OR):
(0.87,
(0.91,
(0.96,
(1.04,
(1.04,
(1.04,
(1.08,
1.16)
1.22)
1.28)
1.37)
1.37)
1.38)
1.43)
D-76
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure Matrix,
Levels (ng/mL)a
Outcome
Comparison
Select Resultsb
Decile 9: 1.18 (1.02, 1.36)
Decile 10: 1.25 (1.08, 1.44)
p-trend <0.001
Per ln-unit increase:
1.13 (1.07, 1.18), p-value< 0.001
GGT: No statistically significant
associations
Results: Lowest decile used as the reference group
Confounding: Age, sex, alcohol consumption, socioeconomic status, fasting status, month of blood sample collection, smoking status, BMI,
physical activity, insulin resistance. Additional confounding for ALT, GGT, and direct bilirubin analyses: Race. Additional confounding for
OR analyses: increased serum iron.
Lin et al.
(2010,
1291111)
Medium
van den
Dungen et al.
(2017,
5080340)
Low
United States
1999-2000,
2003-2004
Cross-
sectional
Adults from
NHANES,
Ages >18 years,
N = 2,216
Serum
23.50 (15.50-33.80)
Levels of Regression Bilirubin
bilirubin (|iIVI). coefficient per log- Separate analysis: -0.30
GGT (log-U/I), unit increase in (SE = 0.24), p-value = 0.223
ALT (U/I) PFOS Composite analysis:-1.06
(SE = 0.27), p-value = 0.001
GGT
Separate analysis: 0.01
(SE = 0.03), p-value = 0.808
Composite analysis: -0.06
(SE = 0.03), p-value = 0.025
ALT
Separate analysis: 1.01
(SE = 0.53), p-value = 0.066
Composite analysis: -0.19
(SE = 0.63), p-value = 0.769
Comparison: Logarithm base not specified.
Confounding: Age, gender, race/ethnicity, smoking status, drinking status, education level, BMI, HOMA-IR, metabolic syndrome, iron
saturation status. Additional confounding for composite analyses: PFHxS exposure, PFNA exposure, PFOA exposure.
The
Netherlands
2015
Cross- Men with
sectional habitual eel
consumption,
Ages 40-70,
N = 37
Serum
40 ng/g wet weight
(15-93)
Levels of ALT, Standardized ALT
AST regression 0.01 (-0.32,0.34)
coefficient per unit
increase in PFOS AST
0.19 (-0.17, 0.55)
D-77
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location, _ . Population,
Years DeHgn AgeS,N
Exposure Matrix, „ ,
t i /• / m Outcome
Levels (ng/mL)a
Comparison
Select Resultsb
Confounding: Age, waist-to-hip ratio
Olsen et al.
(2003,
1290020)
Medium
United States,
Belgium
1994-2000
Cross-
sectional
Current and
former workers
at two
fluorochemical
production
plants
Male
N = 421,
Female
N = 97,
Regression
analysis
N = 174
Serum
Antwerp Mean
(SD) = 0.96 ppm
(0.97);
Decatur = 1.40 ppm
(1.15)
Levels of ALT
(IU/L),
(IU/L),
(IU/L),
(IU/L)
ALP
AST
GGT
Comparison of
mean outcome by
PFOS quartile
Males
Elevated (p < 0.05) ALT for
employees in Q4 compared to Q1
Elevated (p < 0.05) ALP for
employees in Q3 and Q4 compared
to Q1
No significant differences in mean
AST or GGT by PFOS exposure
quartile
Females
Elevated (p < 0.05) ALP for
employees in Q4 compared to Q1
and Q2, and in Q3 compared to Q2
Elevated (p < 0.05) GGT for
employees in Q4 compared to Q1
No significant differences in mean
ALT or AST by PFOS exposure
quartile
Confounding: Sex
Olsen et al.
(2001,
10228462)
Medium
United States, Cohort
Belgium
1994-2000
Male 3M
fluorochemical
plant workers in
Antwerp,
Belgium and
Decatur,
Alabama
N = 175
Antwerp (2000) Levels of ALT Regression
Mean (SD): 1.16 (ln-IU/L), ALP
ppm (1.07); Decatur (ln-IU/L), AST
(2000): 1.67 ppm (ln-IU/L), GGT
(1.39) (ln-IU/L)
coefficient per unit
increase in PFOS
ALT
0.010 (SE = 0.016), p-value = 0.54
PFOS x Years of observation
interaction p-value <0.001
AST
0.010 (SE = 0.011), p-value = 0.39
PFOS x Years of observation
interaction p-value = 0.79
D-78
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location, _ . Population, Exposure Matrix, „ ^ ^ .
„ ~, _7 Design x , / i x \a Outcome Comparison Select Results
Commence Years Ages, N Levels (ng/mL)a
ALP
0.002 (SE = 0.009), p-value = 0.87
PFOS x Years of observation
interaction p-value = 0.47
GGT
-0.004 (SE = 0.020), p-
value < 0.001
PFOS x Years of observation
interaction p-value = 0.42
Confounding: Years of observation, PFOS x Years of observation, age, BMI, drinks/day, cigarettes/day, location, entry period, baseline years
worked, triglycerides
Olsen et al.
United States Cohort 3M Serum Levels of ALT
Regression
ALT
(2012,
2008-2010 fluorochemical Mean change from (IU/L), AST
coefficient per unit
-0.045 (SD = 0.015),
2919185)
plant employees baseline, (IU/L)
increase in PFOS
p-value = 0.005
Low
and contractors, Employees:
N = 179 -101.3 ng/mL;
AST
Contractors: 1
-0.007 (SD = 0.009)
Confounding: Sex, age at baseline, BMI at baseline, alcohol consumption at baseline
Rantakokko et
Finland Cross- Morbidly obese Serum Lobular
OR per log unit
<2 foci: 0.52 (0.13,2.09)
al. (2015,
2005-2011 sectional adults 3.2 (5th-95th inflammation
increase in PFOS
2-4 foci: 0.14 (0.01, 1.66)
3351439)
undergoing percentile: 0.89,
by level of lobular
Medium
bariatric 10.3)
inflammation
surgery,
N = 160
Comparison: Logarithm base not specified.
Results: No foci used as the reference group. Foci measured per 200x field.
Confounding: Age, sex, BMI, serum lipids, fasting insulin
Children and Adolescents
Gleason et al.
United States Cross- Adolescents Serum Levels of ALT
Regression
ALT
(2015,
2007-2010 sectional fromNHANES, 11.3(7.0-18.0) (ln-U/L), GGT
coefficient per ln-
(0.013) (-0.009, 0.034)
2966740)
Ages > 12, (ln-U/L), AST
unit increase in
Medium
N = 4,333 (ln-U/L), ALP
PFOS
GGT
(ln-U/L)
0.036 (0.001, 0.071)
AST
D-79
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location, _ . Population, Exposure Matrix, „ ^ ^ .
„ ~, _7 Design x , / i x \a Outcome Comparison Select Results
Commence Years Ages, N Levels (ng/mL)a
0.004 (-0.010,0.018)
ALP
-0.010 (-0.027, 0.007)
Confounding: Age, gender, race/ethnicity, BMI group, smoking, alcohol consumption "if statistically significant associated with both the
exposure and outcome in univariate analysis."
Mora et al.
Cohort
Children from
Plasma
Levels of ALT
Regression
Prenatal exposure: -0.4 (-1.1, 0.2)
(2018,
Project VIVA,
Prenatal exposure:
(U/L)
coefficient per
Mid-childhood exposure: -0.3
4239224)
United States
1999-2010
N, prenatal
24.6 (17.9-34.0)
IQR increase in
(-0.9, 0.2)
Medium
exposure = 508,
N, mid-
childhood
exposure = 630
Mid-childhood
exposure:
6.2 (4.2-9.7)
PFOS
Confounding: Maternal education, prenatal smoking, gestational age at blood draw, and child's sex, race/ethnicity, age at lipids/ALT
measurements
Attanasio
United States
Cross-
Adolescents
Serum
Levels of ALT
Regression
ALT
(2019,
2013-2016
sectional
from NHANES,
Boys:
(ln-IU/L), AST
coefficient per ln-
Boys,
5412069)
Ages 12-19,
GM = 3.68
(ln-IU/L)
unit increase in
(-0.09,0.10)
Medium
N, boys = 354
N, girls = 305
(SE = 0.12)
Girls:
GM = 2.76
(SE = 0.14)
PFOS or by
quartiles
Q2: -0.05 (-0.21,0.11)
Q3: 0.07 (-0.05,0.18)
Q4: -0.01 (-0.14,0.13)
Girls,
0.09 (-0.01,0.18)
Q2: -0.02 (-0.17,0.14)
Q3: 0.01 (-0.11, 0.13)
Q4: 0.11 (-0.02, 0.24)
AST
Boys,
-0.02 (-0.11,0.06)
Q2: -0.02 (-0.11,0.08)
Q3: 0.01 (-0.07, 0.10)
Q4: -0.01 (-0.12,0.10)
Girls,
0.07 (0.00, 0.013)
Q2: 0.03 (-0.08,0.14)
Q3: 0.05 (-0.04,0.13)
D-80
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure Matrix,
Levels (ng/mL)a
Outcome
Comparison
Select Resultsb
Q4: 0.12 (0.03, 0.21),
p-value = 0.01
Results: Lowest quartile used as the reference group.
Confounding: Age, race/ethnicity, body weight status, education, poverty income ratio, exposure to smoking
Khalil et al.
(2018,
4238547)
Low
United States Cross- Obese children, Serum Levels of ALT Regression ALT
2016 sectional Ages 8-12, 2.79 (IQR = 2.10) (U/L), AST coefficient per unit 0.16 (-1.84, 2.15)
N = 48 (U/L) increase in PFOS
AST
-0.28 (-1.22, 0.65)
Confounding: Age, sex, race
Children and Adolescents - Other hepatic outcomes
Jin et al. (2020,
6315720)
Medium
United States
2007-2015
Cross-
sectional
Children and
adolescents
diagnosed with
nonalcoholic
fatty liver
disease,
Ages 7-19,
N = 74
Plasma
3.59 (2.35-6.81)
Ballooning,
Grade of
steatosis,
Liver fibrosis,
Lobular
inflammation,
Nonalcoholic
steatohepatitis,
Portal
inflammation
OR per IQR
increase in PFOS
Ballooning
Few balloon cells: 1.11 (0.52,
2.37)
Many cells/prominent ballooning:
1.12(0.26, 4.95)
Grade of steatosis
34-66% steatosis: 1.37 (0.54,
3.51)
> 66% steatosis: 0.88 (0.39, 1.97)
Liver fibrosis
Mild (stage 1): 1.71 (0.73, 4.03)
Significant (stages 2-4): 1.51
(0.53,4.35)
Lobular inflammation
<2 foci: 0.50 (0.21, 1.22)
2-4 foci: 2.92 (0.92, 9.23)
Nonalcoholic steatohepatitis
3.32 (1.40, 7.87), p-value < 0.05
Portal inflammation
D-81
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure Matrix,
Levels (ng/mL)a
Outcome
Comparison
Select Resultsb
Mild: 1.85 (0.82,4.21)
Moderate-to-severe: 2.26 (0.75,
6.79)
Results: For ballooning, none was used as the reference group. For grade of steatosis < 5-33% was used as the reference group. For liver
fibrosis, none was used as the reference group. For lobular inflammation, no foci used as the reference group. Foci measured per 200x field.
For portal inflammation, none was used as the reference group.
Confounding: Age, sex, ethnicity, and BMI z-score
Notes: ALT = alanine aminotransferase; AST = aspartate aminotransferase; BMI = body mass index; GF = glomerular filtration; GGT = y-glutamyltransferase; GM = geometric
mean; HDL = high-density lipoprotein; HOMA-IR = homeostasis model assessment of insulin resistance; HR = hazard ratio; IQR = interquartile range; LDL = low-density
lipoprotein; NHANES = National Health and Nutrition Examination Survey; OR = odds ratio; Q1 = quartile 1; Q2 = quartile 2; Q3 = quartile 3; Q4 = quartile 4; SD = standard
deviation; SE = standard error; SMR = standardized mortality ratio; T1 = tertile 1; T2 = tertile 2; T3 = tertile 3.
a Exposure levels reported as median (25th-75th percentile) unless otherwise noted.
b Results reported as effect estimate (95% confidence interval) unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
D.4 Immune
Table D-7. Associations between PFOS Exposure and Vaccine Response in Recent Epidemiological Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Children
Grandjean et al. Faroe Islands,
(2012, 1248827) Denmark
Medium Recruitment
1997-2000,
Follow-up
through 2008
Cohort
Children
Maternal serum
Antibody
Percent change
Child serum
followed from
(prenatal)
concentrations
per doubling in
Anti-diphtheria, prebooster, age 5
birth to age 7
Geometric
(log-IU/mL) for
age 5 and
-16 (-34.9, 8.3)
Birth and
mean = 27.3
tetanus and
maternal PFOS
Anti-diphtheria, postbooster, age 5
infancy:
(23.2-33.1)
diphtheria
-15.5 (-31.5,4.3)
N = 587
Anti-diphtheria, age 7
Prebooster
Child serum
-27.6 (-45.8, -3.3)
(mean age 5.0)
(5 years)
Anti-diphtheria, age 7 adjusted for
examination:
Geometric
age 5 Ab
N = 532
mean = 16.7
-20.6 (-38.2, 2.1)
Postbooster
(13.5-21.1)
(mean age 5.2)
Maternal serum
examination:
Anti-diphtheria, prebooster, age 5
N = 456
-38.6 (-54.7, -16.9)
D-82
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Age 7 (mean Anti-diphtheria, postbooster, age 5
age 7.5) -20.6 (-37.5, 0.9)
examination: Anti-diphtheria, age 7
N = 464 -19.7 (-41.8, 10.7)
Anti-diphtheria, age 7 adjusted for
age 5 Ab
-10 (-32.6, 20)
Child serum
Anti-tetanus, prebooster, age 5
-11.9 (-30, 10.9)
Anti-tetanus, postbooster, age 5
-28.5 (-45.5,6.1)
Anti-tetanus, age 7
-23.8 (-44.3, 4.2)
Anti-tetanus, age 7 adjusted for age
5 Ab
-11.4 (-30.5, 12.8)
Maternal serum
Anti-tetanus, prebooster, age 5
-10.1 (-31.9, 18.7)
Anti-tetanus, postbooster, age 5
-2.3 (-28.6, 33.6)
Anti-tetanus, age 7
35.3 (-3.9, 90.6)
Anti-tetanus, age 7 adjusted for age
5 Ab
33.1 (1.5,74.6)
Confounding: Age, sex. Additional confounding for postbooster analyses: time since vaccination, booster type. Additional confounding
for year 7 analyses: booster type. Additional confounding for year 7 analyses adjusted for age 5 Ab: booster type, child's specific antibody
concentration at age 5 years
Granum et al. Norway
(2013, 1937228) 1999-2008
Medium
Cohort Mother-infant Maternal serum Levels (OD) of Regression Rubella antibody
pairs from with three days rubella anti- coefficient per -0.08 (-0.14,-0.02)
MoBa at 3-year of delivery vaccine unit increase p-value = 0.007
follow-up 5.5(3.8-7.1) antibodies PFOS
D-83
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
N = 56
Confounding: maternal allergy, paternal allergy, maternal education,
child's gender, and/or age at 3-year follow-up.
Mogensen et al.
(2015, 3981889)
Medium
Faroe Islands, Cohort Children aged Serum
Denmark 5-7 years 15.5 (12.8-19.2)
2002-2007 N = 443 at age 7
Antibody
concentrations
(log2-IU/mL)
for diphtheria or
tetanus
Percent change
per doubling of
PFOS
Anti-diphtheria, age 7
-30.3 (-47.3, -7.8)
Anti-tetanus, age 7
-9.1 (-32.8, 23)
Confounding: Age, sex, booster type0
Grandjean et al. Faroe Islands,
Cohort and
Children
Serum
Levels of
Percent change
Diphtheria antibody
(2017, 3858518) Denmark
cross-sectional
followed up at
13 years: 6.7
diphtheria
per doubling of
Age 7:-23.8 (-43.2, 2.3)
Medium Enrollment:
7 years and
(5.2-8.5)
antibody (log2-
PFOS
p-value = 0.07
1997-2000
13 years
7 years: 15.3
IU/mL), tetanus
Age 13:-8.6 (-27.7, 15.6)
(12.4-19.0)
antibody (log2-
p-value = 0.454
N = 505
IU/mL)
(13 years)
Tetanus antibody
N = 427
Age 7: 30 (-16.1, 101.4)
(7 years)
p-value = 0.24
Age 13:22.2 (-12.4, 70.3)
p-value = 0.237
Confounding: Sex, age at antibody assessment, booster type at age 5
Grandjean et al. Faroe Islands,
Cross-sectional
Infants 2 weeks
Serum
Levels of
Percent change
2007-2009 cohort
(2017, 4239492) Denmark
after expected
18 months: 7.1
tetanus antibody per doubling of
Tetanus antibody
Medium 1997-2000 and
term date,
(4.5-10.0)
(IU/mL),
PFOS
Birth: -10.84 (-28.34, 10.94)
2007-2009
followed up at
5 years: 4.7
diphtheria
p-value = 0.3
(year of birth)
18 months and
(3.5-6.3)
antibody
18 mo:-7.027 (-21.63, 10.3)
5 years
(IU/mL)
p-value = 0.4
All: N = 490,
18 months:
N = 275,
5 years: N = 349
5 yr:-9.076 (-28.1,
p-value = 0.43
14.98)
Diphtheria antibody:
Birth: -14 (-31.59, 8.11)
p-value = 0.20
18 mo: 17.55 (-0.84, 39.34)
p-value = 0.062
5 yr: 17.17 (-8.66, 50.31)
D-84
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
p-value = 0.21
Combined cohort
Tetanus antibody
Birth: -10.55 (-24.63,6.16
p-value = 0.2
18 mo: -7.08 (-21.29, 9.70)
p-value = 0.39
5 yr: -10.52 (-24, 5.35)
p-value = 0.18
Diphtheria antibody
Birth: -24.47 (-36.90, -9.60)
p-value = 0.002
18 mo: 15.07 (-2.49, 35.79)
p-value = 0.096
5 yr:-1.34 (-17.05, 17.34)
p-value = 0.88
Confounding: Age, sex
Abraham et al. Berlin,
Cross-sectional Children, 1 year
Plasma
Levels of Hib
Spearman
Hib antibody: -0.05
(2020,6506041) Germany
old
antibody,
correlation
Medium Enrollment:
Formula fed:
tetanus antibody
coefficient
Tetanus antibody IgG: -0.02
1997-1999
All: N = 101,
mean = 6.8
IgG,
formula fed:
(range = 2.8-
tetanus antibody
Tetanus antibody IgGl: -0.07
N = 21,
19.3)
IgGl,
breastfed:
diphtheria
Diphtheria antibody: -0.02
N = 80
Breastfed:
mean = 15.2
(range =1.9-
34.8)
antibody
Confounding: Time since last vaccination
Timmermann et Guinea-Bissau Cohort Infants enrolled Maternal blood Measles Percent
al. (2020, 2012-2015 at 4-7 months 0.77 (0.53-1.02) antibody difference per
6833710) old (inclusion), concentration doubling of
Medium followed up at (mlU/mL) PFOS
Inclusion (no measles vaccination):
-13 (-26, 4)
9-month visit
D-85
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
9 months and
Control (no measles vaccination):
2 years
-27 (-44, -4)
Intervention (1 measles
Inclusion:
vaccination): -21 (-37, -2)
N = 236
9-month
2-year visit
Unvaccinated
Control (1 measles vaccination): -6
controls:
(-25, 18)
o
o
II
z;
Intervention (2 measles
Intervention:
vaccinations): -3 (-20, 17)
II
£
2-year
Unvaccinated
controls:
o
o
II
z;
Intervention:
N = 91
Confounding: Weight and age at inclusion, maternal education, breastfeeding without solids, maternal measles antibody concentration, sex,
and time from vaccination to blood sampling
Timmerman et Greenland
al. (2022,
9416315)
Medium
Recruitment:
1999-2005,
Examination:
2012-2015
Cohort and Vaccinated
cross-sectional children ages 7- from pregnancy
12 years and 19.16(15.20-
their mothers at 24.06)
pregnancy
Maternal serum Levels (IU/mL) Percent
of diphtheria
and tetanus
antibody
Maternal serum
N = 57
Child serum
N = 169
Child serum
8.68 (6.52-
12.23)
difference per
unit increase in
PFOS
OR per log 10-
unit increase in
PFOS
Diphtheria antibody
Child serum
Percent difference: 9 (-16, -2)
OR: 1.14 (1.04, 1.26)
Maternal serum
Percent difference: 1 (-4, 6)
Tetanus antibody
Child serum
Percent difference: -3 (-8, 3)
Maternal serum
Percent difference: 2 (-3, 6)
Confounding: Area of residence (Nuuk, Maniitsoq, Sisimiut, Ilulissat, Aasiaat, Qeqertarsuaq, Tasiilaq). Additional confounding for percent
difference analyses: duration of being breastfed (< 6 months, 12 months, >1 year). Additional confounding for child serum analyses: time
since vaccine booster (only children with known vaccination date were included).
Zeng et al. China
(2019, 5081554) 2013
Cohort
Infants from Cord blood HFMD antibody Percent change CA16
Guangzhou 3.17 (1.88-4.94) titers (CA16 or or OR (below Cord blood: -20.6 (-30.0, -9.9)
D-86
-------
DRAFT FOR PUBLIC COMMENT
Reference, Location, . Population,
Confidence Years esign Ages, N
Low Birth Cohort
Study at birth
and 3 months
Birth N= 194
(91 girls, 103
boys)
3-month
N = 180 (89
girls, 91 boys)
MARCH 2023
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
EV71) in serum clinical
of cord blood or protection) per
at 3 months doubling of
PFOS
Resultsb
Girls: -14.0 (-27.5, 1.9)
Boys:-24.7 (-37.6, -9.1)
3 months: -6.9 (-13.9,0.7)
Girls: -2.8 (-10.9, 6.2)
Boys: -12.2 (-23.7, 1.1)
CA16 below clinical protection
Cord blood: 1.75 (1.16,2.63);
p-value = 0.007
Girls: 1/43 (0.80, 2.56)
Boys: 1.98 (1.03, 3.81)
p-value for interaction by
sex = 0.311
3 months: 1.71 (1.12, 2.60);
p-value = 0.013
Girls: 0.97 (0.88, 1.08)
Boys: 2.29 (1.20, 4.36)
p-value for interaction by
sex = 0.318
EV71
Cord blood: -23.6 (-33.9,-11.8)
Girls: -23.5 (-37.9, -5.8)
Boys: -23.4 (-37.2, -6.6)
3 months: -10.6 (-16.9, -3.9)
Girls: -8.6 (-17.1,0.9)
Boys: -12.2 (-21.3, -1.9)
EV71 below clinical protection
Cord blood: 1.66(1.12,2.45);
p-value = 0.011
Girls: 1.48 (0.92,2.37)
Boys: 2.01 (1.03, 3.90)
p-value for interaction by
sex = 0.265
D-87
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
3 months: 2.25 (1.44, 3.51);
p-value < 0.05
Girls: 2.05 (1.11,3.79)
Boys: 2.35 (1.19,4.65)
p-value for interaction by
sex = 0.579
Outcome: Clinical protection threshold defined as titers > 1:8 in modified cytopathogenic effect assay.
Confounding: Sex, age, parental education, parental occupation, family income, parity, and birth weight
Adults and Adolescents
Looker et al. United States
(2013, 2850913) Baseline:
Medium 2005-2006,
Follow-up:
2010
Cohort
Adults near
water districts
of Ohio and
West Virginia
with
contaminated
drinking water
N = 403
Serum
GM (95%
CI) = 8.32
(7.65-9.05)
Influenza
antibodies (titer
ratio and titer
rise, loglO-
transformed):
A/H1N1,
A/H3N2, type B
Regression Influenza type B titer rise
coefficient per Per loglO-unit: 0.5 (-0.11, 0.21), p-
loglO-unit value = 0.56
increase, or by Q2: 0.02 (-0.13, 0.18), p-
quartiles value = 0.76
Q3: -0.03 (-0.19,0.14), p-
value = 0.73
Q4: 0.04 (-0.14, 0.21), p-
value = 0.68
Influenza type B titer ratio
Per loglO-unit: 0.05 (-0.09, 0.18),
p-value = 0.52
Q2: 0.004 (-0.14, 0.14), p-
value = 0.96
Q3:-0.02 (-0.16, 0.12), p-
value = 0.78
Q4: 0.03 (-0.12,0.18), p-
value = 0.71
Influenza A/H3N2 titer rise
Per loglO-unit: 0.09 (-0.13, 0.32),
p-value = 0.42
Q2: 0.03 (-0.19,0.26), p-
value = 0.78
Q3: 0.18 (-0.06, 0.41), p-
value = 0.14
D-88
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Q4:-0.04 (-0.28, 0.21), p-
value = 0.77
Influenza A/H3N2 titer ratio
Per loglO-unit: -0.005 (-0.20,
0.19), p-value = 0.96
Q2:-0.06 (-0.26, 0.14), p-
value = 0.56
Q3: 0.02 (-0.18, 0.23), p-
value = 0.84
Q4: -0.03 (-0.24,0.19), p-
value = 0.82
Results: Lowest quartile used as reference group
Confounding: Age (cubic spline), gender, mobility, and history of previous influenza vaccination
Influenza A/H1N1 titer rise
Per loglO-unit: 0.15 (-0.02, 0.32),
p-value = 0.08
Q2:-0.04 (-0.21, 0.14), p-
value = 0.68
Q3: 0.13 (-0.04,0.31), p-
value = 0.14
Q4: 0.10 (-0.09, 0.29), p-
value = 0.30
Influenza A/ H1N1 titer ratio
Per loglO-unit: 0.10 (-0.11, 0.3), p-
value = 0.36
Q2:-0.07 (-0.28, 0.13), p-
value = 0.47
Q3: 0.03 (-0.18,0.24), p-
value = 0.78
Q4: 0.03 (-0.19,0.26, p-
value = 0.77
Pilkerton et al. United States
(2018, 5080265) 1999-2000
Medium for
youth
Cross-sectional
Adults and
adolescents
12 years and
older
Serum
Rubella IgA
titers (log-IU)
Regression
coefficient by
quartiles or per
quartile increase
Adolescents:
Per quartile increase:
F-value = 1.44, p-value = 0.251
D-89
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Low for adult Women: Adults:
Youths: mean = 22.1, Per quartile increase: F-
N = 1,012 SE = 0.9 value = 3.44, p-value = 0.030
Women
Adults: Men: Q2:0.05 (-0.34,0.43)
N = 542 mean = 28.1 p-value = 0.81
women, 613 SE = 1.3 Q3: 0.04 (-0.51, 0.6)
men p-value = 0.87
Q4:-0.17 (-1.13, 0.8)
p-value = 0.73
Men
Q2: -0.20 (-0.62, 0.23)
p-value = 0.35
Q3:-0.32 (-0.69, 0.05)
p-value = 0.08
Q4: 0.01 (-0.54, 0.56)
p-value = 0.97
Outcome: Logarithm base not reported
Results: Lowest quartile used as reference group
Confounding: Women: age, ethnicity, BMI, educational level, number of live births; men: age, ethnicity, BMI, educational level
Bulka et al. Unites States
(2021, 7410156) 1999-2000,
Medium 2003-2016
Cross-sectional
NHANES
Serum
Persistent
Persistent
Cytomegalovirus
adolescents and
12-19 years:
infections of
infections:
12-19 years: 0.92 (0.77,
1.09), p-
adults aged 12-
GM (SE) = 7.54
cytomegalovirus
Prevalence ratio
value = 0.36
49 years
(0.26)
, Epstein-Barr
per doubling in
20-49 years: 0.99 (0.92,
1.05), p-
12-19 years:
20-49 years:
virus, hepatitis
PFOS
value = 0.70
N= 3,189
GM (SE) = 8.67
C, hepatitis E,
20-49 years:
(0.24)
herpes simplex
Pathogen
Epstein-Barr virus
N= 5,589
virus 1, herpes
burden: Relative
12-19 years: 1.01 (0.96,
1.05), p-
simplex virus 2,
difference per
value = 0.74
Toxoplasma
log2-unit
gondii, and
increase in
Hepatitis C virus
Toxocara
PFOS
20-49 years: 0.96 (0.71,
1.29), p-
species;
value = 0.77
pathogen burden
Hepatitis E virus
D-90
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
20-49 years: 1.00 (0.83, 1.20), p-
value = 0.99
Herpes simplex virus 1
12-19 years: 1.05 (0.99, 1.11), p-
value = 0.13
20-49 years: 1.04(1.01, 1.06), p-
value < 0.01
Herpes simplex virus 2
20-49 years: 1.04 (0.99, 1.09), p-
value = 0.1
Toxoplasma gondii
12-19 years: 1.15 (0.90, 1.48), p-
value = 0.27
20-49 years: 1.1 (0.97, 1.26), p-
value = 0.15
Toxocara species
12-19 years: 1.12(0.66, 1.91), p-
value = 0.68
20-49 years: 1.57(1.26, 1.96), p-
value < 0.01
Pathogen burden
12-19 years: 1.30 (1.25, 1.36)
20-49years:l.10 (1.07, 1.12)
Outcome: Pathogen burden defined as the sum of pathogens for which an individual was seropositive (including any pathogens with a
seroprevalence < 1.0%)
Confounding: Age, race/ethnicity, sex, ratio of family income to the federal poverty threshold, educational attainment, serum cotinine
concentrations, and BMI
Lopez-Espinosa United States
etal. (2021, 2005-2006,
7751049) 2010
Medium
Cohort and Adults from
cross-sectional C8HP
2005-2006:
N = 42,782
Serum Levels (ln-
2005-2006: cells/^L or
19.7(13.3-28.4) percentage of
white blood
Counts: Precent
difference per
IQR increase in
PFOS
White blood cells, total
2005-2006: -0.55 (-0.84, -0.26)
2010:0.55 (-1.35,2.49)
D-91
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
2010: N = 526
2010: 9.60
(6.10-14.9)
Percentages:
Difference per
IQR increase in
PFOS
Likelihood ratio test p-
value < 0.001 for the comparison
between the two time periods
cells/lymphocyt
es) of white
blood cells,
neutrophils,
monocytes,
eosinophils,
lymphocytes,
CD3+ T cells,
CD3+CD4+ T-
helper cells,
CD3+CD4+CD
8+ double
positive T cells,
CD3+CD8+ T-
cytotoxic cells,
CD3-
CD16+CD56+
natural killer
cells, CD3-
CD19+B cells;
CD4+/CD8+
ratio
Outcome: All cell types reported as cell counts; eosinophils, lymphocytes, monocytes, and neutrophils additionally reported as percentage of
white blood cells; CD3+ T cells, CD3+CD4+ T-helper cells, CD3+CD4+CD8+ double positive T cells, CD3+CD8+ T-cytotoxic cells, CD3-
CD16+CD56+ natural killer cells, and CD3-CD19+ B cells additionally reported as percentage of lymphocytes
Confounding: Gender, age, smoking, month of sampling, alcohol intake, and educational level
Shih et al. Faroe Islands,
Cohort and Faroe Island
Cord blood at
Levels (IU/mL)
Percent change
Hepatitis Type B
(2021, 9959487) Denmark
cross-sectional residents at
birth
of hepatitis A
per log2-unit
Cord blood: -23.24 (-46.77, 10.69)
Medium Recruitment:
birth, 7, 14, 22,
5.96
antibody,
increase in
7-year serum: -4.65 (-45.87,
1986-1987,
and 28 years
(IQR =3.09)
hepatitis B
PFOS
67.87)
Follow-up
N = 399
antibody,
14-year serum: 22.17 (-34.09,
through 2015
Serum
diphtheria
126.46)
7 year: 31.89
antibody,
22-year serum: 15.26 (-22.88,
(IQR = 13.37)
tetanus
72.26)
14 year: 31.29
antibody;
28-year serum: 6.12 (-23.36,
(IQR = 9.62)
Hepatitis A
46.93)
D-92
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Confounding: Sex
22 year: 12.55
(IQR = 7.24)
28 year: 6.85
(IQR =5.29)
antibody signal-
to-cutoff ratio
Hepatitis Type A
Cord blood: 0.11 (-0.36, 0.59)
7-year serum: 0.21 (-0.54, 0.96)
14-year serum: -0.14 (-1.01, 0.74)
22-year serum: -0.1 (-0.63, 0.44)
28-year serum: -0.23 (-0.66, 0.21)
Diphtheria
Cord blood: 28.26 (-5.7, 74.44)
7-year serum: 5.04 (-36.45, 73.59)
14-year serum: -3.5 (-42.87,
63.01)
22-year serum: 5.29 (-21.69,
41.56)
28-year serum: 6.91 (-14.26,
33.31)
Tetanus
Cord blood: 2 (-20.24, 30.44)
7-year serum: 8.91 (-25.85, 59.95)
14-year serum: -19.44 (-48.36,
54.7)
22-year serum: -9.1 (-28.42,
15.44)
28-year serum: -2.1 (-17.77,
16.56)
Stein et al. United States Cohort
(2016, 3860111) 2010
Low
Adults enrolled Serum
at 18-49 years,
followed up at
day 30
Anti-A-HINI RRbytertiles HAI anti-A-HlNl antibody
GM = 5.22
(95% CI: 4.52-
6.02)
Total
population:
N = 75, low
antibody
response
measured by
HAI or by IHC
Total population
T2: 2.6 (0.4, 15.1)
T3: 1.3 (0.2,7.3)
p-value for trend = 0.81
Low baseline Ab
T2: 6.7 (1.2, 37.9)
T3: 1.6 (0.3, 9.7)
D-93
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
baseline Ab:
p-value for trend = 0.81
N = 29
IHC anti-A-HlNl antibody
Total population
T2: 2.6 (0.9, 7.4)
T3: 2.4 (0.9,6.6)
p-value for trend = 0.12
Low baseline Ab
T2: 4.5 (1,20.3)
T3: 3.1 (1, 10.2)
p-value for trend = 0.13
Results: Lowest tertile used as the reference group.
Confounding: Age, sex, and race/ethnicity
Zeng et al. China
(2020,6315718) 2015-2016
Low
Cross-sectional
Adults from the
Isomers of C8
Health Project
N = 605
Serum Hepatitis B Regression
10.7 (6.82-16.2) surface antibody coefficient or
(HBsAb) (log-
mlU/mL) or
surface antigen
(HBsAg) (mlU-
mL); HBsAb
seronegative
(< 10 mlU/mL)
OR (HBsAb
seronegative)
per loglO-unit
increase in
linear or
branched PFOS
HBsAb concentration
Linear: -0.51 (-0.84, -0.18);
p-value = 0.002
Branched: -0.31 (-0.7, 0.07);
p-value = 0.114
HBsAb seronegative
Linear: 1.96 (1.37, 2.81);
p-value < 0.001
Branched: 1.64 (1.05, 2.56);
p-value = 0.03
HBsAg concentration
Linear: 0.74 (-0.02, 1.49);
p-value = 0.056
Branched: 1.08 (0.06, 2.09);
p-value = 0.037
Confounding: Age, gender, BMI, career, income, alcohol drinking, smoking, regular exercise; education for HBsAb concentration alone
Notes: Ab = antibody; C8HP = C8 Health Project; CI = confidence interval; GM = geometric mean; BMI = body mass index; HAI = hemagglutinin inhibition;
ICH = immunohistochemistry; HFMD = hand, foot, and mouth disease; MoBa = Norwegian Mother and Child Cohort Study; OD = optical density; Q2 = quartile 2; Q3 = quartile
3; Q4 = quartile 4; RR = risk ratio; SE = standard error; T2 = tertile 2; T3 = tertile 3.
a Exposure levels reported as median (25th-75th percentile) unless otherwise noted.
b Results reported as effect estimate (95% confidence interval) unless otherwise noted.
D-94
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
: Confounding indicates factors the models presented adjusted for.
Table D-8. Associations between PFOS Exposure and Infectious Disease in Recent Epidemiological Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Children
Fei et al. (2010;
1290805)
Medium
Denmark,
Recruitment:
1996-2003;
Follow up: 2008
Cross-sectional
and cohort
Mother infant
pairs with
follow-up to
11 years
(DNBC)
N = 1,400
Maternal plasma Infectious
Mean
(range) = 35.3
(6.4-106.7)
disease
hospitalizations
IRR by quartiles Girls
or per quartile Q2: 1.14 (0.73, 1.79)
increase in Q3: 1.61 (1.05, 2.47)
PFOS Q4: 1.59(1.02,2.49)
Per quartile increase: 1.18 (1.03,
1.36)
Boys
Q2: 0.8 (0.57, 1.13)
Q3: 0.61 (0.42,0.89)
Q4: 0.77 (0.54, 1.12)
Per quartile increase: 0.90 (0.80,
1.02)
All children
Q2: 0.93 (0.71, 1.21)
Q3: 0.90 (0.68, 1.18)
Q4: 1.0 (0.76, 1.32)
Per quartile increase: 1.0 (0.91,
1.09)
Results stratified by age not
statistically significant
Results: Lowest quartile used as reference group
Confounding: Parity, maternal age, pre-pregnancy BMI, breastfeeding, smoking during pregnancy, socio-occupational status, home density,
child's age, sibling age difference, gestational age at blood drawing, birth year, and birth season
Confounding: Maternal age, maternal allergic history, distance from home to highway, parity, birth season, and blood sampling period
Gourdazi et al. Hokkaido, Cohort Children, early Maternal blood Infectious OR by quartiles Girls
(2017, 3859808) Japan pregnancy 4.93 (3.67-6.65) diseases, total Q2: 1.42 (0.91,2.23)
Medium 2003-2009 followed up at (including Otitis Q3: 1.32 (0.86,2.06)
4 years media, Q4: 1.71 (1.08, 2.72)
D-95
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
N = ,1558 (793
boys, 765 girls)
Pneumonia, RS
vims, Varicella)
p-value for trend = 0.036
Boys
Q2: 1.45 (0.95, 2.22)
Q3: 1.25 (0.83, 1.91)
Q4: 1.59 (1.03,2.46)
p-value for trend = 0.071
All
Q2: 1.44 (1.06, 1.96)
Q3: 1.28 (0.95, 1.73)
Q4: 1.61(1.18,2.21)
p-value for trend = 0.008
Results: Lowest quartile used as reference group.
Confounding: Maternal age, maternal educational level, number of elder siblings, child sex, breast-feeding period, and smoking during
pregnancy0
Manzano- Spain, Cohort Children ages Maternal blood LRTI
Salgado et al. 2003-2008 1.5, 4, or 7 years 6.06 (4.25-7.82)
(2019, 5412076) Age 1.5:
Medium N= 1,188
Age 4:
N = 1,184
Age 7:
N = 1,071
Confounding: OR assessment: Age-at-follow-up of the child; RR assessment: Maternal
pregnancy BMI, region of residence, and country of birth
OR or RR per
log2-unit
increase in
PFOS
OR
1.5 years: 0.99 (0.83, 1.18)
4 years: 0.95 (0.79, 1.16)
7 years: 0.83 (0.57, 1.2)
RR, 1.5-7 years
All: 0.96 (0.85, 1.09)
Boys: 0.97 (0.81, 1.15)
Girls: 0.94 (0.77, 1.14)
age at delivery, parity, previous breastfeeding, pre-
Ait Bamai et al.
(2020, 6833636)
Medium
Hokkaido, Cohort Children, early Maternal blood Chicken pox,
Japan pregnancy 5.12 (3.75-7.02) RSV, otitis
Enrollment: followed up at media,
2003-2012 7 years pneumonia,
wheeze, eczema
N = 2,689
OR or RR per Pneumonia: OR: 1.14 (0.93, 1.38);
ln-unit increase p-value = 0.21
in PFOS
Otitis media: OR: 1 (0.83, 1.2);
p-value = 0.989
Chicken pox: OR: 1.1 (0.91, 1.32);
p-value = 0.348
D-96
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
RSV: OR: 0.72 (0.56,0.91);
p-value = 0.007
Wheeze: RR: 0.93 (0.82, 1.06);
p-value = 0.255
Eczema: RR: 0.86 (0.76, 0.98);
p-value = 0.02
Confounding: Sex, maternal age, parity, maternal smoking during pregnancy, BMI pre-pregnancy, annual household income during
pregnancy, duration nursing, and presence of siblings
Huang et al. China
(2020, 6988475) Recruitment:
Medium 2011-2013,
Follow-up at
5 years
Cohort
Children ages Cord blood Respiratory tract Recurrent
1-5 years
N= 344 (182
boys, 162 girls)
2.44(1.74-3.22) infections (total
and recurrent)
respiratory tract
infections: OR
for > 75th
percentile
vs. < 75th
percentile PFOS
Total respiratory tract infections
-0.64 (—4.38, 3.1), p-value = 0.738
Recurrent respiratory tract
infections
0.91 (0.51, 1.65), p-value = 0.762
Confounding: Infant sex, maternal age, maternal education level, birth weight
Results stratified by age and sex not
statistically significant
Grandjean et al. Denmark
(2020, 7403067) 2020
Medium
Cross-sectional Adults, ages 30-Plasma Covid-19 OR per unit
70 years, with 4.86 (2.85-8.29) severity increase in
known SARS- PFOS
CoV-2 infection
N = 323
Covid-19 severity
0.97 (0.92, 1.02)
Covid-19 severity (hospitalization
vs no hospitalization)
0.96 (0.84, 1.10)
Covid-19 severity (intensive care
unit and/or deceased vs
hospitalization)
1.08 (0.94, 1.24)
Confounding: Age, sex, kidney disease, other chronic disease, national origin, place of testing, and days between blood sampling and
diagnosis
D-97
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Dalsager et al. Denmark
(2021, 7405343) Recruitment:
Medium 2010-2012,
Follow-up until
2015
Cohort
Pregnant
women and
their children
from the OCC,
followed up to
4 years
N = 1,472
Maternal serum Hospitalization Hazard ratio per Any infection
7.52 (0.49-27.5) from infection
(any infection,
upper
respiratory tract,
lower
respiratory tract,
gastrointestinal,
other)
log2-unit 1.23 (1.05, 1.44)
increase in Boys: 1.36 (1.10, 1.67)
PFOS Girls: 1.04 (0.85, 1.28)
Upper respiratory infection
1.25 (0.97, 1.61)
Lower respiratory infection
1.54(1.11,2.15)
Gastrointestinal infection
0.77 (0.46, 1.29)
Other infection
1.17(0.98, 1.40)
Confounding: Maternal age, parity, maternal educational level, child sex, child age
Ji et al. (2021,
7491706)
Medium
Wang et al.
(2022,
10176501)
Medium
China
2020
Case-control
Adults
N = 160
COVID-19 OR per log2-SD COVID-19
infection change inPFOS 1.94 (1.39,2.96)
Urine
Controls: 42.4
(25.5-61.3) ng/g
creatinine
Cases: 67.6
(41.0-96.5) ng/g
creatinine
Confounding: Age, gender, body mass index, diabetes, cardiovascular diseases, and urine albumin-to-creatinine ratio
China
Recruitment:
2010-2013,
Follow-up after
1 year
Cohort
Pregnant
women and
their children at
1 year from
LWBC
N = 235
Maternal serum Common cold,
at delivery bronchitis/pneu
4.58(3.31-6.14) monia, diarrhea
OR per log 10-
unit increase in
PFOS
IRR per loglO-
unit increase in
PFOS
Common cold
OR: 1.86 (0.53,6.50), p-
value = 0.334
IRR: 1.24 (0.76, 2.02), p-
value = 0.382
Bronchitis/pneumonia
OR: 1.54 (0.30,7.78), p-
value = 0.602
D-98
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
IRR: 0.76 (0.23, 2.46), p-
value = 0.644
Diarrhea
OR: 2.6 (0.67, 10.09), p-
value = 0.167
IRR: 1.89 (1.08,3.32), p-
value = 0.027
Confounding: Maternal age, pre-pregnancy BMI, smoking during pregnancy, maternal education level, and parity
Dalsager et al. Odense,
(2016, 3858505) Denmark
Low 2010-2012
Cohort
Children,
pregnancy
followed up at
1-4 years
N = 346
Maternal serum
8.07
(range = 2.36-
25.10)
Fever, cough,
nasal discharge,
diarrhea,
vomiting
OR (of
proportion of
days with
symptoms) by
tertiles
Fever
T2: 1.41 (0.81,2.44)
T3: 2.35 (1.34,4.11);
p-value < 0.05
Cough
T2: 1.16(0.67,2.01)
T3: 1.03 (0.59, 1.79)
Nasal discharge
T2: 1.11 (0.65, 1.93)
T3: 1.07 (0.62, 1.85)
Diarrhea
T2: 0.89 (0.51, 1.56)
T3: 1.04 (0.59, 1.82)
Results: Lowest tertile used as reference group.
Confounding: Maternal age, maternal educational level, parity, and child age.
Vomiting
T2: 1.47 (0.86,2.54)
T3: 0.78 (0.45, 1.35)
Impinen et al. Oslo, Norway Cohort, Nested
(2018, 4238440) Recruited 1992- case-control
Low 1993, followed
up for 10 years
Infants followed Cord blood
up at 2 and 5.2 (4.0-6.6)
10 years of age
N = 641
Common cold Regression
episodes from coefficient per
0-2 years, LRTI log2-unit
episodes from increase in
0-10 years PFOS
Common cold 0-2 years
-0.03 (-0.08,0.01)
p-value = 0.173
LRTI 0-10 years
D-99
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
0.5 (0.42, 0.57)
p-value < 0.0001
Confounding: Child sex
Impinen et al. Oslo, Norway Cohort
(2019, 5080609) Enrollment:
Low 1999-2008
Pregnant
women and
their infants
followed up at 3
and 7 years
0-3 years:
N = 1,207
6-7 years:
N = 921
Maternal blood
12.87 (9.92-
16.63)
Common cold,
bronchitis/pneu
monia, throat
infection with
strep,
pseudocroup,
ear infection,
diarrhea/gastric
flu, urinary tract
infection
OR per IQR Common cold
increase in 0-3 years: 0.94 (0.92, 0.97);
PFOS p-value < 0.05
Bronchitis/pneumonia
0-3 years: 1.20 (1.07, 1.34);
p-value < 0.05
6-7 years: 0.77 (0.50, 1.19)
Throat infection with strep
0-3 years: 0.90 (0.78, 1.04)
Other throat infections
0-3 years: 0.90 (0.81, 1.01)
Pseudocroup
0-3 years: 1.07 (0.96, 1.20)
Ear infection
0-3 years: 0.88 (0.82, 0.94);
p-value < 0.05
6-7 years: 1.13 (0.92, 1.40)
Diarrhea/gastric flu
0-3 years: 0.98 (0.93, 1.03)
6-7 years: 1.12(1.01, 1.24)
Urinary tract infection
0-3 years: 0.78 (0.70, 0.87);
p-value < 0.05
6-7 years: 0.91 (0.63, 1.31)
D-100
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location, Design Population, Matrix, Levels Outcome Comparison
Years Ages,N (ng/mL)a
Resultsb
Confounding: Maternal age, maternal BMI, maternal education, parity, smoking during pregnancy
Kvalem et al. Norway
(2020,6316210) Enrollment:
Low 1992-1993
Follow-up:
2002-2009
Cohort and
cross-sectional
Children,
10 years, all:
378, boys: 193,
girls: 185
Children, 10-
16 years, all:
375, boys: 191,
girls: 184
Children,
16 years, all:
330, boys: 170,
girls: 160
Serum
All: 19.4 (IQR:
9.23)
Boys: 21.7
(IQR: 8.86)
Girls: 17.52
(IQR: 8.02)
Common cold,
LRTI
Colds: OR
(reference: 1-2
colds)
LRTI: RR per
IQR increase in
PFOS
Colds, 10-16 years
3-5 colds
All: 1.26 (0.34,4.55)
p-value = 0.73
Boys: 2.54 (0.38, 17.3)
p-value = 0.34
Girls: 0.86 (0.16,4.75)
p-value = 0.86
> 5 colds
All: 1.16 (0.33,4.07)12.54
p-value = 0.82
Boys: 1.99 (0.3, 13.2)
p-value = 0.48
Girls: 1.07 (0.21, 5.45)
p-value = 0.93
Confounding: Puberty status at 16 years, mother's education, physical activity level at 16 years
LTRI
10-16 years
All: 1.34 (1.17, 1.55)
p-value < 0.001
Boys: 1.33 (1.26, 1.39)
p-value < 0.001
Girls: 1.23 (0.91, 1.66)
p-value = 0.17
16 years
All: 0.82 (0.4, 1.69)
p-value = 0.6
Boys: 0.62 (0.22, 1.78)
p-value = 0.38
Girls: 1.11 (0.41,3)
p-value = 0.84
Notes: BMI = body mass index; CI = confidence interval; DNBC = Danish National Birth Cohort; IQR = interquartile range; IRR = incidence rate ratio; LRTI = lower respiratory
tract infection; LWBC = Laizhou Wan Birth Cohort; OCC = Odense Child Cohort; OR = odds ratio; RR = risk ratio; Q2 = quartile 2; Q3 = quartile 3; Q4 = quartile 4;
RSV = respiratory syncytial virus; SE = standard error; T2 = tertile 2; T3 = tertile 3.
D-101
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
a Exposure levels reported as median (25th-75th percentile) unless otherwise noted.
b Results reported as effect estimate (95% confidence interval) unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
Table D-9. Associations Between PFOS Exposure and Asthma in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Dong et al. Taiwan, 2009- Case control
(2013, 1937230) 2010 and cross-
Medium sectional
Children from
GBCA with
(cases) or
without
(controls)
asthma, ages
10-15 years,
N = 231 (cases),
N = 225
(controls)
Serum
Cases: 33.9
(19.6-61.1)
Controls: 28.9
(14.1-43.0)
Asthma,
Asthma Control
Test score,
asthma severity
score, IgE in
serum (IU/mL),
AEC (106/L),
ECP in serum
(Hg/L)
Asthma: OR by
quartiles of
PFOS
Asthma Control
Test score,
asthma severity
score, IgE,
AEC, ECP:
mean values by
quartiles
Asthma
Q2: 1.96 (1.11,3.47)
Q3: 1.32 (0.75,2.32)
Q4: 2.63 (1.48,4.69)
p-trend = 0.003
IgE
Ql:517.9 (336.7, 699.2)
Q2: 686.2 (501.3,871.1)
Q3: 658.1 (475.2, 841.1)
Q4: 877.3 (695.2, 1,059.5), p-
value < 0.05
p-trend = 0.008
AEC
Ql: 329.4 (255.8, 403.0)
Q2: 368.6 (293.9, 443.3)
Q3: 431.3 (358.1, 504.6)
Q4: 453.4 (379.4, 527.3)
p-trend = 0.009
ECP
Ql: 25.9 (10.4, 41.3)
Q2: 37.4 (21.9, 52.8)
Q3: 43.5 (27.5,59.4)
Q4: 62.4 (46.3, 78.4), p-
value < 0.05
p-trend = 0.001
Asthma severity score
Ql: 3.33 (2.36,4.31)
D-102
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Q2: 4.18 (3.19, 5.17)
Q3: 4.49 (3.52,5.45)
Q4: 4.57 (3.61,5.54)
p-trend = 0.045
Asthma Control Test score: trends
across quartiles not statistically
significant
Results: Lowest quartile used as reference group
Confounding: age, sex, BMI, parental education, ETS exposure, and month of survey
Humblet et al. Unites States,
(2014,2851240) 1999-2008
Medium
Cross-sectional
Adolescents,
ages 12-
19 years old
from NHANES
N = 1,877
Serum
Never asthma
16.8 (10.8-26.2)
Ever asthma
17.0 (10.8-25.8)
No current
asthma
16.8 (10.8-26.2)
Current asthma
16.7 (10.3-25.3)
No wheezing
16.8 (10.8-26.2)
Wheezing
17.2 (10.9-25.4)
Asthma, wheeze OR per
doubling in
PFOS or per
unit increase in
PFOS
Ever asthma
Per doubling: 0.88 (0.74, 1.04), p-
value = 0.13
Per unit increase: 0.99 (0.98, 1.0),
p-value = 0.07
Current asthma
Per doubling: 0.88 (0.72, 1.09), p-
value = 0.24
Per unit increase: 0.99 (0.98, 1.01),
p-value = 0.34
Exposure: No
Confounding:
Wheeze
Per doubling: 0.83 (0.67, 1.02), p-
value = 0.08
Per unit increase: 0.99 (0.98, 1.01),
p-value = 0.37
wheezing defined as no wheezing in the past 12 months. Wheezing defined as history of wheezing in the past 12 months.
Sex, smoking, age, race/ethnicity, survey cycle, poverty income ratio, health insurance
Smit et al. Ukraine and
(2015, 2823268) Greenland,
Medium Exposure:
2002-2004,
Outcome:
2010-2012
Cohort
Mother-child
pairs with
follow-up when
the children
were 5-9 years
of age,
N = 1,024
Maternal blood Asthma
Ukraine:
GM = 4.88 (P5-
P95: 2.34-9.94)
OR per SD Asthma ever (combined): 0.86
increase in (0.67,1.10)
PFOS Ukraine: 0.75 (0.39, 1.42)
Greenland: 0.88 (0.67, 1.15)
D-103
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Greenland:
GM = 20.6 (P5-
P95: 10.2-49.6)
Maternal allergy, smoking during pregnancy, education level, maternal age, child sex, child age at follow-up, gestational age at
parity, breastfeeding, and birthweight0
Confounding
blood sample,
Impinen et al.
(2018, 4238440)
Medium
Oslo, Norway,
1992-2002
Cohort, Nested
case-control
Infants followed Cord blood
up at 2 and 5.2 (4.0-6.6)
10 years of age,
N = 641
Asthma ORperlog2- Current asthma (10 y):
unit increase in 1.14 (0.84, 1.54); p-value = 0.392
PFOS Asthma ever (10 y):
1.32 (0.89,1.97); p-value = 0.167
Confounding: Sex
Beck et al.
(2019, 5922599)
Medium
Denmark,
Enrollment:
2010-2012
Children, early
pregnancy to
5 years
N = 970 (507
boys, 363 girls)
Cohort Children, early Maternal blood Wheeze, self- OR per Wheeze
7.73 (5.68- reported asthma, doubling in All: 1.01 (0.79, 1.30)
10.44) doctor- maternal serum Boys: 1.02 (0.74, 1.39)
diagnosed PFOS Girls: 1.01 (0.67, 1.52)
asthma
Self-reported asthma
All: 1.22 (0.65,2.28)
Boys: 2.39 (0.92,6.21)
Girls: 0.67 (0.29, 1.53)
Doctor-diagnosed asthma
All: 0.83 (0.52, 1.31)
Boys: 0.74 (0.46, 1.20)
Girls: 1.60 (0.46, 5.59)
Confounding: Parity, maternal education level, maternal pre-pregnancy BMI, asthma predisposition, child sex
Gaylord et al. New York City
(2019, 5080201) NY
Medium 2014-2016
Case-control Children with Serum
(cases) or Cases: 3.72
without (Range: 1.01—
(controls) 14.2)
asthma aged Controls: 2.75
13-22, (Range: 0.60-
N= 118 (cases), 27.8)
N = 169
(controls)
Comparison: Logarithm base not specified.
Asthma OR per log-unit 0.89(0.45,1.76)
increase in
PFOS
D-104
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Confounding: Sex, race/ ethnicity, age, BMI, tobacco smoke exposure
Impinen et al. Oslo, Norway, Cohort Pregnant Maternal blood Asthma
OR per IQR
Current asthma:
(2019,5080609) Enrollment: women and 12.87(9.92-
increase in
Total: 1.11 (0.72, 1.69);
Medium 1999-2008 their infants 16.63)
PFOS
p-value = 0.643
(followed to age
Boys: 1.17(0.64,2.15);
7),
p-value = 0.616
N = 921
Girls: 1.03 (0.56,1.91);
p-value = 0.927
Ever asthma:
Total: 0.93 (0.68, 1.26);
p-value = 0.0.631
Boys: 0.94 (0.63, 1.40);
p-value = 0.744
Girls: 0.92 (0.57, 1.49);
p-value = 0.745
Confounding: Maternal age, maternal BMI, maternal education, parity, smoking during pregnancy
Manzano- Spain, Cohort Children, Maternal blood Asthma
OR or RR per
4-year follow-up: OR = 0.72 (0.45,
Salgado et al. 2003-2008 4 years, 6.06(4.52-7.82)
log2-unit
1.13)
(2019,5412076) N= 1,184
increase in
Medium
maternal PFOS
7-year follow-up: OR = 0.84 (0.57,
7 years,
1.25)
N = 1,068
4 and 7 years
Girls: RR = 0.68 (0.38. 1.22)
Boys: RR = 0.91 (0.58, 1.41)
Confounding: OR assessment: Age at follow-up of the child; RR assessment: Maternal age at delivery, parity, previous breastfeeding, pre-
pregnancy BMI, region of residence, and country of birth
Zeng et al.
(2019, 5412431)
Medium
Shanghai,
China,
2012-2015
Cohort
Enrolled in
pregnancy,
follow up at
5 years
N = 358 (187
boys, 171 girls)
Cord blood
Boys: 2.49
(1.81-3.51)
Girls: 2.38
(1.73-3.13)
Asthma
OR per log 10-
unit increase in
PFOS
All: 1.49 (0.29,7.54),
p-value = 0.63
Boys: 4.69 (0.51,42.77),
p-value = 0.17
Girls: 0.17 (0.01,4.15),
p-value = 0.27
Confounding: Child weight at age 5, gestational age, breastfeeding during the first 6 months, maternal education, maternal pre-pregnancy
BMI, and annual household income
D-105
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Jackson-Browne NHANES,
et al. (2020, United States,
6833598) 2013-2014
Medium
Cross-sectional Children, ages
3-11 years,
N = 607
Serum Asthma
GM =3.7 (2.6-
5.5)
OR per ln-SD 1.2(0.8,1.7)
increase in
PFOS
By age:
3-5 y: 1.7(1.0,3.0)
6-11 y: 1.1 (0.7, 1.6)
p-value for interaction by
age = 0.03
By sex:
Females: 1.1 (0.7, 1.7)
Males: 1.2 (0.8, 2.0)
p-value for interaction by
sex = 0.82
By race/ethnicity:
White, non-Hispanic: 1.4 (0.8, 2.6)
Black, non-Hispanic: 1.3 (0.8, 2.2)
Hispanic: 1.3 (0.8,2.0)
Other: 1.1 (0.7, 1.7)
p-value for interaction by
race = 0.35
Confounding: Sex, age, race/ethnicity, serum cotinine, poverty to income ratio
Kvalem et al. Norway
Cohort and Children,
Serum Asthma RR per IQR
10 years
(2020,6316210) Enrollment:
cross-sectional 10 years
increase in
All: 1.01 (0.86, 1.19)
Medium 1992-1993;
N = 378
All: 19.4 (IQR: PFOS
Boys: 1.06 (0.89, 1.26)
Follow-up:
(193 boys,
9.23)
Girls: 0.76 (0.52, 1.12)
2002-2009
185 girls)
Girls: 17.52
(IQR: 8.02)
10-16 years
Children, 10-
Boys: 21.7
All: 0.94 (0.74, 1.20)
16 years
(IQR: 8.86)
Boys: 0.96 (0.71, 1.31)
N = 375
Girls: 0.85 (0.54, 1.31)
(191 boys,
184 girls)
16 years
All: 1.00 (0.79, 1.27)
Children,
Boys: 1.01 (0.76, 1.36)
16 years
Girls: 0.91 (0.60, 1.38)
D-106
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
N = 375 (191
boys, 184 girls)
Confounding: 10 y: Age at follow-up, physical activity, mothers' education; 16 y: BMI at 16 years, puberty status at 16 years, mothers'
education, physical activity level at 16 years
Huang et al. China
(2020, 6988475) Recruitment:
Medium 2011-2013,
Follow-up at
5 years
Cohort
Children ages
1-5 years
N = 344 (182
boys, 162 girls)
Cord blood IgG (ng/mL),
2.44(1.74-3.22) IgE (ng/mL)
Regression
coefficient per
loglO-unit
increase in
PFOS
IgG
-0.01 (-0.06,0.04), p-
value = 0.643
IgE
-0.04 (-0.35, 0.27), p-
value = 0.805
Results stratified by age and sex not
statistically significant
Confounding: Infant sex, maternal age, maternal education level, birth weight
Xu et al. (2020, United States
6988472) 2007-2012
Medium
Cross-sectional
Adults from
NHANES, ages
20-79 years
N = 3,630
Serum
Mean
(SD) = 13.33
(12.92) (ig/L
Fractional
exhaled nitric
oxide (ppb)
Percent change
per doubling in
PFOS, or by
tertile
Fractional exhaled nitric oxide
2.03 (0.11, 4.00), p-value < 0.05
T2: 1.80 (-1.53, 5.25)
T3: 5.02 (1.40, 8.77), p-
value < 0.01
p-trend < 0.006
Results: Lowest tertile used as reference group
Confounding: Age, sex, race/ethnicity, BMI, annual family income, education level,
smoking status
serum cotinine, recent respiratory symptom, and
Zhou et al. Taiwan
(2017, 3981296) 2009-2010
Low
Case-control
Serum
Children with
(cases) or
without
(controls)
asthma ages 10- (13.9-46.0)
15 from the Control boys:
Asthma
Case boys: 36.9
(22.6-67.8)
Case girls: 28.2
Asthma:
Comparison of
PFOS
distributions
(Wilcoxon rank-
sum test)
Asthma: Increased PFOS among
asthmatics, p-value = 0.002
GBCA
N = 456
Case boys: 158
Case girls: 73
29.9 (13.0-43.8)
Control girls:
28.8 (14.8-42.6)
D-107
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Control boys:
102
Control girls:
123
Confounding: Cases and controls were matched on age and sex
Zhu et al. (2016, Taiwan
Case-control
Children with
Serum Asthma
OR for highest
Boys: 4.24 (1.81, 9.42); p-value for
3360105) 2009-2010
(cases) or
Case boys:
vs. lowest
trend = 0.001
Low
without
36.94
quartiles of
Girls: No statistically significant
(controls)
Case girls:
PFOS
associations or trends
asthma ages 10-
28.16
15 from the
Control boys:
GBCA
26.24
N = 456
Control girls:
Case boys: 158
30.12
Case girls: 73
Control boys:
102
Control girls:
123
Confounding: Age, BMI, parental education, environmental tobacco smoke, parental asthma, month of survey
Zhou et al. Taiwan
Case-control
Children with
Serum Asthma
OR per unit
Females with high testosterone:
(2017,3858488) 2009-2010
(cases) or
Cases: 33.94
increase in
0.58 (0.36, 0.93)
Low
without
(19.59-61.10)
PFOS
Females with low testosterone: 1.32
(controls)
Controls: 28.91
(0.88, 1.99)
asthma ages 10-
(14.06-42.02)
p-value for interaction by low/high
15 from the
testosterone = 0.010
GBCA
N = 456
Males with high testosterone: 1.04
Case boys: 158
(0.87, 1.25)
Case girls: 73
Males with low testosterone: 2.54
Control boys:
(1.40, 4.60)
102
p-value for interaction by low/high
Control girls:
testosterone = 0.005
123
D-108
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Sexes evenly
divided into
high/low
hormone
classifications
Females with high estradiol: 1.25
(0.84, 1.86)
Females with low estradiol: 0.65
(0.42, 0.99)
p-value for interaction by low/high
estradiol = 0.026
Males with high estradiol: 1.25
(0.90, 1.72)
Males with low estradiol: 1.06
(0.87, 1.30)
p-value for interaction by low/high
estradiol = 0.407
Confounding: Age, sex, BMI, parental education, environmental tobacco smoke exposure, physical activity, month of survey
Timmermann et Faroe Islands,
al. (2017, recruitment:
3858497) 1997-2000
Low
Cohort
Pregnant
women and
infants, follow
up at ages 5, 7,
and 13 years,
N = 559
Maternal serum Asthma
Prenatal/At
birth: 27.4
(23.3-33.3)
Age 5/7: 16.8
(13.5-21.1)
Age 13:6.7
(5.2-8.5)
OR per Asthma (age 5): Total: 1.21 (0.64,
doubling of 2.29)
maternal PFOS No MMR vaccine before age 5:
3.96 (0.55, 28.39)
Yes MMR vaccine before age 5:
0.98 (0.55, 1.76)
Confounding: Family history of eczema in children,
pregnancy, sex, duration of breastfeeding, fish intake
history of chronic bronchitis/asthma
Asthma (age 13):
Total: 0.69 (0.43, 1.09)
No MMR vaccine before age 5:
5.41 (0.62, 47.16)
Yes MMR vaccine before age 5:
0.94 (0.51, 1.74)
allergic eczema, and hay fever, maternal pre-pregnancy BMI, maternal smoking during
at age 5, number of siblings, daycare attendance at age 5, birth weight, and family
Averina et al. Norway
(2019,5080647) 2010-2011
Low
Cohort
Adolescents in
their first year
of high school
from TFF1 and
TFF2
Serum
Girls: GM=5.8
(IQR = 2.7)
Boys: GM = 6.8
(IQR =3.0)
Asthma self-
reported doctor
diagnosed
OR by quartiles
of PFOS
TFF1
Q2: 1.51 (0.72,3.18)
Q3: 2.75 (1.36,5.57);
p-value = 0.005
D-109
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
N = 675
Q4: 2.11 (1.02,4.37);
p-value = 0.044
p-value for trend = 0.02
TFF2
Q2: 2.00 (0.96,4.15);
p-value = 0.064
Q3: 2.56 (1.24, 5.30);
p-value = 0.011
Q4: 1.43 (0.65,3.12)
Trend not statistically significant
Results: Lowest quartile used as reference group.
Confounding: Sex, age, BMI, physical activity, unemployment/disability of parents, living with adoptive parents, fish intake
Workman et al. Canada
(2019, 5387046) 2010-2012
Low
Cohort
Mothers and
their infants
N = 85
Maternal
plasma
2.2 (Range:
0.18-21)
Recurrent
wheezing
episodes
Difference in
prenatal PFOS
levels for
wheezing vs. no
wheezing
(Mann-Whitney
test)
No significant differences
Confounding: None reported
Notes: AEC = absolute eosinophil counts; BMI = body mass index; CI = confidence interval; ECP = eosinophilic cationic protein; GBCA = Genetic and Biomarker study for
Childhood Asthma; GM = geometric mean; IQR = interquartile range; NHANES = National Health and Nutrition Examination Survey; MMR = measles, mumps, rubella;
OR = odds ratio; RR = risk ratio; SD = standard deviation; TFF1 = Tromse Fit Futures.
a Exposure levels reported as median (25th-75th percentile) unless otherwise noted.
b Results reported as effect estimate (95% confidence interval) unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
Table D-10. Associations Between PFOS Exposure and Allergies in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Study Design PoPulati«n'
J b Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Children
Wang et al. Taiwan Cohort and Pregnant Cord blood Atopic Atopic Atopic dermatitis
(2011,1424977) 2004 cross-sectional women and dermatitis, IgE dermatitis: Q2:0.68 (0.20,2.3)
D-110
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Medium
their children at 5.50 (0.11- levels (log-
age 2 48.36) KU/L)
N = 244 (133
boys, 111 girls)
OR by quartiles
ofPFOS
IgE:
Regression
coefficient per
ln-unit increase
inPFOS
Q3: 2.34 (0.86, 6.41)
Q4: 2.19 (0.78, 6.17)
IgE in cord blood at birth
All: 0.161 (SE = 0.147),
p-value = 0.017
Boys: 0.175 (SE = 0.179),
p-value = 0.053
Girls: 0.151 (SE = 0.165),
p-value = 0.616
IgE in serum at age 2
All: 0.251 (SE = 0.179),
p-value = 0.147
Boys: 0.359 (SE = 0.255),
p-value = 0.238
Girls: 0.095 (SE= 0.325),
p-value = 0.723
Results: Lowest quartile used as reference group.
Confounding: Gender, gestational age, maternal age. Additional confounding for atopic dermatitis: maternal history of atopy, duration of
breast feeding, pre-natal ETS exposure. Additional confounding for IgE: parity.
Okada et al. Japan
Cohort
Pregnant
Maternal serum
Food allergy,
OR and
Food allergy
(2012, 1332477) 2002-2005
women and
5.2 (3.4-7.2)
eczema, otitis
regression
3.72 (0.81, 17.10)
Medium
children from
media, and
coefficients per
the Hokkaido
wheezing
loglO-unit
Eczema
Study on
IgE levels
increase in
0.87 (0.15, 5.08)
Environment
(loglO-IU/mL)
PFOS
and Children's
Otitis media
Health; follow
1.40 (0.33, 6.00)
up at 18 months
N = 343
Wheezing
2.68 (0.39, 18.30)
IgE:
Linear regression
D-lll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
-0.342 (-1.230, 0.546)
Quadratic regression
-0.681 (-2.50, 1.137)
Cubic regression
1.464 (-5.354, 8.282)
Results stratified by gender not
statistically significant for boys and
combined
Confounding: maternal age, maternal educational level, pre-pregnancy BMI, allergy of parents, parity, infant gender, breast-feeding period,
environmental tobacco exposure, day care attendance and blood sampling period; for IgE: maternal age, maternal allergic history, distance
from home to highway, parity, birth season, and blood sampling period
Buser et al.
(2016, 3859834)
Medium
United States
2005-2016
Cross-sectional
Adolescents
aged 12-
19 years from
NHANES
Nby cycle:
2005-2006: 637
2007-2010: 701
Serum
2005-2006:
GM = 14.98
(10.65-22.69)
2007-2010:
GM = 8.74
(5.96-13.75)
Food allergy or
sensitization
OR by quartiles
ofPFOS
Food allergy, 2007-2010 cycle
Q2: 2.22 (0.85, 5.77)
Q3: 2.43 (1.05,5.59)
Q4: 2.95 (1.21,7.24)
p-value for trend = 0.27
Food sensitization, 2005-2006
cycle: No statistically significant
associations or trends
Outcome: Food sensitization defined as at least 1 food specific IgE level >0.35 kU/L.
Results: Lowest quartile used as reference.
Confounding: Age, sex, race/ethnicity, BMI, serum cotinine0
Goudarzi et al.
(2016, 3859523)
Medium
Japan
2003-2013
Cohort
Children at age
4 from the
Hokkaido Study
N = 1,558 (765
girls, 793 boys)
Maternal blood
4.93 (3.67-6.
65)
Allergic
diseases, total
OR by quartiles
ofPFOS
Results: Lowest quartile used as reference.
Confounding: Maternal age, maternal educational level, sex, parental allergic history,
attendance, environmental tobacco smoke exposure
Q2: 0.66 (0.48, 0.90)
Q3: 0.79 (0.58, 1.07)
Q4: 0.82 (0.60, 1.11)
p-value for trend = 0.391
No statistically significant
associations, trends, or interactions
by sex
number of older siblings, breast-feeding, day-care
D-112
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Timmermann et
al. (2017,
3858497)
Medium
Faroe Islands,
Recruitment:
1997-2000
Cohort
Pregnant
women and
infants, follow
up at ages 5, 7,
and 13 years,
N = 559
Maternal serum Allergy, allergic OR per
Prenatal/At
birth: 27.4
(23.3-33.3)
Age 5/7: 16.8
(13.5-21.1)
Age 13:6.7
(5.2-8.5)
rhino-
conjunctivitis in
past 12 months,
positive skin
prick test, IgE
doubling of
PFOS
IgE: Percent
change per
doubling of
PFOS
Allergy (age 5)
OR = 0.73 (0.38,1.41)
Allergic rhino-conjunctivitis in past
12 months, age 13
1.01 (0.54, 1.89)
Positive skin prick test, age 13
1.15 (0.75, 1.77)
IgE, age 7: -9.38 (-37.17, 30.71)
Confounding: Maternal parity, family history of eczema in children, allergic eczema and hay fever, maternal pre-pregnancy BMI, maternal
smoking during pregnancy, maternal fish intake during pregnancy, and duration of breastfeeding; for IgE: family history of eczema in
children, allergic eczema, and hay fever, maternal pre-pregnancy BMI, maternal smoking during pregnancy, sex, duration of breastfeeding,
fish intake at age 5, number of siblings, and daycare attendance at age 5
Impinen et al. Oslo, Norway,
(2018, 4238440) 1992-2002
Medium
Cohort, Nested
case-control
Infants followed Cord blood
up at 2 years 5.2 (4.0-6.6)
and 10 years of
age,
N = 641
Rhinitis, rhino-
conjunctivitis,
SPT
OR per log2-
unit increase in
PFOS
Rhinitis, current, 10 y
1.00 (0.72, 1.40); p-value = 0.983
Rhinitis, ever, 10 y
1.05 (0.74, 1.48); p-value = 0.775
Rhino-conjunctivitis, ever, 10 y
1.02 (0.72, 1.45); p-value = 0.905
Rhino-conjunctivitis, ever, spes
IgE >0.35, 10 y
1.02 (0.71, 1.47); p-value = 0.905
SPT, any pos, 10 y
0.87 (0.65, 1.17); p-value = 0.359
SPT + and/pr slgE > 0.35, 10 y
0.91 (0.69, 1.19); p-value = 0.476
Confounding: Sex
D-113
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Impinen et al. Oslo, Norway, Cohort
(2019, 5080609) Enrollment:
Medium 1999-2008
Pregnant Maternal blood Allergy, food or OR per IQR Allergy, food, current
women and 12.87 inhaled increase in All: 1.02 (0.73, 1.41);
their infants (9.92-16.63) PFOS p-value = 0.928
(followed to age Boys: 1.09(0.68, 1.74);
7), p-value = 0.72
N = 921 Girls: 0.95 (0.59,1.51);
p-value = 0.815
Allergy, food, ever
All: 0.99 (0.72, 1.37);
p-value = 0.969
Boys: 1.11 (0.69, 1.77);
p-value = 0.671
Girls: 0.91 (0.58, 1.42);
p-value = 0.676
Allergy, inhaled, current
All: 1.11 (0.0.72, 1.69);
p-value = 0.643
Boys: 0.86 (0.44, 1.71);
p-value = 0.669
Girls: 1.17 (0.55,2.48);
p-value = 0.679
Allergy, inhaled, ever
All: 1.27 (0.93,1.74);
p-value = 0.135
Boys: 1.2 (0.79, 1.84);
p-value = 0.39
Girls: 1.33 (0.84,2.12);
p-value = 0.224
Confounding: Maternal age, maternal BMI, maternal education, parity, smoking during pregnancy, nursery attendance
Ait Bamai et al. Hokkaido,
(2020, 6833636) Japan,
Medium 2003-2012
Cohort Early pregnancy Maternal blood Rhino- RRperln-unit 0.96 (0.79, 1.15); p-value = 0.626
to 7 years, 5.12 (3.75-7.02) conjunctivitis increase in
N = 2,689 PFOS, from
D-114
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
birth to 7 years
old
Confounding: Sex, parity, maternal age at delivery, maternal smoking during pregnancy, pre-pregnancy BMI, and annual household income
during pregnancy
Kvalem et al. Norway,
(2020,6316210) Enrollment:
Medium 1992-1993;
Follow-up:
2002-2009
Cohort and
cross-sectional
Children, age
10 years:
N = 377
Age 16 years:
N = 375
Serum
All: 19.4 (IQR:
9.23)
Girls: 17.52
(IQR: 8.02)
Boys: 21.7
(IQR: 8.86)
Rhinitis, skin Change in RR
prick test (SPT) per IQR
increase in
PFOS
Rhinitis
10 years
All: 0.98 (0.74,1.30);
p-value = 0.92
Boys: 0.90 (0.66, 1.23);
p-value = 0.52
Girls: 0.97 (0.58, 1.62);
p-value = 0.92
16 years
All: 1.03 (0.90,1.19);
p-value = 0.69
Boys: 0.92 (0.72, 1.19);
p-value = 0.55
Girls: 1.15 (0.91, 1.45);
p-value = 0.24
SPT
10 years
All: 1.10 (0.95, 1.26);
p-value = 0.21
Boys: 0.98 (0.96, 1.01);
p-value = 0.17
Girls: 0.97 (0.65, 1.44);
p-value = 0.0.86
16 years
All: 1.09 (1.03, 1.15);
p-value = 0.001
Boys: 1.07 (0.97, 1.17);
p-value = 0.18
D-115
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Girls: 0.99 (0.80, 1.23);
p-value = 0.93
Confounding: 10 years: Physical activity at 10 years, mothers' education, BMI at 10 years; 16 years: BMI at 16 years, puberty status at
16 years, mothers' education, physical activity level at 16 years
Notes: BMI = body mass index; CI = confidence interval; IgE = immunoglobulin E; IQR = interquartile range; MMR = measles, mumps, rubella; NHANES = National Health and
Nutrition Examination Survey; OR = odds ratio; RR = risk ratio; SD = standard deviation; SPT = skin prick test.
a Exposure levels reported as median (25th-75th percentile) unless otherwise noted.
b Results reported as effect estimate (95% confidence interval) unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
Table D-ll. Associations Between PFOS Exposure and Eczema in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels Outcome
(ng/mL)a
Comparison
Resultsb
General Population
Goudarzi et al.
(2016, 3859523)
Medium
Japan Cohort Children at age Maternal blood Eczema
2003-2013 4 from the 4.93 (3.67-
Hokkaido Study 65654)
N = 1,558 (765
girls, 793 boys)
OR by quartiles
of PFOS
Results: Lowest quartile used as reference.
Confounding: Maternal age, maternal educational level, sex, parental allergic history,
attendance, environmental tobacco smoke exposure0
Q2: 0.64 (0.44, 0.93)
Q3: 0.65 (0.45,0.95)
Q4: 0.85 (0.591, 1.22)
p-value for trend = 0.427
No statistically significant
associations, trends, or interactions
by sex
number of older siblings, breast-feeding, day-care
Timmermann et
al. (2017,
3858497)
Medium
Denmark
1997-2000
Cohort
Pregnant
women and
infants from the
CHEF study at
ages 5, 7, and
13 years
N = 559
Serum Atopic eczema OR per
Prenatal at birth: at age 13 doubling of
16.8 (13.5-21.1) PFOS at age 13
Age 5/7: 27.4
(23.3-33.3)
Age 5: 0.75 (0.42, 1.34)
Age 13: 0.8 (0.46, 1.39)
MMR vaccination before age 5
Yes: 8.94 (0.27, 299.11)
No: 0.82 (0.53, 1.28)
D-116
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Confounding: Confounding: Family history of eczema in children., allergic eczema and hay fever, maternal pre-pregnancy BMI, maternal
smoking during pregnancy, sex, duration of breastfeeding, and fish intake at age 13, birth weight, and family history of chronic
bronchitis/asthma, maternal parity
Chen et al.
(2018, 4238372)
Medium
China Cohort Infants followed
2012-2015 up at 6, 12, and
24 months
N = 687
children (328
female and 359
male)
Cord blood
All: 2.48
(Range = 0.39-
65.61)
Female: 2.47
(Range = 0.39-
18.68)
Male: 2.49
(Range = 0.62-
65.61)
Atopic
dermatitis
OR per log-unit
increase in
PFOS, or by
quartiles
All: 1.23 (0.85, 1.76)
Q2: 0.93 (0.56, 1.58)
Q3:l (0.59, 1.7)
Q4:1.31 (0.78, 2.2)
Female: 1.1 (0.64, 1.87)
Q2: 0.73 (0.33, 1.61)
Q3:0.71 (0.32, 1.6)
Q4: 1.08 (0.5,2.35)
Male: 1.42 (0.84, 2.42)
Q2: 1.34(0.64,2.8)
Q3: 1.3 (0.61,2.75)
Q4: 1.65 (0.79,3.41)
Comparison: Logarithm base not specified.
Results: Lowest quartile used as reference group.
Confounding: Maternal age, maternal pre-pregnancy BMI, gestational week at delivery, birth weight, maternal education, paternal education,
parity, mode of delivery, family history of allergic disorders, infant sex, family income, maternal ethnicity, paternal smoking, breastfeeding
Impinen et al.
(2018, 4238440)
Medium
Norway
1992-2002
Cohort, Nested
case-control
Children from
the ECA study
at 0, 2, and
10 years
N = 641
Cord blood
5.2 (4.0-6.6)
Atopic
dermatitis
diagnosed
anytime
between 0-
2 years old, or
between 0-
10 years old
OR per log2-
unit increase
PFOS
Ages 0-2: 1.15 (0.88, 1.52)
Ages 0-10:0.68 (0.38, 1.2)
Confounding: Sex
Manzano- Spain
Salgado et al. 2003-2015
(2019, 5412076)
Medium
Cohort
Pregnant
women and
children
followed up at
ages 1.5, 4, and
7 from the
INMA study
Maternal plasma Eczema
6.06 (4.52-7.82)
OR or RR per Age 1.5: 1.02 (0.83, 1.27)
log2-unit Age 4: 0.8 (0.65, 0.99)
increase in Age 7: 0.82 (0.68, 0.99)
PFOS Boys at ages 1.5, 4, and 7: 0.91
(0.75, 1.11)
Girls at ages 1.5, 4, and 7: 0.77
(0.64, 0.94)
D-117
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
From ages 1.5 to 7 years: 0.86
(0.75, 0.98)
N = 1,188 at
1.5, N= 1,184
at 4 years,
N = 1,066 at
7 years
Confounding: Age at follow-up of the child, maternal age at delivery, parity, previous breastfeeding, pre-pregnancy BMI, region of residence,
and country of birth
Wen et al.
(2019, 5081172)
Medium
Taiwan
2001-2005
Cohort
Children at age
2 years
N = 839
Cord blood Atopic
3.49 (2.18-5.05) dermatitis
OR by tertiles of T2: 1.33 (0.57, 3.20)
PFOS T3: 1.86 (0.84,4.36)
Results: Lowest tertile used as reference.
Confounding: Sex, family income, maternal atopy, breast feeding, and maternal age at childbirth
Wen et al.
(2019, 5387152)
Medium
Taiwan
2001-2005
Cohort
Cord blood Atopic
3.49 (2.18-5.05) dermatitis
Hazard ratio for
PFOS > 5.05 ng
/mL vs.
< 5.05 ng/mL
General
population,
children, and
adolescents <18
yrs.; Infants
followed from
birth up to
5 years of age
N = 863
Confounding: Sex, parental education, parental atopy, breast feeding, and maternal age at childbirth
1.43 (0.82,2.43)
No statistically significant
associations
Notes: CD = Crohn's disease; CIS = clinically isolated serum syndrome; OR = odds ratio; RRMS = relapsing remitting multiple sclerosis; UC = ulcerative colitis;
a Exposure levels are reported as median (25th-75th percentile) unless otherwise noted.
b Results reported as effect estimate (95% confidence interval) unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
Table D-12. Associations Between PFOS Exposure and Autoimmune Health Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Population, Exposure
Design Ages, Matrix, Levels Outcome Comparison
N (ng/mL)a
Resultsb
Gay lord et al.,
2020,6833754
Medium
United States Case-control
Children and Serum Celiac disease ORperln-unit 2.20(0.78,6.18)
adolescents Cases: 2.02 change inPFOS Girls: 12.8 (1.17, 141);
younger than (IQR=1.85) p-value<0.05
D-118
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Controls: 1.59
(IQR = 1.64)
Boys: 1.02 (0.24, 4.21)
21 years with
(cases) and
without
(controls) celiac
disease
N = 88 (42 girls,
46 boys)
Confounding: Genetic susceptibility score, albumin, BMI, age, race (non-Hispanic white vs. other race/ethnicity) and sex0
Steenland et al.,
United States Case-control Patients with Serum UC
Change in
UC vs. CD:
2018,5079806
1999-2012 UC, CD, or UC: 3.95
log(PFOS)
0.05 (0.16), p-value = 0.77
Low
healthy controls CD: 3.32
comparing cases
UC vs. control:
Neither: 4.21
and controls
-0.40 (0.21), p-value = 0.06
N = 114 UC, 60
CD, 75 controls
Comparison: Logarithm base not specified.
Results: Lowest quintile used as reference.
Confounding: Age, sex, ethnic group (white or non-white), year of sample
Sinisalu et al.
Finland Cohort
Pregnant
Cord blood
Celiac disease
Comparison of
No significant differences in
(2020, 7211554)
1999-2005
women and
Case: 2.21
mean PFOS
exposure between cases and control
Low
infants at birth
(min-max:
exposure levels
at birth or 3 months
and 3 months
0.27-8.17)
from the Type 1
Control: 2.25
Diabetes
(min-max:
Prediction and
0.27-5.32)
Prevention
Study in Finland
3-month serum
(DIPP)
Case: 2.93
N = 33 (17
(min-max:
celiac disease,
0.27-7.66)
16 controls)
Control: 3.40
(min-max:
0.71-6.70)
Ammitzboll et
Denmark Case-control
Adults with
Serum
Relapsing
Percent change
-17 (-27, -6); p-value = 0.004
al., 2019,
2019
(cases) or
Cases: 7.14
remitting
in PFOS
Females: -14 (-28, 3);
5080379
without
(5.76-9.93)
multiple
comparing MS
p-value = 0.093
D-119
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Low
(controls)
RRMS or CIS
N = 162 (92
women, 70
men)
Confounding: Age, sex, breastfeeding
Controls: 9.41 sclerosis cases vs. healthy Males:-19 (-32,-3);
(6.41-13.0) (RRMS) controls p-value = 0.023
Notes: CD = Crohn's disease; CIS = clinically isolated serum syndrome; OR = odds ratio; RRMS = relapsing remitting multiple sclerosis; UC = ulcerative colitis.
a Exposure levels are reported as median (25th-75th percentile) unless otherwise noted.
b Results reported as effect estimate (95% confidence interval) unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
D.5 Cardiovascular
D.S.I Cardiovascular Endpoints
Table D-13. Associations Between PFOS Exposure and Cardiovascular Effects in Recent Epidemiological Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels Outcome
(ng/mL)a
Comparison
Resultsb
Children and Adolescents
Li et al., 2021,
7404102
High for
gestation, birth,
and childhood
exposures (3-
year and 8-year)
Medium for
exposure at 12-
year follow-up
United States Cohort
2003-2006
Pregnant women
and their children
followed up at
birth and ages 3,
8, and 12 from
HOME Study
Gestation:
N = 203
At birth: N= 124
Age 3: N = 137
Age 8: N = 165
Age 12: N = 190
Maternal serum
Gestation: 12.9
(8.9-18.0)
Cord serum
At birth: 4.2
(3.0-6.5)
Serum
At age 3: 6.2
(4.5-9.9)
At age 8: 3.6
(2.8-4.7)
SBP (z-score),
mean of SBP
and DBP (z-
score)
Regression coefficient
per log2-unit IQR
increase in PFOS
SBP (z-score)
Gestation: 0.1 (-0.1, 0.2)
At birth: 0.2(0.0,0.4)
Age 3: 0.1 (-0.1, 0.4)
Age 8: 0.1 (-0.3, 0.4)
Age 12:0.2 (-0.1,0.5)
Mean of SBP and DBP (z-
score)
Gestation: 0.1 (-0.1, 0.2)
At birth: 0.1(0.0,0.3)
Age 3: 0.1 (-0.1, 0.3)
Age 8: 0.1 (-0.2, 0.4)
D-120
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
At age 12: 2.4 Age 12: 0.2 (0.0, 0.4)
(1.8-3.2)
Confounding0: visit, visit*PFAS, maternal age, maternal education, maternal pre-pregnancy BMI, gestational serum cotinine concentrations,
and parity; and child age, sex, race, and pubertal stage. Additional confounding for analyses at age 3, age 8, and age 12: Breastfeeding
duration.
Ma et al., 2019, United States Cross-
Adolescents aged Serum
5413104
Medium
2003-2012
sectional
12-20 from
NHANES
N = 2,251 (1,048
female, 1,203
male)
median = 11.1
(6.2-18.0)
DBP, SBP Regression coefficient
per loglO-unit increase
inPFOS
DBP
Total cohort: 0.014 (-0.001,
0.030)
Females: 0 (-0.02, 0.02)
Males: 0.025 (0.001, 0.049);
p-value < 0.05
SBP
Total cohort: 0.002 (-0.004,
0.009)
Females: -0.001 (-0.009,
0.008)
Males: 0.003 (-0.006, 0.012)
Warembourg et
al., 2019,
5881345
Medium
France, Spain, Cohort Pregnant women
Lithuania, and their children
Norway, at ages 6 and 11
Greece, from the HELIX
United Project
Kingdom N= 1,277
1999-2015 Prenatal exposure
Postnatal
exposure
Maternal blood:
6.4 (4.1-9.6)
Plasma: 2.0 (1.3-
3.2)
DBP, SBP
Regression coefficient
per log2-unit IQR
increase in PFOS
DBP
Maternal PFOS: 0.46 (-0.34,
1.27)
Childhood PFOS: 0.48 (-1.06,
0.62)
SBP
Maternal PFOS: -0.22 (-1.06,
0.62)
Childhood PFOS: 0.23 (-0.56,
1.03)
Confounding: Cohort of inclusion, maternal age, maternal education level, maternal pre-pregnancy BMI, parity, parental country of birth,
child age, child sex, child height0
Canovaetal., Italy Cross- Adolescents aged Serum
2021,10176518 2017-2019 sectional 14 to 19 years and Adolescents: 3.3
Medium children aged 8 to (2.2-4.9)
11 years from
DBP, SBP Regression coefficient
per ln-unit increase in
PFOS, or by quartiles
DBP
Adolescents
Per ln-unit increase: -0.44
(-0.82, 0.05)
D-121
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population,
Design Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
health
surveillance
program in
Veneto Region
Adolescents:
N = 6,669
Children:
N = 2,693
Children: 2.2
(1.6-3.0)
Q2: -0.54 (-1.15,0.08)
Q3: -0.66 (-1.30,-0.02)
Q4: -0.78 (-1.45,-0.10)
Children
Per ln-unit increase: 0.03
(-0.54, 0.61)
Q2: 0.67 (-0.15, 1.54)
Q3: 0.91 (0.05, 1.77)
Q4:-0.10 (-0.95, 0.75)
SBP
Adolescents
Per ln-unit increase: -0.47
(-1.02, 0.08)
Q2: -0.67 (-1.54,0.20)
Q3: -0.96 (-1.87,-0.06)
Q4: -1.34 (-2.30, -0.38)
Children
Per ln-unit increase: -0.42
(-1.18,0.33)
Q2
Q3
Q4
-0.13 (-1.22,0.95)
0.18 (-0.95, 1.31)
-0.80 (-1.92,0.33)
Results: Lowest quartile used as the reference group.
Confounding: Age, gender, country of birth, data on food consumption, degree of physical activity, salt intake, smoking status (for
adolescents only), time-lag between the beginning of the study and the date of enrollment.
Papadopoulou et United
al., 2021,
9960593
Medium
Kingdom,
France, Spain,
Lithuania,
Norway,
Greece
Recruitment
1999-2010,
Follow-up:
2013-2015
Cohort Mother-child
pairs from the
HELIX Project,
children followed
up around age 8
(range 6-12)
N = 1,101
Maternal plasma
(prenatal)
6.15 (3.99-9.16)
Plasma
(childhood)
1.93 (1.22-3.11)
DBP (z-score),
SBP (z-score)
Regression coefficient
per doubling in PFOS,
or by quartiles
DBP
Maternal PFOS: 0.04 (-0.06,
0.14)
Q2: -0.06 (-0.23,0.11)
Q3: 0.03 (-0.16,0.23)
Q4: -0.04 (-0.29,0.21)
p-trend = 0.922
Childhood PFOS: 0.01 (-0.06,
0.08)
Q2: -0.02 (-0.18,0.13)
D-122
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Q3: -0.01 (-0.19,0.17)
Q4: 0.01 (-0.20,0.23)
p-trend = 0.827
SBP
Maternal PFOS: 0.03 (-0.08,
0.14)
Q2: -0.06 (-0.25,0.13)
Q3: 0.10 (-0.12, 0.13)
Q4: -0.05 (-0.32, 0.23)
p-trend = 0.980
Childhood PFOS: -0.01
(-0.08, 0.07)
Q2
Q3
Q4
-0.04 (-0.21,0.13)
-0.03 (-0.23,0.16)
-0.03 (-0.27,0.21)
p-trend = 0.763
Comparison: Maternal PFOS quartiles are defined as follows: Ql: 0.28-3.98; Q2: 3.99-6.15; Q3: 6.15-9.15; Q4: 9.16-47.98; childhood
PFOS quartiles are defined as follows: Ql: 0.00-1.22; Q2: 1.22-1.92; Q3: 1.93-3.10; Q4: 3.11-33.83.
Results: Lowest quartile used as the reference group.
Confounding: Maternal age and education, pre-pregnancy BMI, parity, cohort, child ethnicity, age, child gender, PFHxS, PFNA, PFOA
Manzano-
Salgado et al.,
2017,4238509
Medium
Spain
2003-2008
Cohort
Pregnant women Maternal blood Blood Pressure Regression coefficient BP
and their children
at ages 4 and 7
from INMA study
Age 4 N = 839
(412 girls, 427
boys)
Age 4 N = 386
(197 girls, 189
boys) for CMR
score
measurements
GM = 5.80
(4.52-7.84)
(BP) (z-
score)
Cardiometaboli
c Risk Score
(CMR)
per log2-unit increase
in PFOS
All age 4: -0.05 (-0.15,0.06)
Girls: -0.06 (-0.22, 0.09)
Boys:-0.02 (-0.18,0.14)
All age 7: 0.06 (-0.04,0.15)
Girls: 0.06 (-0.09, 0.20)
Boys: 0.04 (-0.08,0.17)
CMR
All age 4: 0.28 (-0.33, 0.89)
Girls: 0.10 (-0.73, 0.93)
Boys: 0.47 (-0.44, 1.37)
D-123
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Age 7 N = 1,086
(535 girls, 551
boys)
Confounding: Maternal region of residence, country of birth, previous breastfeeding, age, pre-pregnancy BMI; age/sex of child
Linetal., 2013,
2850967
Medium for
CIMT
Low for Systolic
BP
Geiger et al.,
2014,2851286
Medium
Taiwan
2006-2008
Cross-
sectional
Adolescents and Serum SBP, CIMT
young adults ages 8.65 (5.4-13.52)
12-30
N = 637
Mean by quartiles
SBP: No associations across
quartiles; p-trend = 0.177
CIMT:
Significant associations across
exposure groups;
p-trend < 0.002
Females: significant
associations across exposure
groups; p-trend < 0.001
Males: no associations across
exposure groups;
p-trend = 0.401
Ages 12-19: significant
associations across exposure
groups; p-trend < 0.001
Ages 20-30: no associations
across exposure groups;
p-trend = 0.084
Confounding: Age, gender, smoking status, alcohol drinking, body mass index; for CIMT, also includes systolic blood pressure, low density
lipoprotein cholesterol, triglyceride, high sensitivity C-reactive protein, homeostasis model assessment of insulin resistance
United States
1999-2000,
2003-2008
Cross-
sectional
Children
ages <18 years
from NHANES
N = 1,655
Serum
Mean
(SE) = 18.4 (0.5)
Hypertension
OR per ln-unit increase Hypertension
in PFOS, or by quartile Per ln-unit increase: 0.83
(0.58, 1.19)
Q2: 0.99 (0.55, 1.78)
Q3: 0.73 (0.36, 1.48)
Q4: 0.77 (0.37, 1.61)
p-trend = 0.3625
Results: Lowest quartile used as the reference group.
Confounding: Age, sex, race-ethnicity, BMI categories, annual household income categories, moderate activity, total cholesterol, and serum
cotinine
D-124
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Averina et al.,
Norway Cross- First level high Serum Hypertension OR by quartiles
Hypertension
2021,7410155
2010-2011 sectional school students Girls: GM
Q2: 1.40 (0.78,2.51), p-
Medium
ages 15-19 years (IQR) = 5.71
value = 0.261
fromTFFl (2.64)
Q3: 1.01 (0.56, 1.80), p-
N = 940 Boys: GM
value = 0.980
(IQR) = 6.52
Q4: 1.86 (1.08,3.19), p-
(3.09)
value = 0.025
Outcome: Hypertension defined as systolic blood pressure > 130 mmHg and/or diastolic blood pressure
> 80 mmHg.
Comparison: PFOS quartiles are defined as follows: Ql: 1.28-4.86; Q2: 4.87-6.21; Q3: 6.22-7.80; Q4:
1.28-1.86.
Results: Lowest quartile used as the reference group.
Confounding: Sex, age, BMI and physical activity outside school
Linetal., 2016,
3981457
Medium
Taiwan
1992-2000
Cross-
sectional
Adolescents and
young adults ages
12-30
N = 848
Serum
GM = 6.44 (95%
CI: 6.05-6.89)
8-OHDG (log-
l-ig/g
creatinine)
CIMT
CD31+ /
CD42a-
(log coiint/|iL)
CD31+ /
CD42a+
(log count/|iL)
CD62E
(log coiint/|iL)
CD62P
(log count/|iL)
Mean by quartiles
8-OHDG: No associations
across exposure groups;
p-trend = 0.102
CIMT
Ql: 0.433 (0.423,0.442)
Q2: 0.437 (0.428, 0.446)
Q3: 0.456 (0.447,0.465)
Q4: 0.453 (0.444, 0.463)
p-trend <0.001
CD31+ / CD42a~: Statistically
significant increase across
exposure groups, 4.65-5.30
(Q3); p-trend = 0.010
CD31+ / CD42a+: Statistically
significant increase across
exposure groups, 8.02-8.54
(Q3); p-trend = 0.010
CD62E, CD62P: No
statistically significant
associations across exposure
groups
D-125
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Confounding: Age, gender, smoking status, BMI, systolic blood pressure, low density lipoprotein, triglyceride, homeostasis model
assessment of insulin resistance, and high sensitivity C-reactive protein
Khalil et al.,
United States Cross- Obese children
Serum
DBP, SBP
Regression coefficient
DBP: 1.17 (-0.40,2.74)
2018,4238547
2016 sectional ages 8-12
2.79
per unit increase in
SBP: 1.53 (-0.46, 3.51)
Low
00
-t
II
z;
(IQR = 2.10)
PFOS
Confounding: Age, race, sex
Koshy et al.,
United States Cross- Children and
Serum
Augmentation
Regression coefficient
Al: -0.24 (-2.02, 2.41)
2017,4238478
2011-2012 sectional adolescents from
3.72
Index (Al)
per ln-unit increase in
BAD: 0.30 (-0.01, 0.62)
Low
the World Trade
(IQR = 2.82)
Brachial Artery
PFOS
PWV:-0.06 (-0.23,0.11)
Center Health
Comparison:
Distensibility
Registry
2.78
(BAD)
(WTCHR)
(IQR = 2.18)
Pulse Wave
N = 308
Velocity
(PWV)
Confounding: BMI category, caloric intake, cotinine concentration, physical activity, race, sex
Pregnant Women
Matilla-
Spain Cohort Pregnant women
Plasma
CRP
Percent median change
CRP
Santander et al.,
2003-2008 from INMA study 6.05(4.51-7.81)
(log 10 mg/dL
by quartiles and per
-8.41 (-18.4, 3.35)
2017,4238432
N = 1,240
)
loglO-unit increase in
By quartile:
Medium
PFOS
Q2: 6.18 (-11.3, 28.4)
Q3: -6.76 (-22.9, 11.6)
Q4: -5.82 (-22.9, 12.7)
Results: Lowest quartile used as the reference group.
Confounding: Sub-cohort, country of birth, pre-pregnancy body mass index, previous breastfeeding, parity, gestational week at blood
extraction, physical activity, relative Mediterranean Diet Score
General Population
Liao et al., 2020, United States Cross- Adults ages
Serum
DBP, SBP,
DBP and SBP:
DBP
6356903
2003-2012 sectional 20+from
12.8 (7.2-22.0)
hypertension
Regression coefficient
Levels <8.20 ng/mL: -2.62
High
NHANES
per loglO-unit increase
(-4.73,-0.51)
N = 6,967 (3,439
in PFOS or around
Levels > 8.20 ng/mL: 1.23
females, 3,528
inflection point
(-0.42, 2.88)
males)
(8.20 ng/mL)
Hypertension: OR by
tertiles
SBP
Per loglO-unit change: 1.35
(0.18,2.53)
D-126
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Hypertension: No statistically
significant associations or
trends by tertiles or age groups
Males
T2: 1.17 (0.93, 1.47)
T3: 1.07 (0.85, 1.34)
Females
T2: 1.08 (0.87, 1.34)
T3: 1.18(0.92, 1.51)
p-value for interaction by
sex = 0.016
Outcome: Hypertension defined as average SBP > 140 mmHg and average DBP > 90 mmHg, or self-reported use of prescribed anti-
hypertensive medication.
Comparison: Tertiles are defined as follows (in ng/mL PFOS): T1 < 8.9; 8.9 < T2 < 18.1; 18.1 < T3.
Results: Lowest tertile used as the reference group.
Confounding: Age, sex, education level, race, diabetes mellitus, consumption of at least 12 alcohol drinks/year, current smoking status, body
mass index, waist circumference, hemoglobin, total cholesterol, estimated glomerular filtration rate (eGFR), dietary intake of sodium, dietary
intake of potassium, and dietary intake of calcium
Mattsson et al.,
Sweden Case-
Rural men
Serum
CHD
OR by quartiles
CHD
2015,3859607
1990-1991, control
N = 462
Cases: 22.8
Q2: 0.82 (0.46, 1.45)
High
2002-2003
(IQR = 10.0)
Q3: 1.30 (0.74,2.26)
Controls: 22.0
Q4: 1.07 (0.6, 1.92)
(IQR = 10.1)
Results: Lowest quartile used as reference.
Confounding: BMI, systolic blood pressure, total cholesterol, HDL, tobacco use
Mobacke et al.,
Sweden Cross-
Adults aged 70
Serum
Left
Regression coefficient
LVEDD: 0.47 (0.08, 0.87)
2018, 4354163
Years not sectional
from the
Mean
Ventricular
per ln-unit increase in
LVMI: 0.12 (-0.73, 0.97)
High
reported
Prospective
(SD) = 14.9
End-Diastolic
PFOS
RWT: -0.01 (-0.01, -0.001)
Investigation of
(8.88)
Diameter
the Vasculature in
(LVEDD)
Uppsala Seniors
(mm)
(PIVUS) study
Left
N = 801
Ventricular
Mass Index
D-127
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
(LVMI)
(g/m27)
Relative Wall
Thickness
(RWT)
Confounding: Sex, systolic blood pressure, antihypertensive medication, high density lipoprotein (HDL) and low-density lipoprotein (LDL),
cholesterol, blood glucose, waist circumference, triglycerides, body mass index (BMI), education levels, exercise habits, smoking, energy,
alcohol intake
Bao et al., 2017, China Cross-
3860099 2015-2016 sectional
Medium
Adults aged 22-
96
N = 1,612 (408
females, 1,204
males)
Serum
24.2 (14.6-37.2)
DBP, SBP,
hypertension
Regression coefficient
per ln-unit change in
PFOS
DBP
Total: 2.70 (1.98, 3.42)
Females: 2.86 (1.51,4.20)
Males: 0.45 (-0.47, 1.36)
Hypertension: OR per In- p-value for interaction by
unit increase in PFOS sex = 0.001
SBP
Total: 4.84 (3.55,6.12)
Females: 6.65 (4.32, 8.99)
Males: 1.50 (-0.17,3.18)
p-value for interaction by
sex < 0.001
Hypertension
Total: 1.24 (1.08, 1.44)
Females: 1.63 (1.24,2.13)
Males: 1.08 (0.90, 1.29)
p-value for interaction by
sex = 0.016
Outcome: Hypertension defined as mean SBP > 140 mmHg and/or DBP > 90 mmHg, and/or use of antihypertensive medications.
Confounding: Age, sex, BMI, education, income, exercise, smoking, drinking, family history of hypertension
Liu et al., 2018,
4238396
Medium
United States
2004-2007
Controlled
trial
Overweight and
obese adults ages
30-70 in the
POUNDS-Lost
study
Plasma
Females: 22.3
(14.3-34.9)
Males: 27.2
(19.9-45.2)
DBP, SBP
Partial Spearman
correlation coefficient
DBP: 0.15; p-value < 0.05
SBP: 0.07
D-128
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
N = 621 (384
females, 237
males)
Confounding: Age, sex, race, education, smoking status, alcohol consumption, physical activity, menopausal status (women only), hormone
replacement therapy (women only), dietary intervention groups
Linetal., 2020,
6311641
Medium
United States
1996-2014
Cohort
Serum
Baseline: 26.7
(17.4-40.3)
Year 2: 27.6
(19.6-38.9)
Year 14: 9.8
(5.9-14.8)
DBP, SBP,
pulse
pressure
(mmHg), and
hypertension
DBP, SBP: Regression
coefficient per log2-
unit increase in PFOS,
or by quartiles
SBP: lifestyle arm, baseline to
year 2: -2.13 mmHg/year
(-3.54, -0.71)
Adults from the
Diabetes
Prevention
Program (DPP)
and Outcomes
Study (DPPOS)
N = 957 at
baseline, 956 at
year 2, and 346 at
year 14
Outcome: Hypertension defined as SBP > 140 mmHg and DBP > 90 mmHg in those without diabetes, SBP > 130 mmHg, and
DBP > 80 mmHg in those with diabetes, self-reported hypertension diagnosis, or use of antihypertensive medication.
Confounding: Sex, age, race/ethnicity, treatment assignment, education, income, marital status, alcohol intake, smoking, and DASH diet
score
DBP, pulse pressure,
hypertension: No statistically
significant associations by
in PFOS or by quartiles timepoint, by quartiles, or by
sex
Hypertension: HR or RR
per log2-unit increase
Mitroetal., United States Cohort Pregnant women
2020,6833625 1999-2005 and their children
Medium at age 3 from
Project Viva
N = 761 mothers
(496 ages < 35,
265 ages >35)
Plasma
24.7 (18.1-33.9)
DBP, SBP, Regression coefficient SBP: (3 = 1.2 (0.3, 2.2);
CRP (mg/L) per log2-unit increase p-value<0.01
Ages <35: 0.6% (-0.7, 1.8)
Ages >35: 2.3% (0.9, 3.6);
p-value < 0.01
DBP, CRP: No statistically
significant associations
Population: For measurements of C-reactive protein, N = 454 mothers (247 ages < 35, 207 ages > 35).
Confounding: age, pre-pregnancy BMI, marital status, race/ethnicity, education, income, smoking, parity; breastfeeding in a prior pregnancy
for BP measurements only
Regression coefficient
per log2-unit increase
in PFOS
Percent difference (%)
per log2-unit increase
PFOS
Pitter et al.,
2020,6988479
Medium
Italy
2017-2019
Cross-
sectional
Adults aged 20-
39 years from
Veneto Region
with PFAS
Serum
3.7 (2.5-5.6)
Male: 4.8 (3.3-
6.9)
Female: 3 (2-4.4)
DBP, SBP,
hypertension
risk
DBP, SBP: Regression
coefficient per ln-unit
increase in PFOS, or by
quartiles
DBP
0.44 (0.20, 0.68)
Q2: 0.32 (-0.08, 0.72)
Q3: 0.30 (-0.12, 0.71)
Q4: 0.57 (0.13, 1.02)
D-129
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
contaminated
drinking water
DBP and SBP:
N = 15,380 (7,428
males, 7,952
females)
Hypertension risk:
N = 15,786 (7,667
males, 8,119
females)
Hypertension risk: OR
per ln-unit increase in
PFOS, or by quartiles
Males: 0.29 (-0.07, 0.64)
Females: 0.51 (0.17,0.84)
SBP
0.57 (0.24, 0.90)
Q2: -0.01 (-0.56,0.53)
Q3: 0.27 (-0.29,0.84)
Q4: 0.60 (0.00, 1.21)
Males: 0.98 (0.47, 1.48)
Females: 0.32 (-0.13,0.77)
Hypertension risk
1.12(1.02, 1.22)
Q2: 0.99 (0.85, 1.16)
Q3: 1.06 (0.91, 1.24)
Q4: 1.12 (0.95, 1.32)
Males: 1.17 (1.05, 1.31)
Females: 1.06 (0.91, 1.24)
Outcome: Hypertension defined as any self-reported diagnosis, use of antihypertensive drugs, or elevated systolic blood pressure
(SBP > 140 mmHg)/diastolic blood pressure (DBP > 90 mmHg).
Results: Lowest quartile used as the reference group.
Confounding: Age, BMI, time-lag between the enrolment and the beginning of the study, gender, physical activity, smoking habits, food
consumption, salt habit, country of birth, alcohol consumption, education level and center in charge of the BP measurement
Liu et al„ 2018, United States Cross-
Adults ages 18+ Serum
Hypertension OR per ln-unit increase Hypertension: 1.08 (0.:
in PFOS
1.33)
4238514 2013-2014 sectional fromNHANES GM(SE) = 5.28
Medium N= 1,871 (1.02)
Outcome: Hypertension defined as average SBP >130 mmHg and average DBP > 85 mmHg, or self-reported use of prescribed anti-
hypertensive medication.
Confounding: Age, gender, ethnicity, lifestyle variables (smoking status, alcohol intake and household income), medications (anti-
hypertensive, anti-hyperglycemic, and anti-hyperlipidemic agents), other components of the metabolic syndrome
Christensen et
al., 2019,
5080398
Medium
United States
2007-2014
Cross-
sectional
Adults ages 20+
from NHANES
N = 2,975
Serum
8.4 (4.8-14.0)
Hypertension OR by quartiles
Hypertension
No statistically significant
associations
D-130
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Outcome: Hypertension defined as SBP >130 mmHg and/or DBP > 85 mmHg, or use of antihypertensive drug in a patient with a history of
hypertension.
Results: Lowest quartile used as the reference group.
Confounding: Age, alcohol intake, family income, MP AH, PFDE, PFHxS, PFOA, PFUnDA, race/ethnicity, smoking status, survey cycle
Donat-Vargas et Sweden
al., 2019,
5080588
Medium
Jeddi et al.,
2021,7404065
Medium
1990-2013
Cohort Adults aged 30-
60 at baseline
N = 187
Plasma
Baseline: 20 (15-
26)
Follow-up: 15
(9.7-21)
Hypertension OR by tertiles or per SD- Hypertension
unit increase in PFOS Baseline OR per increase: 0.71
(0.56, 0.89)
No other statistically
significant associations
Prospective: No statistically
significant associations
Outcome: Hypertension defined as SBP > 140 mmHg or DBP > 90 mmHg, self-reported diagnosis, or use of antihypertensive drugs
Results: Lowest tertile as the reference group.
Confounding: Gender, age, education, sample year, body mass index, smoking habit, alcohol consumption, physical activity, healthy diet
score
Serum Elevated blood OR per ln-unit increase Elevated blood pressure: 1.10
GM (range): 4.54 pressure inPFOS (1.03, 1.17), p-value < 0.05
(< LOQ-142)
Italy Cross- Residents aged
2017-2019 sectional 20-39 from the
PFAS-
contaminated
Veneto region
N = 15,876
Outcome: Elevated blood pressure defined as SBP >130 mmHg or DBP >85 mmHg.
Confounding: Age, gender, time-lag between the beginning of the study and blood sampling center where BP has been measured, education,
number of deliveries, physical activity, country of birth, diet, alcohol intake, and smoking status, and other components of metabolic
syndrome
Fry and Power,
2017,4181820
Medium
Lind et al., 2017,
3858504
Medium
United States Cohort Adults ages Serum Mortality by HR per SD-unit increase
2003-2006 60+from 4.3 ng/g cerebrovascul inPFOS
NHANES (SE = 0.2 ng/g) ar or heart
N= 1,036 diseases
Confounding: Age, education, gender, race/ethnicity, smoking status
Sweden
2001-2004
Cross-
sectional
Adults ages
70+ in Uppsala,
Sweden
Plasma
13.23 (9.95-
17.77)
CIMT, carotid
artery intima-
media
complex grey
CIMT, CIM-GSM:
Regression coefficient
per ln-unit increase in
PFOS
Mortality
0.85 (0.65, 1.12);
p-value = 0.24
CIMT, CIM-GSM,
atherosclerotic plaque: no
statistically significant
associations
D-131
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
N = 1,016 (509
females and 507
males)
Plaque: OR per ln-unit
increase in PFOS
scale median
(CIM-GSM),
carotid artery
atheroscleroti
c plaque
Confounding: Sex, HDL- and LDL- cholesterol and serum triglycerides, BMI, blood pressure, smoking exercise habits, energy and alcohol
intake, diabetes, educational level
Huang et al.,
2018,5024212
Medium
United States
1999-2014
Cross-
sectional
Adults from
NHANES ages
18+
N = 10,859
Serum
12.40 (6.40-
22.60)
CVD, angina
pectoris,
congestive
heart disease,
CHD, heart
attack, stroke,
CRP (mg/L)
OR by quartiles
CRP: Spearman
correlation coefficient
CVD
Q2: 1.04 (0.78, 1.40)
Q3: 1.36 (1.07, 1.74)
Q4: 1.25 (0.92, 1.69)
p-trend = 0.0681
Females: No statistically
significant associations or
trends
Males
Q2: 1.76(1.11,2.80)
Q3: 2.19 (1.37, 3.51)
Q4: 1.92 (1.20, 3.07)
p-trend = 0.0290; p-trend for
sex interaction = 0.0326
Ages < 50: No statistically
significant associations or
trends
Ages > 50
Q2: 1.01 (0.74, 1.38)
Q3: 1.39 (1.08, 1.78)
Q4: 1.27 (0.92, 1.75)
p-trend = 0.0491; p-trend for
age interaction = 0.1228
Angina pectoris: No
association by quartiles, no
significant trend;
p-trend = 0.4211
D-132
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Congestive heart disease: No
association by quartiles, no
significant trend;
p-trend = 0.9462
CHD: No association by
quartiles, no significant trend;
p-trend = 0.0910
Heart attack
Q2: 1.30 (0.90, 1.87)
Q3: 1.56 (1.01,2.43)
Q4: 1.53 (0.96,2.45)
p-trend = 0.1026
Stroke: No association by
quartiles, no significant trend;
p-trend = 0.3084
CRP: -0.006;
p-value = 0.6062
Comparison: Age groups were defined as < 50 years and > 50 years.
Results: Lowest quartile used as the reference group.
Confounding: Age, sex, race/ethnicity, family poverty income ratio, education levels, physical activity levels, BMI, alcohol drinking status,
smoking status, diabetes, hypertension, family history of CVD, total energy intake, log-transformed levels of serum cotinine, log-transformed
levels of serum total cholesterol
Cardenas et al.,
2019,5381549
Medium
United States
1996-2014
Controlled
trial
Prediabetic adults
ages 25+ from
DPP and DPPOS
N = 877
Plasma
GM
(IQR) = 26.38
(22.8)
MVD,
nephropathy,
neuropathy,
retinopathy
OR per log2-unit
increase baseline PFOS
MVD: lifestyle arm: 1.37
(1.04, 1.84)
Nephropathy, neuropathy,
retinopathy: No statistically
significant associations
Confounding: Sex, race/ethnicity, baseline age, marital status, education, income, smoking history, BMI, maternal diabetes, paternal
diabetes, treatment assignment; baseline fasting glucose and HbAlc levels for microvascular disease only
D-133
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Hutcheson et al.,
2020,6320195
Medium
Osorio-Yanez et
al., 2021,
7542684
Medium
United States
2005-2006
Cross-
sectional
Adults from C8
Health Project
N = 48,206
Serum
With diabetes:
21.4(13.8-31.9)
Without diabetes:
20.1 (13.5-29.0)
Stroke OR per ln-unit increase
PFOS
0.90 (0.82, 0.98);
p-value = 0.02
Confounding: Age, BMI, C-reactive proteins, diabetes duration, eGFR, HDL, LDL, history of smoking, race, sex
United States
1999
Cohort Prediabetic adults Plasma CAC
ages 25+ enrolled 27.55 (Agastston
in the DPP trial (IQR = 19.30) score), AsAC
N = 666
OR per doubling in
PFOS
CAC (11-400): 1.20 (0.94,
1.53)
CAC (>400): 1.49(1.01,
2.21), p-value < 0.05
AsAC: 1.67(1.10, 2.54), p-
value < 0.05
Results: CAC
Confounding:
<11 used as reference group.
Sex, age, body mass index, race/ethnicity, cigarette smoking, education, treatment assignment, statin use.
He et al., 2018,
4238388
Low
United States
2003-2012
Cross-
sectional
Adults ages 20+
from NHANES
N= 3,948
(females) and
3,956 (males)
Serum
Female Mean
(SE)= 14.51
(0.26)
Male Mean
(SE) = 20.80
(0.32)
DBP, SBP Percent difference in log-
transformed outcome per
interquartile ratio
increase PFOS by
quartiles
DBP
Females:
Q2: -1.12 (-2.55, 0.34)
Q3: 0.00 (-1.45, 1.59)
Q4: 1.47 (-0.11,3.08)
p-trend = 0.022
Males: No statistically
significant associations;
p-trend = 0.119
SBP:
Females:
Q2: 0.11 (-0.90, 1.02)
Q3: 0.34 (-0.56, 1.36)
Q4: 1.13 (0.23,2.16)
Males: No statistically
significant associations;
p-trend = 0.171
Comparison: Logarithm base not specified.
Results: Lowest quartile used as the reference group. Interquartile ratio = 75th/25th percentiles of serum PFOS: 3.08 ng/mL.
Confounding: None listed
D-134
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Yang et al.,
2018,4238462
Low
China
Years not
reported
Cross-
sectional
Adult men
N = 148
Serum
3.00 (Range:
0.3-14.6)
DBP, SBP,
hypertension
Regression coefficient
per log-unit increase in
n-PFOS
DBP, SBP, hypertension: no
statistically significant
associations
Outcome: Hypertension evaluated by individual BP components
Comparison: Logarithm base not specified.
Confounding: Age
Hypertension: OR
comparing above or
below median
Chenetal., Croatia Cross- Adults aged 44- Plasma DBP, SBP Regression coefficient DBP: 1.42 (-0.95,3.79)
2019,5387400 2007-2008 sectional 56 GM=8.91 per ln-unit increase
Low N = 122 (Range = 2.36- PFOS SBP: 1.40 (-3.46,6.25)
33.67)
Confounding: Age, sex, education, socioeconomic status, smoking, dietary pattern, physical activity
Graber etal.,
United States Cross-
Members of
Serum Cardiovascular
OR per unit increase in
Any condition
2019, 5080653
2016-2017 sectional
community with
5.66 (3.09-9.28) conditions,
PFOS
1.08 (0.98, 1.21)
Low
Confounding: Age, BMI
exposed water
supply
(Paulsboro, NJ)
ages 12+
N = 105
self-reported
Occupational Populations
Christensen et
United States Cross-
Male anglers ages
Serum Cardiovascular
OR per unit increase in
Any condition: 1.00 (0.98,
al., 2016,
2012-2013 sectional
50+
19.00 (9.80- condition
PFOS
1.02)
3858533
N = 154
28.00) (any), CHD,
CHD: 1.01 (0.98, 1.03)
Low
hypertension
Hypertension: 0.99 (0.96,
1.01)
Outcome: Hypertension was self-reported
Confounding: Age, BMI, work status, and alcohol consumption
Notes: AI = augmentation index; BAD = brachial artery distensibility; BMI = body mass index; CAC = coronary artery calcium; CHD = coronary heart disease; CI = confidence
interval; CIM-GSM = carotid artery intima-media complex grey scale median; CIMT = carotid artery intima-media thickness (mm); CMR = cardiometabolic risk score; CRP = C-
reactive protein; CVD = cardiovascular disease; DBP = diastolic blood pressure (mmHg); DPP = Diabetes Prevention Program; DPPOS = Diabetes Prevention Program
Outcomes Study; GM = geometric mean; HDL = high density lipoprotein cholesterol; HELIX = Human Early-Life Exposome; IQR = Interquartile range; HOME = Heath
D-135
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Outcomes and Measures of the Environment; LDL = low-density lipoprotein-cholesterol; LVEDD = left ventricular end-diastolic diameter (mm); LVMI = left ventricular mass
index (g/m2); MVD = microvascular disease; NHANES = National Health and Nutrition Examination Survey; PFOA = perfluorooctanoic acid; PFDE = perfluorodecanoic acid;
PFHxS = perfluorohexane sulfonic acid; MP AH = 2-(N-methyl-PFOSA) acetate; PFNA = perfluorononanoic acid; PFUnDA = perfluoroundecanoic acid; PWV = pulse wave
velocity; OR = odds ratio; RWT = relative wall thickness; SBP = systolic blood pressure (mmHg); SD = standard deviation; SE = standard error; TFF1 = Tromse Fit Futures 1
a Exposure reported as median (25th-75th percentile) in ng/mL unless otherwise specified.
b Results reported as effect estimate (95% confidence interval) unless otherwise specified.
c Confounding indicates factors the models presented adjusted for.
D-136
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.5.2 Serum Lipids
D.5.2.1 Forest Plots
Confidence Exposure Study
ftfcfl flafciMKM iMfcfe ftaq* Exposure Uiiii Sub-papulaton Comparison EE
Effect Estimate
¦02 0X1 02 0.4 0-6 0.8 1.0 1.2
. , _ Geometric mean-8.91 naftnL.
i Chen at al.. Cross - Coefficient (per
Sd,™ »'• ,m™ra ttSKSt. " ™f"-~
I
1
1"
1
Regression
H.S. Liu eJ aL. _ Cress Geornetnc mean-S.28 ngimL Coefficient (per
2016 iMnm sectional (SE-1.02) ~ fc^uret increase lzz
in PFOS)
J
1
| •
1
Regression
He et al., 2016 Serum Cross Fernafc mean - 14.51 n(jlmL Females QWier T' 1 Qra ikT ^
sectional (S£=0^r^nL)
t
1
t"
1
Regressicri
coefficient (03 KB
01 per 1 10 ratio 0
ngiVnL increase)
|
1
t"
1
Regression
coefficient (G4 «t
Ol per 1IO ratio 0 01
n^WiL increase)
(
1
I*
1
Regression
Maie mean - 20.83 ngi'niL M^es oT^Tlo^lin -0-01
(SE =
\
1
-t
I
Regression
coefficient |03 ks
Ol per 1 IO ratio 0
i
i
T
1
Regression
coefficient
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Interactive figure and additional study details available on Tableau.
D.5.2.2 Tables
Table D-14. Associations Between PFOS Exposure and Serum Lipid Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels"
Outcome
Comparison
Select Resultsb
Li et al., 2021,
7404102
High for
gestation, birth,
and childhood
exposures (3-
year and 8-year)
Medium for
exposure at 12-
year follow-up
Pregnant women
and their children
followed-up at
birth and ages 3,
8, and 12 years
from HOME
Study
Gestation:
N = 203
At birth: N= 124
Age 3: N = 137
Age 8: N = 165
Age 12: N = 190
Children
Maternal
serum
Gestation:
12.9 (8.9-
18.0)
Cord serum
At birth: 4.2
(3.0-6.5)
Serum
At age 3: 6.2
(4.5-9.9)
At age 8: 3.6
(2.8-1.7)
At age 12: 2.4
(1.8-3.2)
Levels (mg/dL) of
triglycerides and HDL;
triglycerides to HDL
ratio
Regression
coefficient per
log2-unit IQR
increase in PFOS
United States Cohort Pregnant women Maternal Levels (mg/dL) of Regression Triglycerides
2003-2006 and their children serum triglycerides and HDL; coefficient per Gestation: 0.0 (-0.2,0.2)
At birth: 0.1 (-0.1, 0.3)
Age 3:-0.1 (-0.3, 0.1)
Age 8: 0.1 (-0.1, 0.3)
Age 12: 0.1 (-0.1,0.3)
HDL
Gestation: 0.9 (-2.3, 4.1)
At birth: 0.9 (-2.6, 4.3)
Age 3: 0.4 (-3.5,4.4)
Age 8: 3.8 (-0.2, 7.7)
Age 12: 6.0 (1.9, 10)
Triglycerides to HDL
ratio
Gestation: 0.0 (-0.2, 0.2)
At birth: 0.1 (-0.1, 0.3)
Age 3:—0.1 (-0.3,0.1)
Age 8: 0.1 (-0.1, 0.3)
Age 12: 0.1 (-0.1,0.3)
HOME = Heath Outcomes and Measures of the Environment
Confounding: visit, visit*PFAS, maternal age, maternal education, maternal pre-pregnancy BMI, gestational serum cotinine
concentrations, and parity; and child age, sex, race, and pubertal stage. Additional confounding for analyses at age 3, age 8, and age 12:
Breastfeeding duration.
Lin et al., 2009, United States Cross-sectional Adolescents ages Serum
1290820 1999-2000 12-20 years from Mean
Medium and 2003- NHANES (SEM) = 3.11
2004 N = 474
Metabolic syndrome
HDL cholesterol and
metabolic syndrome
triglycerides
OR per log 10-
unit increase in
PFOS
Metabolic syndrome
HDL cholesterol
Model 4: 0.89 (0.51,
1.55)
D-138
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(0.05) loglO-
ng/mL
Model 5: 1.38 (0.61,
3.14)
Metabolic syndrome
triglycerides
Model 4: 0.95 (0.50,
1.80)
Model 5: 0.78 (0.41,
1.49)
Outcome: Metabolic syndrome HDL cholesterol defined as HDL < 1.04 mmol/L; metabolic syndrome triglycerides defined as
triglycerides > 1.24 mmol/L.
Confounding: Model 4: Age, sex, race, health behaviors (smoking status, alcohol intake, and household income), measurement data
(CRP and HOMA/insulin) and medications; additional confounding for model 5: Other components of the metabolic syndrome.
Nelson et al.,
2010,1291110
Medium
United States Cross-sectional
2003-2004
Level (mg/dL) of HDL Regression
coefficient by
quartiles
HDL
Q4: 3.7 (-0.5, 7.9)
Adolescent girls Serum
ages 12-19 years Level not
from NHANES reported
N not reported
Results: Lowest quartile used as the reference group. Quartile analyses discussed in-text only and quantitative values provided for Q4
only.
Confounding: Not reported.
Geiger et al.,
2014,2850925
Medium
United States
1999-2008
Cross-sectional
Adolescents ages
12-18 years from
NHANES
N = 815
Plasma
Mean
(SE) = 17.7
(0.7)
Levels (mg/dL) of TC,
LDL, HDL, and
triglycerides; elevated
TC; elevated LDL;
depressed HDL;
elevated triglycerides
Lipid levels:
Regression
coefficient per
ln-unit increase
in PFOS, Mean
change by
tertiles
Elevated or
depressed: OR
per ln-unit
increase in
PFOS, or by
tertiles
TC: 0.06 (0.02,0.1)
T2: 3.37 (-1.39, 8.13)
T3: 5.85 (0.1, 11.61)
p-trend = 0.051
HDL
T2: 1.62 (-0.54, 3.78)
T3: -0.01 (-2.06,2.04)
p-trend = 0.970
LDL: 4.28 (1.6, 6.95)
T2: 2.7 (-1.39, 6.78)
T3: 6.99 (1.99, 11.98)
p-trend = 0.0081
TG: -1.85 (-5.61, 1.91)
T2: -4.79 (-11.09, 1.5)
T3: -5.55 (-12.26, 1.16)
p-trend = 0.110
D-139
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Elevated TC: 1.35 (1.11,
1.64)
T2: 1.35 (0.94, 1.95)
T3: 1.53 (1.07,2.19)
p-trend = 0.018
Depressed HDL: 1.03
(0.7, 1.53)
T2: 0.88 (0.52, 1.5)
T3: 0.99 (0.58, 1.7)
p-trend = 0.987
Elevated LDL: 1.48
(1.15, 1.9)
T2: 1.43 (0.91,2.24)
T3: 1.76 (1.1,2.82)
p-trend = 0.018
Elevated TG: 0.9 (0.56,
1.43)
T2: 0.82 (0.46, 1.45)
T3: 0.64 (0.3, 1.37)
p-trend = 0.242
Outcome: Elevated TC defined as TC > 170 mg/dL; elevated LDL defined as LDL >110 mg/dL; depressed HDL defined as
HDL < 40 mg/dL; elevated triglycerides defined as triglycerides > 150 mg/dL.
Results: Lowest tertile used as the reference group. Regression coefficient for continuous analysis of HDL not reported.
Confounding: Age, sex, race-ethnicity, BMI categories, annual household income categories, activity level, and serum cotinine
Frisbee et al.,
2010,1430763
Medium for TC,
GDL-C, fasting
TG; low for LDL
United States
2005-2006
Cross-sectional
Children and
adolescents ages
1.0 to 17.9 years
in the C8 Health
Project
N = 12,470
Serum
Mean
(SD) = 22.7
(12.6)
Abnormal TC,
abnormal HDL,
abnormal LDL, and
abnormal fasting
triglycerides
OR by quintiles Abnormal TC
Q2: 1.3 (1.1, 1.4)
Q3: 1.3 (1.2, 1.5)
Q4: 1.3 (1.2, 1.6)
Q5: 1.6 (1.4, 1.9)
Abnormal HDL
Q2: 0.9 (0.8, 1.1)
Q3: 0.8 (0.7, 1.0)
Q4: 0.8 (0.7, 0.9)
Q5: 0.7 (0.6,0.9)
D-140
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Abnormal LDL
Q2: 1.2 (1.0, 1.5)
Q3: 1.2 (1.0, 1.5)
Q4: 1.3 (1.1, 1.6)
Q5: 1.6 (1.3, 1.9)
Abnormal fasting
triglycerides
Q2: 1.3 (0.9, 1.8)
Q3: 1.0 (0.7, 1.4)
Q4: 1.1 (0.7, 1.6)
Q5: 1.2 (0.8, 1.5)
Outcomes: Abnormal TC defined as TC > 170 mg/dL; abnormal HDL defined as HDL < 40 mg/dL; abnormal LDL calculated for
participants with a triglyceride level < 400 mg/dL regardless of fasting status and defined as LDL >110 mg/dL; fasting triglycerides
defined as self-reported fasting > 6 hours before phlebotomy, and abnormal fasting triglycerides defined as fasting
triglycerides >150 mg/dL.
Results: Lowest quintile used as the reference group.
Confounding: Age, estimated time of fasting, BMI z-score, sex, regular exercise
Timmermann et Denmark
al., 2014, 1997
2850370
Medium
Cross-sectional Children
ages 8-10 from
Danish
component of
EYHS
N = 400 normal
weight, N = 59
overweight
Plasma
41.5
(Range = 6.2-
132.5)
Triglycerides (mmol/L) Percent change Normal weight: -0.5
per 10-unit
increase PFOS
(-3.2,2.4), p-
value = 0.75
Overweight: 8.6 (1.2,
16.5), p-value = 0.02
p-value for PFOS-BMI
interaction = 0.02
Confounding: Sex, age, ethnicity, paternal income, fast-food consumption, and fitness
Maisonet et al.,
2015,3981585
Medium for TC
and HDL at age 7
and all lipids at
age 15
Low for
Triglycerides and
LDL at age 7
United
Kingdom
1991-1992
Case-control
Pregnant women
and their
daughters
followed-up at
ages 7 and 15
from ALSPAC
Age 7: N = 111
Age 15: N = 88
Serum
20.5
(Range = 7.6-
38.2)
Levels (mg/dL) of TC,
LDL, HDL, and
triglycerides (ln-
mg/dL)
Regression
coefficient per
unit increase in
PFOS in each
tertile of
exposure
TC
Age 7
Tl: 0.30 (-3.10, 3.70)
T2: 2.09 (-0.64, 4.82)
T3:-0.10 (-0.73, 0.54)
Age 15
Tl: 1.64 (-2.20, 5.48)
T2: 3.41 (0.37,6.45)
T3: -0.77 (-1.40,-0.13)
LDL
D-141
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Age 7
Tl: 0.37 (-2.34, 3.08)
T2: 1.02 (-1.15, 3.19)
T3: 0.02 (-0.48, 0.53)
Age 15
Tl: 1.91 (-1.34, 5.17)
T2: 2.09 (-0.50, 4.67)
T3: -0.54 (-1.08,
-0.003)
HDL
Age 7
Tl: 0.76 (-0.79, 2.31)
T2: 0.22 (-1.03, 1.46)
T3: -0.04 (-0.33,0.25)
Age 15
Tl
T2
T3
-0.55 (-2.34, 1.24)
1.15 (-0.27, 2.57)
-0.18 (-0.47, 0.12)
ALSPAC = Avon Longitudinal Study of Parents and Children
Confounding: Previous live births, maternal education, and maternal age at delivery
Triglycerides
Age 7
Tl: -0.031 (-0.085,
0.023)
T2: 0.008 (-0.035, 0.052)
T3:-0.004 (-0.015,
0.006)
Age 15
Tl: 0.012 (-0.032,0.056)
T2: 0.016 (-0.019,0.051)
T3:-0.004 (-0.011,
0.004)
Zeng et al., 2015, Taiwan Cross-sectional Children
2851005 2009-2010 ages 12-15
Medium N = 225
Serum Levels (ng/dL) of TC, Regression
Median = 28.8 LDL, HDL, and coefficient per
among males, triglycerides ln-unit increase
29.9 among PFOS
females
TC: 0.31 (0.18,0.45)
p-value <0.001
LDL: 0.28 (0.18, 0.38)
p-value <0.001
HDL: -0.01 (-0.07,
0.05)
D-142
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Domazet et al.,
2016,3981435
Medium
p-value = 0.72
Triglycerides: 0.19 (0,
0.38)
p-value = 0.05
Confidence: Results for TG and LDL considered low confidence because of a lack of fasting prior to blood sample collection.
Confounding: Age, gender, BMI, parental education level, exercise, environmental tobacco smoke exposure0
Denmark
1997-2009
Cohort
Members of the
Plasma
Levels (mmol/L) of TG Percent change
Age 9 to 15:
EYHS evaluated
Median at
in TG at age 15
-0.7 (-5.03, 3.77)
at ages 9 and 15
9 = 44.5
or 21 per 10 unit
Age 9 to 21: -1.98
(N = 260), 9 and
(male) or 39.9
increase in PFOS
(-8.17,4.75)
21 (N= 175), or
(female)
at age 9 or 15
Age 15 to 21: 0.77
15 and 21
Median at
(-8.28, 10.71)
(N = 171)
15 = 22.3
(male) or 20.8
(female)
Median at
21 = 11.9
(male) or 9.1
(female)
Confounding: Sex, age, and TG levels at baseline age; ethnicity, maternal parity, and maternal income in 1997 (9 years of age). Waist
circumference was adjusted for height in order to account for body size.
Manzano-
Salgado et al.,
2017,4238509
Medium
Jain et al., 2018,
5079656
Medium
Spain
2003-2008
Cohort
Pregnant women Maternal Levels (z-score) of TC.
and their children plasma during LDL, HDL, and TG
(age 4) from 1st trimester
INMA study GM = 5.80
N = 627
Confidence: Results for TG and LDL considered low confidence because of a lack of fasting prior to blood sample collection.
Confounding: Maternal region of residence, country of birth, previous breastfeeding, age, pre-pregnancy BMI; age/sex of child
Regression
coefficient per
log2-unit
increase PFOS
TC: 0.02 (-0.10,0.15)
LDL: 0.02 (-0.10,0.15)
HDL: -0.03 (-0.14,
0.09)
TG: 0.05 (-0.06,0.17)
United States
2013-2014
Cross-sectional
Children ages 6-
11
N = 458
Serum Levels (loglO-mg/dL) Regression
GM = 2.67 for of TC, HDL, and non- coefficient per
linear PFOS, HDL loglO-unit
1.35 for lm- increase PFOS
PFOS
Linear PFOS
TC: 0.02738
p-value = 0.03
Non-HDL: -0.00357
p-value = 0.4
HDL: 0.04631
p-value = 0.1
lm-PFOS
TC: 0.01241
p-value = 0.22
D-143
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Confounding: Gender, race/ethnicity, age, poverty income ratio, body mass index percentiles,
smoke
Non-HDL: -0.00661
p-value = 0.04
HDL: 0.04612
p-value = 0.05
fasting time, and exposure to secondhand
Kang et al..
4937567
Medium
2018,
Mora et al..
4239224
Medium
Jensen et al.,
2020,6833719
Medium
Korea
2012-2014
Cross-sectional
Serum Levels of TC (mg/dL),
Median = 5.68 LDL (mg/dL), and TG
(ln-mg/dL)
Children aged 3-
18 from Korea
Environmental
Health Survey in
Children and
Adolescents
(KorEHS-C)
N = 147
Results: LDL and TG evaluated at ages 7-18 only (N = 117)
Confounding: Age, sex, BMI z-score, household income, second-hand smoking
Regression
coefficient per
ln-unit increase
PFOS
TC: -0.45 (-10.67, 9.77)
LDL: 2.51 (-6.88, 11.89)
TG: -0.020 (-0.19,0.15)
All p-value >0.5
2018,
United States Cohort and Pregnant women Prenatal
1999-2010 cross-sectional and their children maternal
from Project Viva plasma
N = 512 prenatal, Median = 24.6
596 mid-
childhood Mid-
childhood
plasma
Median = 6.2
Levels (mg/dL) of TC,
HDL, LDL, and TG
Regression
coefficient per
IQR increase in
PFOS
Prenatal:
TG: -1.4 (-4.6, 1.8)
Boys: 1.0 (-2.2,4.2)
Girls: -4.2 (-9.2, 0.8)
p-value for interaction by
sex = 0.04
Mid-childhood:
TC: 1.8 (-0.2, 3.7)
HDL: 1.5 (0.4, 2.5)
TG: -2.5 (-4.3, -0.6)
Boys: 0.5 (-1.8, 2.9)
Girls: 4.0 (0.3, 7.8)
No other statistically
significant associations
Confounding: maternal education, prenatal smoking, gestational age at blood draw (for prenatal data), and child's sex, race/ethnicity, and
age at lipids/ALT measurements
Denmark
2010-2012
Cohort
Pregnant women
and their children
assessed at
3 months and
18 months
Maternal
serum
Median = 8.04
Levels (standard
deviation score) of TC,
LDL, HDL, and TG
Regression
coefficient per
unit increase in
PFOS
All associations were
between -0.07 and 0.05,
all with p-values > 0.05
D-144
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
N = 260 at
3 months, 83 at
18 months
Confounding: Maternal age, parity, pre-pregnancy BMI, pre-pregnancy BMI2, education, smoking, sex, and lipid outcome at 3 months
Spratlen et al.,
United States Cross-sectional Pregnant women
Cord blood
Levels (mg/dL) of TC,
Percent change
TC: 0.062
2020,5915332
2001-2002 and their children
Median = 6.32 total lipids, and TG in
per 1% increase
(-0.004,0.13)
Medium
from the
cord blood
inPFOS
Columbia
Total lipids: 0.067
University World
(0.005,0.129)
Trade Center
p-value <0.05
birth cohort
N = 222
TG: 0.086
(-0.036, 0.21)
Confounding: Maternal age, child sex, maternal education, maternal race, parity, pre-pregnancy BMI, marital status, family smoking, and
gestational age
Averina et al.,
Norway Cross-sectional First level high
Serum
Levels (mmol/L) of
Regression
TC: 0.38 (0.10,0.66), p-
2021,7410155
2010-2011 school students
Girls: GM
TC, HDL, LDL, and
coefficient per
value = 0.008
Medium
ages 15-19 years
(IQR) = 5.71
TG
loglO-unit
fromTFFl
(2.64)
increase in PFOS
HDL: 0.08 (-0.03, 0.20),
N = 940
Boys: GM
p-value = 0.152
(IQR) = 6.52
(3.09)
LDL: 0.30 (0.05, 0.55),
p-value = 0.021
TG: 0.006 (-0.18,0.20),
p-value = 0.947
TFF1 = Tromso Fit Futures 1
Confounding: Sex, age, BMI, and lifestyle and diet variables
Blomberg et al.,
Faroe Islands Cohort and Children from the
Serum
Levels (mmol/L) of
Regression
TC, age 9 (PFOS age 9)
2021,8442228
Recruitment: cross-sectional Faroese Birth
Birth: 2.87
TC, HDL
coefficient per
0.15 (0.025,0.27), p-
Medium for HDL
2007-2009 Cohort 5 at birth,
(2.13-4.04)
log2-unit
value < 0.05
and TC
18 months, and
Female: 2.82
increase in PFOS
Females: 0.25 (0.077,
Low for LDL
9 years
(2.04-3.86)
0.43), p-value < 0.05
and TG
Birth: N = 459
Male: 2.93
Males: 0.05 (-0.12,0.22)
(219 female, 240
(2.19-4.10)
p-value for interaction by
male)
sex = 0.104
18 months:
18 months:
N = 334
6.81 (4.38-
HDL, age 9 (PFOS
9 years: N = 366
9.82)
age 9)
D-145
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
9 years: 3.08
(2.42-4.31)
0.077 (0.03,0.12), p-
value < 0.05
Females: 0.07 (0.0017,
0.14), p-value < 0.05
Males: 0.083 (0.018,
0.15), p-value < 0.05
p-value for interaction by
sex = 0.788
Levels at
5 years and by
sex at
18 months and
9 years not
reported
Confounding: Child sex and maternal education; analyses except PFAS at 9 years additionally adjusted for maternal smoking during
pregnancy, maternal pre-pregnancy BMI, and parity
Canova et al.,
Italy
Cross-sectional Adolescents aged Serum
Levels (ng/mL) of TC, Regression
TC
2021,10176518 2017-2019
Medium for TC,
HDL; Low for
LDL, TG
14 to 19 years
and children aged
8 to 11 years
from health
surveillance
program in
Veneto Region
Adolescents:
N = 6,669
Children:
N = 2,693
Adolescents:
3.3 (2.2-4.9)
Children: 2.2
(1.6-3.0)
HDL, LDL,
triglycerides
coefficient per Adolescents: 3.32 (2.20,
ln-unit increase 4.45)
in PFOS Children: 6.22 (4.32,
8.13)
HDL
Adolescents: 1.17 (0.71,
1.63)
Children: 1.91 (1.10,
2.73)
LDL
Adolescents: 2.66 (1.70,
3.62)
Children: 4.52 (2.80,
6.23)
Triglycerides
Adolescents: -0.02
(-0.04, 0.00)
Children: -0.01 (-0.04,
0.02)
Confounding: Age, gender, country of birth, data on food consumption, degree of physical activity, salt intake, smoking status (for
adolescents only), time-lag between the beginning of the study and the date of enrollment.
D-146
-------
DRAFT FOR PUBLIC COMMENT
Papadopoulou et
United Cohort
Mother-child
Maternal
al., 2021,
Kingdom,
pairs from the
plasma
9960593
France, Spain,
HELIX Project,
(prenatal)
Medium
Lithuania,
children
6.15 (3.99-
Norway,
followed-up
9.16)
Greece
around age 8
Recruitment
(range 6-12)
Plasma
1999-2010,
N = 1,101
(childhood)
Follow-up:
1.93 (1.22-
2013-2015
3.11)
D-147
MARCH 2023
Levels (z-scores) of
HDL, LDL, and
triglycerides
Regression
coefficient per
doubling in
PFOS, or by
quartiles
HDL
Maternal PFOS: 0.06
(-0.06,0.18)
Q2:-0.13 (-0.33, 0.07)
Q3: -0.06 (-0.29, 0.17)
Q4:-0.18 (-0.47, 0.12)
p-trend = 0.577
Childhood PFOS: 0.00
(-0.08, 0.08)
Q2: 0.23 (0.04,0.41)
Q3: 0.33 (0.11,0.54)
Q4: 0.37 (0.11, 0.63)
p-trend = 0.009
LDL
Maternal PFOS: -0.03
(-0.15,0.09)
Q2: -0.05 (-0.26, 0.15)
Q3: -0.11 (-0.35,0.12)
Q4: 0.09 (-0.21,0.39)
p-trend = 0.990
Childhood PFOS: 0.05
(-0.03,0.13)
Q2: 0.06 (-0.13,0.25)
Q3: 0.15 (-0.06,0.37)
Q4: 0.12 (-0.14, 0.38)
p-trend = 0.210
Triglycerides
Maternal PFOS: -0.07
(-0.19,0.05)
Q2: -0.07 (-0.27, 0.14)
Q3:-0.19 (-0.43, 0.04)
Q4:-0.14 (-0.44, 0.16)
p-trend = 0.191
Childhood PFOS: 0.04
(-0.04,0.12)
Q2: 0.02 (-0.17,0.21)
Q3: 0.03 (-0.19,0.24)
Q4: 0.13 (-0.14,0.39)
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
p-trend = 0.256
Comparison: Maternal PFOS quartiles are defined as follows: Ql: 0.28-3.98; Q2: 3.99-6.15; Q3: 6.15-9.15; Q4: 9.16-47.98; childhood
PFOS quartiles are defined as follows: Ql: 0.00-1.22; Q2: 1.22-1.92; Q3: 1.93-3.10; Q4: 3.11-33.83.
Results: Lowest quartile used as the reference group.
Confounding: Maternal age and education, pre-pregnancy BMI, parity, cohort, child ethnicity, age, child gender, PFHxS, PFNA, PFOA
Tian et al., 2021, China
7026251 2012
Medium
Cohort
Pregnant women Maternal
and their newborn plasma
children from the 10.5 (7.3 7-
S-MBCS 16.3)
N = 306
Levels (ln-mg/dL) of
TC, LDL, HDL, and
triglycerides
Regression
coefficient per
ln-unit increase
in PFOS, or by
tertile
TC
Per ln-unit: -0.10 (-0.18,
-0.02), p-value = 0.018
T2: -0.09 (-0.20, 0.03)
T3:-0.15 (-0.27,-0.03),
p-value <0.05
p-trend < 0.05
LDL
Per ln-unit: -0.07 (-0.18,
0.03), p-value = 0.164
T2:-0.12 (-0.27, 0.03)
T3: -0.09 (-0.24,0.06)
HDL
Per ln-unit: -0.11 (-0.21,
-0.02), p-value = 0.021
T2: -0.11 (-0.25,0.03)
T3:-0.17 (-0.31,
-0.031), p-value < 0.05
p-trend < 0.05
Results: Lowest tertile used as reference group.
Confounding: Maternal age, pre-pregnancy BMI, household income, infant sex, gestational age.
Triglycerides
Per ln-unit: -0.05 (-0.14,
0.04), p-value = 0.287
T2: -0.08 (-0.21,0.06)
T3: -0.02 (-0.16,0.11)
Pregnant Women
Starling et al.,
2014,2850928
Medium for TC,
HDL, and LDL
Norway
2003-2004
Cross-sectional
Women in mid
pregnancy
(median = 18
weeks of
Plasma
13.03 (10.31-
16.60)
Levels (mg/dL) of TC,
HDL, LDL, and
triglycerides (ln-
mg/dL)
Regression
coefficient per
ln-unit or IQR
increase in
TC
Per ln-unit: 8.96 (1.70,
16.22)
D-148
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Low for gestation) from PFOS, or by Per IQR: 4.25 (0.81,
Triglycerides MoBa quartiles 7.69)
N = 891 Q2:-3.35 (-10.34, 3.64)
Q3: 3.06 (-4.93, 11.05)
Q4: 7.59 (-0.42, 15.60)
HDL
Per ln-unit: 4.39 (2.37,
6.42)
Per IQR: 2.08(1.12,
3.04)
Q2: 1.96 (-0.39,4.31)
Q3: 2.49 (0.00, 4.97)
Q4: 4.45 (2.04, 6.86)
LDL
Per ln-unit: 6.48 (-0.07,
13.03)
Per IQR: 3.07 (-0.03,
6.18)
Q2: -3.23 (-9.28, 2.83)
Q3: 2.60 (-4.49,9.70)
Q4: 5.51 (-1.62, 12.64)
Triglycerides
Per ln-unit: -0.02 (-0.09,
0.04)
Per IQR: -0.01 (-0.04,
0.02)
Q2: 0.00 (-0.06, 0.07)
Q3: -0.03 (-0.10, 0.05)
Q4: 0.00 (-0.09, 0.04)
Results: Lowest quartile used as reference group.
Confounding: Age, pre-pregnant body mass index, nulliparous or inter-pregnancy interval, duration of breastfeeding previous child,
education completed, current smoking at mid-pregnancy, gestational weeks at blood draw, and oily fish consumed daily.
Skuladottiretal., Denmark Cross-sectional Pregnant women Serum
2015,3749113 1988-1989 N= 854 Mean = 22.3
Medium
Levels (mmol/L) of TC Regression
coefficient by
quintile
Q2: 0.24 (-0.04, 0.53)
Q3: 0.22 (-0.07,0.50)
Q4: 0.35 (0.06, 0.64)
Q5: 0.44 (0.15,0.74)
p-trend = 0.004
D-149
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Matilla-
Santander et al.,
2017,4238432
Medium
Starling et al.,
2017,3858473
Medium
Results: Lowest quintile used as reference group.
Confounding: Age, parity, education, smoking and pre-pregnancy BMI, total caloric intake, and intake of vegetables, meat, and meat
products
TC: 0.88 (-0.53, 2.37)
TG: -5.86 (-9.91, -1.63)
Spain Cohort Pregnant women Plasma Levels of TC (mg/dL), Percent change
2003-2008 from the Spanish Median =6.05 TG (loglO-mg/dL), and in median lipid
INMA birth C-reactive protein level per loglO-
cohort (loglO-mg/dL) unit increase in
N = 1240 PFOS
Confidence: TG results considered low confidence because of a lack of fasting prior to blood sample collection.
Confounding: Sub-cohort, country of birth, pre-pregnancy body mass index, previous breastfeeding, parity, gestational week at blood
extraction, physical activity, and relative Mediterranean Diet Score
United States
2009-2014
Cohort
Pregnant women
ages 16-45 from
the Healthy Start
study
N = 598
Serum
Median = 2.4
Levels of HDL
(mg/dL) and TG (ln-
mg-dL)
Regression
coefficient per
ln-unit increase
PFOS
HDL: 0.79 (-0.68, 2.27)
TG: 0.004 (-0.033,
0.041)
Confounding: Maternal age, race/ethnicity, pre-pregnancy body mass index, education, gravidity, smoking, and gestational age at
blood draw
Yang et al., 2020,
7021246
Medium
China
2013-2014
Cohort
Pregnant women
ages 20-40 years
in early
pregnancy
N = 436
Serum
6.78 (5.08-
9.60)
Levels (ln-mmol/L) of
TC, triglycerides,
HDL, and LDL;
LDL/HDL ratio
Regression
coefficient per
ln-unit increase
in PFOS, or by
quartiles
TC
Per ln-unit: -0.090
(-0.274, 0.093)
Q2
Q3
Q4
0.26 (-0.33, 0.85)
-0.04 (-0.44, 0.36)
-0.10 (-0.52, 0.32)
p-trend = 0.832
Triglycerides
Per ln-unit: -0.084
(-0.307,0.138)
Q2: -0.03 (-0.48, 0.42)
Q3: 0.07 (-0.38,0.52)
Q4: 0.09 (-0.35,0.53)
p-trend = 0.478
HDL
Per ln-unit: 0.025
(-0.030, 0.081)
Q2: 0.06 (-0.05,0.17)
Q3: 0.00 (-0.05,0.17)
D-150
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Q4: 0.04 (-0.06,0.14)
p-trend = 0.600
LDL
Per ln-unit: -0.116
(-0.262, 0.027)
Q2: 0.02 (-0.22, 0.26)
Q3: -0.05 (-0.28, 0.18)
Q4:-0.11 (-0.36, 0.14)
p-trend = 0.532
LDL/HDL ratio
Per ln-unit: -0.039
(-0.084, 0.007)
Q2: -0.02 (-0.08, 0.04)
Q3: 0.00 (-0.07,0.07)
Q4: -0.08 (-0.18, 0.02)
p-trend = 0.240
Results: Lowest quartile as reference group.
Confounding: Age, body mass index (BMI) at baseline, husband smoking, GDM, parity (nulliparous, muciparous), education, career,
income, energy intake and physical activity in the late term of pregnancy, gestational weeks, carbohydrate, protein, SFA, MUFA, and
PUFA intake in the late term of pregnancy.
Dalla Zuanna et
al., 2021,
7277682
Medium for TC
HDL; low for
LDL
value < 0.05
Q3: 4.81 (0.49, 9.14), p-
value < 0.05
Q4: 9.20 (4.65, 13.76), p-
value < 0.05
D-151
Italy
2017-2020
Cross-sectional Pregnant women Serum
Levels (mg/dL) of TC, Regression
ages 18-44 from
an area exposed
to PFAS through
drinking water
N = 319
I Trimester:
N = 101
II Trimester:
N = 88
III Trimester:
N = 130
2.7(1.9-3.8)
I Trimester:
2.9 (2.2-3.9)
II Trimester:
2.5 (1.8-3.5)
III Trimester:
2.9 (1.8-4.2)
HDL, and LDL
coefficient per
ln-unit increase
inPFOS, or by
quartiles
TC
Per ln-unit: 3.01 (-4.51,
10.53)
Q2: 4.42 (-8.21, 17.05)
Q3: -1.65 (-13.80,
10.50)
Q4: 9.89 (-2.82, 22.59)
HDL
Per ln-unit: 4.84 (2.15,
7.54), p-value < 0.05
02: 8.60 C4.07. 13.14V o-
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
LDL
Per ln-unit: -2.50 (-8.99,
3.98)
Q2: -2.76 (-13.73, 8.21)
Q3:-5.10 (-15.63, 5.43)
Q4: 0.01 (-11.04, 11.06)
First Trimester
TC: 15.34 (-1.08, 31.78)
HDL: 8.31 (1.07, 15.55),
p-value <0.05
LDL: 6.65 (-5.90, 19.20)
Second Trimester
TC: -2.86 (-17.86,
12.13)
HDL: 3.76 (-3.35, 10.87)
LDL: -3.51 (-14.72,
7.69)
Third Trimester
TC: -4.51 (-18.13, 9.09)
HDL: 4.25 (0.26, 8.24),
p-value <0.05
LDL: -10.05 (-22.71,
2.61)
Results: Lowest quartile as the reference group.
Confounding: Age, number of previous deliveries, BMI, physical activity, smoking habits, country of birth, education level, laboratory in
charge of the analyses of serum lipids, gestation weeks and reported fish consumption (in tertiles)
General Population
Lin et al., 2009
1290820
Medium
2.26), p-value < 0.05
United States
1999-2000
and 2003-
2004
Cross-sectional
Adults ages 20+
years from
NHANES
N = 969
Serum
Mean
(SEM) = 3.19
(0.04) loglO-
ng/mL
Metabolic syndrome
HDL cholesterol and
metabolic syndrome
triglycerides
OR per log 10-
unit increase in
PFOS
Metabolic syndrome
HDL cholesterol
Model 4: 1.47 (1.07,
2.00), p-value < 0.05
Model 5: 1.61 (1.15,
Metabolic syndrome
triglycerides
Model 4: 0.97 (0.73,
1.27)
D-152
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Nelson et al.,
2010,1291110
Medium
Model 5: 0.86 (0.65,
1.16)
Outcome: Metabolic syndrome HDL cholesterol defined as HDL < 1.03 mmol/L in men and HDL < 1.29 mmol/L in women; metabolic
syndrome triglycerides defined as triglycerides > 1.69 mmol/L.
Confounding: Model 4: Age, sex, race, health behaviors (smoking status, alcohol intake, and household income), measurement data
(CRP and HOMA/insulin) and medications; additional confounding for model 5: Other components of the metabolic syndrome.
United States
2003-2004
Cross-sectional
Adults ages 20-
80 years from
NHANES
N = 860
Serum
21.0
(Range = 1.4-
392.0)
Levels (mg/dL) of TC,
HDL, non-HDL, LDL
Regression
coefficient per
unit increase in
PFOS, or by
quartiles
TC
Per unit increase: 0.27
(0.05, 0.48)
Q4: 13.4 (3.8, 23.0)
p-trend by
quartiles = 0.01
HDL
Per unit increase: 0.02
(-0.05, 0.09)
Non-HDL
Per unit increase: 0.25
(0.00, 0.50)
Liu et al., 2018,
4238514
Medium
LDL
Per unit increase: 0.12
(-0.17,0.41)
Results: Lowest quartile used as the reference group.
Confounding: Age, sex, race/ethnicity, SES, saturated fat intake, exercise, time in front of a TV or computer, BMI, alcohol consumption,
and smoking.
United States
2013-2014
Cross-sectional
Adults ages 18+
from NHANES
N = 1871
Serum
GM = 5.28
Levels of TC (mg/dL), Regression TC: 1.22 (1.91)
LDL (mg/dL), HDL coefficient (SE) LDL: 0.88 (1.75)
(mg/dL), TG (In- per ln-unit HDL: 0.91 (0.70)
mg/dL) increase in PFOS TG: -0.08 (0.05)
Confounding: Age, gender, ethnicity, smoking status, alcohol intake, household income, waist circumference, and medications (anti-
hypertensive, anti-hyperglycemic, and anti-hyperlipidemic agents)
Dongetal., United States Cross-sectional Adults age 20-80 Serum Levels (mg/dL) of TC, Regression
2019,5080195 2003-2014 from NHANES Mean =15.6 LDL, HDL coefficient per
Medium N=8814 unit increase
PFOS
TC all cycles: 0.4 (0.06,
0.6)
p-value <0.05
Inconsistent associations
with LDL or HDL across
NHANES cycles.
D-153
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Confounding: Age, gender, race, family income index, BMI, waist circumference, physical activities, diabetes status, smoking status,
number of alcoholic drinks per day
Jain et al., 2019,
5080642
Medium
United States
2004-2015
Cross-sectional
Members of
NHANES
Non-obese
N = 1053 females
(NF) and 1237
males (NM)
Obese N = 699
females (OF) and
640 males (OM)
Serum
GMs:
Female = 7.4
Male = 11.5
Levels (mg/dL) of TC, Regression
LDL, HDL, TG coefficient per
loglO-unit
increase PFOS
Confounding: Race/ethnicity, smoking status, age, poverty income
exercise, survey year, daily dietary intake of total cholesterol, daily
TC: No clear associations
LDL
OF: 0.0375 (0.0024,
0.0727)
p-value = 0.04
No clear associations in
NF. NM, or OM
HDL: No clear
associations
TG
OF:-0.0912 (-0.153,
-0.0294)
p-value <0.01
No clear associations in
NF, NM, or OM
ratio (PIR), fasting time, use of lipid lowering medicine, physical
intake of total saturated fat, calories, caffeine, alcohol, protein intake
Fanetal., 2020,
7102734
Medium
United States
2011-2014
Cross-sectional
Adults age 20+
from NHANES
N = 1067
Serum Levels (mg/dL) of TC,
Median = 5.14 LDL, HDL, and TG
ng/mL
Regression TC: 3.85 (1.27, 6.42)
coefficient per p-value = 0.003
loglO-unit LDL: 3.02 (0.75, 5.29)
increase in PFOS p-value = 0.009
HDL: 1.24 (0.32,2.16)
p-value = 0.009
TG: -0.01 (-0.04,0.02)
p-value = 0.505
Confounding: Age, gender, race, education level, PIR, BMI, smoking status, alcohol use, energy intake levels, screen time
Jain and
Ducatman, 2020,
6988488
Medium
United States
2007-2014
Cross-sectional
Adults age 20+
from NHANES
Non-diabetic
non-LLM users:
N = 2,872
Diabetic non-
LLM users:
N = 316
Non-diabetic
LLM users:
N = 519
Serum
Levels not
reported
Apolipoprotein B
(loglO-mg/dL)
Regression
coefficient per
loglO-unit
Apolipoprotein B
Non-diabetic non-LLM
users: 0.02027, p-
increase in PFOS value = 0.02
Diabetic non-LLM users:
0.01547, p-value = 0.41
Non-diabetic LLM
users: -0.01327, p-
value = 0.40
D-154
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Diabetic LLM
users: N = 293 Diabetic LLM users:
0.02001, p-value = 0.19
Confounding: Gender, age, age squared, race/ethnicity, poverty income ratio, fasting time in hours, loglO-transformed BMI, smoking
status, survey year, daily intake of cholesterol, caffeine, alcohol, total calories, total protein, and total fat
Steenland et al., United States Cross-sectional
Adults ages 18+ Serum
Levels (ln-mg/dL) of
Lipid levels,
TC
2009,1291109 2005 2006
from the C8 19.6 (Range:
TC, LDL, HDL, non-
ratios:
0.0266 (SD = 0.0014)
Medium for TC,
Health Project, 0.25-759.2)
HDL cholesterol, and
Regression
HDL
current or former
triglycerides; TC/HDL
coefficient per
HDL
Low for TG,
residents from
ratio; high TC
ln-unit increase
0.00355 (SD = 0.00173)
LDL
areas supplied
in PFOS
with
LDL
contaminated
High TC:
0.04172 (SD = 0.00221)
water
OR by PFOS
N = 46494
quartiles
Triglycerides
0.01998 (SD = 0.00402)
TC/HDL ratio
0.02290 (SD = 0.00202)
Non-HDL
0.03476 (SD = 0.0019)
High TC
Q2: 1.14 (1.05, 1.23)
Q3: 1.28 (1.19, 1.39)
Q4: 1.51 (1.40, 1.64)
p-trend < 0.0001
Outcome: High TC defined as > 240 mg/dL.
Results: Lowest quartile used as the reference group; lowest decile used as the reference group.
Confounding: Age, male gender, smoking status, education level, drinks alcohol, currently exercises, and BMI
Chateau-Degat et Canada
al., 2010, 2004
2919285
Medium
Cross-sectional Nunavik Inuit Plasma
adults GM (95%
Quartile analyses: confidence
N = 716 (395 interval): 18.6
women, 325 men) (17.8-19.5)
TC, TC/HDL
ratio: N = 663
LDL: N = 651
Levels (mmol/L) of
TC, LDL, HDL, non-
HDL cholesterol, and
triacylglycerols;
TC/HDL ratio
Regression
coefficient per
unit increase in
PFOS or
adjusted mean by
quartiles
TC
0.0009, p-value = 0.086
Ql: 4.781 (4.704, 4.864)
Q2: 4.869 (4.804, 4.940)
Q3: 4.969 (4.901,5.041)
Q4:5.301 (5.221, 5.381)
p-trend < 0.0001
D-155
-------
DRAFT FOR PUBLIC COMMENT
Non-HDL:
N = 670
HDL: N = 384
women, 309 men
Triacylglycerols:
N = 365 women,
284 men
D-156
MARCH 2023
LDL
-0.002, p-value = 0.242
Ql: 2.750 (2.680,2.819)
Q2: 2.780 (2.730, 2.830)
Q3: 2.831 (2.770, 2.891)
Q4: 2.871 (2.801, 2.942)
p-trend = 0.58
HDL
Women: 0.0042, p-
value = 0.001
Men: 0.0016, p-
value < 0.001
Ql: 1.539 (1.510, 1.572)
Q2: 1.619(1.580, 1.660)
Q3: 1.630 (1.580, 1.660)
Q4: 1.831 (1.788, 1.868)
p-trend < 0.0001
Non-HDL
-0.0011, p-value = 0.315
Ql: 3.241 (3.160, 3.321)
Q2: 3.241 (3.182, 3.301)
Q3: 3.341 (3.271, 3.412)
Q4: 3.469 (3.388, 3.549)
p-trend = 0.09
Triacylglycerols
Women: -0.0014, p-
value = 0.04
Men: -0.0009, p-
value = 0.162
Ql: 1.051 (1.009, 1.092)
Q2: 1.067 (1.038, 1.096)
Q3: 0.941 (0.910, 0.970)
Q4: 1.000 (0.968, 1.030)
p-trend = 0.42
TC/HDL ratio
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
-0.0035, p-value < 0.001
Ql: 3.250 (3.181
Q2: 3.210(3.140
Q3: 3.240 (3.170
Q4: 3.130 (3.049
p-trend = 0.75
Results: Adjusted means presented with lower and upper bounds of standard error in parentheses.
Confounding: Means adjusted for age, gender, BMI, and smoking status. All regression analyses adjusted for lipid-lowering drugs.
Additional regression analyses adjustments: TC: gender, smoking status, age and n-3 PUFAs; LDL: age, BMI, smoking status, and
insulinaemia; HDL: PFOS and n-3 PUFAs; non-HDL cholesterol: smoking status, age and gender; triacylglycerols: PFOS, smoking
status, BMI, stratified by gender; TC/HDL ratio: smoking status and gender
3.320)
3.281)
3.311)
3.211)
Eriksen et al.,
2013,2919150
Medium
Fisher et al.
2013,2919156
Medium
Denmark
1993-1997
Cross-sectional Adults ages 50- Plasma
65 from DCH Mean =36.1
N = 753
Levels of TC (mg/dL) Regression 4.6(0.8,8.5)
coefficient per p-value = 0.02
IQR increase in
PFOS
Confounding: Sex, education, age, BMI, smoking status, intake of alcohol, egg, and animal fat and physical activity
Canada
2007-2009
Cross-sectional
Adults ages 18-
74 years from
CHMS, cycle 1
N = 2,700
TC, HDL, Non-
HDL, TC/HDL
ratio: N = 2,345
LDL,
triglycerides:
N = 1,168
High cholesterol:
N = 1,042
Plasma
GM
(SD) = 8.40
(2.04)
Levels (ln-mmol/L) of
TC, HDL, LDL, non-
HDL, triglycerides;
TC/HDL ratio (In-
transformed); high
cholesterol
Lipid levels,
TC/HDL ratio:
Regression
coefficient per
ln-unit increase
in PFOS
High cholesterol:
OR per ln-unit
increase in
PFOS, or by
quartiles
TC
0.014 (-0.019, 0.05)
HDL
-0.02 (-0.07, 0.02)
LDL
0.02 (-0.03,
Non-HDL
0.03 (-0.11,
0.08)
0.07)
Triglycerides
-0.02 (-0.12, 0.07)
TC/HDL ratio
0.04 (-0.008, 0.08)
High cholesterol
per ln-unit increase: 1.15
(0.89, 1.59)
Q2: 0.97 (0.58, 1.62)
Q3: 0.94 (0.58, 1.54)
D-157
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Q4: 1.36(0.87,2.12)
p-trend = 0.13
Outcome: High cholesterol defined as TC > 5.2 mmol/L.
Results: Lowest quartile used as the reference group.
Confounding: Lipid levels, TC/HDL ratio: Age, sex, marital status, BMI alcohol, smoking status and physical activity index; High
cholesterol: Age, gender and alcohol consumption
Fitz-Simon et al.,
United States Cohort
Adults ages 20-
Serum
Levels (mg/dL) of TC,
Percentage
TC: 3.20 (1.63, 4.76)
2013, 2850962
Baseline:
60 from C8
Baseline GM
LDL, HDL, and
decrease (loglO
R2 = 0.04
Medium for TC,
2005-2006;
Short-Term
(SD) = 18.5
triglycerides
of final and
LDL: 4.99 (2.46, 7.44)
HDL
Follow-up:
Follow-up Study
(13.5)
initial ratio
R2 = 0.07
Low for TG,
2010
living in West
change per log 10
HDL: 1.28 (-0.59,3.12)
LDL
Virginia and Ohio
Follow-up
of ratio change in
R2 = 0.04
with PFOA-
GM
PFOS)
Triglycerides: 2.49
contaminated
(SD) = 8.2
(-2.88, 7.57)
drinking water
(7.1)
R2 = 0.08
N = 560 (N = 521
for LDL analysis)
Confounding: Age, sex, interval between measurements, and fasting status
Donat-Vargas et
Sweden Cohort
Non-diabetic
Plasma
Levels (mmol/L) of TC Regression
Per change in PFOS
al., 2019,
1990-2013
adults ages 30-60
Baseline
and TG
coefficient per 1-
TC
5080588
at baseline in
median = 20
SD change PFOS Baseline: -0.21 (-0.39,
Medium
Vasterbotten
Median at 10-
or by tertiles
-0.04)
Intervention
year follow-
Follow-up: 0.01 (-0.19,
Programme (VIP)
up = 15
0.21)
N= 187 Prospective: 0.05 (-0.15,
0.21)
TG
Baseline: -0.05 (-0.16,
0.06)
Follow-up: -0.15
(-0.28, -0.03)
Prospective: -0.14
(-0.27, -0.02)
Confounding: Gender, age, education, sample year, body mass index, smoking habit, alcohol consumption, physical activity and healthy
diet score
Lin etal., 2019, United States Cohort and Prediabetic adults Plasma Levels (mg/dL) of TC, Regression Cross-sectional
5187597 1996-2014 cross-sectional age 25+from the Median = 27.2 LDL, HDL, coefficient per TC: 2.53 (-0.10, 5.16)
Medium Diabetes triglycerides, non- doubling PFOS LDL: 1.38 (-1.02,3.77)
D-158
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Prevention
Program (DPP)
and Outcomes
Study (DPPOS)
N = 940 (888 not
on metformin)
HDL, and very low
density lipids (VLDL);
hypercholesterolemia,
hypertriglyceridemia
HR or OR for
hypercholesterol
emia or
hypertriglyceride
mia per doubling
ofPFOS
HDL: -0.40 (-1.19,
0.39)
Triglycerides: 7.75 (0.63,
14.88)
VLDL: 1.57 (0.24, 2.89)
Hypercholesterolemia at
baseline OR: 1.02 (0.85,
1.21)
Hypertriglyceridemia at
baseline OR: 1.23 (1.03,
1.46)
Prospective
Hypercholesterolemia
HR: 1.01 (0.91, 1.12)
Hypertriglyceridemia
HR: 1.09 (0.93, 1.27)
Greater effect in the
placebo group
Confounding: Age, sex, race and ethnicity, marital status, educational attainment, drinking, smoking, percent of daily calorie from fat
intake, daily fiber intake, physical activity level, and waist circumference at baseline
Canova et al.,
2020,7021512
Medium
Italy
2017-2019
Cross-sectional
Residents of
PFAS "Red
Area" with
contaminated
public water
supply ages 20-
39
N = 15720(7620
female, 8100
male)
Serum
Median = 3.7
Female = 3
Male = 4.8
Levels (mg/dL) of TC,
LDL, HDL, non-HDL,
and triglycerides
Regression
coefficient per
ln-unit increase
PFOS or by
quartile, or by
decile
TC
4.99 (4.12, 5.86)
p-value for interaction by
sex = 0.39
Consistently increased
associations by deciles,
from 4.33 to 11.77
LDL
3.97 (3.21,4.73)
Males: 5.07 (3.87,6.27)
Females: 2.43 (1.47,
3.39)
p-value for interaction by
sex = 0.003
Associations for deciles
2-10 consistently
increase from 2.94 to
9.67
D-159
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HDL
1.43 (1.1, 1.76)
Males: 0.91 (0.47, 1.36)
Females: 1.95 (1.46,
2.45)
p-value for
associations = 0.001
Associations for deciles
2-10 moderately increase
from 1.13 to 3.43
Triglycerides
0 (-0.01, 0.01)
p-value for
associations = 0.954
Associations for deciles
2-10 inconsistently vary
from 0 to 0.02
Results: Lowest quartile or decile used as reference group.
Confounding: Age, BMI, time-lag between enrollment and beginning of study, physical activity, smoking habits, country of birth,
alcohol consumption, education level, laboratory in charge of analyses, reported food consumption
Lin et al., 2020,
6988476
Medium
Taiwan
2016-2017
Cross-sectional
Adults aged 55 to
75 that resided in
the study area for
more than
10 years and not
taking lipid-
lowering
medication
N = 352
Serum
16.2(10.1-
24.1)
Levels (mg/dL) of TC,
HDL, LDL, and
triglycerides
Regression
coefficient by
quartiles
TC
Q2: 15.06 (4.66, 25.46),
p-value <0.05
Q3: 11.47 (1.03,21.91),
p-value <0.05
Q4: 10.18 (-0.59, 20.94)
p-trend = 0.11
HDL
Q2: 3.23 (-0.79,7.24)
Q3: 1.92 (-2.11,5.95)
Q4: -2.68 (-6.84, 1.47)
p-trend = 0.19
LDL
Q2: 13.43 (4.05, 22.80),
p-value <0.05
D-160
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Q3: 12.32 (2.91,21.73),
p-value <0.05
Q4: 15.29 (5.59, 24.99),
p-value <0.05
p-trend = 0.004
Triglycerides
Q2: 8.93 (-9.74, 27.59)
Q3: 7.58 (-11.16, 26.31)
Q4: 6.76 (-12.55, 26.07)
p-trend = 0.53
Results: Lowest quartile used as the reference group.
Confounding: Age, sex, smoking status, and drinking status
Liu et al., 2020, United States Randomized Adults from Plasma Levels (mg/dL) of TC, Least-squared TC
6318644 2004-2007 clinical trial POUNDS Lost 23.5 triglycerides, and means (LSM) by Tl: 180.9 (8.0)
Medium study ages 20+ apolipoproteins loglO- tertile PFOS T2: 189.3 (7.9)
N = 326 ApoB, ApoE, and T3: 190.7 (7.3)
ApoC-III p-trend = 0.21
Triglycerides
Tl: 126.8 (11.6)
T2: 132.4 (11.4)
T3: 126.1 (10.5)
p-trend = 0.80
Results: LSM are presented with standard error in parentheses.
Confounding: Age, sex, race, educational attainment, smoking status, alcohol consumption, physical activity, BMI, regular lipid -
lowering medication use, dietary intervention groups
Han et al., 2021,
7762348
Medium
China
2016-2017
Case-control
Adults ages 25 to
74 including type
2 diabetes cases
and healthy
controls
N = 304
Serum
Cases: 7.60
(4.47-10.55)
Controls: 8.45
(5.40-11.95)
Levels (loglO-mmol/L)
of TC, HDL, LDL, and
triglycerides
Regression TC: 0.06 (-0.01, 0.12)
coefficient per HDL -0.02 (-0.09, 0.05)
loglO-unit LDL: 0.12 (0.03, 0.21),
increase in PFOS p-value < 0.05
Triglycerides: 0.03
(-0.13,0.18)
Confounding: Age, sex, BMI.
Jeddi et al., 2021, Italy
Cross-sectional Residents aged
Serum
Reduced HDL,
OR per ln-unit Reduced HDL: 0.79
7404065 2017-2019
20-39 from the
GM (range):
elevated triglycerides
increase in PFOS (0.73, 0.86), p-
Medium
PFAS-
4.54 (
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Outcome: Reduced HDL defined as HDL< 40 mg/L for male or HDL < 50 mg/L for female; elevated triglycerides defined as
triglycerides > 175 mg/dL.
Confounding: Age, gender, time-lag between the beginning of the study and blood sampling center where BP has been measured,
education, number of deliveries, physical activity, country of birth, diet, alcohol intake, and smoking status, and other components of
metabolic syndrome
Occupational Populations
Olsen et al.
United States, Cross-sectional Current and
Serum
Levels of cholesterol Comparison of No significant
(2003, 1290020) Belgium
Medium 1994-2000
former workers at Antwerp
two
Mean
(ln-mg/dL), HDL
(mg/dL)
mean outcome by differences between
PFOS exposure mean cholesterol or HDL
fluorochemical (SD) = 0.96
production plants ppm (0.97);
Male Decatur =1.4
N = 421, 0 ppm (1.15)
Female
N = 97,
Regression
analysis
N = 174
Confounding: Age, BMI, drinks/day, cigarettes/day, location, entry period, baseline years worked
quartile
Regression
coefficient per
unit increase in
PFOS
by quartile among male
and female employees
Cholesterol
0.01 (-0.005, 0.025)
Notes: ALSPAC = Avon Longitudinal Study of Parents and Children; APFO = ammonium perfluorooctanoate; ApoB = Apolipoprotein B; ApoE = Apolipoprotein E; ApoC-
III = Apolipoprotein C-III; CHMS = Canadian Health Measures Survey; DCH = Diet, Cancer and Health; EYHS = European Youth Study; HDL = high density lipids; KorEHS-
C = Korea Environmental Health Survey in Children and Adolescents; LDL = low density lipids; HELIX = Human Early-Life Exposome; HOME = Heath Outcomes and
Measures of the Environment; HR = hazard ratio; IQR = interquartile range; S-MBCS = Shanghai-Minhang Birth Cohort Study; MoBa = Norwegian Mother and Child Cohort
Study; NHANES = National Health and Nutrition Examination Survey; SE = standard error; TC = total cholesterol; OR = odds ratio; VLDL = very low-density lipoprotein.
a Exposure reported as median (25th-75th percentile) in ng/mL unless otherwise specified.
b Results reported as effect estimate (95% confidence interval) unless otherwise specified.
c Confounding indicates factors the models presented adjusted for.
D.6 Endocrine
Table D-15. Associations Between PFOS Exposure and Endocrine Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels Outcome
(ng/mL)a
Comparison
Select Resultsb
General Population
Lebeaux et al. United States Cohort Mother-infant Cord serum Levels of TSH Regression Cord serum
(2020, 6356361) 2003-2007 pairs from 14.3 (jilU/L). TT4 coefficient per TSH: 0.09 (-0.06, 0.25)
D-162
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
High for cord
serum thyroid
hormones;
Medium for
maternal thyroid
hormones
Health Outcome
Measures of the
Environment
(HOME) Study
N = 256 for
cord serum
N = 185 for
maternal serum
Maternal serum
5.5
(jig/dL). TT3
(ng/dL), FT4
(ng/dL), and FT3
(pg/mL)
log2-unit
increase in
PFOS
TT4
TT3
FT4
FT3
0.01 (-0.04, 0.07)
-0.02 (-0.10,0.06)
-0.02 (-0.06, 0.02)
-0.03 (-0.07, 0.02)
Maternal serum
TSH: 0.02 (-0.24, 0.28)
TT4
TT3
FT4
FT3
0.02 (-0.08, 0.08)
-0.02 (-0.07, 0.03)
0.02 (-0.02, 0.07)
-0.03 (-0.06, 0.00)
Confounding: Individual PFAS, maternal age at delivery, race/ethnicity, marital status at baseline, maternal education level, household
income, mean loglO-transformed cotinine, maternal alcohol usage during pregnancy, nulliparity, maternal BMI based on pre-pregnancy
weight in pounds, child's sex, gestational week at blood draw for PFAS measurement, and (for cord serum only) delivery mode
Blake et al. Fernand, Ohio,
Cohort
FCC
Drinking water
Levels of
Percent
TSH
(2018, 5080657) USA
Median age
Serum
TSH (ln-nIU/mL),
change per
9.75 (1.72, 18.4), p-value = 0.02
Medium 1991-2008
38 years at
28.4
TT4 (In- ng/dL)
IQR increase
Males: 21.4 (6.55, 38.3)
enrollment,
in PFOS
p-value = 0.01
N = 122 for
Females: 5.13 (-5.29, 16.7)
TSH
p-value = 0.36
measurements;
47 male and 75
TT4
female
-0.51 (-4,3.1), p-value = 0.78
N = 144 for TT4
Males: -5.29 (-10.1,-0.26),
measurements;
p-value = 0.04
63 males and 81
Females: 1.69 (-3.28,6.91),
females
p-value = 0.52
Confounding: Age, year of measurement, sex, education, income, marital status, BMP
Jain and United States
Cross-sectional
Adults from
Serum
Levels of
Regression
TT4
Ducatman 2007-2012
NHANES aged
Levels not
TSH (log-
coefficient per GF-1: 0.002, p-value = 0.76
(2019, 6315816)
20+
reported
|iIU/mL).
loglO-unit
GF-2: -0.008, p-value = 0.47
Medium
Glomerular
TGN (log-ng/mL),
increase in
GF-3A: 0.058, p-value = 0.02
filtration (GF)
TT4 (log-ng/dL),
PFOS
GF-3B/4: -0.002, p-value = 0.94
status:
FT4 (log-ng/dL),
GF-1 = 1,653
TT3 (log-ng/dL),
GF-2 = 720
FT3 (log-pg/mL)
GF-3A = 114
GF-3B/4 = 62
D-163
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
GF Stages: GF-1: GFR > 90 mL/min/1.73 m2; GF-2: GFR between 60 and 90 mL/min/1.73 m2; GF- 3A: GFR between 45 and
60 mL/min/1.73 m2; GF- 3B/4: GFR between 15 and 45 mL/min/1.73 m2
Confounding: Gender, race/ethnicity, iodine deficiency status, age, BMI, fasting time, poverty income ratio, total calories consumed during
the last 24h, smoking status, use of drugs
Jain (2013,
United States Cohort Adults and
Serum
Levels of
Regression
TSH, FT3, FT4, TT3, TT4, TGN:
2168068)
2007-2008 children from
Total cohort
TSH (nIU/L),
coefficient per No statistically significant
Low
NHANES aged
FT3 (pg/L),
loglO-unit
associations
12+
TT3 (fg/dL),
increase in
N = 1,540
FT4 (pg/L),
PFOS, or by
including
TT4 (pg/L),
tertiles
children
TGN
Results: Lowest tertile used as the reference group.
Confounding: Gender, race, age, iodine deficiency, iodine replete
Lewis et al.
United States Cross-sectional Men and
Serum
Levels of
Percent
TSH
(2015,3749030) 2011-2012 womenfrom
Males 20-40:
TSH (nIU/mL),
change per
Males
Low
NHANES ages
7.75
TT3 (ng/dL),
doubling of
20 to <40:-2.9 (-8.6, 3.2)
20-80
Males 40-60:
FT3 (pg/mL),
PFOS
40 to <60:-1.3 (-8.9,7.1)
699 men
9.28
TT4 (jig/mL).
60 to 80: -2.3 (-9.4, 5.3)
680 women
Males 60-80:
FT4 (ng/dL)
Females
11.1
20 to <40:-1.0 (-7.9,6.4)
Females 20-40:
40 to <60: 0.0 (-7.1,7.7)
4.20
60 to 80: -1.5 (-9.6,7.3)
Females 40-60:
4.93
FT4
Females 60-80:
Females
9.50
20 to <40: 2.2 (0.5, 3.9)
p-value < 0.05
40 to <60: 1.3 (-0.5,3.2)
60 to 80: -0.5 (-2.5, 1.5)
Males: No statistically significant
associations
TT3, FT3, TT4: No statistically
significant associations
Confounding: Age, BMI, poverty income ratio, serum cotinine, and race/ethnicity
Li et al. (2017,
China Cross-sectional Residents of
Serum
Levels of
Regression
TSH: 0.41 (0.05, 0.76),
3856460)
2013-2014 Southern China,
1.3
TSH (nIU/mL),
coefficient per p-value = 0.024
Low
ages 1 month to
FT3 (pmol/L),
log-unit IQR
FT3:-0.14 (-0.24,-0.04),
90 years, 70%
FT4 (pmol/L)
p-value = 0.007
D-164
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
with thyroid
condition
N = 202
Comparison: Logarithm base not specified.
Confounding: Age, sex
increase in FT4: -0.13 (-0.22, -0.04),
PFOS p-value = 0.004
Byrne et al. St. Lawrence
(2018, 5079678) Island, Alaska,
Low USA
2013-2014
Cross-sectional
Alaska Natives,
aged 18-45
N = 85
38 men
47 women
Serum
4.55
Males: 6.81
Females: 3.35
Levels of
TSH (ln-nIU/mL),
TT3 (pg/mL),
FT3 (ng/dL),
TT4 ( ng/dL).
FT4 (ng/dL)
Regression TSH
coefficient per Males: -0.06 (-0.62, 0.51),
ln-unit
increase in
PFOS
p-value = 0.085
Females: No association
TT3
Males: -10.54 (-22.28, 1.20),
p-value = 0.08
Females: No association
FT3
Males: -0.30 (-0.53, 0.07),
p-value = 0.01
Females: 0.35 (0.05, 0.65)
p-value for sex interaction = 0.02
Confounding: Age, sex, smoking status
TT4, FT4: No statistically
significant associations
Zhang et al. China
(2018, 5079665) 2013-2016
Low
Cross-sectional
Women aged
20-40 years,
with (cases) or
without
(controls) POI
N = 120
Plasma Levels (ng/mL) of Regression TSH
Cases: 8.18 TSH, FT3, FT4 coefficient per POI cases: 1.57 (0.65, 2.5)
Controls: 6.02 log-unit POI controls: 0.67 (0.08, 1.26)
increase in
PFOS
FT3
POI cases -0.88 (-1.64, -0.09)
FT4
POI cases -2.99 (-4.52, -1.46)
Comparison: Logarithm base not specified.
Confounding: Age, BMI, education, income, sleep, and parity
FT3 and FT4 in POI controls: No
associations
Children
D-165
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Xiao et al.
(2019, 5918609)
High
Faroe Islands,
Denmark
1994-1995
Cohort Pregnant
women and
their infant
children
N = 172 and
153 for
measurements
in maternal and
cord serum,
respectively
Confounding: Child sex (in detailed results), parity,
alcohol during pregnancy, total PCB, mercury
Maternal blood
Geometric
mean = 20.86 n
Cord serum levels
of TSH (log-
IU/L),
T4 (log-pmol/L),
FT3 (log-pmol/L),
FT4, (log-pmol/L)
FT3 resin uptake,
FT4 index (FTI)
(log-IU/L)
Regression TSH
coefficient per All children: 39.7 (7.9, 80.9)
log2-unit
increase in
PFOS
Boys: 39.5 (0.4, 94.1)
Girls: 39.9 (-4.1, 104.2)
FTI
All children: 6.7 (-1.5, 15.6)
Boys: 2.1 (-7.7, 13)
Girls: 13.2(0.9, 27.1)
T4, FT3, FT4, FT3 resin uptake: No
statistically significant associations
maternal BMI, maternal height, maternal education, maternal age, smoking and drinking
Kim et al.
(2020, 6833758)
High
South Korea
2012-2017
Cohort
Children, aged
2, 4, 6 years
N = 511 for age
6 (268 boys)
Serum
Age 2: 4.530
Age 4: 4.050
Age 6: 3.980
Levels of
TSH (ln-nIU/mL),
FT4 (ln-ng/dL),
and T3 (ln-ng/dL)
at age 6
Subclinical
hypothyroidism
Regression
coefficient per
ln-unit
increase in
PFOS
Subclinical
hypothyroidis
m: OR per
increase in
PFOS
T3 at age 6
All: 0.04 (0.017), p-value < 0.05
Boys: 0.04 (0.018), p-value < 0.05
No interaction with sex
Subclinical hypothyroidism at age 6
All: 0.36(0.41,0.96)
Boys: 0.24 (0.07, 0.92)
No interaction with sex
TSH, FT4: No statistically
significant associations between or
within age groups
Results: Comparisons for T3 are presented with standard error in parentheses.
Confounding: Age, sex, dietary iodine intake
Kato et al.
(2016, 3981723)
Medium
Japan Cross-sectional Pregnant
2002-2005 women and
their children
N = 392
Male
children =180
Female
children = 212
Maternal serum Levels of
Male: 5.2 TSH (loglO-
Female: 5.3
|iU/mL).
FT4 (loglO-
ng/mL)
Regression TSH
coefficient per All infants: 0.18, p-value = 0.001
loglO-unit Increasing trend in LSM by
increase in quartiles p-trend = 0.024
PFOS Males: 0.21, p-value = 0.014
Females: 0.17, p-value = 0.021
Least square
means (LSM) FT4: No statistically significant
by quartile associations
D-166
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Confounding: Maternal age at delivery, BMI, parity, educational level, thyroid antibody, intake of seaweed, blood sampling period
before/after delivery for PFOS and PFOA, and gestational week at which blood sampling was obtained for TSH and FT4
Preston et al. United States
(2018, 4241056) 1999-2002
Medium
Cohort
Pregnant
women and
their children
N = 465
neonates (236
male, 229
female)
Maternal plasma Levels of
23.5 T4 (ng/dL)
Regression
coefficient by
quartiles
T4, all neonates:
Q2
Q3
Q4
Q2
Q3
Q4
-0.63 (-1.64,0.37)
-0.36 (-1.36, 0.67)
-1.1 (-2.13,-0.07)
T4, males:
-1.56 (-3.04, -0.08)
-1.7 (-3.28,-0.12)
-2.2 (-3.74, -0.66)
No associations in newborn females
Results: Lowest quartile used as the reference group.
Confounding: Maternal age, race/ethnicity, smoking status, fish intake, parity, and gestational week at blood draw
Aimuzi et al. China
(2019, 5387078) 2012-2013
Medium
Cross-sectional
Levels of
TSH (ln-mlU/L),
FT3 (pmol/L),
FT4 (pmol/L)
Pregnant Cord blood
women and 2.51
their children
N = 567
Male
children = 305
Female
children = 262
Confounding: Maternal age, fish intake, parity infant sex, gestational age at delivery, and maternal pre-pregnancy BMI
Regression TSH
coefficient per All children: -0.05 (-0.08, -0.02)
ln-unit Boys: -0.047 (-0.097, 0.003)
increase in Girls: -0.048 (-0.093, -0.003)
PFOS
Itoh et al. (2019, Japan
5915990)
Medium
2003-2005
Cohort Pregnant Plasma Levels of
women and 6.21 TSH (ln-nU/mL),
their children FT3 (ln-pg/mL),
365 male FT4 (ln-pg/mL),
children TPOAb (ln-
336 female IU/mL),
children TgAb (ln-IU/mL)
Regression TSH
coefficient per All boys: 0.23 (0.07, 0.39),
ln-unit
increase in
PFOS
p-value = 0.004
Boys with TA-negative mothers:
0.39 (0.12, 0.66), p-value = 0.005
No significant association among
TA-positive mother-infant pairs
Confounding: Age at delivery, parity, educational level, alcohol consumption, smoking during pregnancy, pre-pregnancy BMI, logFT4
Tsai(2017, Taiwan Cross-sectional Newborns from Cord blood Levels of Regression TSH, all newborns:
3860107) 2004-2005 Taiwan Birth Mean = 7.24 TSH (nIU/mL), coefficient by Q2:0.21 (-0.20,0.63)
Low Panel Study T3 (In- ng/dL), quartiles or Q3:0.19 (-0.22,0.61)
(TBPS) T4 (ng/dL) per ln-unit Q4:0.65 (0.02, 1.28)
Per increase: 0.35 (0.10, 0.59)
D-167
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
N = 118(64
boys, 54 girls)
increase in
PFOS
TSH, boys:
Q2: 0.63 (0.04, 1.22)
Q3: 0.30 (-0.33, 0.94)
Q4: 0.75 (0.13, 1.62)
Per increase: 0.33 (0.01,
T4, all newborns:
0.68)
Q2
Q3
Q4
-0.50 (-1.29,0.29)
-0.28 (-1.08,0.51)
-1.03 (-2.17,-0.12)
Per increase: -0.46 (-0.92, -0.001)
T4, boys:
Q2
Q3
Q4
-0.30 (-1.40, 0.80)
0.19 (-0.99, 1.36)
-2.12 (-3.62,-0.618)
Per increase: -0.67 (-1.28, -0.05)
Results: Lowest quartile used as the reference group.
Confounding: Maternal age at delivery, newborn sex, maternal BMI, maternal education, gestational age, and delivery type
Pregnant Women
Dreyer et al. Denmark
(2020,6833676) 2010-2012
High
Cohort
Pregnant
women from
Odense Child
Cohort (OCC)
N = 1,048
Serum
7.64
Levels of diurnal
urinary (dU)
Cortisol (nmol/24-
hours), dU-
cortisone
(nmol/24-hours),
dU-
cortisol/cortisone,
serum Cortisol
(nmol/L)
Percent dU-cortisone: -9.1 (-14.7, -3.0), p-
change per 2- value < 0.05
fold increase T2: -5.7 (-14.7, 4.2)
inPFOS T3: -16.0 (-23.9, -7.2), p-
value < 0.05
p-trend <0.01
dU-cortisol/cortisone: 9.3 (3.3,
15.6), p-value < 0.05
T2: 11.0(1.8,21.1), p-value <0.05
T3: 16.6 (6.9, 27.1), p-value < 0.05
p-trend <0.01
dU-cortisol and serum Cortisol: No
statistically significant associations
Confounding: Age, parity, and offspring sex
D-168
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Xiao et al. Faroe Islands,
(2019, 5918609) Denmark
High 1994-1995
Cross-sectional
Pregnant
women and
their children
Maternal age 28
(SD = 5.6)
Maternal blood
Geometric
mean = 20.86 n
N = 172 and
153 for
measurements
in maternal and
cord serum,
respectively
Confounding: Child sex (in detailed results), parity,
alcohol during pregnancy, total PCB, mercury
Maternal serum
levels of TSH
(log-IU/L),
T4 (log-pmol/L),
FT3 (log-pmol/L),
FT4 (log-pmol/L)
FT3 resin uptake
FT4 index
Regression TSH in maternal serum
coefficient per All children: 16.4 (-7.5, 46.5)
log2-unit
increase in
PFOS
Boys: -6 (-29.6, 25.4)
Girls: 54.2(11.3, 113.8)
T4, FT3, FT4, FT3 resin uptake,
FT4 index: No statistically
significant associations
maternal BMI, maternal height, maternal education, maternal age, smoking and drinking
Berg (2017,
Norway Cohort Pregnant
Serum
Levels of
Regression
TSH
3350759)
2007-2009 or women and
8.03
TSH (mlU/L),
coefficient by
Q2: 0.04 (-0.03,0.11)
Medium
until 3 days children from
FT3 (pmol/L),
quartiles
Q3: 0.08 (0.01,0.15)
after birth the Norway
T3 (nmol/L),
Q4: 0.10 (0.02, 0.17)
Mother and
FT4 (pmol/L),
Child
T4 (nmol/L)
T3,T4, FT3, orFT4: No
Contaminant
statistically significant associations
Cohort Study
(MISA)
N = 370
Results: Lowest quartile used as reference group.
Confounding: Parity, t-uptake
Preston et al.
United States Cross-sectional Pregnant
Maternal plasma Levels of
Percent
TSH among TPOAb-positive
(2018,4241056) 1999-2002 women and
24.0
TSH (mlU/mL),
difference in
mothers: -16.4 (-29.8, -0.38)
Medium
their children
T4 ((ig/dL),
hormone level
p-value for effect modification by
FT4 index
per IQR
TPOAb status = 0.05
N = 718 women
increase in
(98 TPOAb-
PFOS
FT4, TT4: No statistically
positive and 620
significant associations
TPOAb-
negative)
Confounding: Maternal age, race/ethnicity, smoking status, fish intake, parity, and gestational week at blood draw
Reardon et al.
Canada Cohort Pregnant
Maternal blood
Levels of
Regression
TSH, linear PFOS
(2019, 5412435) 2019-2012 women
Total PFOS:
TSH (log-
coefficient per Main effect: 0.01 (-0.03, 0.04)
Medium
recruited prior
4.77
mlU/mL),
unit increase
D-169
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
to 18 weeks of Linear PFOS:
gestation 2.49
N = 478 XBr-PFOS:
1.08
FT3 (log-pmol/L), in total,
FT4 (log-pmol/L) linear, or lm-
by gestation status PFOS
and 3 months
post-partum
3 months post-partum: 0.06 (0.01,
0.12)
TSH, XBr-PFOS
Main effect: 0.29 (0.02, 0.56)
FT3, FT4: No statistically
significant associations
Confounding: Maternal age, ethnicity, history of smoking, history of drug and alcohol use
Kato et al. Japan
(2016, 3981723) 2002-2005
Low
Cross-sectional
Pregnant
women and
their children
N = 392
Male
children =180
Female
children = 212
Maternal serum
Male: 5.2
Female: 5.3
Levels of
TSH (loglO-
|iU/mL).
FT4 (loglO-
ng/mL)
Regression TSH
coefficient per All mothers: -0.21, p-value < 0.001
loglO-unit
increase
PFOS
Least square
means (LSM)
by quartile
Decreasing trend in LSM by
quartiles: p-trend < 0.001
Male: -0.25, p-value = 0.002
Female: -0.21, p-value = 0.005
FT4: No statistically significant
associations
Confounding: Maternal age at delivery, BMI, parity, educational level, thyroid antibody, intake of seaweed, blood sampling period
before/after delivery for PFOS and PFOA, and gestational week at which blood sampling was obtained for TSH and FT4
Notes: BMI = body mass index; FCC = Fernald Community Cohort; GF = glomerular filtration; GFR = glomerular filtration rate; TSH = thyroid stimulating hormone;
T3 = triiodothyronine; T4 = thyroxine; FT3 = free triiodothyronine; FT4 = free thyroxine; POI = premature ovarian insufficiency TgAb = thyroglobulin antibody;
TPOAb = thyroid peroxidase antibody; TT3 = total triiodothyronine; TT4 = total thyroxine; TGN = thyroglobulin.
a Exposure levels are reported as median unless otherwise noted.
b Results reported as effect estimate (95% confidence interval), unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
D.7 Metabolic/Systemic
Table D-16. Associations Between PFOS Exposure and Metabolic Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
„ , Exposure
°')U a I"n' Matrix, Levels Outcome
geS' (ng/mL)a
Comparison
Resultsb
Children and Adolescents
Ashley-Martin Canada,
etal. (2018, Recruitment
3981371) 2008-2011
Cohort
Pregnant
women and
their children,
Maternal blood
4.6
Adiponectin,
leptin
Regression
coefficient per
loglO-unit
Adiponectin, leptin: No statistically
significant associations
D-170
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location, . Population, .. ^xj)osurc
Design ' .. Matrix, Levels Outcome
Years Ages,N (ng/mL)a
Comparison
Resultsb
High
from the
MIREC Study
N = 1,175
Confounding: Maternal age, pre-pregnancy body mass index, sex, and parity0
increase in
PFOS
Buck et al.
(2018, 5080288)
High
United States
2003-2006
Cohort
Maternal serum
14
Adiponectin,
leptin
Pregnant
women and
their children in
the HOME
study
N = 230
Confounding: Maternal age, race, education, income, parity, maternal body mass index,
Percent change
per doubling of
PFOS
Adiponectin, leptin: No statistically
significant associations
serum cotinine, delivery mode, and infant sex
Chen et al.
(2019, 5080578)
High
China,
2012-2017
Infants followed Cord blood
up at age 5, 2.44
N = 404
BMI, WC, body
fat, waist-to-
height ratio
BMI, waist circumference, body
fat, waist to height ratio: No
statistically significant association
Confounding: Maternal
pregnancy, and parity
Cohort Infants followed Cord blood BMI, WC, body Regression
coefficient per
ln-unit increase
in PFOS, or by
tertile
age, maternal pre-pregnancy BMI, gestational week at delivery, maternal education, paternal smoking during
Jensen et al.
(2020, 6833719)
High
Denmark,
2010-2012
Cohort
Maternal serum
8.04
BMI z-score,
WC
Regression
coefficient per
unit increase in
PFOS
BMI z-score, WC: No statistically
significant associations
Minatoya et al.
(2017, 3981691)
High
Pregnant
women and
their infants
assessed at
birth, 3 months,
and 18 months,
Odense Child
Cohort
N = 593
Confounding: Maternal age, parity, pre-pregnancy BMI, pre-pregnancy BMI2, education, smoking, sex, visit, adiposity marker at birth
Japan, Cohort Pregnant Serum Adiponectin, Regression Adiponectin: 0.12 (0.01,0.22),
2002-2005 women and 5.1 leptin coefficient per p-value = 0.028
their children loglO-unit
N = 168 increase in Leptin: No statistically significant
maternal serum association
PFOS
Confounding: Maternal BMI, parity, smoking during pregnancy, blood sampling period, gestational age, infant sex
D-171
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Alderete et al.
(2019, 5080614)
Medium
United States
2001-2012
Cohort
Obese Hispanic
children (8-
14 years),
SOLAR Project
N = 38
Plasma
12.22
Regression
coefficient per
In- unit increase
inPFOS
Blood glucose,
insulin, 2-hour
glucose
(mg/dL)), 2-
hour insulin,
insulin
resistance,
insulin levels
Confounding: Sex, baseline social position (categorical), baseline outcome, baseline and change in age at follow-up, pubertal status
(categorical), baseline and change in body fat percent at follow-up.
Glucose (2-hour)
6.2 (-2.3, 14.8)
Blood glucose, insulin, 2-hour
insulin, insulin resistance, insulin
levels: No statistically significant
associations
Braun et al.
(2016, 3859836)
Medium
United States,
2003-2006,
follow up at age
Cohort
Maternal serum
13
Pregnant
women and
their children in
8 the HOME
study
N = 204
Confounding: Maternal age, race, education, income, parity, marital
fruit/vegetable consumption, fish consumption, prenatal vitamin use,
Overweight,
obesity, BMI z-
score, waist
circumference,
body fat
Percent change Overweight, obesity, BMI z-score,
per doubling of waist circumference, body fat: No
PFOS statistically significant associations
status, employment, depressive symptoms, BMI at 16 weeks gestation,
maternal serum cotinine concentrations, and child age in months
Conway et al.
(2016, 3859824)
Medium
United States, Cross-Sectional Children Serum
2005-2006 working or Mean = 86.5
living in six
PFOS-
contaminated
water districts,
C8 Health
Project
N = 47
Confounding: Age, sex, race, BMI, eGFR, hemoglobin, iron
Type 1 Diabetes OR per ln-unit
increase in
PFOS
Children with T1D: 0.52 (0.54,
0.87)
Domazet et al. Denmark,
(2016, 3981435) 1997-2009
Medium
Cohort
Children from
EYHS followed
through ages 9,
15, and 21,
N = 176
Plasma
Age 21
Males: 11.9
Females: 9.1
Age 15
Males: 22.3
Females: 20.8
WC, HOMA-
Beta, HOMA-
IR, insulin,
glucose,
skinfold
thickness, BMI
Percent change WC:
at 15 or 21 years Age 15 from age 9:
old per 10-unit
increase in
PFOS at 9 years
old
1.18(0.42, 1.84)
Age 21 from age 9:
1.52 (0.05,2.91)
Skinfold thickness:
D-172
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Age 9 Age 15 from age 9:
Males: 44.5 4.03 (1.33,6.67)
Females: 39.9 Age 21 from age 9:
5.67 (0.6, 10.93)
BMI:
Age 15 from age 9:
1.54 (0.62,2.4)
HOMA-Beta age 21, BMI age 21,
HOMA-IR, insulin, glucose: No
statistically significant associations
Confounding: Sex, age, and outcome levels at baseline (9 years of age), and ethnicity, maternal parity, and maternal income in 1997 (9 years
of age). Waist circumference was adjusted for height in order to account for body size.
Domazet et al.
(2020, 6833700)
Medium
Denmark, 1997 Cross-sectional
Children from Plasma Body fat (mm), Percent change
EYHS, 9-year- Boys: 42.9 adiponectin per 10%
old Girls: 42.0 (ng/mL), leptin increase in
N = 242 (pg/mL) PFOS
Body fat: -0.59 (-2.88, 1.24),
p-value = 0.552
Adiponectin: 0.24 (-1.70, 2.21),
p-value = 0.811
Leptin: -3.65 (-8.23, 1.16),
p-value = 0.134
Confounding (Adiponectin and leptin): Sex, age, parity, maternal income level
Confounding (Body fat): Sex, age, accelerometer wear time, parity, maternal income level
Gyllenhammar
et al. (2018,
4238300)
Medium
Sweden,
1996-2011,
children
followed up at
age 5
Cohort
Mothers and
their children
from the
POPUP Study
N = 381
Maternal serum BMI z-score
13
Regression
coefficient per
IQR increase in
maternal PFOS
Confounding: Sampling year, maternal age, pre pregnancy BMI, maternal weight gain during pregnancy,
delivery, years of education, and total time of breastfeeding
BMI z-score:
Ages 36 Non-significant positive
association (numeric results not
provided)
Ages 48 and 60 months: Positive
statistically significant associations,
maternal weight loss after
Hartman et al.
(2017, 3859812)
Medium
United
Kingdom,
recruitment
1991-1992
Cohort
Pregnant
women and
their daughters,
ALSPAC
N = 319
Maternal serum
19.8
Waist
circumference
(WC)(cm),
Trunk fat (%),
BMI (kg/m2),
Regression
coefficient per
unit increase in
PFOS
WC:-0.12 (-0.20,-0.04),
p-value = 0.005
Trunk fat:
-0.06 (-0.12, 0.01), p-value = 0.02
D-173
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Total body fat
(%) per high,
medium, and
low educational
status
Confounding: Sampling design, pre-pregnancy BMI (kg/m2) and maternal educational status
BMI:
-0.04 (-0.07, 0.0), p-value = 0.03
Total body fat (%), WC, Trunk fat,
and BMI for overall, low, and
medium education status: No
statistically significant associations
Kang et al.
(2018, 4937567)
Medium
Korea,
2012-2014
Cross-sectional
Plasma
5.68
Fasting blood
glucose (mg/dL)
Children from
KorEHS-C
Seoul and
Gyeonggi, 3-
18 years of age,
N = 147
Confounding: Age, sex, BMI z-score, household income, second-hand smoking
Regression
coefficient per
ln-unit increase
inPFOS
Blood glucose:
0.707 (-1.921,3.336),
p-value = 0.595
Karlsen et al.
(2017, 3858520)
Medium
Faroe Islands,
recruited 2007-
2009 (at birth);
follow up at
child ages
18 months,
5 years
Serum, BMI z-score,
Maternal serum Overweight
5 years: 4.7
18 months: 8.25
BMI z-score
18 months: 0.2(0.1,0.4),
p-value < 0.05
OW
18 months: 1.29 (1.01, 1.64),
p-value < 0.05
Cohort Children,
5 years (BMI)
N = 349
Children,
5 years
(overweight)
N = 371
Children,
18 months
(overweight)
N = 444
Results: Lowest tertile used as reference.
Confounding: Maternal nationality, age at delivery, pre-pregnancy BMI, smoking during pregnancy, child sex, exclusive breastfeeding
duration, child's fish intake at age 5 years
Risk Ratio
(OW), or
Regression
coefficient per
loglO-unit
increase in
maternal PFOS,
or by tertiles
(BMI)
Kobayashi et al. Japan,
(2017, 3981430) 2002-2005
Medium
Cross-sectional Children from Maternal serum Ponderal index Regression
Hokkaido Study 5.3 coefficient per
on Environment
-1.07 (-1.79, -0.36),
p-value = 0.004
D-174
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
and Children's ln-unit increase
Health in PFOS
N = 176
Confounding: Maternal age, pre-pregnancy BMI, parity, maternal education, maternal smoking during pregnancy, gestational age, infant sex,
and maternal blood sampling period
Lauritzen et al. Norway and
(2018,4217244) Sweden,
Medium Recruitment
1986-1988
Cohort Pregnant Serum BMI, triceps Regression
women and Norway: 9.62 skinfold, coefficient or
their children at Sweden: 16.3 subscapular OR per ln-unit
5-year follow up skinfold, increase in
N = 412 overweight maternal PFOS
Regression coefficient
BMI: 0.18(0.01,0.35)
Triceps skinfold: 0.15 (0.02, 0.27)
Odds ratio
Overweight: 2.04 (1.11, 3.74)
Subscapular skinfold: No
statistically significant association
Confounding: Age, education, smoking at conception, pre-pregnancy BMI, weight gain at 17 weeks, interpregnancy interval, previous
breastfeeding duration and country of residence
Lopez-Espinosa United States,
etal. (2016, 2005-2006
3859832)
Medium
Cohort
Children, ages
6-9 years from
the C8 Health
Project
N= 1123 girls
and 1169 boys
Serum
Girls: 20.9
Boys: 22.4
Insulin-like
growth factor 1
(IGF-l)P(ln-
ng/mL)
Percent
difference
for 75th vs. 25th
percentile of
ln(PFOS), or by
quartiles
IGF-1
Girls: -5.6 (-8.2, -2.9)
Q4:-11.4 (-16.5,-6.0)
Boys: -5.9 (-8.3, -3.3)
Q3:-6.3 (-11.6, -0.6)
Q4:-11.5 (-16.6,-6.1)
Results: Lowest quartile used as reference.
Confounding: Age and month of sampling
Boys Q2; Girls Q2, Q3: No
statistically significant associations
Manzano- Spain,
Salgado et al. Recruitment
(2017,4238509) 2003-2008
Medium
Cohort Mother-child Maternal blood BMI, WC,
pairs, followed GM=5.80 overweight,
for 8 years, waist-to-hip
INMA Study ratio
N = 1230
Confounding: Maternal characteristics (i.e., region of residence, country of birth, previous breastfeeding, age, pre-pregnancy BMI), age of
child
Regression BMI, waist circumference,
coefficient per- overweight, waist-to-hip ratio: No
log2-unit statistically significant associations
increase in
PFOS
D-175
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Martinsson et al. Sweden,
(2020,6311645) 2003-2008
Medium
Case-control Pregnant Serum
women and 16.6
their children at
age 4, Southern
Sweden
Maternity
Cohort
N = 1,048
Results: Lowest quartile used as reference
Confounding: Risk strata, difference from strata-specific mean, sex
Overweight OR by quartiles OW
Q4: 1.57 (1.07,2.3)
Q2 and Q3: No statistically
significant association
Mora et al. United States,
(2017, 3859823) 1999-2002
Medium
Cohort
Early childhood
N = 992
Mid-childhood
N = 871
Maternal
Plasma
Early childhood:
24.8
Mid-childhood:
24.7
WC (cm), Sum
of subscapular
and triceps
skinfold
thickness (mm),
BMI, waist-to-
hip ratio,
obesity,
overweight,
total fat mass
index, total fat-
free mass index
Regression
coefficient per
IQR increase in
PFOS
All:
Sum of subscapular and triceps
skinfold thickness:
-0.41 (-0.77, -0.05)
Boys:
Waist-to-hip ratio:
-0.76 (-1.47, -0.05)
Early childhood:
BMI, obesity, overweight, total fat
mass index, total fat-free mass
index: No statistically significant
association
Mid-childhood:
Waist circumference (cm), Sum of
subscapular and triceps skinfold
thickness (mm), BMI, waist-to-hip
ratio, obesity, overweight, total fat
mass index, total fat-free mass
index:
No statistically significant
association
D-176
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Confounding: Maternal age, race/ethnicity, education, parity, pre-pregnancy BMI, timing of blood draw, household income, child sex, age at
outcome assessment
Scinicariello et
al., 2020,
6391244
Medium
United States,
2013-2014
Cross-sectional
Children aged Serum
3-11 years from
NHANES
N = 600
BMI z-score Regression
(BMIZ), height- coefficient per
GM = 3.90 (SE
= 0.17)
Girls: GM =
3.69 (SE = 0.15)
Boys: GM =
4.12 (SE = 0.27)
for-age z-score
(HAZ), weight-
for-age z-score
(WAZ)
ln-unit increase
in PFOS or by
tertiles
BMIZ: -0.09 (-0.30,0.13)
T2: -0.19 (-0.41,0.03)
T3: -0.21 (-0.53,0.11)
p-value for trend = 0.17
Girls: -0.20 (-0.48, 0.07)
Boys: -0.02 (-0.29, 0.24)
HAZ: -0.29 (-0.49,-0.10)
T2: -0.32 (-0.60, -0.04)
T3: -0.39 (-0.72,-0.06)
p-value for trend = 0.06
Girls: -0.34 (-0.73, 0.05)
Boys: -0.22 (-0.41, -0.03)
T3: -0.28 (-0.53,-0.03)
WAZ: -0.25 (-0.47, -0.03)
T2: -0.32 (-0.60, -0.04)
T3: -0.40 (-0.76, -0.04)
p-value for trend = 0.06
Girls: -0.35 (-0.72, 0.03)
Boys:-0.17 (-0.37, 0.03)
No other statistically significant
associations or trends by quartiles
stratified by sex
NHANES = National Health and Nutrition Examination Survey
Results: Lowest tertile used as reference
Confounding: Age, quadratic age, race/ethnicity, poverty income ratio, serum cotinine, birthweight, maternal smoking during pregnancy,
hematocrit, sex
Fleisch et al. United States,
(2017, 3858513) Pregnant
Medium for women
metabolic recruited 1999-
function 2002, outcome
Cohort Pregnant Plasma Leptin, Percent change HOMA-IR:
women and GM = 6.2 Adiponectin, per IQR Per IQR increase -10.1% (-16.4,
their children HOMA-IR increase in -3.3)
from PFOS, or by Q4:-24.7 (-37.8,-8.8)
Project Viva quartiles Females:
D-177
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Low for
assessed at mid-
N = 584
-16.7 (-25.7, -6.7)
HOMA-IR
childhood
Q4: -30.7 (-47.5, -8.4)
follow-up
Median age at
follow-
Leptin, adiponectin: No statistically
up = 7.7 years
significant associations
Results: Lowest quartile used as reference; Q4 (9.8-51.4 ng/mL), Q1 (< 0.1-4.2 ng/mL) PFOS.
Confounding: Characteristics of child (age, sex, race/ethnicity), mother (age, education), and neighborhood census tract at mid-childhood
(median household income, percent below poverty)
Pregnant Women
Jensen et al.
Denmark, Cohort
Pregnant Serum
Blood glucose,
Percent change
Blood glucose, insulin, c-peptide,
(2018,4354143)
recruitment
women, Odense 8.37
insulin, c-
per log2-unit
2-hour glucose, insulin resistance,
High
2010-2012,
Child Cohort
peptide, 2-hour
increase in
beta cell function, insulin
outcome
N = 158
glucose, insulin
PFOS
sensitivity: No statistically
assessed 12-
resistance, beta
significant association
20 weeks later
cell function,
insulin
sensitivity
Confounding: Age, parity, education level, pre-pregnancy BMI
Mitro et al.
United States, Cohort
Pregnant Plasma
WC (cm),
Percent
Skinfold thickness
(2020, 6833625) Recruitment
women, 24.8
BMI (kg/m2),
difference per
High
1999-2002
Project Viva
Adiponectin
log2-unit
All: 1.2 (0.1, 2.2), p-value < 0.05
N = 786
(ug/mL),
increase in
Skinfold
PFOS
Women < 35 at pregnancy: 1.5
thickness, Arm
(0.1,3), p-value <0.05
circumference,
HbAlc, Leptin WC, BMI, Adiponectin, arm
circumference, HbAlc, leptin: No
statistically significant associations
Confounding: Age, pre-pregnancy BMI, marital status, race/ethnicity, education, income, smoking, parity, breastfeeding in a prior pregnancy
Preston et al.
(2020, 6833657)
High
United States,
1999-2002
Cohort
Pregnant
women from
Project Viva
N = 1,533
Serum
25.7
Gestational
diabetes,
glucose
tolerance,
hyperglycemia,
Regression
coefficient by
quartiles
Glucose blood level,
All
Q4: 4.3 (0.5, 8.0)
<35 years
Q4: 6.5 (2.1, 10.9)
D-178
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
glucose blood
level
Q3: 5.2 (0.8, 9.7)
Q2: 5.2 (0.8, 9.6)
Gestational diabetes, glucose
tolerance, hyperglycemia: No
statistically significant association
Results: Lowest quartile used as reference; Q1 (0.1-18.8 ng/mL), Q2 (18.9-25.7 ng/mL), Q3 (25.8-34.9 ng/mL), Q4 (35.0-185.0 ng/mL).
Confounding: Pre-pregnancy BMI, prior history of gestational diabetes/parity, race/ethnicity, smoking, and education, maternal age (Full
group only)
Starling et al.
(2017, 3858473)
High
United States
2009-2014
Cohort
Maternal serum
2.4
Pregnant
women and
their children in
the Healthy
Start study
N = 628
Confounding: Maternal age, pre-pregnancy body mass index (BMI)
gestational age at blood draw
Maternal
glucose
Regression
coefficient per
unit increase in
PFOS and by
tertile
Maternal glucose: No statistically
significant associations
race/ethnicity, education, smoking during pregnancy, gravidity, and
Ashley-Martin Canada, Cohort Pregnant
etal. (2016, Pregnant women from
3859831) women MIREC
Medium recruited 2008- N= 1,609
2011, outcome
assessed at birth
Confounding: Age, income, parity
Serum
0.15
GWG (kg) Regression Underweight/normal BMI: 0.39
coefficient per (0.02, 0.75)
log2-unit
increase in Overweight and obese BMI: No
PFOS statistically significant association
Jaacks et al.
(2016, 3981711)
Medium
United States,
2005-2007
Cohort
Pregnant
women
N = 218
Serum
Mean= 14.81
GWG (kg)
Regression
coefficient and
OR per SD-unit
increase in
PFOS
GWG
0.26 (-0.66, 1.18)
OR for excessive GWG: 1.01
(0.72, 1.4)
Confounding: Pre-pregnancy non-fasting serum lipids, BMI
Liu et al. (2019,
5881135)
Medium
China, 2013-
2015
Case-control
Pregnant Serum
women without 3.13
history or
family history
of diabetes
Gestational Regression
diabetes coefficient per
(GDM), glucose ln-unit increase
homeostasis or by tertiles
GDM:
m-PFOS
Per ln-unit increase: 1.36 (Oi
2.11)
D-179
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
N = 189
summ-PFOS or T2: 1.53 (0.7, 3.34)
L-PFOS T3: 1.23 (0.56, 2.72)
L-PFOS
Per ln-unit increase: 1.58 (0.89,
2.79)
T2: 1.34 (0.62, 2.93)
T3: 1.37 (0.62,3.02)
Glucose homeostasis: No
statistically significant association
Results: Lowest tertile used as reference.
Confounding: Maternal age, BMI in early pregnancy, fetal sex, serum triglyceride, total cholesterol
Marks et al.
(2019, 5381534)
Medium
United
Kingdom
1991-1992
Cohort
Mothers from
ALSPAC
N = 905
GWG (absolute) Regression
coefficient per
10% increase in
log-unit PFOS
GWG: No statistically significant
associations
Serum
Mothers of
sons: 13.8
Mothers of
daughters: 19.8
Comparison: Logarithm base not specified.
Confounding: Maternal education, prenatal smoking, maternal age at delivery, parity, pre-pregnancy BMI, gestational age at delivery
gestational age at sample
Rahman et al.
(2019, 5024206)
Medium
United States,
2009-2013
Cohort
Pregnant
women with
singleton
pregnancies
N = 2,292
Plasma
GM = 5.21
GDM
Risk Ratio per GDM: No statistically significant
SD-unit increase associations
in PFOS
Confounding: Maternal age, enrollment BMI, education, parity, race/ethnicity, serum cotinine
Ren et al. (2020,
6833646)
Medium
China, 2012
Cross-sectional
Pregnant
women,
Shanghai-
Minhang Birth
Cohort Study
N = 705
Plasma Glucose (1 hour, Regression
10.7 fasting) coefficient per
ln-unit increase
in PFOS
Glucose (1 hour tolerance test):
0.31 (0.11,0.50), p-value = 0.003
Glucose after fasting, glucose after
1 hour tolerance test by gestational
weeks: No statistically significant
association
D-180
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Outcome Comparison
Resultsb
Exposure
Matrix, Levels
(ng/mL)a
Confounding: Maternal age at enrollment, pre-pregnancy BMI, per capita household income, education level, passive smoking, pregnancy
complication, history of abortion and stillbirth, parity
Shapiro et al.
(2016, 3201206)
Medium
Canada,
2008-2011
Cohort
Pregnant
women
N = 1,195
Urine
Normal glucose
GM = 4.58
Gestational
impaired
glucose
tolerance
GM = 4.29
Women with
GDM
GM = 4.74
Confounding: Maternal age, race, pre-pregnancy BMI, and education
GDM,
gestational
impaired
glucose
tolerance
OR per quartile Gestational diabetes, gestational
PFOS impaired glucose tolerance: No
statistically significant association
Valvi et al.
(2017, 3983872)
Medium
Faroe Islands
1997-2000
Cohort
Maternal serum
27.2
Gestational
diabetes
Pregnant
women and
their children
N = 604
Results: Lowest tertile used as the reference group
Confounding: Maternal age at delivery, education, parity, pre-pregnancy BMI, smoking during pregnancy
OR per
doubling of
PFOS, or by
tertiles
Gestational diabetes:
Per doubling: 0.86 (0.43, 1.7)
T2: 0.85 (0.43, 1.7)
T3: 0.56 (0.26, 1.19)
Wang et al. China
(2018, 5079666) 2013
Medium
Case-control
Pregnant
Serum
women with
n-PFOS
(cases) and
Cases: 2.70
without
Controls: 2.81
(controls) GDM
lm-PFOS
N = 242
Cases: 0.14
Controls: 0.14
3m+4m-PFOS
Cases: 0.44
Controls: 0.42
5m-PFOS
Cases: 0.36
Controls: 0.36
6m-PFOS
Cases: 0.29
Controls: 0.31
Fasting blood
glucose, GDM
Fasting blood
glucose: OR by
tertiles of PFOS
isomer
GDM: OR per
unit increase in
PFOS isomer
Fasting blood glucose
n-PFOS
T2: 1.94 (1.05,3.58),
p-value < 0.05
T3: 1.59 (0.85,2.96)
lm-PFOS
T2: 1.86 (1.00, 3.48),
p-value < 0.05
T3: 2.07 (1.09, 3.93),
p-value < 0.05
3m+4m-PFOS
T2: 1.81 (0.98,3.33)
T3: 1.88 (1.00,3.52),
p-value < 0.05
5m-PFOS
D-181
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
T2: 1.94 (1.05,3.80),
p-value < 0.05
T3: 2.45 (1.24,4.64),
p-value < 0.05
6m-PFOS
T2: 1.24 (0.67,2.28)
T3: 1.42 (0.83,2.77)
GDM: No statistically significant
associations
Results: Lowest tertile used as reference.
Confounding: Fasting blood glucose: BMI, age, GDM status; GDM: BMI, GWG, ethnic groups, maternal education, parity, maternal
drinking during pregnancy, household income
Wang et al. China,
(2018, 5080352) 2013-2014
Medium
Cohort
Pregnant
women aged
20-40
N = 385
Serum
5.4
Fasting blood LSM by tertiles Fasting blood glucose:
glucose, fasting
insulin, HOMA-
IR, gestational
diabetes, oral
glucose
tolerance
T2: 1.47 (1.45, 1.48),
p-value < 0.05
T3: 1.47 (1.45, 1.48),
p-value < 0.05
Oral glucose tolerance: 1.88 (1.84,
1.91), p-value < 0.05
Fasting insulin, HOMA-IR,
gestational diabetes: No statistically
significant association
Results: Lowest tertile used as reference.
Confounding: Pregnant age, diabetes mellitus history of relatives, husband smoking status, family per capita income, baby sex, averaged
intake of meat, vegetable, and aquatic products, averaged physical activity, and averaged energy intake, pre-pregnant maternal BMI
Xu et al.(2020,
China, Nested case- Pregnant Serum GDM
OR per unit
GDM
6833677)
2017-2019 control women Cases: 6.69
increase in
Q2: 0.69 (0.34, 2.07)
Medium
N = 165 cases, Controls: 6.45
PFOS; OR per
Q3: 0.72 (0.48, 1.90)
330 controls
loglO-unit
Q4: 1.07 (0.51, 1.32)
increase in
p-trend = 0.27
PFOS
log-PFOS: 0.61 (0.42, 1.65), p-
value = 0.21
Confounding: Maternal age, sampling time, parity, BMI, educational level, and serum lipids
D-182
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels Outcome
(ng/mL)a
Comparison
Resultsb
General Population
Cardenas et al. United States,
(2017, 4167229) Recruitment
High July 1996-May
1999, outcome
assessed
annually until
May 2001
Adults at high
risk of Type 2
diabetes
N = 956
Plasma
GM = 26.38
Cohort Adults at high Plasma Adiponectin
(ug/mL),
HbAlc (%),
Insulin (fasting)
(uU/mL),
Glucose
(fasting)
(uU/mL),
HOMA-IR,
Insulin (30 min,
uU/mL),
Proinsulin
(fasting, pM),
HOMA-B,
Insulin
(corrected
response),
Insulinogenic
index, Diabetes,
HOMA-IR,
glucose (30
mins), glucose
(2 hours), BMI
Confounding: Sex, race/ethnicity, BMI, age, marital status, education, smoking history.
Regression
coefficient per
doubling of
PFOS
HbAlc: 0.03 (0.002, 0.07),
p-value = 0.04
Insulin (fasting): 1.37 (0.41, 2.34),
p-value = 0.005
Glucose (fasting): 0.55 (0.03, 1.06),
p-value = 0.04
HOMA-IR: 0.39 (0.13, 0.66),
p-value = 0.004
Insulin (30 min): 4.63 (0.89, 8.36),
p-value = 0.02
Proinsulin (fasting): 1.37(0.5,
2.25), p-value = 0.002
HOMA-B: 9.62 (1.55, 17.7),
p-value = 0.02
Diabetes, glucose (30 mins),
glucose (2 hours), BMI,
adiponectin, insulin (corrected),
insulinogenic index: No statistically
significant association
Blake et al. United States,
(2018, 5080657) 1991-2008
Medium
Cohort
Adults living in
a community
with water
supply from a
PFAS-
contaminated
aquifer
N = 192
Serum
28.4
BMI
Percent change
per IQR
increase in
PFOS
BMI: No statistically significant
associations
D-183
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location, . Population, Exposure
Design ' .. Matrix, Levels
Years Ages,N (ng/mL)a
Outcome
Comparison
Resultsb
Confounding: Age, year of measurement, sex, education, income, marital status, and BMI
Cardenas et al.
United States, Controlled trial Adults older Plasma
T2D
Hazard ratio per T2D:
(2019,5381549) 1996-2014 than 25 without GM = 26.38
log2-unit
HR: 1.05 (0.94, 1.18)
Medium
diabetes and
increase in
T2: 0.94 (0.75, 1.17)
with elevated
baseline PFOS
T3: 0.94 (0.75, 1.18)
fasting and
postload
glucose,
Diabetes
and by PFOS
tertiles
Prevention
Program
N = 956
Confounding: Sex, race/ethnicity, baseline age, marital status, education, income, smoking history, BMI, maternal diabetes, paternal diabetes,
treatment assignment
Christensen et
United States, Cross-sectional Male anglers Serum
Diabetes, pre-
OR per-unit in
Diabetes, pre-diabetes: No
al. (2016,
2011-2013 N = 154 19.0
diabetes
PFOS
statistically significant associations.
3350721)
Medium
Confounding: Age, BMI, employment status, number of alcoholic drinks consumed per month
Conway et al.
United States, Cross-sectional All individuals Serum
T1D,
OR per ln-unit
T1D: 0.73 (0.67, 0.79)
(2016,3859824) 2005-2006 working or All participants
Medium living in six mean =86.5
PFOS-
T2D,
Uncategorized
Diabetes
increase in
PFOS
T2D: 0.92 (0.88, 0.96)
Children with T1D: 0.52 (0.54,
0.87)
contaminated
Adults with TID: 0.77 (0.71, 0.84)
water districts
with diabetes
Uncategorized diabetes: No
N = 6,460
statistically significant association
Confounding: Age, sex, race, BMI, eGFR, hemoglobin, iron
Donat-Vargas et Sweden, Case-control Adults with Plasma
T2D
OR per SD
T2D
al. (2019,
1990-2003, (cases) and Cases:
loglO-unit
OR: 0.7 (0.47, 1.03)
5083542)
2001-2012 without 19.0
increase in
T2: OR: 0.79 (0.34, 1.87)
Medium
(controls) type 2 Controls:
baseline PFOS,
diabetes living 20.0
or by tertiles
HOMA-B and HOMA-IR: No
in Sweden
N = 248
statistically significant associations
Results: Lowest tertile used as reference; T1 (13, 11-16 ng/mL), T2 (21, 19-23 ng/mL)
D-184
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location, Design Population, Matrix, Levels Outcome Comparison
Years Ages,N (ng/mL)a
Resultsb
Confounding: Gender, age, sample year, red and processed meat intake, fish intake, BMI
Duan et al.
(2020, 5918597)
Medium
China, 2017
Cross-sectional
Adults, 19 to
87 years old
N = 252
Serum
14.24
Fasting glucose
(nmol/L),
HbAlc
HbAlc < 55, fasting glucose: No
statistically significant association
Confounding: Sex, age, body mass index, smoking and alcohol-drinking status, exercising status, education level, and family history of
diabetes
Regression
coefficient per
1% increase in
serum PFOS
HbAlc 55+: 0.02819 (0.00557,
0.04965)
Jain et al. (2019, United States, Cohort
Adults from
Serum Obesity
Comparison of Obesity: p-value = 0.01
5080621)
2011-2014
NHANES, 20
Non-obese
GM of PFOS
Medium
and older
GM = 2.2
levels for non-
N = 2,883
Obese GM = 2.0
obese vs obese
Confounding: Not reported
Jeddy et al. England,
(2018, 5079850) mothers
Medium recruited 1991—
2002, outcome
assessed at
age 17
Fat mass: No statistically
significant association
Nested case- Pregnant Maternal serum Fat mass Regression
control studies mothers and 20.2 coefficient per
their 17-year old unit increase in
daughters, PFOS
ALSPAC
N = 221
Confounding: Maternal pre-pregnancy BMI, maternal education, maternal age at delivery, gestational age at sample collection, and ever
breastfed status at 15 months
Liu et al. (2018, Boston,
4238396) Massachusetts
Medium for and Baton
adiposity/weight Rouge,
change Louisiana,
Uninformative 2004-2007
for insulin
resistance
Controlled Trial
Overweight and
obese patients
from the
POUNDS-Lost
Trial, Ages 30-
70,
N = 621
Plasma, glucose
Males: 27.2
Females: 22.3
Body weight
(kg), Resting
metabolic rate
(RMR)
(kcal/24h),
HbAlc, insulin,
glucose, fat
mass, WC,
leptin, HOMA-
IR
Partial
Spearman
correlation with
baseline PFOS
(insulin, leptin)
Regression
coefficient per
loglO-unit
increase in
PFOS, or
by tertile
Spearman correlations
Body weight: 0.8, p-value < 0.05
Body weight, months 6-24
All:
Tl: 1.5, p-trend = 0.007
T2: 3.5, p-trend = 0.007
T3: 3.2, p-trend = 0.007
Women:
Tl: 2.1, p-trend = 0.01
T2: 4.1, p-trend = 0.01
T3: 4.0, p-trend = 0.01
Per loglO-umit increase in PFOS
D-185
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
0.8, p-value < 0.05
RMR
First 6 months, all
Tl: -5.0, p-trend = 0.005
T2: -24.7, p-trend = 0.005
T3: -45.4, p-trend = 0.005
Months 6-24, all
Tl: 94.6, p-trend < 0.001
T2: 67.3, p-trend < 0.001
T3: 0.9, p-trend < 0.001
First 6 months, women
Tl: -19.2, p-trend = 0.01
T2: -29.7, p-trend = 0.01
T3: -60.4, p-trend = 0.01
Months 6-24, men
Tl: 46.8, p-trend = 0.05
T2: 60.8, p-trend = 0.05
T3: -40.2, p-trend = 0.05
Months 6-24, women
Tl: 141.6, p-trend = 0.001
T2: 90.1, p-trend = 0.001
T3: 47.7, p-trend = 0.001
HbAlc, glucose, fat mass, WC,
leptin: No statistically significant
association
Results: Lowest tertile used as reference; Tertile 1 (< 19.2 ng/mL), tertile 2 (19.2-32.1 ng/mL), tertile 3 (> 32.1 ng/mL) PFOS.
Confounding: Age, sex, race, education, smoking status, alcohol consumption, physical activity, menopausal status (women only), hormone
replacement therapy (women only), and dietary intervention groups.
Liu et al. (2018,
4238514)
Medium
United States,
2013-2014
Cross-sectional
Adults from
NHANES
N = 1,871
Serum
GM =
5.28
Fasting blood
glucose, 2-hour
glucose, HbAlc,
insulin levels,
HOMA-IR, beta
cell function,
Regression
coefficient per
ln-unit increase
in PFOS
Fasting blood glucose: 1.96
(SE = 0.79)
2-hour glucose, HbAlc, insulin
levels, HOMA-IR, beta cell
function, metabolic syndrome, WC:
D-186
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
metabolic No statistically significant
syndrome, WC associations
Confounding: Age, gender, ethnicity, smoking status, alcohol intake, household income, WC, and medications (anti-hypertensive, anti-
hyperglycemic, and anti-hyperlipidemic agents)
Mancini et al. France,
(2018, 5079710) 1990-2012
Medium
Cohort
Su et al. (2016,
3860116)
Medium
Women aged
40-60, E3N
Cohort
N = 71294
Food T2D
Mean =0.49 ng/
kg body
weight/day
Hazard ratio per T2D: No statistically significant
decile PFOS association
Confounding: Smoking status, physical activity, education level, hypertension, hypercholesterolemia, family history of diabetes, energy
intake, alcohol intake, adherence to the Western diet and adherence to the Mediterranean diet, water consumption, dairy product consumption
Taiwan,
2009-2011
Cross-Sectional
Adults aged 20- Plasma
60 living in 8.0
Taiwan
N = 571
Diabetes,
Fasting blood
glucose
(ng/mL),
blood glucose
(120 mins) (In)
(ng/mL),
glucose AUC
(ng/mL),
HbAlc (In) (%)
OR and GM Diabetes:
ratio (GMR) per OR: 2.39 (1.52, 3.76)
doubling of ORQ4: 3.37 (1.18, 9.56)
PFOS, or by
quartiles
Glucose (Fasting):
GMR: 1.03 (1.01, 1.04)
GMRQ4: 1.05 (1.02, 1.09)
Glucose (120 min)
GMR: 1.08 (1.05, 1.12)
GMRQ4: 1.17(1.08, 1.25)
Glucose AUC:
GMR: 1.06 (1.04, 1.09)
GMRQ4: 1.12(1.06, 1.19)
Results: Lowest quartile used as reference; Q1 (< 2.4 ng/mL); Q4 (> 4.8 ng/mL).
Confounding (Diabetes): Age, sex, education, smoking (ever vs never), alcohol (evervs never), BMI, hypertension, total cholesterol, regular
exercise
Confounding (Other): Age, sex, education, smoking, alcohol, BMI, hypertension, total cholesterol, regular exercise
Sunetal. (2018,
4241053)
Medium
United States,
recruitment
1989, blood
sample
collection 1995—
2000, outcome
Case-control
Female nurses
drawn from the
Nurses' Health
Study II cohort
study,
N = 1586
Plasma T2D Regression T2D
Cases: hemoglobin, coefficient SD Per SD increase: 1.15 (0.98, 1.35),
35.7 insulin, loglO-unit p-value = 0.008
Controls: adiponectin increase in OR for T2: 1.63 (1.25, 2.12)
33.1 PFOS OR for T3: 1.62 (1.09,2.41)
D-187
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Partial Spearman correlation
coefficient for hemoglobin, insulin,
and adiponectin: No statistically
significant association
assessed during OR by tertiles
biennial follow
up through June
2011
Results: Lowest tertile used as reference.
Confounding: Age, month of sample collection, fasting status, menopausal status, postmenopausal hormone use, family history of diabetes,
oral contraceptive use, breastfeeding duration at blood draw, number of children delivered after 1993, states of residence, smoking status,
alcohol intake, physical activity, baseline BMI, and Alternative Healthy Eating Index (AHEI) score
Metabolic syndrome: 1.89 (0.93,
3.86); p-value = 0.08
All other outcomes: No statistically
significant associations
Chen et al. Croatia
Cross-sectional Residents of
Plasma
BMI, fasting
Metabolic
(2019, 5387400) 2007-2008
Hvar ages 44-
GM = 8.91
insulin
syndrome: OR
Medium for
56 years
(Range: 2.36-
(nIU/mL),
per ln-unit
metabolic
N = 122
33.67)
fasting plasma
increase in
syndrome
glucose
PFOS
Low for all
(mmol/L),
other outcomes
glycated HbAlc
All other
(%), hip
outcomes:
circumference
regression
(cm),
coefficient per
homeostatic
ln-unit increase
model
in PFOS
assessment of
beta-cell
function
(HOMA-(3),
homeostatic
model
assessment of
insulin
resistance
(HOMA-IR),
metabolic
syndrome
defined by the
ATP III criteria,
waist
D-188
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location, Design Population, Matrix, Levels Outcome Comparison
Years Ages,N (ng/mL)a
Resultsb
circumference
(cm)
Confounding: Age, sex, education, socioeconomic status, smoking, dietary pattern, and physical activity
Notes: ALSPAC = Avon Longitudinal Study of Parents and Children; AUC = area under the curve; BMI = body mass index; DM = diabetes mellitus; EYHS = European Youth
Heart Study; GDM = gestational diabetes mellitus; GM = geometric mean; GWG = gestational weight gain; HbAlc = Hemoglobin Ale; HOMA = Homeostatic model
assessment; HOME = Health Outcomes and Measures of the Environment; IGF = insulin-like growth factor; IQR = interquartile range; IR = insulin resistance; KorEHS-C: Korea
Environmental Health Survey in Children and Adolescents; LSM = least square mean; MIREC = Maternal Infant Research on Environmental Chemicals;OR = odds ratio;
OW = overweight; POPUP = Persistent Organic Pollutants in Uppsala Primiparas; RR = risk ratio; SD = standard deviation; SOLAR = Study of Latino Adolescents at Risk of
Type 2 Diabetes; T1D = type 1 diabetes; WC = waist circumference.
a Exposure levels are reported as median in ng/mL unless otherwise noted.
b Results are reported as effect estimate (95% confidence interval) unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
D.8 Nervous
Table D-17. Associations Between PFOS Exposure and Neurological Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population, Exposure
Ages, Matrix, Levels Outcome
N (ng/mL)a
Comparison
Resultsb
Children and Adolescents
Harris et al. United States,
(2018, 4442261) Recruitment:
High 1999-2002;
Follow-up at
early- and mid-
childhood
Cohort
Pregnant
women and
their children
from Project
Viva
N = 853
Plasma
Maternal: 24.9
(18.4-34.4)
Child: 6.2 (4.2-
9.7)
Both age
groups: Wide
Range
Assessment of
Visual Motor
Abilities
(WRAVMA)
score
Early childhood
only: Peabody
Picture
Vocabulary Test
(PPVT-III)
score
Mean difference
by quartiles of
PFOS exposure
Visual-Motor
Mid-childhood (maternal plasma)
Q2
Q3
Q4
-1.6 (-4.7, 1.6)
-1.4 (-4.7, 1.8)
-3.2 (-6.6, 0.2)
Mid-childhood (child plasma)
Q2
Q3
Q4
-1.6 (-5.5, 2.2)
-4.6 (-8.7, -0.5)
-2.0 (-6.3, 2.2)
Non-Verbal IQ
Mid-childhood (maternal plasma)
Q2
Q3
Q4
-0.7 (-3.8, 2.3)
-1.8 (-5.0, 1.4)
1.6 (-1.8, 4.9)
D-189
-------
DRAFT FOR PUBLIC COMMENT
T Population,
Reference, Location, . .
^ «, Design Ages9
Confidence Years
MARCH 2023
Exposure
Matrix, Levels Outcome Comparison Resultsb
(ng/mL)a
Mid-childhood (child plasma)
Mid-childhood
Q2: -0.4 (-4.0, 3.2)
only: Kaufman
Q3: 1.6 (-2.3, 5.4)
Brief
Q4: -0.1 (-4.1, 3.8)
Intelligence Test
Second Edition
Verbal IQ
(KBIT-2) non-
Mid-childhood (maternal plasma)
verbal and
Q2: -2.1 (-4.5,0.2)
verbal IQ,
Q3: -1.7 (-4.2, 0.7)
(WRAML2)
Q4: 0.8 (-1.8, 3.4)
design memory
Mid-childhood (child plasma)
and picture
Q2: 0.9 (-2, 3.8)
memory
Q3: -0.4 (-3.4,2.7)
Q4: -0.2 (-3.4, 3.0)
Design memory
Mid-childhood (maternal plasma)
Q2: -0.1 (-0.7, 0.4)
Q3: 0.3 (-0.3,0.8)
Q4: 0.6 (0, 1.2)
Mid-childhood (child plasma)
Q2: 0.1 (-0.5,0.7)
Q3: 0.1 (-0.6,0.7)
Q4: -0.2 (-0.9, 0.5)
Picture memory
Mid-childhood (maternal plasma)
Q2: -0.3 (-0.9, 0.2)
Q3: -0.1 (-0.7, 0.5)
Q4: 0.4 (-0.2, 1.0)
Mid-childhood (child plasma)
Q2: -0.1 (-0.8, 0.5)
Q3: 0.1 (-0.6,0.9)
Q4: 0 (-0.7, 0.8)
D-190
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Early childhood: No statistically
significant associations
Results: Lowest quartile used as reference.
Confounding: Year of pregnancy blood collection gestational age at time of pregnancy blood collection, estimated glomerular filtration rate at
blood draw, maternal race/ethnicity, age, education, KB IT-2 score, pre-pregnancy BMI, smoking status, paternal education, annual household
income in mid-childhood, HOME-SF score, child's sex and age at mid-childhood cognitive testing, proxy for breastfeeding of a prior child0
Niuetal. (2019,
5381527)
High
China,
Recruitment:
2012; Follow-up
at age 4 years
Cohort
Pregnant
women and
their children
from the
Shanghai-
Minhang Birth
Cohort
N = 533 (236
Females; 297
Males)
Maternal serum ASQ-3 skill RR per ln-unit Communication
10.8(7.6-15.8) scales: increase in Overall: 1.01 (0.77, 1.34)
communication, PFOSandby Females: 1.04 (0.65, 1.68)
gross motor, tertiles T2: 0.52 (0.26, 1.04); p-value <0.10
fine motor, T3: 1.10 (0.63, 1.92)
problem Males: 1.00 (0.70, 1.44)
solving, T2: 1.16(0.76, 1.77)
personal-social T3: 0.89 (0.53, 1.51)
p-value for interaction by
sex = 0.350
Gross Motor
1.22 (0.79, 1.89)
No statistically significant
associations, trends, or interactions
by sex
Fine Motor
Overall: 1.25 (0.79, 1.96)
No statistically significant
associations, trends, or interactions
by sex
Problem Solving
Overall: 1.02 (0.71, 1.47)
Females: 1.16 (0.63,2.15)
T2: 0.55 (0.15,2.07)
T3: 2.00 (0.77,5.17)
Males: 0.93 (0.59, 1.47)
T2: 1.21 (0.65,2.28)
D-191
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
T3: 0.66 (0.29, 1.48)
p-value for interaction by
sex = 0.010
Personal-Social Skills
Overall: 1.34 (0.91, 1.96)
Females: 2.56 (1.2, 5.45)
T2: 0.32 (0.04, 2.77)
T3: 2.97 (0.90, 9.84);
p-value <0.10
p-trend < 0.10
Males: 1.05 (0.67, 1.64)
T2: 1.47 (0.76,2.84)
T3: 1.18 (0.57,2.44)
p-value for interaction by
sex = 0.039
Outcome: Neuropsychological problems defined as scores < 10th percentile.
Results: Lowest tertile used as reference
Confounding: Maternal age at enrollment, pre-pregnancy BMI, maternal education, paternal education, parity, per capita household income,
maternal passive smoking, maternal prenatal depressive symptoms, gestational age, child's sex
Oulhote et al. Faroe Islands,
(2016, 3789517) Recruitment:
High 1997-2000,
Follow-up at
ages 5 and 7
Cohort
Children at
5 years
(N = 508) and
7 years
(N = 491)
Serum
Maternal: 27.35
(23.19-33.13)
5 years: 16.78
(13.52-21.05)
7 years: 15.26
(12.38-18.99)
Strengths and
Difficulties
Questionnaire
(SDQ) scores:
Total score
(hyperactivity/in
attention,
conduct
problems, peer
relationship
problems,
emotional
symptoms),
prosocial
behavior,
internalizing
Mean difference
(autism,
internalizing,
externalizing,
total) or mean
ratio
(hyperactivity/in
attention,
conduct, peer
relationship,
emotional,
prosocial) per
doubling of
PFOS
SDQ total score
Prenatal: 0.46 (-0.78, 1.7),
p-value = 0.47
5-year serum: 0.51 (-0.5, 1.52),
p-value = 0.32
7-year serum: 0.18 (-0.95, 1.31),
p-value = 0.76
Hyperactivity /Inattention
Prenatal: 1.03 (0.80, 1.31),
p-value = 0.84
5-year serum: 1.05 (0.86, 1.29),
p-value = 0.64
7-year serum: 0.88 (0.70, 1.11),
p-value = 0.27
D-192
-------
DRAFT FOR PUBLIC COMMENT
T Population,
Reference, Location, . .
^ «, Design Ages9
Confidence Years
MARCH 2023
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
problem,
externalizing
problems,
autism
screening (peer-
problems minus
pro-social)
Conduct
Prenatal: 1.03 (0.81, 1.32),
p-value = 0.80
5-year serum: 1.00 (0.81, 1.23),
p-value = 0.98
7-year serum: 1.01 (0.80, 1.26),
p-value = 0.95
Peer Relationship
Prenatal: 1.31 (0.87, 1.96),
p-value = 0.19
5-year serum: 1.28 (0.91, 1.80),
p-value = 0.15
7-year serum: 1.17 (0.82, 1.69),
p-value = 0.39
Emotional
Prenatal: 1.10(0.84, 1.44),
p-value = 0.49
5-year serum: 1.14 (0.90, 1.45),
p-value = 0.26
7-year serum: 1.22 (0.94, 1.58),
p-value = 0.13
Prosocial
Prenatal: 1.00 (0.91, 1.09),
p-value = 0.96
5-year serum: 0.98 (0.91, 1.06),
p-value = 0.70
7-year serum: 1.01 (0.92, 1.10),
p-value = 0.88
Internalizing
Prenatal: 0.35 (-0.35, 1.05),
p-value = 0.32
D-193
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
5-year serum: 0.44 (-0.15, 1.02),
p-value = 0.15
7-year serum: 0.48 (-0.16, 1.13),
p-value = 0.14
Externalizing
Prenatal: 0.11 (-0.68, 0.89),
p-value = 0.79
5-year serum: 0.08 (-0.58, 0.73),
p-value = 0.82
7-year serum: -0.31 (-1.03, 0.42),
p-value = 0.41
Autism screening
Prenatal: 0.2 (-0.37, 0.77),
p-value = 0.49
5-year serum: 0.33 (-0.14, 0.8),
p-value = 0.17
7-years serum: 0.06 (-0.46, 0.58),
p-value = 0.82
Confounding: Age, sex, maternal age, pre-pregnancy BMI, parity, socio-economic status, alcohol, and smoking during pregnancy
Braun et al. United States
(2014, 2345999) Recruitment:
Cohort
Maternal Serum
13 (9.3-18)
2003-2006;
Follow-up at
ages 4-5 years
Social
Responsiveness
Scale (SRS)
total score
Regression
coefficient per
loglO-unit/2SD
increase in
PFOS
Pregnant
women and
Medium 2003-2006; their children
from the HOME
study
N = 175 (80
Females; 95
Males)
Confounding: Maternal race, maternal age, maternal education, marital status, annual household income, maternal depressive symptoms,
maternal IQ, child sex, caregiving environment score, maternal serum
SRS
2.1 (0.2, 3.9)
Females: 0.9 (-1.5, 3.3)
Males: 3.8 (1.3,6.3)
p-value for interaction by
sex = 0.08
Chen et al. Taiwan,
(2013, 2850933) Recruitment:
Medium 2004-2005;
Follow-up at
age 2 years
Cohort Pregnant Cord blood
women and Mean =7.0
their children (SD = 5.8)
from the Taiwan
Comprehensive
Developmental
Inventory (CDI)
skill quotients:
cognitive, fine-
Regression
coefficient per
IQR increase in
ln-transformed
PFOS
Cognitive: -0.8 (-2.8, 1.1)
Fine-Motor: -1.8 (-3.8, 0.1)
Gross-Motor: -3.7 (-6.0, -1.5)
Language: -0.9 (-2.9, 1.2)
Self-Help: -2.2 (-4.8, 0.3)
D-194
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Social: -1.0 (-3.7, 1.6)
Whole Test: -2.1 (-4.1,-0.2)
Birth Panel motor, gross-
Study motor,
N = 239 language, self-
help, social,
whole test
Confounding: Maternal education, family income, infant sex and gestational age, breastfeeding, HOME score at 24 months of age, cord blood
cotinine levels, postnatal environmental tobacco smoke exposure
Ghassabian et United States,
al. (2018, 2008-2010
5080189)
Medium
Cohort
Children aged
7 years from
Upstate KIDS
Study
N = 788
Blood
1.74
(IQR = 1.33)
SDQ scores:
total behavioral
difficulties-total
score,
borderline
problems;
hyperactivity,
conduct, peer,
or emotional
problems;
difficulties in
prosocial
behavior
Regression
coefficient (total
behavioral
difficulties,
problem scores)
and OR
(borderline
behavioral
difficulties,
problem scores,
difficulties in
prosocial
behavior) per
log-SD increase
in PFOS and by
quartiles
Total Behavioral Difficulties ((3)
0.04 (-0.02,0.10)
Q2: 0.14 (-0.01, 0.28)
Q3: 0.04 (-0.11,0.19)
Q4: 0.17 (0.01, 0.32)
Conduct problems (OR)
1.22 (0.97, 1.52)
Q2: 1.78 (0.97, 3.27)
Q3: 0.86 (0.43, 1.74)
Q4: 2.22 (1.18, 4.15)
Conduct problems ((3)
0.02 (-0.08, 0.13)
Q2: 0.14 (-0.10, 0.39)
Q3: -0.07 (-0.33,0.19)
Q4: 0.19 (-0.07, 0.46)
Emotional problems (OR)
1.31 (1.04, 1.63)
Q2: 2.08 (1.13, 3.80)
Q3: 0.89 (0.47, 1.68)
Q4: 2.28 (1.24, 4.18)
Emotional problems ((3)
0.09 (0,0.18)
Q2: 0.24 (0.03, 0.45)
Q3: 0.01 (-0.20,0.22)
Q4: 0.27 (0.05, 0.49)
D-195
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Borderline Behavioral Difficulties
(OR)
1.30 (1.03, 1.65)
Q2: 1.67 (0.84, 3.34)
Q3: 1.73 (0.87,3.43)
Q4: 2.47 (1.29, 4.72)
Difficulties in Prosocial Behavior
(OR)
1.26 (0.92, 1.72)
Q2: 0.86 (0.35,2.15)
Q3: 1.72 (0.65,4.52)
Q4: 1.87 (0.70,4.98)
Hyperactivity problems, peer
problems: No statistically
significant associations
Comparison: Logarithm base not specified.
Results: Lowest quartile used as reference.
Confounding: Child's age and sex, maternal age, pre-pregnancy BMI, race/ethnicity, education, marital status, history of smoking in
pregnancy, having private insurance, parity, and infertility treatment
Goudarzi et al. Japan,
Cohort Pregnant
Maternal serum
Bayley Scales
Regression
MDI
(2016, 3981536) 2002-2005
women and
5.7 (4.4-7.4)
of Infant
coefficient
6 Months: 0.018 (-4.52, 5.59)
Medium
their infants at 6
Development,
loglO-unit
Females: 0.072 (-5.19, 9.38)
and 18 months
Second Edition
increase in
Males:-0.141 (-11.26,3.45)
from the
(BSID-II)
PFOS
18 Months: 0.052 (-9.91, 16.66)
Hokkaido Study
mental
on Environment
development
PDI
and Children's
index (MDI),
6 Months: 0.039 (-6.38, 10.37)
Health
psychomotor
Females: 0.031 (-11.66, 15.09)
N = 173 (90
development
Males: 0.120 (-5.24, 15.60)
Females; 83
index (PDI)
18 Months: -0.023 (-13.45, 10.72)
Males)
Confounding: Gestational age, parity, maternal age, smoking during pregnancy, alcohol consumption during pregnancy, caffeine intake
during pregnancy, maternal education level, blood sampling period, breast feeding, total dioxin levels
D-196
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Jeddy et al. Great Britain.
(2017, 3859807) Recruitment:
Medium 1991-1992;
Follow-up at
ages 15 and
18 months
Mothers and
daughters aged
15 and
38 months from
ALSPAC
N = 353
Maternal serum
19.8(15.0-
24.95)
MacArthur
Communicative
Development
Inventories
(MCDI):
communicative,
intelligibility,
language,
nonverbal
communication,
social
development,
verbal
comprehension,
and vocabulary
comprehension
scores
Regression
coefficient per
unit increase in
PFOS
Cohort Mothers and Maternal serum MacArthur Regression Nonverbal, 15 mo.: 0.02 (-0.01,
0.05)
Social, 15 mo.: 0.02 (-0.03, 0.08)
Verbal, 15 mo.: 0.03 (0.01, 0.05)
Maternal age < 30: No statistically
significant associations
Maternal age > 30: 0.04 (0.01,
0.08)
Vocabulary, 15 mo.: 0.02 (-0.39,
0.44)
Communicative, 38 mo.: 0 (-0.01,
0.01)
Intelligibility, 38 mo.: -0.01
(-0.01, 0)
Maternal age < 25: 0.02 (0.01,
0.03)
Maternal age > 25: No statistically
significant associations
Language, 38 mo.: -0.29 (-0.54,
-0.05)
Nonverbal, social, vocabulary,
communicative, language: No
statistically significant associations
stratified by maternal age at
delivery
Confounding: Parity, maternal age, maternal education, maternal smoking status, gestational age at sample collection, total maternal Crown-
Crisp Experiential Index
Liew et al.
(2015,2851010)
Denmark,
Case-control
Mother-child
pairs from
Maternal plasma ADHD, ASD
RRandOR
(stratified by
ADHD: 0.87 (0.74, 1.02)
Q4: 0.79 (0.64, 0.98)
D-197
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Medium
Recruitment:
1996-2002;
Follow-up at
average age
10.7 years
Danish National
Birth Cohort
Cases: 25.40
(18.73-32.40)
Controls: 27.40
(20.40-35.60)
quartile or by
sex) per ln-unit
increase in
PFOS or by
quartiles
ASD: 0.92 (0.69, 1.22)
No other statistically significant
associations by quartiles or by sex
215 Cases (39
Females; 176
Males)
545 Controls
(33 Females;
180 Males)
Results: Lowest quartile used as reference
Confounding: Maternal age at delivery, SES, parity, smoking and drinking during pregnancy, psychiatric illnesses, gestational week of blood
drawn, child's sex, birth year
Liew et al.
(2018, 5079744)
Medium
Denmark,
Recruitment:
1996-2002;
Follow-up at
age 5 years
Cohort
Pregnant
women and
their children
from the Danish
National Birth
Cohort
N = 1,592
Maternal plasma Wechsler
28.10(21.60- Primary and
35.80)
Preschool
Scales of
Intelligence-
Revised
(WPPSI-R) full
scale IQ,
performance IQ,
verbal IQ
Regression
coefficient for
mean difference
per ln-unit
increase in
PFOS and by
quartiles
Full Scale IQ
Q2: -0.4 (-3.2, 2.5)
Q3: 1.1 (-1.8,4.0)
Q4: -0.5 (-3.5, 2.6),
Performance IQ
Q2: 0.6 (-2.3, 3.5)
Q3: 1.6 (-1.2, 4.5)
Q4: -0.1 (-3.1,2.8),
Verbal IQ
p-trend = 0.87
p-trend = 0.93
Q2
Q3
Q4
-1.0 (-3.9,
-0.2 (-3.3,
-0.7 (-3.9,
1.9)
2.9)
2.4),
p-trend = 0.76
Results: Lowest quartile used as reference.
Confounding: Maternal age at childbirth, parity, maternal socioeconomic status, maternal IQ, maternal smoking during pregnancy, maternal
alcohol consumption during pregnancy, maternal pre-pregnancy BMI, gestational week of blood draw
Long et al.
(2019, 5080602)
Medium
Denmark,
Recruitment:
1982-1999;
Follow-Up:
1993-2009
Case-control
Pregnant
women and
their children
from the
Historic Birth
Cohort at
Amniotic fluid
Cases: 0.61
(Range: 0.61-
2.98)
Controls: 1.44
(Range: 0.61-
ASD
ORperunit 0.410(0.174,0.967),
increase in p-value = 0.042
PFOS Females: 0.027 (0, 4.755),
p-value = 0.171
Males: 0.586 (0.192, 1.782),
p-value = 0.346
Statens Serum 4.22)
Institute
D-198
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
37 Cases (7
Females; 29
Males)
50 Controls (15
Females; 35
Males)
Confounding: Child's birth year, child sex, mother's age at delivery, father age at childbirth, birth weight, gestational week at sampling,
gestational age at birth, Apgar score, parity, congenital malformation
Lyall et al.
(2018, 4239287)
Medium
United States,
2007-2009
Case-control
Children and Maternal serum
adolescents
aged 4.5-
9 years from
EMA study
985 (553 Cases;
432 Controls)
Cases:
GM = 17.5
(95%
CI = 16.8-18.3)
Controls:
GM = 17.9
(95%
CI = 17.0-18.7)
ASD measured
by Diagnostic
and Statistical
Manual of
Mental
Disorders,
Fourth Edition
(DSM-IV-TR),
intellectual
disability
OR per ln-unit
increase in
PFOS and by
quartiles
ASD: 0.77 (0.58, 1.01)
Q2: 0.85 (0.58, 1.23)
Q3: 0.74 (0.50, 1.09)
Q4: 0.64 (0.43, 0.97),
p-trend = 0.03
Intellectual Disability: 0.67 (0.45,
0.98)
1.05)
1.38)
1.09),
Q2: 0.61 (0.36
Q3: 0.80 (0.46
Q4: 0.59 (0.32
p-trend = 0.17
Results: Lowest quartile used as reference.
Confounding: Matching factors, parity, maternal age, race/ethnicity,
weight at sample collection, and maternal birthplace
Oulhote et al.
(2019, 6316905)
Medium
Faroe Islands,
Recruitment:
1997-2000;
Follow-up at
age 7 years
Cohort
Children
N = 419
Blood
Maternal: 27.69
(23.22-33.35)
5 Years: 16.8
(13.5-21.13)
Boston Naming
Test with and
without cues,
SDQ total score
Regression
coefficient per
IQR increase in
PFOS
Boston Naming Test
With Cues
Prenatal:-0.11 (-0.27, 0.01)
5-year serum: 0.00 (-0.08, 0.07)
Without Cues
Prenatal: -0.04 (-0.19, 0.06)
5-year serum: 0.00 (-0.06, 0.06)
SDQ
Prenatal: 0.15 (0.08, 0.23)
5-year serum: 0.02 (-0.03, 0.08)
Confounding: None reported
D-199
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Quaak et al. Netherlands,
(2016, 3981464) Recruitment:
Medium 2011-2013;
Follow-up
through age
18 months
Cohort
Pregnant
women and
their children
from the LINC
cohort
54 (20 Females;
34 Males)
Cord blood
1,600.0 ng/L
(Range: 570-
3,200 ng/L)
Child Behavior
Checklist 1.5-5
(CBCL 1.5-5)
measures of
ADHD,
externalizing
behavior
Regression
coefficient by
tertiles
Results: Lowest tertile used as reference.
Confounding: Alcohol use, smoking, family history of ADHD, education
ADHD
T2: -0.33 (-1.75, 1.17),
p-value = 0.66
T3: -0.87 (-2.06,0.42),
p-value = 0.19
Females
T2: 0.17 (-1.50, 1.67),
p-value = 0.85
T3: -0.73 (-2.36,0.90),
p-value = 0.43
Males
T2: -0.55 (-2.84, 1.57),
p-value = 0.64
T3: -0.99 (-3.03,0.92),
p-value = 0.35
Externalizing Behavior
T2: -1.23 (-5.68, 3.85),
p-value = 0.62
T3: -2.43 (-6.55, 1.93),
p-value = 0.31
Females
T2: -2.63 (-8.21, 4.33),
p-value = 0.44
T3: -2.98 (-8.08,2.23),
p-value = 0.31
Males
T2: 0.72 (-5.77, 6.59),
p-value = 0.81
T3: -0.94 (-6.72, 5.12),
p-value = 0.74
Shin et al. United States,
(2020, 6507470) Recruitment:
Medium 2002-2009;
Case-Control Mother-child Maternal serum ASD measured OR per increase
pairs from 5.81 (3.86-9.11) by Autism (ln-transformed
CHARGE with Diagnostic or linear scale)
By modeled prenatal exposure
ln-transformed: 1.18 (0.77, 1.80)
D-200
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
children aged 2-
5 years
N = 453 (239
Cases; 214
Controls; 88
Females; 365
Males)
Follow-up: children aged 2- Interview- in modeled, No statistically significant
2009-2017 5 years Revised (ADI- maternal, associations or interactions by sex
R) prenatal PFOS Linear: 1.03 (0.99, 1.08);
or measured, p-value <0.10
maternal, Females: 0.96 (0.85, 1.08)
postnatal PFOS Males: 1.05 (1.00, 1.10),
and by quartiles p-value < 0.05
Interaction p-value = 0.38
By measured postnatal levels
ln-transformed: 1.21 (0.80, 1.83)
Linear: 1.05 (0.97, 1.13);
p-value <0.10
No statistically significant
associations or trends by quartiles
Confounding: Child's age, child's sex, regional center, child's birth year, parity, gestational age at delivery, maternal race/ethnicity, maternal
birthplace, mother's age at delivery, maternal pre-pregnancy BMI, periconceptional maternal vitamin intake, homeownership, breastfeeding
duration
Skogheim et al.
(2019, 5918847)
Medium
Norway,
Recruitment:
1999-2008;
Follow-up:
2007-2011
Cohort
Mother-child
pairs from
MoBa
N = 943
Maternal plasma Nonverbal and
11.51 (8.77- Verbal Working
14.84)
Memory
measured by
Stanford Binet
Intelligence
Scales
Regression
coefficient per
unit increase in
PFOS and by
quintiles
Nonverbal Working Memory
Q2
Q3
Q4
Q5
0.06 (-0.14,0.26)
-0.10 (-0.30, 0.10)
-0.02 (-0.22,0.18)
-0.26 (-0.48, -0.06)
Verbal Working Memory
Q2
Q3
Q4
Q5
-0.05 (-0.27,0.17)
0.09 (-0.14,0.31)
0.10 (-0.12, 0.33)
-0.01 (-0.24,0.22)
Results: Lowest quintile used as reference.
Confounding: Maternal education, age, parity, fish intake, child sex, child age at testing, maternal ADHD symptoms
Spratlen et al.
(2020, 6364693)
Medium
United States,
Recruitment:
2001-2001;
Follow-up at
Cohort
Pregnant
women and
their children
from the
Cord blood
GM = (Range:)
BSID-II scores:
Mental and
Psychomotor
Development
Regression
coefficient of
mean difference
per log-unit
MDI
Year 1: -0.61 (-3.17, 1.95)
Year 2: 2.36 (-1.23, 5.94)
Females: 5.52 (0.64, 10.4)
D-201
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
age 1, 2, and
3 years
Columbia
University Birth
Cohort
N= 302 (150
Females; 152
Males)
Index (MDI and increase in
PDI), Full IQ, maternal PFOS
Performance IQ,
Verbal IQ
Males: -1.35 (-7.09,4.39)
Interaction p-value = 0.04
Year 3: 1.96 (-1.24,5.16)
PDI
Year 1
Year 2
Year 3
-0.07 (-4.56, 4.43)
-1.34 (-4.26, 1.57)
-0.55 (-5.34, 4.23)
Full IQ
Year 4: -0.41 (-4.25, 3.43)
Year 6: 2.81 (-1.84,7.46)
Performance IQ
Year 4:-0.05 (-4.56, 4.46)
Year 6: 2.81 (-2.29,7.91)
Verbal IQ
Year 4:-0.19 (-4.50, 4.12)
Year 6: 2.67 (-2.56,7.90)
No other statistically significant
associations or interactions by sex
Comparison: Logarithm base not specified.
Confounding: Maternal age, material hardship, parity, pre-pregnancy BMI, maternal IQ, maternal race, maternal education, family smoking
status, child age at testing, child's gestational age at birth, maternal demoralization, trimester on 9/11, child's sex, child's breastfeeding history
Strom et al. Denmark
(2014, 2922190) Recruitment:
Medium 1988-1999
Follow-up: 2010
Cohort
Pregnant
women and
their children,
from the
DaF088 cohort
N = 876
Maternal serum Depression,
Median = 21.4
(IQR = 9.0)
ADHD,
scholastic
achievement
Depression, Depression
ADHD: Hazard T2: 1.61 (0.99, 2.61)
ratio (depression T3: 1.16 (0.69, 1.95)
and ADHD) by p-value = 0.14
tertile
ADHD
Scholastic T2: 1.05 (0.43, 2.53)
achievement: T3: 0.54 (0.19, 1.53)
Regression p-value = 0.38
coefficient per
D-202
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
unit increase in
PFOS and by
tertiles
Scholastic Achievement: -0.01
(-0.03,0.01), p-value = 0.57
T3: -0.11 (-0.50,0.28),
p-trend = 0.59
Results: Lowest tertile used as reference.
Confounding: Maternal age, pre-pregnancy BMI, parity, maternal smoking during pregnancy, maternal education, maternal cholesterol,
maternal triglycerides, offspring sex
Vuong et al. United States,
(2016, 3352166) Recruitment:
Medium 2003-2006;
Follow-up at
ages 5 and
8 years
Children ages 5
and 8 years
from the HOME
study
N = 218
Cohort Children ages 5 Serum BRIEF All outcomes: Behavioral Regulation: 3.14 (0.68,
13.2(8.8-17.8) measures of OR for 5.61)
Metacognition: 3.10 (0.62, 5.58)
Global Executive Function: 3.38
(0.86, 5.90)
No statistically significant
interactions by age; no statistically
significant trends by quartiles
Inhibit: 2.59 (1.23, 5.41)
Shift: 1.50 (0.72,3.11)
Emotional control: 1.97 (0.84, 4.64)
Working memory: 1.87 (1.01, 3.48)
Plan/organize: 3.54 (1.65, 7.60)
Initiate: 1.89 (0.80, 4.45)
Organization: 1.84 (0.82, 4.13)
Monitor: 3.39 (1.42, 8.08)
Confounding: Maternal age, race, education, income, maternal serum cotinine, maternal depression, HOME score, maternal IQ, marital
status, child sex
BRIEF
All outcomes:
measures of
OR for
behavioral
score > 60 per
regulation,
unit increase in
metacognition,
PFOS
global executive
composite
Index and
indices, inhibit,
compositive
shift, emotional
scores only:
control, working
Regression
memory,
coefficient per
plan/organize,
ln-unit increase
initiate,
in PFOS and by
organization of
quartiles
materials,
monitor
Vuong et al. United States,
(2018, 5079675) Recruitment:
Medium 2003-2006;
Follow-up at
age 3 and
8 years
Cohort
Children from
the HOME
study
N = 204
Serum
3 years: 6.2
(4.5-10.0)
8 years: 3.6
(2.7-4.9)
BRIEF
measures of
behavioral
regulation,
metacognition,
global executive
composite
indices
OR per ln-unit Behavioral Regulation
increase in 3 years: 0.66 (0.29, 1.51)
PFOS 8 years: 0.40 (0.14, 1.14)
Metacognition
3 years: 0.83 (0.42, 1.63)
8 years: 1.53 (0.67, 3.52)
D-203
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Global Executive Function
3 years: 0.95 (0.45, 2.01)
8 years: 1.04 (0.41,2.68)
Confounding: Maternal age, race/ethnicity, household income, maternal smoking status, maternal alcohol consumption, maternal depression,
HOME Score, marital status, maternal marijuana use, maternal IQ, maternal serum PCBs, maternal blood lead levels, child sex
Vuong et al. United States,
(2018, 5079693) Recruitment:
Medium 2003-2006;
Follow-up at
ages 3 and
8 years
Cohort
Mother-child
dyads from the
HOME study
204
Serum
Prenatal: 12.9
(8.8-17.6)
3 years: 6.2
(4.5-9.9)
8 years: 3.6
(2.7-4.8)
Conners'
Continuous
Performance
Test II
commissions t-
score, omissions
t-score, hit
reaction time,
tau (ms)
Virtual Morris
Water Maze
(VMWM)
scores for
visual-spatial
learning
distance (pool
units), learning
time (s),
memory
retention
distance (%),
and memory
retention time
(s)
Regression Conners'
coefficient per Commissions
ln-unit increase Prenatal: -0.1 (-2.0, 1.8)
inPFOS 3 Years: 1.0 (-1.5, 3.5)
8 Years: 1.3 (-1.0,3.6)
Omissions
Prenatal: -0.8 (-5.2, 3.5)
3 Years: -0.1 (-4.4,4.2)
8 Years:-0.8 (-5.3, 3.8)
Females: 4.3 (-1.2, 9.9)
Males: -7.3 (-13.0, -1.7)
Hit reaction time
Prenatal: -1.5 (-4.2, 1.2)
3 years: -0.4 (-3.2, 2.5)
8 years: -2.5 (-6.0, 1.1)
Tau
Prenatal: 6.0 (-23.2, 35.2)
3 years: 13.4 (-9.8, 36.5)
8 years: 5.8 (-22.1, 33.7)
Visual-spatial scores (VMWM)
Learning distance
Prenatal: 0.2 (-1.6, 1.7)
3 years: -0.7 (-2.2, 0.7)
8 years:-0.2 (-1.7, 1.3)
Learning time
Prenatal: -0.1 (-2.8, 2.6)
3 years: -1.1 (-3.5, 1.2)
8 years: -2.1 (-4.9,0.6)
Memory retention distance
Prenatal: 2.8 (-1.3, 6.8)
D-204
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
3 years: 0.3 (-4.7, 5.4)
8 years: 2.1 (-2.9, 7.0)
Memory retention time
Prenatal: 0.4 (-1.1, 1.9)
3 years:-0.4 (-2.1, 1.3)
8 years: 0.5 (-1.3,2.3)
Confounding: Maternal age, race/ethnicity, household income, maternal smoking status, maternal alcohol consumption, maternal depression,
HOME Score, marital status, maternal marijuana use, maternal IQ, maternal serum IPCBs. maternal blood lead levels, child sex
Vuong et al. United States,
(2019, 5080218) Recruitment:
Medium 2003-2006;
Follow-up at
ages 3 and
8 years
Pregnant
women and
their children
from the HOME
study
N = 221
Serum
Maternal:
GM = 12.4
8 Years:
GM = 3.9
Wechsler Regression
Intelligence coefficient per
Scale for ln-unit increase
Children-Fourth inPFOS
Edition (WISC-
IV): full scale
IQ, perceptual
reasoning,
processing
speed, verbal
comprehension,
working
memory
Cohort Pregnant Serum Wechsler Regression Full Scale IQ
Prenatal: 2.2 (-0.9, 5.2)
3 Years: 0.8 (-2.4, 4.0)
8 Years: 1.6 (-2.7, 5.8)
Perceptual Reasoning
Prenatal: 1.4 (-1.8, 4.7)
3 Years: 1.0 (-2.6, 4.5)
8 Years: 2.8 (-2.1,7.7)
Processing Speed
Prenatal: 1.3 (-2.0, 4.7)
3 Years: 1.6 (-1.9, 5.1)
8 Years: 3.7 (-1.2, 8.5)
Verbal Comprehension
Prenatal: 1.4 (-1.7, 4.5)
3 Years: 0.1 (-3.3, 3.5)
8 Years: -1.7 (-5.2, 1.8)
Working Memory
Prenatal: 2.6 (-0.8, 5.9)
3 Years:-0.1 (-3.4,3.2)
8 Years: 2.9 (-0.8, 6.5)
Confounding: Maternal age, race/ethnicity, household income, maternal marijuana use, maternal blood lead, maternal serum IPCBs and
cotinine, maternal depression, vitamin use, maternal IQ, marital status, HOME score, child sex, breastfed
Vuong et al. United States,
(2020, 6833684) Recruitment:
Cohort
Mother-child
pairs with
Maternal serum
Wide Range
Achievement
Regression
coefficient per
7.0 (-2.9, 16.9)
D-205
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Medium
2003-2006; children aged Mean = 13.9 Test4(WRAT- loglO-unit
Follow-up at 8 years from the (SD = 7.9) 4) reading increase in
age 8 years HOME study composite score PFOS
N = 161
Confounding: Maternal age, race/ethnicity, education, household income, marital status, maternal depression, maternal serum cotinine,
maternal blood lead levels, maternal fish consumption, maternal IQ, child sex, HOME score
Wang et al.
Taiwan, Cohort
Pregnant
Serum
Full Scale IQ, Regression
Full Scale IQ
(2015, 3860120) Recruitment:
women and
5 Years: 13.25
Performance IQ, coefficient per
5 Years: -1.9 (-4.3, 0.5)
Medium
2000-2001;
their children
(9.75-17.50)
Verbal IQ log2-unit
8 Years: -1.9 (-4.3, 0.4)
Follow-up at
aged 5 and
8 Years: 12.28
increase in
ages 5 years
8 years from
(9.50-16.30)
PFOS
Performance IQ
TMICS
5 Years: -2.2 (-4.7, 0.3)
N = 120
8 Years: -1.6 (-4, 0.7)
Verbal IQ
5 Years: -1.7 (-4, 0.7)
8 Years: -1.3 (-3.6, 1.1)
Confounding: Maternal education, family annual income, children's
age, sex, HOME score at IQ assessment
Zhang et al. United States,
(2018, 4238294) Recruitment:
Medium 2003-2006;
Follow-up at
ages 3, 5, and
7 years
Cohort
Pregnant
Serum
Basic reading,
women and
Maternal: 13.0
brief reading,
their children
(9.1-17.8)
letter word
aged 3, 5, and
3 years: 6.6
identification,
7 years from the
(4.6-10.2)
passage
HOME study
8 years: 3.6
comprehension
N = 167
(2.7-4.9)
measured by
Woodcock
Johnson Test of
Achievement-Ill
(WJ-III)
Reading
composite, word
reading,
sentence
Comprehension
measured by
Regression
coefficient per
ln-unit increase
PFOS
Basic Reading
Maternal Serum: 3.2 (-2.0, 8.3)
Year 3 Serum: 1.1 (-4.8, 7.0)
Brief Reading
Maternal Serum: 2.9 (-2.2, 8.1)
Year 3 Serum: 3.2 (-2.6, 9.1)
Letter Word Identification
Maternal Serum: 2.0 (-2.7, 6.8)
Year 3 Serum: 2.1 (-3.4,7.5)
Passage Comprehension
Maternal Serum: 1.7 (-1.9, 5.3)
Year 3 Serum: 3.5 (-0.5, 7.6)
Word Attack
Maternal Serum: 4.1 (-1.2, 9.5)
D-206
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Wide Range
Achievement
Test 4 (WRAT-
4)
Year 3 Serum: 2.8 (-2.8, 8.4)
Reading Composite
Maternal Serum: 3.1 (-1.3, 7.5)
Year 3 Serum: 1.6 (-3.1, 6.4)
Year 8 Serum: 2.6 (-1.7, 6.9)
Word Reading
Maternal Serum: 3.1 (-1.0, 7.3)
Year 3 Serum: -0.3 (-4.8, 4.3)
Year 8 Serum: 4.4 (0.3, 8.4)
Sentence Comprehension
Maternal Serum: 3.2 (-1.8, 8.2)
Year 3 Serum: 2.5 (-3.1, 8.1)
Year 8 Serum: 1.6 (-3.3, 6.5)
Confounding: Maternal age, race, education, household-income, parity, smoking (serum cotinine concentration), maternal IQ, breastfeeding
duration, HOME score
General Population
Ding and Park United States,
(2020,6711603) 2003-2016
Medium
Cross-sectional
Adults aged 20- Serum
69 years from 6.2 (3.5-10.5)
NHANES
N = 2,731
High and low
frequency
hearing
impairment
(HFHI and
LFHI)
OR per log2-
unit increase in
PFOS and
for > 90th
percentile
vs. < 90th
percentile
Confounding: Age, age square, sex, race/ethnicity, education level, poverty-income ratio, smoking status
(occupational, recreational, firearm noise), NHANES cycles
HFHI
OR (per doubling): 0.96 (0.85,
1.10)
OR (90th percentiles): 1.31 (0.75,
2.27)
LFHI
OR (per doubling): 0.87 (0.73,
1.03)
OR (90th percentiles): 0.72 (0.29,
1.75)
, BMI, noise exposures
Gallo et al. United States,
(2013,2272847) 2005-2006
Medium
Cross-sectional Adults aged 50+ Serum Memory OR per
years from the Range = 0.25- impairment doubling of
759.2 (self-reported)
0.93 (0.90, 0.96)
Q2: 0.96 (0.87, 1.07)
Q3: 0.86 (0.78,0.96)
D-207
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Q4: 0.87 (0.78, 0.96)
Q5: 0.85 (0.76,0.94)
p-trend <0.001
C8 Health PFOS and by
Project quintiles
N = 21,024
Comparison: Logarithm base not specified.
Results: Lowest quartile used as reference.
Confounding: Age, ethnicity, gender and school level, household income, physical activity, alcohol consumption, cigarette smoking
Lenters et al. Norway,
(2019, 5080366) Recruitment:
Medium 2003-2009;
Follow-up:
2008-2016
Cohort
Children and
adults from
HUMIS
N = 1,199
Breast milk
117.732 ng/L
(80.000-
160.000 ng/L)
ADHD
OR per IQR
increase in ln-
unit PFOS
1.75 (1.11,2.76), p-value = 0.017
Confounding: Maternal age, childbirth year, maternal education, parity, smoking during pregnancy, small-for-gestational age, preterm birth,
maternal pre-pregnancy BMI, single mother around perinatal period, maternal fish intake
Li (2020,
6833686)
Medium
United States,
1999-2016
Cross-sectional
Adults aged
20+ years from
NHANES
N = 2,525
Serum
8.00 (Range:
0.14-392)
Hearing
threshold > 25
dB by frequency
OR by quartiles
2,000 Hz
Q2: 0.70 (0.46, 1.06)
Q3: 1.12 (0.76, 1.65)
Q4: 1.60(1.09,2.37), p-
trend < 0.0001
3,000 Hz
Q2: 0.76 (0.53, 1.08)
Q3: 1.00 (0.71, 1.41)
Q4: 1.20 (0.85, 1.71),
p-trend = 0.02
4,000 Hz
Q2: 0.69 (0.50, 0.97)
Q3: 0.89 (0.65, 1.24)
Q4: 1.02 (0.73, 1.44),
p-trend = 0.14
Results: Lowest quartile used as reference.
Confounding: Age, sex, BMI, education, ethnicity group, family income, sample weights
Shrestha et al. United States, Cross-sectional Residents aged Serum Affective state: Regression Depression:
(2017,3981382) 2000-2002 55-74 years 33.7 (23.3-50.8) Beck coefficient per 0.25 (-0.77, 1.26), p-value = 0.63
Medium who lived Depression
D-208
-------
DRAFT FOR PUBLIC COMMENT
Reference, Location,
Confidence Years
Population, Exposure
Design Ages, Matrix, Levels
N (ng/mL)a
adjacent to
Hudson River
N = 126
D-209
MARCH 2023
Outcome Comparison
Inventory (BDI) IQR increase in
total score, ln-unit PFOS
State-Trait
Anxiety
Inventory state
and trait t-scores
Attention: Trail
making test Part
A (In-
transformed
time to
complete)
Executive
function: Stroop
color word test
t-score, Trail
making test part
B (In-
transformed
time to
complete),
Wisconsin card
sorting test
preservative In-
transformed
error and
response
Resultsb
CVLT-Total score:
-0.14 (-0.59, 0.31)
Wisconsin card-sorting test
Perseverative Error:
-0.14 (-0.30, 0.02), p-value = 0.09
Perseverative Response:
-0.16 (-0.34, 0.01), p-value = 0.07
Wechsler Memory Scale
Logical Memory
Immediate Recall: -0.7 (-1.92,
0.52), p-value = 0.26
Delayed Recall: -0.14 (-1.29,
1.01), p-value = 0.81
Visual Reproduction
Immediate Recall: 0.56 (-0.16,
1.29), p-value = 0.13
Delayed Recall: 0.79 (0.03, 1.55),
p-value = 0.04
No statistically significant
associations: State-Trait Anxiety
Inventory, Stroop color word test,
trail-making tests, motor function
outcomes, visuospatial outcomes
Memory and
learning:
California
Verbal Learning
Test total and
subscores,
-------
DRAFT FOR PUBLIC COMMENT
T Population,
Reference, Location, . .
^ «, Design Ages9
Confidence Years
MARCH 2023
Exposure
Matrix, Levels Outcome Comparison Resultsb
(ng/mL)a
Wechsler
Memory Scale
logical memory
and visual
reproduction
immediate and
delayed recall
scores
Motor function
(dominant and
non-dominant
hands): finger
tapping test
average scores,
grooved
pegboard test
ln-transformed
times to
completion,
static motor
steadiness test
ln-transformed
total numbers of
contacts and
times touching
Dominant hand
reaction time
Visuospatial
function:
Wechsler Adult
Intelligence
Scale-Revised
total scores for
D-210
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location, Population, Exposure
Design Ages, Matrix, Levels Outcome Comparison
^cars n (ng/mL)a
Resultsb
block design
and digit
symbol coding
Confounding: Age, sex, education, serum total PCB
Pregnant Women
Vuong et al. United States Cohort
Pregnant Maternal serum Beck
Relative risk
Medium Score Trajectory: 0.9 (0.6,
(2020, 6356876) Recruitment:
women from the 13.3(9.0-17.9) Depression
and OR per ln-
1.5)
Medium 2003-2006
HOME study Inventory-II
unit increase in
High Score Trajectory: 0.6 (0.3,
Follow-up: ~20
N = 355 (BDI-II)
PFOS
1.2)
weeks gestation
and postpartum
OR for score >13 from pregnancy
(4 weeks, 1, 2,
to 8 years postpartum: 1.0 (0.7, 1.5)
3, 4, 5, and
8 years)
Confounding: Age, race/ethnicity, household income, maternal marijuana use, serum cotinine and PCBs,
IQ, marital status, parity
Notes: ADHD = attention deficit hyperactivity disorder; ALSPAC = Avon Longitudinal Study of Parents and Children; ASD = autism spectrum disorder; ASQ-3 = Ages and
Stages Questionnaire-3; BMI = body mass index; BRIEF = Behavior Rating Inventory of Executive Function; CDI = Comprehensive Developmental Inventory;
CHARGE = Childhood Autism Risk from Genetics and Environment; DaF088 = Danish Fetal Origins 1988; EMA = Early Markers for Autism; GM = geometric mean;
HFHI = high frequency hearing impairment; HOME = Health Outcomes and Measures of the Environment; HUMIS = Human Milk Study; ID = intellectual disability;
IQR = interquartile range; LINC = Linking Maternal Nutrition to Child Health; LFHI = low frequency hearing impairment; MoBa = Mother, Father, and Child Cohort Study;
NHANES = National Health and Nutrition Examination Survey; OR = odds ratio; PFOS = perfluorooctane sulfonic acid; RR = risk ratio; SDQ = Strengths and Difficulties
Questionnaire; TMICS = Taiwan Maternal and Infant Cohort Study.
a Exposure levels are reported as median unless otherwise noted.
b Results reported as effect estimate (95% confidence interval), unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
D-211
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.9 Renal
Table D-18. Associations Between PFOS Exposure and Renal Effects in the General Population
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels Outcome
(ng/mL)a
Comparison
Select Resultsb
General Population
Blake et al.
(2018, 5080657)
Low
United States,
1991-2008
Cohort
Adults and
children from
FCC
N = 192(115
females, 77
males)
Serum eGFR
28.4 (21.6-35.7)
Percent change
per IQR
increase in
PFOS
All:
Repeated measures model: -0.68
(-1.9, 0.54); p-value = 0.27
Latent model: -1.72 (-3.29, -0.15);
p-value = 0.03
Females: -1.32 (-3.37, 0.73),
p-value = 0.64
Males: 0.71 (-2.75,4.16),
p-value = 0.69
p-value for interaction by
sex = 0.46
Confounding: Age, year of measurement, sex, education, income, marital status, and BMP
Lin etal. (2013, Taiwan, Cross-sectional Adolescents and Serum Uric acid
2850967) 2006-2008 young adults 8.65(5.41- (mg/dL)
Low fromYOTA 13.52)
study, 12-
30 years,
N = 644
Results: Effect estimates are provided with standard error in parentheses.
Confounding: Age, gender, smoking status, alcohol drinking, BMI
Mean
concentration by
PFOS
percentiles
< 25th percentile: 6.09 (0.13)
25th-50th: 6.13 (0.13)
50th-75th: 6.04 (0.13)
> 75th: 6.12(0.13)
p-value for trend = 0.891
Conway et al.
(2018, 5080465)
Low
United States
2005-2006
Cohort
Adults, C8
Health Project,
Diabetic = 5,21
0, non-
diabetic = 48,44
0
Confounding: Age, sex, BMI, HDLc, LDLc, white blood cell count, CRP, hemoglobin, and iron
Serum
Diabetic: 21.2
(13.7-31.4)
Non-diabetic:
20.2 (13.6-29.1)
CDK (eGFR of OR per ln-unit
<60 mL/min/1. increase in
73 m2) PFOS
Diabetics: 0.81 (0.73, 0.9)
Non-diabetic: 1.09 (1.03, 1.16)
D-212
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Liu et al. (2018,
4238514)
Low
United States
2013-2014
Cross-sectional
Serum
GM = 5.28
(SE= 1.02)
Adults from
NHANES,
18+ years,
N = 1871
Confounding: Age, gender, ethnicity, smoking status, alcohol intake,
hypertensive, anti-hyperglycemic, and anti-hyperlipidemic agents)
Total protein Regression 0.05 (SE = 0.02); p-value < 0.01
(g/dL) coefficient per
ln-unit increase
inPFOS
household income, waist circumference, and medications (anti-
Arrebola et al. Spain,
(2019,5080503) 2009-2010
Low
Adults, Serum
BIO AMBIENT. 7.23 (5.14-
ES study 10.11)
N = 342
Cross-sectional Adults, Serum Uric acid OR Uric acid
(mg/dL), (hyperuricemia), Wet-basis and lipid basis models:
hyperuricemia or regression 0.06 (-0.03, 0.16); p-value = 0.192
coefficient per Wet-basis model with adjustment
log-unit increase for serum lipids:
inPFOS 0.06 (-0.03, 0.157);
p-value = 0.207
Hyperuricemia
Wet-basis and lipid-basis models:
1.70 (0.86, 3.49); p-value = 0.138
Wet-basis model with adjustment
for serum lipids:
1.67 (0.84, 3.41); p-value = 0.151
Outcome: Hyperuricemia defined as at least one of a) serum uric acid levels > 7.0 mg/dL in males or > 6.0 mg/dL in females, at recruitment
or in previous screenings, b) had been prescribed any pharmacological treatment for lowering uric acid levels, and/or c) had been diagnosed
with gout by a clinician.
Comparison: Logarithm base not specified.
Confounding: Sex, age, body mass index, weight loss during the last 6 months, region of recruitment, smoking habit, alcohol consumption,
education, place of residence
Chen et al. Croatia,
(2019,5387400) 2007-2008
Low
Cross-sectional
Adults, 44-
56 years
N = 122
Plasma
GM = 8.91
(range = 2.36-
33.67)
Uric acid
(|imol/L),
creatinine
(|imol/L)
Regression
coefficient per
ln-unit increase
inPFOS
Uric acid: -4.87 (-25.63, 15.89)
Creatinine: -3.36 (-7.96, 1.24)
Confounding: Age, sex, education, socioeconomic status, smoking, dietary pattern, and physical activity
Jain and
Ducatman
(2019, 5381566)
Low
United States, Cross-sectional Adults from Serum Levels of Regression Albumin in urine
2005-2014 NHANES, Levels not albumin in urine coefficient per Per loglO-unit increase: -0.08
>20 years reported (logl0-|ig/mL). loglO-unit p-value <0.01
N = 8,220 creatinine in increase in
D-213
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
urine (loglO- PFOS, or Negative associations across eGFR
mg/dL), percent change stages
albumin-to- per 10% Percent change per 10% increase:
creatinine ratio increase in -0.75, p-value < 0.05
in urine (log 10- PFOS p-value for gender and
mg/g), albumin race/ethnicity interaction = 0.10
in serum (loglO-
mg/dL), Creatinine in urine
creatinine in Per loglO-unit increase: 0.04
serum (loglO- p-value = 0.01
mg/dL) Positive associations across eGFR
stages
Percent change per 10% increase:
0.38
p-value < 0.05
p-value for gender and
race/ethnicity interaction = 0.02
Albumin-to-creatinine ratio in urine
Per loglO-unit increase: -0.12
p-value < 0.01
Negative associations across eGFR
stages
Percent change per 10%
increase: -1.13
p-value < 0.05
p-value for gender and
race/ethnicity interaction = 0.73
Albumin in serum
Per loglO-unit increase: 0.01
p-value < 0.01
Positive associations across eGFR
stages
Percent change per 10% increase:
0.11
D-214
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
p-value < 0.05
p-value for gender and
race/ethnicity interaction = 0.68
Creatinine in serum
Per loglO-unit increase: 0.01
p-value = 0.01
Positive associations in GF-1, GF-
2, GF-3A
Negative association in GF-3B/4
Percent change per 10% increase:
0.11
p-value < 0.05
p-value for gender and
race/ethnicity interaction < 0.01
GF Stages: GF-1: GFR> 90 mL/min/1.73 m2; GF-2: GFR between 60 and 90 mL/min/1.73 m2; GF-3A: GFR between 45 and
60 mL/min/1.73 m2; GF-3B/4: GFR between 15 and 45 mL/min/1.73 m2.
Confounding: Gender, race/ethnicity, age, loglO(BMI), Iogl0(serum cotinine), poverty income ration, NHANES survey period
Jain and
Ducatman
(2019, 5080378)
Low
United States,
2007-2014
Cross-sectional
Adults from
NHANES,
> 20 years,
Males = 3,330,
females = 3,506
Uric acid
(mg/dL) by
glomerular
filtration (GF)
stage
Regression
coefficient per
loglO-unit
increase in
PFOS
Males
GF-1: 0.01, p-value = 0.01
GF-2: 0.02, p-value = 0.05
GF-3A: -0.01, p-value = 0.66
GF-3B: -0.04, p-value < 0.01
Females
GF-1: 0.02, p-value = 0.04
GF-2: 0.01, p-value = 0.52
GF-3A: 0.04, p-value < 0.01
GF-3B: 0.01, p-value = 0.64
GF Stages: GF-1: eGFR > 90 mL/minper 1.73 m2; GF-2: 60 < eGFR < 90 mL/minper 1.73 m2; GF-3A: 45 < eGFR < 60 mL/minper
1.73 m2; GF-3B/4: 15 < eGFR<45 mL/minper 1.73 m2.
Confounding: Gender, race/ethnicity, age, loglO(BMI), Iogl0(serum cotinine), poverty income ration, NHANES survey period
Serum
Males:
GM = 10.51
(9.88-11.18)
Females:
GM = 6.58
(6.22-6.96)
Wang et al. China, 2015-
(2019, 5080583) 2016
Low
Cross-sectional
Adults, Isomers
of C8 Health
Project
Serum
24.22 (14.62-
37.19)
CKD, eGFR
OR (CKD) or
regression
coefficient per
ln-unit increase
CKD (OR)
Per ln-unit increase: 1.71 (0.92,
1.49), pvalue = 0.205
Q2: 1.19 (0.67,2.09)
D-215
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Exposure
Design Ages, Matrix, Levels
N (ng/mL)a
Outcome Comparison
Select Resultsb
N = 1612
(males = 1204,
females = 408)
inPFOS, or by
quartiles
Q3: 1.42 (0.82,2.47)
Q4: 1.34 (0.77,2.33)
p-value for trend = 0.617
eGFR
Per ln-unit increase:
All: -0.91 (-1.83,0),
p-value = 0.05
Males: -0.73 (-1.82,0.37)
p-value = 0.193
Females:-0.62 (-0.24, 1.15)
p-value = 0.491
p-value for interaction by
sex = 0.419
Q2
Q3
Q4
-1.25 (-3.14,0.63)
-1.59 (-3.53, 0.35)
-1.77 (-3.74,0.19)
p-value for trend = 0.086
Outcome: CKD defined as eGFR < 60 mL/min per 1.73 m2.
Results: Lowest quartile used as reference group.
Confounding: Age, sex, BMI, education, annual income, regular exercise, cigarette smoking, drinking alcohol, family history of CKD, total
cholesterol
Zeng et al. China,
(2019, 5918630) 2015-2016
Low
Cross-sectional
Adults, Isomers
of C8 Health
Project
N = 1612
(males = 1204,
females = 408)
Serum
24.22 (14.62-
37.19)
Hyperuricemia,
uric acid
(mg/dL)
OR
(hyperuricemia)
or regression
coefficient (uric
acid) per loglO-
unit increase in
PFOS
Hyperuricemia
All: 1.17 (0.99, 1.39)
Males: 1.11 (0.92, 1.34)
Females: 1.27 (0.8, 2)
p-value for interaction by
sex = 0.118
Uric acid
All: 0.1(0.02, 0.18),
p-value = 0.017
Males: 0.07 (-0.03,0.18)
Females: 0.11 (-0.01,0.18)
p-value for interaction by
sex = 0.209
D-216
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Outcome: Hyperuricemia defined as serum uric acid levels > 7.0 mg/dL in males or > 6.0 mg/dL in females.
Confounding: Age, sex, BMI, income, drinking, smoking, career, exercise, offal consumption, fish and seafood consumption, serum
creatinine
Scinicariello et United States,
al. (2020, 2009-2014
6833670)
Low
Cross-sectional Adults from Serum Uric acid
NHANES GM = 6.98 (mg/dL),
N = 4915 (no (SE = 0.23) hyperuricemia,
CKD = 4103; gout
CKD = 874)
OR
(hyperuricemia,
gout), or
regression
coefficient (uric
acid) by
quartiles
Uric acid
Overall population
Q2: 0.13 (0.01,0.24)
Q3: 0.21 (0.05,0.37)
Q4: 0.29 (0.14,0.44)
p-value for trend = 0.003
Participants with CKD
Q2: 0.6 (0.15, 1.05)
Q3: 0.31 (-0.02,0.7)
Q4: 0.38 (0.06, 0.83)
p-value for trend = 0.08
Participants without CKD
Q2: 0.03 (-0.1,0.15)
Q3: 0.13 (-0.02, 0.28)
Q4: 0.2 (0.06, 0.34)
p-value for trend = 0.02
Hyperuricemia
Overall population
Q2: 1.1 (0.84, 1.45)
Q3: 1.27(0.92, 1.76)
Q4: 1.45 (1.03,2.03)
p-value for trend = 0.15
Participants with CKD
Q2: 1.93 (0.91,4.06)
Q3: 0.85 (0.4, 1.77)
Q4: 1.15 (0.53,2.5)
p-value for trend = 0.12
Participants without CKD
Q2: 0.94 (0.68, 1.3)
Q3: 1.26 (0.89, 1.79)
Q4: 1.35 (0.92, 1.99)
p-value for trend = 0.19
D-217
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Gout
Overall population
Q2: 1.17 (0.54,2.53)
Q3: 1.23 (0.54,2.53)
Q4: 1.46 (0.67, 3.16)
p-value for trend = 0.79
Participants with CKD
Q2: 0.88 (0.26, 2.92)
Q3: 1.08 (0.38,3.07)
Q4: 1.08 (0.39, 2.94)
p-value for trend = 0.97
Participants without CKD
Q2: 1.73 (0.6,4.94)
Q3: 1.56 (0.51,4.78)
Q4: 1.93 (0.71, 5.22)
p-value for trend = 0.58
Outcomes: CKD defined as eGFR < 60 mL/min per 1.73 m2 and/or albuminuria. Hyperuricemia defined as serum uric acid levels
> 7.0 mg/dL in males or > 6.0 mg/dL in females. Gout was self-reported diagnosis from a health professional.
Results: Lowest quartile used as reference group.
Confounding: Race/ethnicity, age, sex, education, alcohol, smoking, serum cotinine, BMI, diabetes, hypertension, CKD
Children and Adolescents
Geiger et al. United States,
(2013, 2919148) 1999-2000;
Low 2003-2008
Cross-sectional Children and Serum Hyperuricemia, OR Hyperuricemia
adolescents Mean =18.4 uric acid (hyperuricemia) Per In increase: 1.37 (1.06, 1.76)
fromNHANES, (SE = 0.5) (mg/dL) or regression Q2: 1.17 (0.8, 1.72)
12-18 years, coefficient (uric Q3: 1.18 (0.74, 1.87)
N = 1,772 acid) per ln-unit Q4: 1.65 (1.1,2.49)
increase in
PFOS or by
quartiles
p-value for trend = 0.022
Uric acid
Per 1-ln increase: 0.09 (0.02, 0.17)
Q2: 0.03 (-0.1,0.16)
Q3: 0.09 (-0.04,0.21)
Q4: 0.12 (-0.01, 0.26)
p-value for trend = 0.058
D-218
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Outcome: Hyperuricemia defined as serum uric acid levels > 6 mg/dL.
Results: Lowest quartile as reference group.
Confounding: Age, sex, race/ethnicity, BMI, annual household income, moderate activity, total cholesterol, serum cotinine
Kataria et al. United States,
(2015, 3859835) 2003-2010
Low
Cross-sectional
Children and
adolescents
from NHANES,
12-19 years,
NHANES
N = 1,962
Serum
3.5 (2.5—4.7)
eGFR
(min/mL/1.73
m2), uric acid
(mg/dL),
creatinine
(mg/dL)
Regression
coefficient by
quartiles
eGFR
Q2: -5.24 (-9.75, -0.73),
p-value < 0.05
Q3: -7.21 (-12.21,-2.21),
p-value < 0.01
Q4: -9.47 (-14.68, -4.25),
p-value < 0.001
Uric acid
Q2: 0.095 (-0.081, 0.27)
Q3: 0.046 (-0.1,0.19)
Q4: 0.19 (0.032, 0.34),
p-value < 0.05
Creatinine
Q2: 0.021 (-0.007, 0.049)
Q3: 0.038 (0.008,0.068),
p-value < 0.05
Q4: 0.04 (0.01,0.071),
p-value < 0.01
Results: Lowest quartile as reference group.
Confounding: Sex, poverty-income ratio, caregiver education, serum cotinine, prehypertension, insulin resistance, BMI,
hypercholesterolemia, race/ethnicity categories
Qinetal. (2016,
3981721)
Low
Taiwan,
2009-2010
Cross-sectional
Children from
GBCA Study,
12-15 years,
N = 225 (123
girls, 102 boys)
Serum
All: 28.9(14.1-
43.0)
Boys: 29.9
(13.0-43.8)
Girls: 28.8
(14.8-42.6)
Uric acid
(mg/dL),
hyperuricemia
Regression
coefficient per
ln-unit increase
in PFOS (uric
acid); OR scaled
with increasing
quartiles
(hyperuricemia)
Uric acid
All: 0.05 (-0.03,0.13)
Boys: 0.05 (-0.04,0.15)
Girls: 0.01 (-0.14,0.16)
Hyperuricemia (OR)
All: 1.35 (0.95, 1.93)
Boys: 1.4 (0.88,2.21)
Girls: 1.51 (0.79,2.89)
Outcome: Hyperuricemia defined as uric acid level > 6 mg/dL.
D-219
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Results: Lowest quartile used as the reference group.
Confounding: Age, gender, BMI, parental education level, exercise, environmental tobacco smoke exposure, and serum creatinine
Khalil et al. United States Cross-sectional
Obese children,
Serum
Creatinine
Regression
0 (-0.02, 0.03)
(2018, 4238547) 2016
8-12 years
2.79
(mg/dL)
coefficient per
Low
N = 40
(IQR = 2.10)
unit increase in
PFOS
Confounding: Age, sex, race
Pregnant Women
Nielsen et al. Sweden, Cohort
Pregnant
Serum
eGFR:
Spearman's
Cross-sectional correlations
(2020,6833687) 2009-2014
women,
Early
LMrev, CKD-
correlation
consistently weak and
Low
PONCH study
pregnancy: 5.6
EPI(creatinine),
coefficient
nonsignificant
N = 73
(5th-95th
CAPA, CKD-
Early to late pregnancy changes:
percentile = 2.6
EPI(cystatin C),
No significant associations
-11.5)
mean of LMrev
Late pregnancy:
and CAPA,
eGFR:
4.8 (5th-95th
mean of CKD-
LMrev: 0.02, p-value = 0.85
percentile =1.9
EPI(creatinine)
CKD-EPI(creatinine): 0.02,
-8.4)
and CKD-
p-value = 0.87
EPI(cystatin C)
CAPA: -0.04, p-value = 0.73
CKD-EPI(cystatin C): -0.05,
Glomerular pore p-value = 0.66
size mean of LMrev and CAPA: -0.04,
p-value = 0.76
mean of CKD-EPI(creatinine) and
CKD-EPI(cystatin C): -0.06,
p-value = 0.63
Glomerular pore size:
CAPA/LMrev: -0.05,
p-value = 0.68
CKD-EPI(cy statin Q/CKD-
EPI(creatinine): -0.06,
p-value = 0.63
Outcome: Glomerular pore size is estimated as the ratio between eGFR (cystatin C) and eGFR(creatinine) and was calculated by the two
ratios provided.
D-220
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Confounding: Number of days between sampling, pregnancy-induced change in BMI
Occupational Populations
Rotander et al.
Australia, 2013 Cross-sectional Firefighters with Serum Uric acid
Regression 0.045 (SE = 0.047), p-value = 0.342
(2015, 3859842)
past exposure to 66 (range = 3.1- (|imol/L)
coefficient per
Low
AFFF, 17- 391)
loglO-unit
66 years old
increase in
N = 137 (97%
PFOS
male)
Confounding: Age, sex, BMI, smoking status, total serum protein, PFOA, PFHxS
Notes: AFFF = aqueous film-forming foam; BMI = body mass index; CAPA = Caucasian Asian Pediatric Adult; CKD = chronic kidney disease; CKD-EPI = Chronic Kidney
Disease Epidemiology Collaboration study; eGFR = estimated glomerular filtration rate (mL/min per 1.73 m2); FCC = Fernald Community Cohort; GBCA = Genetic Biomarkers
Study for Childhood Asthma; GF = glomerular filtration; GM = geometric mean; LMrev = Lund Malmo Revised; NHANES = National Health and Nutrition Examination
Survey; OR = odds ratio; PONCH = Pregnancy Obesity Nutrition and Child Health study; SD = standard deviation; SE = standard error; YOTA = Young Taiwanese Cohort
Study.
a Exposure levels reported as median (25th-75th percentile) unless otherwise noted.
b Results reported as effect estimate (95% confidence interval) unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
D.10 Hematological
Table D-19. Associations Between PFOS Exposure and Hematological Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
_ , ^ . Exposure Matrix,
Population, Ages, Levels
N (ng/mL)
Outcome
Comparison
Select Results"
General Population
D-221
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population, Ages,
N
Exposure Matrix,
Levels
(ng/mL)
Outcome Comparison
Select Results"
Etzel et al. (2019,
5043582)
Medium
United States,
2003-2010
Cross-sectional Children and adults Serum,
fromNHANES, Median = 15.1
> 12 years of age, (9.1-23.8)
N = 7,040
Vitamin D
deficiency
(< 50 ng/mL), 25-
hydroxy Vitamin
D ([25(OH)D],
nmol/L)
Regression
coefficient or
prevalence OR
(POR) per
doubling of
PFOS, or by
quintiles
Per doubling of PFOS:
Vitamin D deficiency
POR: 1.05 (0.97, 1.13)
25-hydroxy Vitamin D
-0.9 (-1.5, -0.2)
Q5: -2.8 (-4.7, -0.8)
60+years: -1.7 (-2.9,
-0.5)
No other statistically
significant associations or
trends
Results: Lowest quintile used as reference group.
Confounding: Gender, race/ethnicity, age, body mass index category, vitamin D supplement use, poverty to income ratio, smoking status,
6-month examination time period0
Chen et al. (2019,
5387400)
Medium
Croatia Cross-sectional Adults, 44-
2007-2008 56 years of age,
N = 122
Plasma, Calcium in serum Regression -0.05 (-0.09,-0.01),
GM = 8.91(min = (mmol/L) coefficient per p-value < 0.05
2.36, max = 33.67) ln-unit
increase in
PFOS
Confounding: Age, sex, education, socioeconomic status, smoking, dietary pattern, and physical activity
Jain (2020,
6333438)
Medium
United States
2003-2016
Cross-sectional
Adults from
NHANES,
> 20 years of age,
N = 11,251
Serum,
Non-anemic
males: GM = 12.0
(95% CI: 11.5,
12.7)
Non-anemic
females: GM = 8.1
(95% CI: 7.7, 8.5)
anemic males:
GM = 10.7 (95%
CI: 9.2, 12.5)
Whole blood
hemoglobin
(WBHGB)
(logl0-g/dL)
Regression
coefficient per
loglO-unit
increase in
PFOS
Non-anemic males: 0.009,
p-value < 0.01
Non-anemic
females:0.006,
p-value < 0.01
Anemic males: 0.023,
p-value < 0.01
Anemic females: 0.024,
p-value < 0.01
D-222
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population, Ages,
N
Exposure Matrix,
Levels
(ng/mL)
Outcome Comparison
Select Results"
anemic females:
GM = 5.0 (95%
CI: 4.4, 5.8)
Confounding: Age, BMI, poverty income ratio, serum cotinine, survey year, daily alcohol intake
Khalil etal. (2018,
4238547)
Low
United States,
2016
Cross-sectional
Children with Serum, 25-hydroxy
obesity, 8-12 years Median = 2.79 vitamin D
of age, N = 47 (IQR = 2.10) (ng/mL)
Regression -0.10 (-1.54, 1.33)
coefficient per
unit increase
inPFOS
Confounding: Age, sex, race
van den Dungen et The
al. (2017, 5080340) Netherlands,
Low 2015
Cross-sectional
Dutch men, 40-
70 years of age,
with habitual eel
consumption of at
least one portion
a month, N = 37
Serum,
Median=40 ng/g
wet weight (15-
93)
Hemoglobin
(Hb),
Hematocrit (Ht),
Retinol (units not
provided)
Regression Hb:-0.112 (-0.477,
coefficient 0.250)
Ht: -0.095 (-0.455, 0.263)
Retinol: 0.205 (-0.146,
0.561)
Confounding: Age, waist-to-hip ratio
Notes: aPTT = activated partial thromboplastin time; BMI = body mass index; GM = geometric mean; HIV = human immunodeficiency virus; IQR = interquartile range
NHANES = National Health and Nutrition Examination Survey; OR = odds ratio; PPT = prothrombin time.
a Exposure levels reported as median (25th-75th percentile) unless otherwise noted.
b Results reported as effect estimate (95% confidence interval) unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
D-223
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.ll Respiratory
Table D-20. Associations Between PFOS Exposure and Respiratory Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location, . Population,
Years gn Ages, N
Exposure Matrix,
Levels (ng/mL)a
Outcome
Comparison
Resultsb
Agier et al.
France, Greece, Cohort Pregnant women
Maternal and child's
FEV1
Regression
Prenatal: 0.1
(2019,
Lithuania, and their
serum, plasma, or whole
coefficient per
(-1.1, 1.3), p-value = 0.89
5043613)
Norway, Spain, children, ages 6-
blood
log2-unit
Medium
United 12 years,
increase in
Postnatal: 0.5
Kingdom N= 1,033
Prenatal (maternal)
PFOS
(-0.6, 1.6),
2003-2009
Median = 6.6
p-value = 0.38
(IQR =5.8)
Postnatal (child)
Median = 2.1
(IQR = 1.9)
Confounding: Centre of recruitment, child's sex, child's age, child's height, parental country of birth, breastfeeding duration, season of
conception, presence of older siblings, parental education level, maternal age, maternal pre-pregnancy body mass index, postnatal passive
smoking status, prenatal maternal active, passive smoking status0
Gaylord et al.
New York, US Cross- Adolescents and
Serum,
FEV1
Regression
No statistically significant
(2019,
2014-2016 sectional young adults,
FVC
coefficient per
differences observed between
5080201)
ages 13-22 years, Comparison group:
FEV1/FVC
log-unit increase
groups for the measured
Medium
N = 287
median = 2.75 (range :
TLC
in PFOS
outcomes, p-values > 0.05
0.60, 27.80)
RV
FRC
WTCHR group:
Resistance at an
median = 3.72 (range:
oscillation
1.01, 14.20)
frequency of
5Hz, 5-20Hz,
20Hz
Comparison: Logarithm base not specified.
Confounding: Sex, race/ethnicity, age, BMI, tobacco smoke exposure
Impinen et al.
Norway Cohort Infants followed
Cord blood,
Oslo Severity
OR per log2-
1.71 (1.16,2.53),
(2018,
1992-2002 up at 2 years and
Median =5.2 (4.0,6.6)
Score (1-5 vs. 0)
unit increase in
p-value = 0.007
4238440)
10,
PFOS
D-224
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure Matrix,
Levels (ng/mL)a
Outcome
Comparison
Resultsb
Medium
N = 641
Oslo Severity
Score (6-12 vs.
0)
1.15 (0.71, 1.84),
p-value = 0.576
Reduced lung 0.86 (0.43, 1.72),
function at birth p-value = 0.680
Outcome: Reduced lung function at birth: Lung function (tPTEF/tE) with standardized z-score, and binary variable of decreased lung function
(cutoff <0.20).
Confounding: Sex
Manzano-
Spain
Cohort
Pregnant women
Maternal blood,
FEV1,
Regression
No statistically significant
Salgado et al.
2003-2015
and their
Median = 6.06 (4.52,
FVC
coefficient per
associations for the measured
(2019,
children,
7.82)
FEV1/FVC,
log2-unit
outcomes
5412076)
followed up at
FEF25-75%
increase in
Medium
ages 1.5, 4, and
7 years,
N = 503 (4 years)
N = 992 (7 years)
PFOS
Confounding: Maternal age at delivery, parity, previous breastfeeding, pre-pregnancy BMI, region of residence, and country of birth
Qin et al.
Taiwan,
Case-
Children with
Serum,
FEV1
Regression
Statistically significant
(2017,
2009-2010
control
asthma and
Children with asthma:
FVC
coefficient per
associations in children with
3869265)
without asthma,
Median =31.51 (19.60,
FEF25-75%
ln-unit increase
asthma:
Medium
ages 10-15,
N = 132 (with
asthma)
91.69)
Children without asthma:
PEF
in PFOS
FEV1:-0.06 (-0.10,-0.02),
p-value < 0.05
N = 168 (without
asthma)
Median = 28.83 (12.39,
42.02)
FVC:-0.06 (-0.10,-0.01),
p-value < 0.05
Confounding: Age, sex, BMI, parental education level, exercise, environmental tobacco smoke exposure, and month of survey
Notes: BMI = body mass index; IQR = Interquartile range; FEF25-75% = Forced Expiratory Flow at 25-75%; FEV1 = Forced Expiratory Volume in 1 s; FRC = Functional
Residual Capacity; FVC = Forced Vital Capacity; PEF = Peak Expiratory Flow rate; RV = residual volume; TLC = total lung capacity; WTCUR = World Trade Center Health
Registry.
a Exposure levels reported as median (25th-75th percentile) unless otherwise noted.
b Results reported as effect estimate (95% confidence interval), unless otherwise noted.
c Confounding indicates factors the models presented adjusted for.
D-225
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.12 Musculoskeletal
Table D-21. Associations Between PFOS Exposure and Musculoskeletal Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix, Levels3
(ng/mL)
Outcome Comparison
Resultsb
Children and Adolescents
Jeddy et
England, 1991- Cohort
Females from Maternal serum Area adjusted
Regression
Height:
al.(2018,
2009
the ALSPAC 20.2 (15.6-25.5) BMC (g), bone
coefficient per
-0.11 (-0.19,-0.02)
5079850)
Study, area (cm2),
unit increase in
Total lean mass:
Medium
Age 17, BMC (g), BMD,
PFOS
-75.61 (-131.12, -20.1)
N = 221 cortical bone
area (cm2),
Bone area: -4.07 (-7.38, -0.76)
cortical BMC
(mg), cortical
BMC: -5.94 (-10.96, -0.92)
BMD (mg/cm2),
cortical
No other statistically significant
thickness (mm),
associations
endosteal
circumference
(mm), height
(cm), periosteal
circumference
(mm), total
femoral neck
BMD (g/cm2),
total hip BMD
(g/cm2), total
lean mass (g)
Confounding: Maternal pre-pregnancy BMI, maternal education, maternal age at delivery, gestational age at sample collection, ever breastfed
status at 15 months0
Cluett et
al.(2019,
5412438)
Medium
United States,
1999-2010
Cross-sectional
Children from
Project Viva,
Ages 6-10,
N = 531
Plasma
6.4 (IQR = 5.6)
Areal bone
mineral density
(aBMD) z-
score, bone
mineral content
(BMC) z-score
Regression
coefficient per
log2-unit
increase in
PFOS
aBMD z-score
-0.08 (-0.16,-0.01)
No statistically significant
associations or interactions by sex
BMC z-score: No statistically
significant associations
D-226
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix, Levels3
(ng/mL)
Outcome Comparison
Resultsb
Confounding: Maternal age, education, census tract median household income, individual household income, and child age, sex,
race/ethnicity, year of blood draw, dairy intake, physical activity
Khalil et
United States Cross-sectional
Obese children,
Serum
BMD measured
Regression
BMD (broadband ultrasound
al.(2018,
2016
ages 8-12
2.79
as broadband
coefficient per
attenuation)
4238547)
N = 23
(IQR = 2.10)
ultrasound
unit increase in
-1.03 (-5.35,3.29)
Low
attenuation
PFOS
(dB/MHz) and
BMD (speed of sound)
speed of sound
-5.22 (-11.2,0.79)
(m/s), stiffness
index (%)
Stiffness index
-2.15 (-5.56, 1.26)
Confounding: Age, sex, race
Di Nisio et
Italy Cross-sectional
Male high
Serum
Arm span (cm)
Mann-Whitney
Arm span
al.(2019,
2017-2018
school students
Controls: 0.82
test (Exposed vs
Controls: 182.75 (178.0, 185.8)
5080655)
N = 100 (50
(0.4-1.3)
Controls)
Exposed: 179.00 (174.2, 187.0)
Low
controls, 50
Exposed: 1.11
Adjusted p-value for comparison of
exposed)
(0.8-1.3)
medians = 0.738
Semen
Controls: 0.11
(0.08-0.13)
Exposed: 0.11
(0.01-0.14)
Results: Values for each outcome are reported as median (25th, 75th percentile).
Confounding: None reported
General Population
Uhl et al.(2013, United States,
1937226) 2003-2008
Medium
Cross-sectional
Adults from
NHANES,
Ages 20-84,
N= 3,809,
Females
N = 1,921
Serum
Adults:
Weighted
mean = 21.23
Females:
Weighted
mean = 18.17
Osteoarthritis
OR per ln-unit
increase in
PFOS or by
quartiles
Adults 20-84
1.15 (0.94, 1.40)
Q2: 1.04 (0.58, 1.85)
Q3: 1.99 (1.14,3.49),
p-value < 0.05
Q4: 1.77(1.05,2.96),
p-value < 0.05
Females 20-49
D-227
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix, Levels3
(ng/mL)
Outcome Comparison
Resultsb
2.37 (1.35, 4.16), p-value < 0.01
Q2: 0.65 (0.19,2.20)
Q3: 1.11 (0.29,4.30)
Q4: 4.99 (1.61, 15.4),
p-value < 0.01
No other statistically significant
associations
Results: Lowest quartile used as the reference group.
Confounding: Age, race/ethnicity, SES, smoking, BMI, vigorous reactional activity, prior wrist, hip, or spine fracture
Linetal.(2014,
United States, Cross-sectional Adults from
Serum
Total BMD
Regression Total BMD in lumbar spine
5079772)
2005-2006, NHANES
GM = 15.32
(g/cm2) in hip or
coefficient per Women not in menopause: -0.022
Medium
2007-2008 Ages > 20,
(SD = 17.58)
lumbar spine;
ln-unit increase (-0.038, -0.007), p-value = 0.006
Males
fractures in hip,
inPFOS
N = 1,192,
wrist, spine, or
Other outcomes: No statistically
Females
all types
significant associations
N = 842,
Females in
menopause
N = 305
Confounding: Age, race/ethnicity, BMI, smoking, drinking, treatment for osteoporosis,
use of prednisone or Cortisol daily
Khalil et
al.(2016,
3229485)
Medium
United States,
2009-2010
Cross-sectional
Adolescents and Serum
adults from Mean = 12.7
NHANES, Ages (SE = 1.20)
12-80,
Males N = 956,
Females
N = 958
BMD (g/cm2) of
total femur,
femoral neck,
lumbar spine;
Osteoporosis
among females
BMD: Total femur
Regression Females: -0.018 (-0.034, -0.002),
coefficient per p-value < 0.05
ln-unit increase Q2: -0.007 (-0.038, 0.023)
inPFOS and by Q3: -0.009 (-0.037, 0.019)
quartiles Q4: -0.044 (-0.074, -0.014),
Osteoporosis: p-value < 0.05
OR per ln-unit Males: Not statistically significant
increase in
PFOS and by Femoral neck
quartiles Females: -0.016 (-0.029, -0.002),
p-value < 0.05
Q2: 0.001 (-0.019, 0.019)
Q3:-0.001 (-0.025, 0.025)
D-228
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix, Levels3
(ng/mL)
Outcome Comparison
Resultsb
Q4: -0.034 (-0.059, -0.009),
p-value < 0.05
Males: -0.013 (-0.024, -0.002),
p-value < 0.05
Q2
Q3
Q4
-0.036 (-0.077, 0.006)
-0.027 (-0.063, 0.009)
-0.046 (-0.078, -0.015),
p-value < 0.05
Lumbar spine, osteoporosis: No
statistically significant associations
Results: Lowest quartile used as the reference group.
Confounding: Age, ethnicity, BMI, serum cotinine, physical activity, milk consumption, blood lead concentration
Hu et al.(2019, United States,
6315798) 2004-2007
Medium
Cohort and
cross-sectional
Adults from the
POUNDS-Lost
study,
Ages 30-70,
N = 294
BMD and 2-yr
ABMD (g/cm2)
of spine, total
hip, femoral
neck, hip
trochanter, hip
intertrochanteric
area, and
Ward's triangle
area
Plasma BMD and 2-yr Regression Spine BMD analyses
Mean = 32.2 ABMD (g/cm2) coefficient per Cross-sectional: -0.02 (-0.037,
(16.8-43.1) of spine, total SD increase in -0.003)
)S
Total hip BMD analyses
2-yr ABMD: -0.005
(-0.009, -0.001), p-value < 0.05
Hip intertrochanteric area BMD
analyses
2-yr ABMD: -0.008 (-0.013,
-0.003), p-value < 0.05
Femoral neck, hip trochanter,
Ward's triangle area: no
statistically significant associations
No statistically significant
associations or interactions by sex
Confounding: For cross-sectional, age, sex, race, alcohol consumption, physical activity, BMI, dietary intervention group; For cohort, age,
sex, race, alcohol consumption, physical activity, BMI, dietary intervention group, baseline BMD, 2-year weight change
D-229
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Notes: aBMD = areal bone mineral density; ALSPAC = Avon Longitudinal Study of Parents and Children; BMD = bone mineral density; BMI = body mass index;
GM = geometric mean; IQR = interquartile range; NHANES = National Health and Nutrition Examination Survey; OR = odds ratio; POUNDS-Lost = Prevention of Obesity
Using Novel Dietary Strategies Lost clinical trial; Q1 = quartile one; Q4 = quartile four; SD = standard deviation; SE = standard error; SES = socioeconomic status.
a Exposure levels reported as median (25th-75th percentile) unless otherwise specified.
b Results reported as effect estimate (95% confidence interval) unless otherwise specified.
c Confounding indicates factors the models presented adjusted for.
D.13 Gastrointestinal
Table D-22. Associations Between PFOS Exposure and Gastrointestinal Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Population,
Exposure
Resultsb
Years
Study Design Ages,
N
Matrix, Levels
(ng/mL)a
Outcome
Comparison
Timmerman et
Guinea-Bissau Cohort Children aged Serum Diarrhea
OR per
At inclusion: 1.14(0.66, 1.96)
al. (2020,
2012-2015 <2 years 0.77 (0.53-1.02)
doubling of
At 9 months: 1.2 (0.62,2.31)
6833710)
previously
PFOS at
Medium
enrolled in a
inclusion or 9-
No statistically significant
RCT for
month visit
associations or interactions by sex
measles
vaccination
N = 236 (113
girls, 123 boys)
Confounding: Weight and age at inclusion, sex, maternal education, breastfeeding without solids0
Dalsager et al.
(2016, 3858505)
Low
Denmark Cohort Pregnant Serum
2010-2015 women and 8.07 (Range:
their children 2.36-25.10)
from the Odense
Child Cohort,
Ages 1-4 years
N = 346
Diarrhea,
vomiting
(number of days
with symptom
or proportion
of days
under/above
median)
Incidence rate
ratio (number
of days) or OR
(proportion
of days) by
tertiles of PFOS
exposure
Diarrhea
Number of days with symptom
T2: 1.41 (0.79,2.51)
T3: 1.19(0.67,2.12)
Proportion of days under/above
median
T2: 0.89 (0.51, 1.56)
T3: 1.04 (0.59, 1.82)
Vomiting
Number of days with symptom
T2: 1.18(0.8, 1.74)
T3: 0.87 (0.58, 1.31)
Proportion of days under/above
median
D-230
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Population,
Location,
Study Design Ages,
Years
N
Exposure
Matrix, Levels Outcome
(ng/mL)a
Comparison Resultsb
Results: Lowest tertile used as reference.
Confounding: Maternal age, maternal educational level, parity, and child age
T2: 1.47 (0.86,2.54)
T3: 0.78 (0.45, 1.35)
Hammer et al. Faroe Islands Cohort
Children and
Blood
Inflammatory
Incidence rate
0.30 (0.08, 1.07)
(2019, 8776815) Enrollment:
adults from
Low exposure:
bowel disease
ratio for highest
Low 1986-2009;
CHEF
GM = 2.33
vs. lowest tertile
follow-up until
N = 2,843
(1.93-2.90)
of PFOS
2017
High exposure:
GM = 26.88
(21.90-32.24)
exposure
Confounding: Age, calendar period
Xu et al. (2020, Sweden Cohort
Residents of
Serum
Inflammatory
Regression
Calprotectin
6315709) 2014-2016
Ronneby
Ronneby panel
bowel disease
coefficient per
Panel study: -0.0008 (-0.0033,
Low
municipality
study: 216
(ln-ng/mL levels
unit increase in
0.0018)
(118-300)
of calprotectin
PFOS
Resampling: -0.0006 (-0.0016,
Ronneby panel
Ronneby
or zonulin)
0.0005)
study: N = 57
resampling: 271
Karlshamn: -0.045 (-0.14, 0.05)
Ronneby
(147-449)
resampling:
Karlshamn: 5
Zonulin
N = 113
(4-7)
Panel study: 0.0007 (-0.0012,
Karlshamn:
0.0025)
N = 19
Resampling: -0.0001 (-0.0008,
0.0005)
Karlshamn: -0.019 (-0.1, 0.063)
Confounding: Age, BMI, gender
Notes: BMI = body mass index; CHEF = Children's Health and the Environment in the Faroes; PFOS = perfluorooctane sulfonate; RR = risk ratio; RCT = randomized controlled
trial.
a Exposure levels reported as median (25th-75th percentile) unless otherwise specified.
b Results reported as effect estimate (95% confidence interval) unless otherwise specified.
c Confounding indicates factors the models presented adjusted for.
D-231
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D. 14 Dg nta I
Table D-23. Associations Between PFOS Exposure and Dental Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Population,
Study Design Ages,
N
Exposure
Matrix, Levels Outcome
(ng/mL)a
Comparison
Resultsb
Puttige Ramesh
United States
Cross-sectional Adolescents
Serum Dental caries
OR per log2-
0.99 (0.92, 1.07)
etal. (2019,
1999-2002
from NHANES
Median =13
unit increase in
Q2: 0.91 (0.72, 1.16)
5080517)
aged 12-
(7.2-22)
PFOS and by
Q3: 1.02 (0.81, 1.31)
Medium
19 years
quartiles
Q4: 0.92 (0.72, 1.17)
N = 2,869
Results: Lowest quartile used as reference
Confounding: Gender, race, education level of parent/guardian, family poverty to income ratio, blood lead level, serum cotinine level0
Wiener &
Waters (2019,
5386081)
United States
2013-2014
Cross-sectional Children from
NHANES aged
3-11 years
Serum Dental caries
GM=3.88 experience
(95% CI: 3.53,
OR per IQR
increase in
PFOS
1.41 (0.97, 2.05); p-value = 0.069
Medium N = 629 4.27)
Confounding: Age, sex, race/ethnicity, ratio of family income to poverty guidelines, tooth brushing frequency, dental visit, percentages of
sugar in the diet, fluoride in the water
Notes: CI = confidence interval; IQR = interquartile range; NHANES = National Health and Nutrition Examination Survey; OR = odds ratio.
a Exposure levels reported as median (25th-75th percentile) unless otherwise specified.
b Results are reported as effect estimate (95% confidence interval).
c Confounding indicates factors the models presented adjusted for.
D.15 Ocular
Table D-24. Associations Between PFOS Exposure and Ocular Effects in Recent Epidemiologic Studies
Reference, Location,
Confidence Years
„ . Population,
Design ' ..
b Ages, N
Exposure Matrix,
Levels (ng/mL)a
Outcome
Comparison
Resultsb
Zeeshan et al. China,
Cross-sectional Adults from the
Serum
Visual impairment,
OR per ln-unit
Visual impairment
(2020,6315698) 2016
Isomers of C8
Median = 24.07
synechia, macula
increase in PFOS
3.11 (2.3,4.2);
Medium
Health Project,
(14.13-36.41)
disorder, corneal
p-value < 0.05
ages 22-
pannus, shallow
96 years,
anterior chamber,
Eye disease, combined
N = 1,202
vitreous disorder,
<65 years: 1.52(1.21,
retinal disorder,
1.91); p-value < 0.05
D-232
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population, Exposure Matrix,
Ages, N Levels (ng/mL)a
Outcome
Comparison
Resultsb
lens opacity,
> 65 years: 0.91 (0.55,
conjunctival
1.51)
disorder, combined
eye disease
All other outcomes: No
statistically significant
associations
Confounding: Age, sex, BMI, education, income, career, exercise time, drinking, smoking0
Notes: BMI = body mass index.
a Exposure levels reported as median (25th-75th percentile) unless otherwise specified.
b Results are reported as effect estimate (95% confidence interval).
c Confounding indicates factors the models presented adjusted for.
D.16 Dermal
Table D-25. Associations Between PFOS Exposure and Dermal Health Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix, Levels
(ng/mL)a
Outcome Comparison
Resultsb
Ernst et al. Denmark Cohort Pregnant
Maternal blood Acne, age at
Regression
Girls: -1.73 (-5.24, 1.77)
(2019,5080529) 1999-2017 women and
(1 st trimester) occurrence
coefficient per
T2: 0.09 (-4.69, 4.87)
Medium their children
Girls Sample 1: (months)
log2-unit
T3: -1.96 (-6.89,2.97)
from the
32.3 (19.3-50.8)
increase in
Puberty Cohort
Girls Sample 2:
PFOS, and by
Boys:-1.52 (-4.52, 1.48)
within the
27.9 (16.5-42.2)
tertiles
T2:-1.33 (-5.02,2.36)
DNBC
Boys Sample 1:
T3:-0.7 (-4.75,3.35)
N = 555 girls,
31.9(19.2-51.2)
565 boys
Boys Sample 2:
27.2 (16.7-45.2)
Results: Lowest tertile used as a reference group.
Confounding: Highest social class of parents, maternal age at menarche, maternal age
at delivery, parity,
pre-pregnancy body mass index,
daily number of cigarettes smoked in first trimester0
Notes: DNBC = Danish National Birth Cohort.
a Exposure levels reported as median (10th-90th percentile).
b Results reported as effect estimate (95% confidence interval).
c Confounding indicates factors the models presented adjusted for.
D-233
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.17 Cancer
Table D-26. Associations Between PFOS Exposure and Cancer in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population, Ages,
N
Exposure
Matrix,
Levels"
Outcome
Comparison
Select Resultsb
Grice et al. (2007,
4930271)
Medium
United States Cohort Employees of a
1997-1998 PFOS-based
chemical and film
manufacturing
plant, 2,083
Modeled
Non-exposed:
GM
range = 0.11—
0.29 ppm;
Low-exposed:
GM
range = 0.39-
0.89 ppm;
High-
exposed:
GM
range = 1.30—
1.97 ppm
Cancers: colon,
melanoma, and
prostate
OR by PFOS
exposure category
Colon cancer:
Ever exposed: 1.21 (0.51, 2.87)
Low or high exposed: 1.37
(0.57, 3.30)
High exposed: 1.69 (0.68,
4.17)
Melanoma:
Ever exposed: 1.08 (0.31, 3.72)
Low or high exposed: 0.90
(0.24, 3.43)
High exposed: 1.01 (0.25,
4.11)
Prostate cancer:
Ever exposed: 1.34 (0.62, 2.91)
Low or high exposed: 1.36
(0.61, 3.02)
High exposed: 1.08 (0.44,
2.69)
Results: Non-exposed used as the reference group
Confounding: Age and gender
Eriksen et al.
(2009, 2919344)
Medium
Denmark
1993-2006
Cohort
Adults with no
Serum
Cancers: prostate, IRR per unit increase Prostate cancer:
previous cancer
Mean (5th-
bladder, in PFOS, or by
Q2: 1.35 (0.97, 1.87)
diagnosis,
95th
pancreatic, liver quartiles
Q3: 1.31 (0.94, 1.82)
Ages 50-65 at
percentile):
Q4: 1.38 (0.99, 1.93)
enrollment,
Cases, men:
Per unit increase: 1.05 (0.97,
Prostate cancer,
35.1 (17.4-
1.14)
1,393;
60.9);
Bladder cancer,
Controls,
Bladder cancer:
1,104;
men: 35.0
Q2: 0.76 (0.50, 1.16)
(16.8-62.4);
Q3: 0.93 (0.61, 1.41)
D-234
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population, Ages,
N
Exposure
Matrix,
Levels"
Outcome
Comparison
Select Resultsb
Pancreatic cancer,
900;
Liver cancer, 839
Cases,
Q4: 0.70 (0.46, 1.07)
women: 32.1
Per unit increase: 0.93
(14.0-58.1);
1.03)
Controls,
women: 29.3
Pancreatic cancer:
(14.2-55.6)
Q2: 1.02 (0.57, 1.84)
Q3: 1.24 (0.67,2.31)
Q4: 0.91 (0.51, 1.65)
Per unit increase: 0.99
1.14)
Liver cancer:
Q2: 0.62 (0.29, 1.33)
Q3: 0.72 (0.33, 1.56)
Q4: 0.59 (0.27, 1.27)
Per unit increase: 0.97
1.19)
Results: Lowest quartile used as the reference group
Confounding: Prostate cancer: years of school attendance, BMI, dietary fat intake, and vegetable intake; Bladder cancer: smoking status,
smoking intensity, smoking duration, years of school attendance, occupation associated with risk for bladder cancer; Pancreatic cancer:
smoking status, smoking intensity, smoking duration, dietary fat intake, and fruit and vegetable intake; Liver cancer: smoking status, years
of school attendance, alcohol intake, and occupation associated with risk for liver cancer
Bonefeld-
Jorgensen et al.
(2011, 2150988)
Medium
Greenland Case-control Greenlandic Inuit Plasma Breast cancer
2000-2003 women with and Cases: 45.6
without breast (Range =11.6
cancer, 76 -124)
Controls: 21.9
(Range = 1.5-
172)
Confounding: Age, BMI, pregnancy, cotinine, breastfeeding, and menopausal status
OR per ln-unit
increase in PFOS
1.030(1.001, 1.070), p-
value = 0.05
Ducatman et al.
(2015, 3859843)
Medium
United States
2005-2006
Cross-
sectional
Men from C8
Health Study ,
Ages 20-49,
9,169;
Ages 50-69, 3,819
Serum
Mean (SD):
22.18(1.97)
Prostate-specific
antigen (PSA)
level
Regression
coefficient ((3) per
ln-unit increase in
PFOS
Age 20-49
(3=1, p-value = 0.71;
GMR = 0.95 (0.71, 1.28)
Age 50-69
D-235
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population, Ages,
N
Exposure
Matrix,
Levels"
Outcome
Comparison
Select Resultsb
GM ratio (GMR)
(PSA < 4.0 ng/mL
vs.
PSA > 4.0 ng/mL)
(3=1, p-value = 0.99;
GMR = 1.1 (0.98, 1.23)
Confounding: Age, smoking status, average alcohol intake, and body mass index0
Ghisari et al.
(2017, 3860243)
Medium
Hurley et al.
(2018, 5080646)
Medium
Denmark
1996-2002
Nested case-
control
Adult women, 283
Serum
Cases: 27.80
Controls:
28.77
Breast cancer
Cohort: 1.15 (0.64,2.08)
CYP19 CC: 6.42 (1.08, 38.3),
p-value <0.05
No significant associations
observed for remaining
genotypes
Relative risk ratio
(RR) per ln-unit
increase in PFOS,
compared across
genotypes:
CYP1A1
(Ile462Val),
CYP1B1
(Leu432Val), COMT
(Vall58Met),
CYP17 (-34T > C),
CYP19 (C > T)
Confounding: Age at blood draw, BMI before pregnancy, total number of gravidities, oral contraceptives use, age of menarche, smoking
status and alcohol intake during pregnancy, physical activity, maternal education
Results: Lowest tertile used as the reference group
Confounding: Age, BMI, cotinine levels, parity, and breastfeeding
OR per loglO-unit
increase in PFOS, or
by tertiles
T3: 0.898 (0.695, 1.161)
T2: 0.883 (0.691, 1.129)
Per unit increase: 0.934 (0.683,
1.277), p-value = 0.67
California, US Nested case- Adult women, Serum Breast cancer
2011-2015 control 1,760 Median (min- (invasive)
max): Cases:
6.695 (0.046-
39.400)
Controls:
6.950 (0.046-
99.800)
Confounding: Age at baseline enrollment, race/ethnicity, region of residence, date of blood draw, season of blood draw, total smoking
pack-years, BMI, family history of breast cancer, age at first full-term pregnancy, menopausal status at blood draw, and pork consumption
Cohn et al. (2020,
5412451)
Medium
United States Nested case- Adult daughters of Perinatal
1959-2013 control women in CHDS serum
cohort, 310 Cases: 3 0.5
controls, 102 cases (14.1-55.8)
Breast cancer
OR per log2-unit
increase in PFOS
0.3 (0.1,0.9), p-value = 0.02
D-236
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population, Ages,
N
Exposure
Matrix,
Levels"
Outcome
Comparison
Select Resultsb
Controls: 32.1
(14.9-58.2)
Confounding: Maternal: cholesterol, age at pregnancy, history of breast cancer, primiparity, overweight at first prenatal visit, serum levels
of DDTs and metabolite DDE, African American status, whether daughter was breastfed
Mancini et al.
(2020, 5381529)
Medium
France Nested case- Postmenopausal Serum Breast cancer ORs by quartiles, Overall:
1990-2013 control women, 17.51(5.83- and by estrogen (ER) Q2: 1.94 (1,3.78)
Ages 40-65 in 85.26) or progesterone Q3:2.03 (1.02,4.04)
1990, 194 cases, receptor (PR) status Q4: 1.72 (0.88, 3.36)
194 controls p-trend = 0.25
ER positive:
ORs of 1.8-2.4
p-trend = 0.04
ER negative:
ORs of 4.7-15
p-trend = 0.72
PR positive:
ORs of 1.8-2.7
p-trend = 0.02
PR negative: ORs of 1.7-3.5
p-trend = 0.93
Results: Lowest quartile used as the reference group
Confounding: Total serum lipids, BMI, smoking status, physical activity, education level, personal history of benign breast disease, family
history of breast cancer, parity/age at first full-term pregnancy, total breastfeeding duration, age at menarche, age at menopause, use of oral
contraceptives, current use of menopausal hormone therapy
Shearer et al.
(2021, 7161466)
Medium
ORs per log2-unit
increase in PFOS or
by quartiles (total
cohort only)
United States Nested case- Adults, 55-74, Serum Renal cell
1993-2002 control 648 38.4 (26.3- carcinoma
Ages 55-59, 190 49.9) ng/L
Ages 60-65,224
Ages 65+, 234
Males 432
Females 216
Results: Lowest quartile used as the reference group
Confounding: BMI, smoking, history of hypertension, estimated glomerular filtration rate, previous freeze-thaw cycle, calendar year of
blood draw; sex, race and ethnicity, study year of blood draw, study center
Q2: 1.67 (0.84, 3.3)
Q3: 0.92 (0.45, 1.88)
Q4: 2.51 (1.28,4.92)
p-trend = 0.009
Per doubling increase: 1.39
(1.04, 1.86)
D-237
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Population, Ages, Exposure
Design Matrix,
^ Levels3
Outcome
Comparison
Select Resultsb
Fry and Power
(2017, 4181820)
Medium
US NHANES Cohort
2003-2006
Adults,
Ages 60+, 1,036
Serum
Median (SE):
4.3 (0.2) ng/g
lipid
Confounding: Age, gender, race/ethnicity, and smoking status
Cancer mortality Hazard ratio per SD 1.01(0.86,1.19),
unit increase in p-value = 0.88
PFOS
CMstensen et al.
(2016, 3858533)
Low
Wisconsin, US, Cross-
2012-2013 sectional
Male anglers, Serum
Ages 50+, 154 19.00 (9.80-
28.00)
Confounding: Age, BMI, work status, alcohol consumption
Cancer (any) OR per unit increase 0.99(0.96,1.01)
in PFOS
Lin et al. (2020,
6835434)
Low
Tsai et al. (2020,
6833693)
Low
China Case-control Children, < 16, 84 Serum Germ cell tumors OR per unit increase 1.08(0.96,1.21)
2014-2017 4.47(2.48- in PFOS
8.26)
Confounding: Infectious disease, cosmetics usage, barbecued food consumption, filtered water use, indoor decorating, living near farmland
(maternal behaviors/factors during pregnancy)
Taiwan Case-control Adult women, 239 Plasma
2014-2016 Age 50 or Mean(GM):
younger, 120 5.64 (4.77)
Age over 50, 119
Breast cancer
OR per ln-unit
increase in PFOS
Total cohort: 1.07 (0.64, 1.79)
Age 50 or younger: 2.34 (1.02,
5.38), p-value < 0.05
ER+: 3.25 (1.29, 8.23)
Age over 50: 0.62 (0.29, 1.29),
p-value >0.05
Confounding: Pregnancy history, oral contraception use, abortion, BMI, menopause, and education level
Itoh et al. (2021
9959632)
Low
Japan
2001-2005
Case-control
Adult women,
Ages 20-74,
802 (401 breast
cancer cases, 401
controls)
Serum
14.27 (10.24-
19.24)
Breast cancer OR by quartiles
Q2: 0.38 (0.18, 0.82), p-
value < 0.05
Q3: 0.31 (0.14,0.69), p-
value < 0.05
Q4: 0.15 (0.06,0.39), p-
value < 0.05
p-trend = 0.0001
Results: Lowest quartile used as the reference group
D-238
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population, Ages,
N
Exposure
Matrix,
Levels"
Outcome
Comparison
Select Resultsb
Confounding: Age, residential area, BMI, height, menopausal status, age at menopause, age at first childbirth, family history of breast
cancer, smoking status, strenuous physical activity in the past five years, moderate physical activity in the past five years, age at menarche,
number of births, breastfeeding duration, alcohol intake, isoflavone intake, education level, serum total concentrations of PCBs, fish and
shellfish intake, vegetable intake, and calendar year of blood sampling
Liu et al. (2021,
China Case-control Adult men, 96 Serum Thyroid cancer OR by quartiles
Total
10176563)
2016-2017 Adult women, 223 Case: 5.5
Q2: 0.81 (0.42, 1.53)
Low
(3.6-8.8);
Q3: 0.26 (0.12,0.57)
Control: 7.5
Q4: 0.28 (0.12,0.66)
(4.7-10.8)
p-trend = 0.001
Male:
Q2: 1.13 (0.30,4.23)
Q3: 0.15 (0.02, 1.04)
Q4: 0.62 (0.15,2.65)
p-trend = 0.284
Female:
Q2: 1.10 (0.52,2.34)
Q3: 0.33 (0.13, 0.80)
Q4: 0.24 (0.09, 0.64)
p-trend = 0.001
Results: Lowest quartile used as the reference group
Confounding: Age, sex, and diabetes status
Omoike et al.
(2021, 7021502)
Low
United States
2005-2012
Cross-
sectional
Adults from
NHANES,
Ages > 20 years,
6,652
Serum
11.40 (6.45-
19.68
Cancers: ovarian,
breast, uterine,
and prostate
OR per unit increase
in PFOS, or by
quartiles
Ovarian cancer:
Q2: 0.08 (0.08, 0.084)
Q3: 1.64(1.62, 1.66)
Q4: 2.25 (2.22, 2.28)
p-trend < 0.001
Per unit increase: 1.012(1.012,
1.013)
Breast cancer:
Q2: 0.87 (0.86, 0.89)
Q3: 1.06 (1.05, 1.06)
Q4: 1.47 (1.46, 1.48)
p-trend < 0.001
D-239
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference. Location. Population, Ages, Exposure
„ .... Design Matrix, Outcome Comparison Select Results
Commence Years N T . »
Levels3
Per unit increase:
1.011 (1.011, 1.011)
Uterine cancer:
Per unit increase:
0.945 (0.944, 0.945)
Prostate cancer:
Per unit increase:
0.994 (0.994, 0.994)
Results: Lowest quartile used as the reference group
Confounding: Age, sex, education, race/ethnicity, PIR, BMI, and serum cotinine
Notes: CHDS = The Child Health and Development Studies; DDE = dichlorodiphenyldichloroethylene; DDT = dichloro-diphenyl-trichloroethane; GM = geometric mean;
IRR = incidence rate ratio; NHANES = National Health and Nutrition Examination Survey; OR = odds ratio; SD = Standard deviation.
a Exposure levels reported as median (25th-75th percentile) in ng/mL unless otherwise noted.
b Results reported as effect estimate (95% confidence interval).
c Confounding indicates factors the models presented adjusted for.
D-240
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix E. Benchmark Dose Modeling
E.l Epidemiology Studies
E.l.l Modeling Results for Immunotoxicity
E.l. 1.1 Modeling Results for Decreased Tetanus Antibody
Concentrations
E.l. 1.1.1 Budtz-J0rgensen and Grandjeon (2018, 5083631) Results for
Decreased Tetanus Antibody Concentrations at Seven Years of Age and PFOS
Exposure Measured at Five Years of Age
Budtz-j0rgensen and Grandjean (2018, 5083631) fit multivariate models of PFOS measured at
age five years, against log2-transformed anti-tetanus antibody concentrations measured at the
seven-year-old examination controlling for sex, exact age at the seven-year-old examination, and
booster type at age five years. Models were evaluated with additional control for PFOA (as
log2[PFOA]), and without PFOA. Three model shapes were evaluated by Budtz-j0rgensen and
Grandjean (2018, 5083631) using likelihood ratio tests: a linear model, a piecewise-linear model
with a knot at the median PFOS concentration, and a logarithmic function. The logarithmic
functions did not fit better than the piecewise-linear functions {Budtz-j0rgensen, 2018,
5083631}. The piecewise-linear model did not fit better than the linear model for the PFOS
exposure without adjustment for PFOA using a likelihood ratio test (p = 0.60; see Budtz-
J0rgensen and Grandjean (2018, 5083631) Table 3), or for the model that did adjust for PFOA
(log2[PFOA]) (p = 0.71).
Table E-l summarizes the results from Budtz-j0rgensen and Grandjean (2018, 5083631) for
PFOS at age five years and tetanus antibodies at age seven years. These regression coefficients
(P) and their standard errors (SE) were computed by EPA from the published BMDs and BMDL
based on a BMR of 5% decrease in the antibody concentration in Table 1 of Budtz-j0rgensen
and Grandjean (2018, 5083631).6 As Budtz-j0rgensen and Grandjean (2018, 5083631) log2-
transformed the outcome variable, the BMR measured in unit of log2[tetanus antibody
concentration] was log2(l-0.05) = 0.074 log2(IU/mL)).
6 Budtz-Jergensen and Grandjean (2018, 5083631) computed BMDs and BMDLs using a BMR of 5% decrease in the antibody
concentrations. Their formula, BMD = log2(l-BMR)/p, can simply be reversed to solve for p = log2(l-BMR)/BMD. For
negative dose-response where more exposure results in lower antibody concentration, the BMDL is based on the lower bound of
P, (Plb). Thus, the Plb = log2(l-BMR)/BMDL. The SE(P) = (P-Plb)/1 .645. The p-value is the two-sided probability that
Z <= SE(P)/p.
E-l
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-l. Results specific to the slope from the linear analyses of PFOS measured at age
five years and log2(tetanus antibody concentrations) measured at age seven years from
Table 1 in Budtz-Jorgensen and Grandjean (2018, 5083631) in a single-PFAS model and in
a multi-PFAS model
Exposure
Model shape
PFOA adjusted Slope ») per
J ng/mL
SE(P)
ng/mL
Slope (P) fit
Lower bound
slope (Plb)
ng/mL
PFOS at Age 5
Linear
No -0.0274
0.0176
p = 0.12
-0.0565
PFOS at Age 5
Linear
Yes -0.0039
0.0198
p = 0.84
-0.0365
Notes: SE = standard error.
Interpretation of results in Table E-l:
• PFOS is a non-significant predictor in the single-PFAS model (P = -0.0274; p = 0.12).
• Effects of PFOS in the single-PFAS model are attenuated when log2[PFOA] is included in
the model (P = -0.0039; p = 0.84).
• Nevertheless, these data can be used to estimate a BMDL for completeness and to allow
comparisons across PFAS.
Selection of the Benchmark Response
The BMD approach involves dose-response modeling to obtain BMDs, i.e., dose levels
corresponding to specific response levels near the low end of the observable range of the data
and the BMDLs to serve as potential PODs for deriving quantitative estimates below the range of
observation {U.S. EPA, 2012, 1239433}. Selecting a BMR to estimate the BMDs and BMDLs
involves making judgments about the statistical and biological characteristics of the data set and
about the applications for which the resulting BMDs and BMDLs will be used. An extra risk of
10% is recommended as a standard reporting level for quantal data for toxicological data.
Biological considerations may warrant the use of a BMR of 5% or lower for some types of
effects as the basis of the POD for a reference value. However, a BMR of 1% has typically been
used for quantal human data from epidemiology studies {U.S. EPA, 2012, 1239433}, although
this is more typically used for epidemiologic studies of cancer mortality within large cohorts of
workers which can support the statistical estimation of small BMRs.
In the 2021 Proposed Approaches draft {U.S. EPA, 2021, 10428576} reviewed by the SAB
PFAS Review Panel, EPA relied on the BMDL modeling approach published in Budtz-
J0rgensen and Grandjean (2018, 5083631), described above. During validation of the modeling,
EPA reevaluated the approach chosen by Budtz-j0rgensen and Grandjean (2018, 5083631) and
determined that a different approach should be used to be consistent with EPA guidance {U.S.
EPA, 2012, 1239433}, which recommends the use of a 1 or V2 SD change in cases where there is
no accepted definition of an adverse level of change or clinical cut-off for the health outcome.
A blood concentration for tetanus antibodies of 0.1 IU/mL is sometimes cited in the tetanus
literature as a 'protective level' and {Grandjean, 2017, 4239492} noted that the Danish vaccine
producer Statens Serum Institut recommended the 0.1 IU/mL "cutoff level "to determine
whether antibody concentrations could be considered protective," and Galazka and
Kardymowicz (1989, 9642152) mentions the same concentration. However, the 2018 WHO
update {WHO, 2018, 10406857} argues that:
E-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
"...the minimum amount of circulating antitoxin that in most cases ensures immunity
to tetanus is assay specific. Within in vivo neutralization tests, modified ELISAsor
bead-based immunofluorescence assays¦, concentrations at or exceeding 0.01 IU/mL
are usually considered protective against disease, whereas antitoxin concentrations of
at least 0.1-0.2 IU/mL are defined as positive when ELISA techniques are used for the
assessment. Cases of tetanus have been documentedhowever¦, in persons with
antitoxin concentrations above these thresholds. Hence, a "protective antibody
concentration" may not be considered a guarantee of immunity under all
circumstances."
In the absence of a clear definition of an adverse effect for a continuous endpoint like antibody
concentrations, a default BMR of 1 or V2 SD change from the control mean may be selected
{U.S. EPA, 2012, 1239433}. As noted above, a lower BMR can also be used if it can be justified
on a biological and/or statistical basis. Figure E-l replicates a figure in the Technical Guidance
(page 23) {U.S. EPA, 2012, 1239433} to show that in a control population where 1.4% are
considered to be at risk of having an adverse effect, a downward shift in the control mean of 1
SD results in a -10% extra risk of being at risk of having an adverse effect.
Standard deviation units
Figure E-l. Difference in population tail probabilities resulting from a one standard
deviation shift in the mean from a standard normal distribution, illustrating the theoretical
basis for a baseline BMR of 1 SD
E-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Statistically, the Technical Guidance additionally suggests that studies of developmental effects
can support lower BMRs. Consistent with EPA's BenchmarkDose Technical Guidance {U.S.
EPA, 2012, 1239433}, EPA typically selects a 5% or 0.5 standard deviation (SD) benchmark
response (BMR) when performing dose response modeling of data from an endpoint resulting
from developmental exposure. Because Budtz-j0rgensen and Grandjean (2018, 5083631)
assessed antibody response after PFAS exposure during childhood, this is considered a
developmental study {U.S. EPA, 1991, 732120} based on EPA's Guidelines for Developmental
Toxicity Risk Assessment, which states that a developmental effect "may result from exposure
prior to conception (either parent), during prenatal development, or postnatally to the time of
sexual maturation" and can be "detected at any point in the lifespan of the organism."
Biologically, a BMR of V2 SD is a reasonable choice as anti-tetanus antibody concentrations
prevent against tetanus, which is a rare, but severe and sometimes fatal infection, with a case-
fatality rate in the U.S. of 13% during 2001-2008 {CDC, 2011, 9998272}. The case-fatality rate
can be more than 80% for early lifestage cases {Patel, 1999, 10176842}. Selgrade (2007,
736210) suggests that specific immunotoxic effects observed in children may be broadly
indicative of developmental immunosuppression impacting these children's ability to protect
against a range of immune hazards—which has the potential to be a more adverse effect that just
a single immunotoxic effect. Thus, decrements in the ability to maintain effective levels of
tetanus antitoxins following immunization may be indicative of wider immunosuppression in
these children exposed to PFOS. By contrast, a BMR of 1 SD may be more appropriate for an
effect that would be considered 'minimally adverse.' A BMR smaller than V2 SD is generally
selected for severe effects (e.g., 1% extra risk of cancer mortality); decreased antibody
concentrations offer diminished protection from severe effects but are not themselves severe
effects.
Following the technical guidance {U.S. EPA, 2012, 1239433}, EPA derived BMDs and BMDLs
associated with both a 1 SD change in the distribution of log2(tetanus antibody concentrations)
and V2 SD change in the distribution of log2(tetanus antibody concentrations). The SD of the
log2(tetanus antibody concentrations) at age 7 years was estimated from the distributional data
presented in Grandjean et al. (2012, 1248827) as follows: the 25th and 75th percentiles of the
tetanus antibody concentrations at age 7 years in IU/mL were (0.65, 4.6). Log2-tranforming these
values provides the 25th and 75th percentiles in log2(IU/mL) as (-0.62, 2.20). Assuming that
these log2-transformed values are reasonably represented by a normal distribution, the IQR
(which is the difference between the 75th and 25th percentiles) is approximately 1.35 SDs
{Rosner, 2015, 10406286}. Thus, SD = IQR/1.35, and the SD of tetanus antibodies in
log2(IU/mL) is (2,20-(-0.62))/1.35 = 2.09 log2(IU/mL).
While there was not a clear definition of the size of an adverse effect for a continuous endpoint
like antibody concentrations, the value of 0.1 IU/mL is sometimes cited. As a check, EPA
evaluated how much extra risk would have been associated with a BMR set at a cutoff value of
0.1 IU/mL. Using the observed distribution of tetanus antibodies at age seven years in
log2(IU/mL), EPA calculated that 2.8% of those values would be below the cutoff value of
0.1 IU/mL (i.e., -3.32 log2(IU/mL)). A BMR of V2 SD resulted in 7.9% of the values being
below that cutoff which is 5.1% extra risk. This demonstrates the generic guidance that a BMR
of V2 SD can provide a reasonably good estimate of 5% extra risk. Figure E-2 shows an example
of this.
E-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
0.25 -T
Control distribution
Distribution for group with mean response 1/2 SD below control mean
LWY1 5.1% excess risk below control distribution percentile
0.20-
1/2 SD shift in mean [1.05 Log2(IU/ml)]
0.15
si
CD
-Q
O
0.10-
0.05-
0.00
Tetanus antibody concentrations in Log2(IU/ml)
Figure E-2. Difference in population tail probabilities resulting from a Vz standard
deviation shift in the mean from an estimation of the distribution of log2(tetanus antibody
concentrations at age seven years)
Table E-2. BMDs and BMDLs for effect of PFOS at age five years on anti-tetanus antibody
concentrations at age seven years {Budtz-Jorgensen, 2018, 5083631} using a BMR of Vz SD
change in log2(tetanus antibodies concentration) and a BMR of 1 SD change in log2(tetanus
antibodies concentration)
Estimated without control of PFOA
Estimated with control of PFOA
BMR BMD (ng/mL) BMDL (ng/mL) BMD (ng/mL) BMDL (ng/mL)
P = -0.274 per ng/mL 0lb = -0.0565 per ng/mL p = -0.0039 per ng/mL 0lb = -0.0365 per ng/mL
!/> SD 38.1 18.5a 268 28.6
1 SD 76.2 37.0 536 57.3
Notes:
a Denotes the selected POD.
The lowest serum PFOS concentration measured at age five years was 3.3 ng/mL, the 5th
percentile was 9.5 ng/mL, and the 10th percentile was 10.7 ng/mL {Grandjean, 2021, 9959716}
so the estimated BMDL for a BMR of V2 SD (BMDL', sd = 18.5 ng/mL) in the single-PFAS
model is well within the observed range. No information was available to judge the fit of the
E-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
model in the range of the BMDLs, but the BMD and BMDL were both within the range of
observed values.
The BMD>/2 sd estimate from the multi-PFAS models is 7-fold higher than the BMD', sd estimate
from the models with just PFOS, and the BMDLy2 sd estimates is 55% higher. The change in
BMD estimates may, or may not, reflect control for any potential confounding of the regression
effect estimates. While it is not clear which PFAS model provided the 'better' estimate of the
point estimate of the effect of PFOS considering potential confounding, the two BMDL', si)
estimates are 55% different (18.5 ng/mL vs. 28.6 ng/mL). EPA advanced the derivation based on
results that did not controls for PFOA because this model appeared to fit PFOS better (p = 0.12
vs. 0.84) and there was moderate uncertainty due to potential confounding in the BMDL.
However, confidence was diminished by the non-significant fit for PFOS (p = 0.12) and stronger
potential confounding in the main effect—even though there was moderate confounding of the
BMDL. Overall confidence in the BMDLs for Tetanus was judged to be low.
For immunotoxicity related to tetanus associated with PFOS exposure measured at age
five years, the POD is based on a BMR of Vi SD and a BMDLy2 sd of 18.5 ng/mL.
E. 1.1.1.2 Budtz-J0rgensen and Grandjeon (2018, 5083631) Results for
Decreased Tetanus Antibody Concentrations at Five Years of Age and PFOS
Exposure Measured Perinatally
Budtz-j0rgensen and Grandjean (2018, 5083631) fit multivariate models of PFOS measured
perinatally in maternal serum, against log2-transformed anti-tetanus antibody concentrations
measured at the five-year-old examination controlling for sex, and exact age at the five-year-old
examination, cohort, and interaction terms between cohort and sex, and between cohort and age.
Models were evaluated with additional control for PFOA (as log2[PFOA]), and without PFOA.
Three model shapes of PFOS were evaluated by Budtz-j0rgensen and Grandjean (2018,
5083631) using likelihood ratio tests: a linear model, a piecewise-linear model with a knot at the
median, and a logarithmic function. The logarithmic functions did not fit better than the
piecewise-linear functions Budtz-j0rgensen and Grandjean (2018, 5083631). Compared to the
linear model, the piecewise-linear model did not fit better than the linear model for either the
PFOS exposure without adjustment for PFOA using a likelihood ratio test (p = 0.43; see Budtz-
J0rgensen and Grandjean (2018, 5083631) Table 3), or for the model that did adjust for PFOA
(log2[PFOA]) (p = 0.98).
Table E-3 summarizes the results from Budtz-j0rgensen and Grandjean (2018, 5083631) for
tetanus in this exposure window. These regression coefficients (P) and their standard errors (SE)
were computed by EPA from the published BMDs and BMDL based on a BMR of 5% change in
tetanus antibody concentrations in Table 2 of Budtz-j0rgensen and Grandjean (2018, 5083631).
Table E-3. Results of the linear analyses of PFOS measured perinatally and tetanus
antibodies measured at age five years from Budtz-Jorgensen and Grandjean (2018,
7276745) in a single-PFAS model and in a multi-PFAS model
Exposure
Model shape
PFOA adjusted
Slope (P) per
ng/mL
SE(P)
ng/mL
Slope (P) fit
Lower bound
slope (Plb)
ng/mL
Perinatal PFOS
Linear
No
-0.0102
0.0095
p = 0.28
-0.0259
E-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Exposure
Model shape
PFOA adjusted
Slope (P) per
ng/mL
SE(P)
ng/mL
Slope (P) fit
Lower bound
slope (Plb)
ng/mL
Perinatal PFOS
Linear
Yes
0.0021
0.0107
p = 0.85
-0.0156
Notes: SE = standard error.
Interpretation of results in Table E-3:
• PFOS is a non-significant predictor in the single-PFAS model (P = -0.0102; p = 0.28).
• Effects are attenuated when log2[PFOA] are included in the model (P = 0.0021; p = 0.85).
• Nevertheless, these data can be used to estimate a BMDL for completeness and to allow
comparisons across PFAS.
Selection of the Benchmark Response
In the 2021 Proposed Approaches draft {U.S. EPA, 2021, 10428576} reviewed by the SAB
PFAS Review Panel, EPA relied on the BMDL modeling approach published in Budtz-
J0rgensen and Grandjean (2018, 5083631), described above. During validation of the modeling,
EPA reevaluated the approach chosen by Budtz-j0rgensen and Grandjean (2018, 5083631) and
determined that a different approach should be used to be consistent with EPA guidance {U.S.
EPA, 2012, 1239433}, which recommends the use of a 1 or V2 SD change in cases where there is
no accepted definition of an adverse level of change or clinical cut-off for the health outcome.
Additionally, consistent with EPA's Benchmark Dose Technical Guidance {U.S. EPA, 2012,
1239433}, EPA typically selects a 5% or 0.5 standard deviation (SD) benchmark response
(BMR) when performing dose response modeling of data from an endpoint resulting from
developmental exposure. Because Budtz-j0rgensen and Grandjean (2018, 5083631) assessed
antibody response after PFAS exposure during childhood, this is considered a developmental
study {U.S. EPA, 1991, 732120} based onEPA's Guidelines for Developmental Toxicity Risk
Assessment, which states that a developmental effect "may result from exposure prior to
conception (either parent), during prenatal development, or postnatally to the time of sexual
maturation" and can be "detected at any point in the lifespan of the organism."
Following the technical guidance {U.S. EPA, 2012, 1239433}, EPA derived BMDs and BMDLs
associated with a 1 SD change in the distribution of log2(tetanus antibody concentrations) and V2
SD change in the distribution of log2(tetanus antibody concentrations). The SD of the
log2(tetanus antibody concentrations) at age five years was estimated from two sets of
distributional data presented from two different cohorts of five-year-olds that were pooled in
Budtz-j0rgensen and Grandjean (2018, 5083631). Grandjean et al. (2012, 1248827) reported on
587 five-year-olds from the cohort of children born during 1997-2000 and Grandjean et al.
(2017, 4239492) reported on 349 five-year-olds from the cohort of children born during 2007-
2009. The means and SDs were computed separately by the authors. EPA then pooled the
summary statistics to describe the common SD. The 25th and 75th percentiles of the tetanus
antibody concentrations in the earlier birth cohort at age five years in IU/mL were (0.10, 0.51).
Log2-tranforming these values provides the 25th and 75th percentiles in log2(IU/mL) as
(-3.32, -0.97). Assuming that these log2-transformed values are similar to the normal
distribution, the IQR is approximately 1.35 SDs, thus SD = IQR/1.35, and the SD of tetanus
antibodies in log2(IU/mL) is (-0.97-(-3.32))/1.35 = 1.74 log2(IU/mL).
E-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The 25th and 75th percentiles of the tetanus antibody concentrations in the later birth cohort at
age five years in IU/mL was (0.1, 0.3). Log2-tranforming these values provides the 25th and 75th
percentiles in log2(IU/mL) as (-3.32, -1.74), and the SD of tetanus antibodies in log2(IU/mL) is
(—1.74—(—3,32))/l .35 = 1.17 log2(IU/mL). The pooled variance is a weighted sum of the
independent SDs, and the pooled SD was estimated as 1.55 log2(IU/mL).7 To show the impact of
the BMR on these results, Table E-4 presents the BMDs and BMDLs at BMRs of V2 SD and
1 SD.
Table E-4. BMDs and BMDLs for effect of PFOS measured perinatally and anti-tetanus
antibody concentrations at age five years {Budtz-Jorgensen, 2018, 5083631}
Estimated without control of PFOA
Estimated with control of 1
BMR
BMD (ng/mL)
BMDL (ng/mL)
BMD (ng/mL) 1
P = -0.0102 per ng/mL
Plb = -0.0259 per ng/mL
P = 0.00207 per ng/mL Plb =
V2SD
75.9
29.9a
_b
1 SD
151.8
59.8
-
Notes:
a Denotes the POD that corresponds to the analyses of PFOS concentrations perinatally and tetanus antibodies at age five years.
b Values cannot be determined.
The lowest perinatal maternal serum PFOS concentration measured was 9.4 ng/mL, the 5th
percentile was 17.1 ng/mL, and the 10th percentile was 19.1 ng/mL {Grandjean, 2021, 9959716}
so the estimated BMDLs for a BMR of V2 SD (BMDL', sd = 29.9 ng/mL) in the single-PFAS
model is well within the observed range. No information was available to judge the fit of the
model in the range of the BMDLs, but the BMD and BMDL were both within the range of
observed values. The BMDLi/2 sd estimate from the single-PFAS models was 29.9 ng/mL. The
BMDL estimates from the multi-PFAS models were about 67% higher than for the single-PFAS
model.
There is low confidence in the BMDLs from the PFOS-only model (29.9 ng/mL) and in the
multi-PFAS model (49.8 ng/mL). Confidence is diminished by the low quality of the model fit
for PFOS in either model compared to the PFOS results from tetanus in the 5-year to 7-year
exposure-outcome window of time and there is some uncertainty regarding potential
confounding.
For immunotoxicity related to tetanus, associated with PFOS measured perinatally, the POD is
based on a BMR of V2 SD and a BMDL , sd of 29.9 ng/mL. Note that this result is based on a
poorly fit PFOS regression parameter (P) estimated as -0.0102 per ng/mL (90% CI: -0.0259,
0.0055; p = 0.28) (Budtz-j0rgensen and Grandjean, 2018, 7276745), and thus this POD is
identified with low confidence.
For immunotoxicity related to tetanus associated with PFOS exposure measured at age
five years, the POD estimated for comparison purposes were based on a BMR of Vz SD and
a BMDLy2 sd of 29.9 ng/mL.
7 Pooled variance for tetanus in five-year-olds = [(502-l)(1.74)A2+(298-l)(1.17)A2]/[502+298-2] = 2.41. The pooled SD is the
square root of 2.41 which is 1.55 log2(IU/mL).
E-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E. 1.1.1.3 Timmermon et al. (2021, 9416315)
Timmerman et al. (2021, 9416315) analyzed data from Greenlandic children ages 7-12 and fit
multivariate models of PFOS and logio-transformed anti-tetanus antibody concentrations
measured at the same time as PFOS, controlling for time since vaccine booster/estimated time
since vaccine booster, and duration of being breastfed (< 6 months, 6-12 months, > lyear) and
area of residence (Nuuk, Maniitsoq, Sisimiut, Ilulissat, Aasiaat, Qeqertarsuaq, Tasiilaq) and
including children with known tetanus-diphtheria booster date only. Estimates from the linear
regression models were subsequently back-transformed to express the percent difference in
antibody concentrations at each ng/mL increase in serum PFOS concentrations in children,
which was -3 (95% CI: -8, 3) (Table 4, Timmerman et al. (2021, 9416315)). Using the equation
provided below, EPA estimated the regression slope as -0.013 (95% CI: -0.036, 0.013).
Percent Difference = (10^ — 1) x 100
Following the approach described previously for Budtz-j0rgensen and Grandjean (2018,
5083631), EPA derived BMDs and BMDLs were derived for both a one SD change in the
distribution of logio (tetanus antibody concentrations) as a standard reporting level, and '/2 SD
change in the distribution of logio (tetanus antibody concentrations). The SD of the logio (tetanus
antibody concentrations) was estimated from the median (25th, 75th percentiles) of 0.92 (0.25,
2.20) tetanus antibody concentrations in IU/mL (Table 1 in Timmerman et al. (2021, 9416315)).
Logio -transforming these values results in 25th and 75th percentiles in logio (IU/mL) as -0.60
and 0.34, respectively. Assuming that these logio -transformed values are reasonably represented
by a normal distribution, the IQR is approximately 1.35 SDs. Thus, SD = IQR/1.35, and the SD
of tetanus antibodies in logio (IU/mL) is (0.34 - (—0.60))/l .35 = 0.70 logio (IU/mL).
Table E-5. BMDs and BMDLs for effect of PFOS on anti-tetanus antibody concentrations
{Timmerman, 2021, 9416315} using a BMR of Vz SD change in logio (tetanus antibodies
concentration) and a BMR of 1 SD change in logio (tetanus antibodies concentration).
BMR
BMD (ng/mL)
BMDL (ng/mL)
P = -0.013 per ng/mL
P = -0.036 per ng/mL
!/2SD
26.4
9.66
1 SD
52.9
19.3
Notes: SD = standard deviation.
As a check, EPA evaluated how much extra risk would have been associated with a BMR set at a
cutoff value of 0.1 IU/mL. Using the observed distribution of tetanus antibodies in logio (IU/mL),
a BMR of V2 SD resulted in 10.6% extra risk. This suggests that, in this case, a BMR of V2 SD
may not be a reasonably good estimate of 5% extra risk.
Note that this BMDL is based on a poorly fit PFOS regression parameter (P) estimated as -0.013
(95% CI: -0.036, 0.013) (Timmerman, 2021, 9416315), and thus this POD is identified with low
confidence.
For immunotoxicity related to tetanus associated with PFOS exposure measured at ages
five to ten years old, the POD estimated for comparison purposes was based on a BMR of
V2 SD and a BMDLy2 sd of 9.7 ng/mL.
E-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E. 1.1.1.4 Summary of Modeling Results for Decreased Tetanus Antibody
Concentrations
Table E-6 summarizes the PODs resulting from the modeling approaches for decreased tetanus
antibody concentrations. The selected and comparison PODs were based on a BMR of V2 SD,
resulting in BMDLs ranging from 9.7 to 29.9, with the selected POD of 18.5 also representing
the median of the BMDLs. The comparisons PODs are considered low confidence because they
are based on a poorly fit PFOS regression parameters.
Table E-6. BMDLs for effect of PFOS on anti-tetanus antibody concentrations using a
BMR of1/! SD {Timmerman, 2021, 9416315}
Study
Effect
BMDLy, sd (ng/mL)
y2SD
Budtz-Jorgcnscn
and Grandjean
(2018, 5083631)
Budtz-Jorgcnscn
and Grandjean
(2018, 5083631)
Timmerman et al.
(2021, 9416315)
PFOS at age five years and anti-tetanus antibody 18.5
concentrations at age seven years
PFOS perinatally and anti-tetanus antibody 29.9
concentrations at age seven years
PFOS and anti-tetanus antibody concentrations at 9.66
ages seven-10 years
1.05 log2(IU/mL)
0.78 log2(IU/mL)
0.35 logio(IU/mL)
E. 1.1.2 Modeling Results for Decreased Diphtheria Antibody
Concentrations
E. 1.1.2.1 Budtz-J0rgensen and Grandjean (2018, 5083631) Results for
Decreased Diphtheria Antibody Concentrations at Seven Years of Age and PFOS
Exposure Measured at Five Years of Age
Budtz-j0rgensen and Grandjean (2018, 5083631) fit multivariate models of PFOS measured at
age five years, against log2-transformed anti-diphtheria antibody concentrations measured at the
seven-year-old examination controlling for sex, exact age at the seven-year-old examination, and
booster type at age five years. Models were evaluated with additional control for PFOA (as
log2[PFOA]), and without PFOA. Three model shapes were evaluated by Budtz-j0rgensen and
Grandjean (2018, 5083631) using likelihood ratio tests: a linear model of PFOS, a piecewise-
linear model with a knot at the median, and a logarithmic function. The logarithmic functions did
not fit better than the piecewise-linear functions {Budtz-j0rgensen, 2018, 5083631}. The
piecewise-linear model did not fit better than the linear model for the PFOS exposure without
adjustment for PFOA using a likelihood ratio test (p = 0.30; see Budtz-j0rgensen and Grandjean
(2018, 5083631) Table 3), or for the model that did adjust for PFOA (log2[PFOA]) (p = 0.34).
Table E-7 summarizes the results from Budtz-j0rgensen and Grandjean (2018, 5083631) for
diphtheria in this exposure window. These P and their SE were computed by EPA from the
published BMDs and BMDL based on a BMR of 5% decrease in diphtheria antibody
concentrations in Table 1 of Budtz-j0rgensen and Grandjean (2018, 5083631).6
Table E-7. Results specific to the slope from the linear analyses of PFOS measured at age
five years and log2(diphtheria antibodies) measured at age seven years from Table 1 in
E-10
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Budtz-Jergensen and Grandjean (2018, 5083631) in a single-PFAS model and in a multi-
PFAS model
Exposure
Model shape
PFOA adjusted
Slope (P) per
ng/mL
SE(P)
ng/mL
Slope (P) fit
Lower bound
slope (Plb)
ng/mL
PFOS at Age 5
Linear
No
-0.0322
0.0163
p = 0.05
-0.0591
PFOS at Age 5
Linear
Yes
-0.0207
0.0184
p = 0.26
-0.0510
Notes: SE = standard error.
Interpretation of results in Table E-7:
• PFOS is a significant predictor in the single-PFAS model (P = -0.0322; p = 0.05).
• Effects are attenuated when log2[PFOA] are included in the model (P = -0.0207;
p = 0.26).
• The point estimate results for PFOS are potentially confounded by PFOA since there was
a 36% reduction in the effect size for PFOS from -0.0322 to -0.0207 when controlling for
PFOA.
• One explanation is that PFOA was a confounder of the PFOS effect.
• Another possibility is physiological confounding which can arise when biomarkers
measured from the same blood test are more highly correlated due to individual's
physiological processes. Physiological confounding can therefore induce confounding bias
by the inclusion of co-measured co-exposures in regression models.
• The reasons for the change in main effect size are not known and remain an uncertainty
because it is not known whether the change in estimate was induced by physiologic
confounding or was the result of controlling for classical confounding. For this reason,
there is uncertainty in knowing which estimate is the best representation of any effect of
PFOS.
• The uncertainty from potential confounding does not have much impact on the RfD which
is defined as allowing for an order of magnitude (10-fold or 1,000%) uncertainty in the
estimate. This is because there is only 36% difference in the BMD and 16% difference in
the BMDL when PFOS is included in the model.
Selection of the Benchmark Response
In the 2021 Proposed Approaches draft {U.S. EPA, 2021, 10428576} reviewed by the SAB
PFAS Review Panel, EPA relied on the BMDL modeling approach published in Budtz-
J0rgensen and Grandjean (2018, 5083631), described above. During validation of the modeling,
EPA reevaluated the approach chosen by Budtz-j0rgensen and Grandjean (2018, 5083631) and
determined that a different approach should be used to be consistent with EPA guidance {U.S.
EPA, 2012, 1239433}, which recommends the use of a 1 or V2 SD change in cases where there is
no accepted definition of an adverse level of change or clinical cut-off for the health outcome.
Additionally, consistent with EPA's Benchmark Dose Technical Guidance {U.S. EPA, 2012,
1239433}, EPA typically selects a 5% or 0.5 standard deviation (SD) benchmark response
(BMR) when performing dose response modeling of data from an endpoint resulting from
developmental exposure. Because Budtz-j0rgensen and Grandjean (2018, 5083631) assessed
antibody response after PFAS exposure during childhood, this is considered a developmental
study {U.S. EPA, 1991, 732120} based onEPA's Guidelines for Developmental Toxicity Risk
E-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Assessment, which states that a developmental effect "may result from exposure prior to
conception (either parent), during prenatal development, or postnatally to the time of sexual
maturation" and can be "detected at any point in the lifespan of the organism."
Following the technical guidance {U.S. EPA, 2012. 1239433}, EPA derived BMDs and BMDLs
associated with a one SD change in the distribution of log2(diphtheria antibody concentrations),
and '/2 SD change in the distribution of log2(diphtheria antibody concentrations). A blood
concentration for diphtheria antibodies of 0.1 IU/mL is sometimes cited in the diphtheria
literature as a 'protective level' (Grandjean et al. (2017, 4239492) noted that the Danish vaccine
producer Statens Serum Institut recommended the 0.1 IU/mL 'cutoff level; and Galazka et al
(1993, 10228565) mentions the same concentration, but also argues:
"However¦, it has also been shown that there is no sharply defined level of antitoxin that
gives complete protection from diphtheria {Ipsen, 1946,10228563}. A certain range of
variation must be accepted; the same degree of antitoxin may give an unequal degree
of protection in different persons. Other factors may influence the vulnerability to
diphtheria including the dose and virulence of the diphtheria bacilli and the general
immune status of the person infected {Christenson, 1986, 9978484}. Thus, an antibody
concentration between 0.01 and 0.09 IU/mL may be regarded as giving basic
immunity, whereas a higher titer may be needed for full protection. In some studies
that used in vitro techniques, a level of 0.1 IU/mL was considered protective {Cellesi,
1989, 9642154; Galazka, 1989, 9642152}."
Statistically, the Technical Guidance suggests that studies of developmental effects can support
lower BMRs. Biologically, a BMR of '/2 SD is a reasonable choice as anti-diphtheria antibody
concentrations prevent against diphtheria, which is very rare in the U.S., but can cause life-
threatening airway obstruction, or cardiac failure {Collier, 1975, 9642066}. Among 13 cases
reported in the U.S. during 1996-2016, no deaths were mentioned {Liang, 2018, 9978483}.
However, diphtheria remains a potentially fatal disease in other parts of the world (Galazka {,
1993, 10228565} mentions a case fatality rate of 5%—10%) and PFOS-related changes in anti-
diphtheria antibody concentrations cannot be considered 'minimally adverse' given the historic
lethality of diphtheria in the absence of vaccination. Selgrade (2007, 736210) suggests that
specific immuno-toxic effects observed in children may be broadly indicative of developmental
immunosuppression impacting these children's ability to protect against a range of immune
hazards—which has the potential to be a more adverse effect that just a single immuno-toxic
effect.
Following the technical guidance {U.S. EPA, 2012, 1239433}, EPA derived BMDs and BMDLs
associated with a one SD change in the distribution of log2(diphtheria antibody concentrations)
as a standard reporting level, and V2 SD change in the distribution of log2(diphtheria antibody
concentrations). The SD of the log2(diphtheria antibody concentrations) at age 7 years was
estimated from the distributional data presented in Grandjean et al. (2012, 1248827) as follows:
the 25th and 75th percentiles of the diphtheria antibody concentrations at age 7 years in IU/mL
were (0.4, 1.6). Log2-tranforming these values provides the 25th and 75th percentiles in
log2(IU/mL) as (-1.32, 0.68). Assuming that these log2-transformed values are similar to the
normal distribution, the IQR is approximately 1.35 SDs, thus SD = IQR/1.35, and the SD of
tetanus antibodies in log2(IU/mL) is (0.68—(—1.32))/l .35 = 1.48 log2(IU/mL). To show the
E-12
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
impact of the BMR on these results, Table E-8 presents the BMDs and BMDLs at BMRs of V2
SD and 1 SD.
Table E-8. BMDs and BMDLs for effect of PFOS at age five years on anti-diphtheria
antibody concentrations at age seven years {Budtz-Jorgensen, 2018, 5083631} using a BMR
of Vz SD change in log2(diphtheria antibodies concentration) and a BMR of 1 SD
log2(diphtheria antibodies concentration)
Estimated without control of PFOA
Estimated with control of PFOA
BMR
BMD (ng/mL) BMDL (ng/mL)
P = -0.0322 per ng/mL 0lb = -0.0592 per ng/mL
BMD (ng/mL) BMDL (ng/mL)
P = -0.0207 per ng/mL 0lb = -0.0510 per ng/mL
V2SD
1 SD
23.0 12.5a
46.0 25.0
35.8 14.5
71.7 29.0
Notes:
a Denotes the selected POD.
The lowest serum PFOS concentration measured at age five years was 3.3 ng/mL, the 5th
percentile was 9.5 ng/mL, and the 10th percentile was 10.7 ng/mL {Grandjean, 2021, 9959716}
so the estimated BMDL for a BMR of V2 SD (BMDLi/2 sd = 12.5 ng/mL) in the single-PFAS
model is well within the observed range. No information was available to judge the fit of the
model in the range of the BMDLs, but the BMD and BMDL were both within the range of
observed values and the model fit PFOS well (p = 0.05).
The BMD>, sd estimate from the multi-PFAS models is 56% higher than the BMD', sd estimate
from the model with just PFOS, and the BMDLusnis 16% higher. This may, or may not, reflect
control for any potential confounding of the regression effect estimates. While it is not clear
which PFAS model provided the 'better' estimate of the point estimate of the effect of PFOS in
light of potential confounding, the two BMDL , sd estimates which serve as the PODs are
comparable (12.5 ng/mL vs. 14.5 ng/mL). EPA advanced POD based on results that did not
controls for PFOA because this model appeared to fit PFOS data better (p = 0.05 vs. 0.26) and
there was low uncertainty due to potential confounding in the BMDL. However, confidence was
diminished by the potential confounding in the main effect—even though there was low
confounding of the BMDL, and overall confidence in the BMDLs for diphtheria was judged to
be medium.
For immunotoxicity related to diphtheria, associated with PFOS measured at age
five years, the POD is based on a BMR of V2 SD and a BMDLy2 sd of 12.5 ng/mL.
E. 1.1.2.2 Budtz-J0rgensen and Grandjean (2018, 5083631) Results for
Decreased Diphtheria Antibody Concentrations at Five Years of Age and PFOS
Exposure Measured Perinatally
Budtz-j0rgensen and Grandjean (2018, 5083631) fit multivariate models of PFOS measured
perinatally, against log2-transformed anti-diphtheria antibody concentrations measured at the
five-year-old examination controlling for sex and age. Models were evaluated with additional
control for PFOA (as log2[PFOA]), and without PFOA. Three model shapes were evaluated by
Budtz-j0rgensen and Grandjean (2018, 5083631) using likelihood ratio tests: a linear model of
PFOS, a piecewise-linear model with a knot at the median, and a logarithmic function. The
E-13
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
logarithmic functions did not fit better than the piecewise-linear functions Budtz-j0rgensen and
Grandjean (2018, 5083631). Compared to the linear model, the piecewise-linear model did not fit
better than the linear model for either the PFOS exposure without adjustment for PFOA using a
likelihood ratio test (p = 0.55; see Budtz-j0rgensen and Grandjean (2018, 5083631) Table 3), or
for the model that did adjust for PFOA (log2[PFOA]) (p = 0.84). Table E-9 summarizes the
results from Budtz-j0rgensen and Grandjean (2018, 5083631) for diphtheria in this exposure
window. These P and their SE were computed by EPA from the published BMDs and BMDL
based on a BMR of 5% change in diphtheria antibody concentrations in Table 2 of Budtz-
J0rgensen and Grandjean (2018, 5083631).6
Table E-9. Results of the linear analyses of PFOS measured perinatally and diphtheria
antibodies measured at age five years from Budtz-Jorgensen and Grandjean (2018,
7276745) in a single-PFAS model and in a multi-PFAS model
Exposure
Model shape
PFOA adjusted
Slope (P) per
ng/mL
SE(P)
Slope (P) fit
Lower bound
slope (Plb)
Perinatal PFOS
Linear
No
-0.0310
0.0100
p = 0.002
0.0475
Perinatal PFOS
Linear
Yes
-0.0241
0.0113
p = 0.033
0.0427
Notes: SE = standard error.
Interpretation of results in Table E-9:
• PFOS is a significant predictor in the single-PFAS model (P = -0.0310; p = 0.002).
• Effects of PFOS are attenuated when PFOA is in the model (P = -0.0241; p = 0.033).
• Results for PFOS are potentially confounded by PFOA since there was a 22% change in
the effect size for PFOS from -0.0310 to -0.0241 when controlling for PFOA.
• One explanation is that PFOA was a confounder of the PFOS effect.
• Another possibility is physiological confounding which can arise when biomarkers
measured from the same blood test are more highly correlated due to individual's
physiological processes. Physiological confounding can therefore induce confounding bias
by the inclusion of co-measured co-exposures in regression models.
• The reasons for the change in main effect size are not known and remain an uncertainty
because it is not known whether the change in estimate was induced by physiologic
confounding or was the result of controlling for classical confounding. For this reason,
there is uncertainty in knowing which estimate is the best representation of any effect of
PFOS.
• The uncertainty from potential confounding does not have much impact on the RfD which
is defined as allowing for an order of magnitude (10-fold or 1,000%) uncertainty in the
estimate. This is because there is only 22% difference in the BMD and 11% difference in
the BMDL when PFOS is included in the model.
Selection of the Benchmark Response
In the 2021 Proposed Approaches draft {U.S. EPA, 2021, 10428576} reviewed by the SAB
PFAS Review Panel, EPA relied on the BMDL modeling approach published in Budtz-
J0rgensen and Grandjean (2018, 5083631), described above. During validation of the modeling,
EPA reevaluated the approach chosen by Budtz-j0rgensen and Grandjean (2018, 5083631) and
determined that a different approach should be used to be consistent with EPA guidance {U.S.
EPA, 2012, 1239433}, which recommends the use of a 1 or V2 SD change in cases where there is
E-14
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
no accepted definition of an adverse level of change or clinical cut-off for the health outcome.
Additionally, consistent with EPA's Benchmark Dose Technical Guidance {U.S. EPA, 2012,
1239433}, EPA typically selects a 5% or 0.5 standard deviation (SD) benchmark response
(BMR) when performing dose response modeling of data from an endpoint resulting from
developmental exposure. Because Budtz-j0rgensen and Grandjean (2018, 5083631) assessed
antibody response after PFAS exposure during childhood, this is considered a developmental
study {U.S. EPA, 1991, 732120} based onEPA's Guidelines for Developmental Toxicity Risk
Assessment, which states that a developmental effect "may result from exposure prior to
conception (either parent), during prenatal development, or postnatally to the time of sexual
maturation" and can be "detected at any point in the lifespan of the organism."
Following the technical guidance {U.S. EPA, 2012, 1239433}, EPA derived BMDs and BMDLs
associated with a one SD change in the distribution of log2(tetanus antibody concentrations) as a
standard reporting level, and V2 SD change in the distribution of log2(tetanus antibody
concentrations). The SD of the log2(diphtheria antibody concentrations) at age five years was
estimated from two sets of distributional data presented from two different birth cohorts of five-
year-olds that were pooled in Budtz-j0rgensen and Grandjean (2018, 5083631). Grandjean et al.
(2012, 1248827) reported on 587 five-year-olds from the cohort of children born during 1997-
2000 and Grandjean et al. (2017, 4239492) reported on 349 five-year-olds from the cohort of
children born during 2007-2009. The means and SDs were computed separately by the author.
EPA then pooled the summary statistics to describe the common SD. The 25th and 75th
percentiles of the diphtheria antibody concentrations in the earlier birth cohort at age five years
in IU/mL were (0.05, 0.4). Log2-tranforming these values provides the 25th and 75th percentiles
in log2(IU/mL) as (-4.32, -1.32). Assuming that these log2-transformed values are similar to the
normal distribution, the IQR is approximately 1.35 SDs, thus SD = IQR/1.35, and the SD of
diphtheria antibodies in log2(IU/mL) is (—1.32—(—4.32))/l.35 = 2.22 log2(IU/mL).
The 25th and 75th percentiles of the diphtheria antibody concentrations in the later birth cohort
at age five years in IU/mL were (0.1, 0.3). Log2-tranforming these values provides the 25th and
75th percentiles in log2(IU/mL) as (-3.32, -1.74), and the SD of diphtheria antibodies in
log2(IU/mL) is (— 1.74—(—3.32))/1.35 = 1.17 log2(IU/mL). The pooled variance is a weighted sum
of the independent SDs, and the pooled SD was estimated as 1.90 log2(IU/mL).8 To show the
impact of the BMR on these results, Table E-10 presents the BMDs and BMDLs at BMRs of V2
SD and 1 SD.
Table E-10. BMDs and BMDLs for effect of PFOS measured perinatally and anti-
diphtheria antibody concentrations at age five years {Budtz-Jorgensen, 2018, 5083631}
Estimated without control of PFOA
Estimated with control of PFOA
BMR
BMD (ng/mL)
P = -0.031 per ng/mL
BMDL (ng/mL)
Plb = -0.0475 per ng/mL
BMD (ng/mL) BMDL (ng/mL)
P = -0.0241 per ng/mL Plb = -0.0427per ng/mL
V2SD
1 SD
30.6
61.3
20.0a
40.0
39.4 22.3
78.9 44.5
Notes:
8 Pooled variance for diphtheria in five-year-olds = [(502-l)(2.22)A2+(298-l)(1.17)A2]/[502+298-2] = 3.60. The pooled SD is
the square root of 3.60 which is 1.90 log2(IU/mL).
E-15
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
a Denotes the selected POD.
The lowest serum PFOS concentration measured perinatally was 9.4 ng/mL, the 5th percentile
was 17.1 ng/mL, and the 10th percentile was 19.1 ng/mL {Grandjean, 2021, 9959716} so the
estimated BMD for a BMR of '/2 SD (BMDLi/2 sd = 20.0 ng/mL) in the single-PFAS model is
well within the observed range. No information was available to judge the fit of the model in the
range of the BMDLs, but the BMD and BMDL were both within the range of observed values
and the model fit PFOS well (p = 0.002).
The BMD>, sd estimate from the multi-PFAS models is 29% higher than the BMD', sd estimated
from the model with just PFOS, and the BMDLusnis 12% higher. This may, or may not, reflect
control for any potential confounding of the regression effect estimates. The BMDLs which
serve as the PODs are comparable (20.0 ng/mL vs. 22.3 ng/mL) and EPA advanced the
derivation based on results that did not control for PFOA because this model appeared to fit
PFOS well (p = 0.002 vs. 0.031) and there was low uncertainty due to potential confounding in
the BMD and moderate uncertainty in the BMDL. Medium confidence in the BMDLs from
PFOS linear model (20.0 ng/mL) with control of PFOA since the model fit reasonably well and
these BMDLs show low uncertainty about confounding.
For immunotoxicity related to diphtheria, associated with PFOS measured at age
five years, the POD is based on a BMR of V2 SD and a BMDLy2 sd of 20.0 ng/mL.
E. 1.1.2.3 Timmerman et al. (2021, 9416315)
Timmerman et al. (2021, 9416315) analyzed data from Greenlandic children ages 7-12 and fit
multivariate models of PFOS against logio-transformed anti-diphtheria antibody concentrations
measured at the same time as PFOS, controlling for time since vaccine booster/estimated time
since vaccine booster, and duration of being breastfed (< 6 months, 6-12 months, > 1 year) and
area of residence (Nuuk, Maniitsoq, Sisimiut, Ilulissat, Aasiaat, Qeqertarsuaq, Tasiilaq) and
including children with known tetanus-diphtheria booster date only. Estimates from the linear
regression models were subsequently back-transformed to express the percent difference in
antibody concentrations at each ng/mL increase in serum PFOS concentrations in children,
which was -9 (95% CI: -16, 2) (Table 4, Timmerman et al. (2021, 9416315)). Using the
equation provided below, EPA estimated the regression slope as -0.04 (95% CI: -0.08, 0.01).
Percent Difference = (10^ — 1) x 100
Following the description provided for Budtz-j0rgensen and Grandjean (2018, 5083631), EPA
derived BMDs and BMDLs for both a one SD change in the distribution of logio (diphtheria
antibody concentrations) as a standard reporting level, and V2 SD change in the distribution of
logio (diphtheria antibody concentrations). The SD of the logio (diphtheria antibody
concentrations) was estimated from the distributional data presented in Table 1 as follows: the
25th and 75th percentiles of the diphtheria antibody concentrations in IU/mL were 0.02 and 0.28,
respectively. Logio -transforming these values provides the 25th and 75th percentiles in logio
(IU/mL) as (-1.7, -0.55). Assuming that these logio -transformed values are reasonably
represented by a normal distribution, the IQR is approximately 1.35 SDs. Thus, SD = IQR/1.35,
and the SD of tetanus antibodies in logio(IU/mL) is 0.85 logio(IU/mL).
E-16
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-ll. BMDs and BMDLs for effect of PFOS on anti-diphtheria antibody
concentrations {Timmerman, 2021, 9416315} using a BMR of Vz SD change in logio (tetanus
antibodies concentration) and a BMR of 1 SD change in logio (tetanus antibodies
concentration).
BMR
BMD (ng/mL)
BMDL (ng/mL)
P = -0.11 per ng/mL
P = -0.28 per ng/mL
!/2SD
10.4
5.61
1 SD
20.7
11.2
Notes: SD = standard deviation.
As a check, EPA evaluated how much extra risk would have been associated with a BMR set at a
cutoff value of 0.1 IU/mL. Using the observed distribution of diphtheria antibodies in
logio(IU/mL), EPA calculated that 57% of those values would be below the cutoff value of
0.1 IU/mL. A BMR of '/2 SD resulted in 75% of the values being below that cutoff which is 18%
extra risk. This suggest that in this case a BMR of V2 SD may not be a reasonably good estimate
of 5% extra risk.
Note that this result is based on a poorly fit PFOS regression parameter (P) estimated as -0.04
(95% CI: -0.08, 0.01) (Timmerman, 2021, 9416315), and thus this POD is identified with low
confidence.
For immunotoxicity related to tetanus associated with PFOS exposure measured at ages
five to ten years old, the POD estimated for comparison purposes were based on a BMR of
V2 SD and a BMDLy2 sd of 5.6 ng/mL.
E. 1.1.2.4 Summary of Modeling Results for Decreased Diphtheria Antibody
Concentrations
Table E-12 summarizes the PODs resulting from the modeling approaches for decreased tetanus
antibody concentrations. The selected and comparison PODs were based on a BMR of V2 SD,
resulting in BMDLs ranging from 5.6 ng/mL to 20.0 ng/mL with the selected POD of 12.5 also
representing the median of the BMDLs. The comparison PODs are considered low confidence
because they are based on a poorly fit PFOS regression parameters.
Table E-12. BMDLs for effect of PFOS on anti-diphtheria antibody concentrations using a
BMR of1/! SD {Timmerman, 2021, 9416315}
Study name
Effect
BMDL (ng/mL)
ViSD
Budtz-Jorgcnscn and
Grandjean (2018, 5083631)
PFOS at age five years on anti-diphtheria
antibody concentrations at age seven years
12.5
0.74 log2(IU/mL)
Budtz-Jorgcnscn and
Grandjean (2018, 5083631)
PFOS perinatally on anti-diphtheria antibody
concentrations at age seven years
20.0
0.95 log2(IU/mL)
Timmerman et al. (2021,
9416315)
PFOS and anti-diphtheria antibody
concentrations at ages 7-10 years
5.6
0.48 logio(IU/mL)
E-17
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.1.1.3 Modeling Results for Decreased Rubella Antibody
Concentrations
Granum et al (2013, 1937228) investigated the effects of prenatal exposure to perfluorinated
compounds on vaccination responses and clinical health outcomes in early childhood in a
subcohort of the Norwegian Mother and Child Cohort Study. A total of 56 mother-child pairs,
for whom both maternal blood samples at delivery and blood samples from the children at
3 years of age, were evaluated. Antibody titers specific to measles, rubella, tetanus, and influenza
were measured. Rubella antibody levels were inversely associated with maternal PFOS
(mean = 5.6 ng/mL), but not with any other outcomes.
EPA considered applying a similar approach to those described above for decreased tetanus
antibody concentrations in Budtz-j0rgensen and Grandjean (2018, 5083631) and Timmerman et
al. (2021, 9416315) to estimate the BMD/BMDL associated with decreased rubella antibody
concentrations in Granum et al. (2013, 1937228). Granum et al (2013, 1937228) reported a
regression coefficient and 95% confidence interval from multivariate regression analyses of
maternal PFOS and anti-Rubella antibody levels (-0.08, 95% CI: -0.14, -0.02). Granum et al
(2013, 1937228) also reported summary statistics of rubella antibodies levels at the age of 3
(median = 1.9; 25th, 75th percentiles: 1.5,2.1). Upon investigation of the extra risk using the
distributional data and a cutoff value of 0.1 IU/mL it was determined that this data did not allow
for application of a BMR of 1 and '/2 SD to provide a reasonably good estimate of 10% and 5%
extra risk. The Benchmark Dose Technical Guidance {U.S. EPA, 2012, 1239433} explains that
in a control population where 1.4% are considered to be at risk of having an adverse effect, a
downward shift in the control mean of one SD results in about 10% extra risk of being at risk of
having an adverse effect. The cut off value of 0.1 IU/mL resulted in 0.003% of the control
population at risk of having an adverse effect, a value much smaller than 1.4% which in turn did
not result in 10% extra risk.
E.1.2 Modeling Results for Decreased Birthweight
Six high confidence studies {Chu, 2020, 6315711; Sagiv, 2018, 4238410; Starling, 2017,
3858473; Wikstrom, 2020, 6311677; Darrow, 2013, 2850966; Yao, 2021, 9960202} reported
decreased birth weight in infants whose mothers were exposed to PFOS. These candidate studies
offer a variety of PFOS exposure measures across the fetal and neonatal window. All six studies
reported their exposure metric in units of ng/mL and reported the P coefficients per ng/mL or
ln(ng/mL), along with 95% confidence intervals, estimated from linear regression models. The
logarithmic transformation of exposure yields a negative value for small numbers, which can
result in implausible results from dose-response modeling (i.e., estimated risks are negative and
unable to determine the responses at zero exposure). EPA first re-expressed the reported /?
coefficients in terms of per ng/mL, if necessary, according to Dzierlenga et al. (2020, 7643488).
Then EPA used the re-expressed P and lower limit on the confidence interval to estimate BMD
and BMDL values using the general equation y = mx + b, where y is birth weight and x is
exposure, substituting the re-expressed P values from these studies for m. The intercept b
represents the baseline value of birth weight in an unexposed population and it can be estimated
through y = mx + b using an average birth weight from an external population as y, an average
exposure as x and re-expressed P from the studies as m.
E-18
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The CDC Wonder site (https://wonder.cdc.eov/natality.html) provides vital statistics for babies
born in the United States. There were 3,791,712 all live births in the United States in 2018
according to final natality data. The mean and standard deviation of birth weight were
3261.6 ± 590.7 g (7.19 ± 1.30 lb), with 8.27% of live births falling below the public health
definition of low birth weight (i.e., 2500 g, or 5.5 lb). The full natality data for the United States
data on birth weight was used as it is more relevant for deriving toxicity values for the US
general public than the study-specific birthweight data. Also, the CDC Wonder database may be
queried to find the exact percentage of the population falling below the cut-off value for clinical
adversity. America's Children and the Environment (ACE) Biomonitoring on Perfluorochemicals
(https://www.epa.gov/americaschildrenenvironment/data-tables-biomonitoring-
perfluorochemicals-pfcs) provides in Table B6b the median blood serum levels of PFOS of
2.6 ng/mL in 2015-2016 in woman ages 16 to 49, using NHANES as data source. These values
are assumed to be representative of women of reproductive age and are subsequently used in the
estimation of BMD and BMDL values from the available four epidemiological studies.
E. 1.2.1 Chu et al. (2020, 6315711)
Chu et al. (2020, 6315711) reported a P coefficient of-83.3 g (95%CI: -133.2, -33.4) per
ln(ng/mL) increase for the association between birth weight and maternal PFOS serum
concentrations (collected within 3 days of delivery) in a China cohort. The reported P coefficient
can be re-expressed in terms of per ng/mL according to Dzierlenga et al. (2020, 7643488). Given
the reported study-specific median (7.2 ng/mL) and the 25th and 75th percentiles (4.4 and
11.9 ng/mL) of the exposure from Chu et al. (2020, 6315711), EPA estimated the distribution of
exposure by assuming the exposure follows a log-normal distribution with mean and standard
deviation as:
PL = ln(q50) = ln{72) = 1.97 (1)
o- = ln(q75/q25)/1349 = Zn(11.9/4.4)/1.349 = 0.75 (2)
Then, EPA estimated the 25th-75th percentiles at 10 percentile intervals of the exposure
distribution and corresponding responses of reported P coefficient. The re-expressed P
coefficient is determined by minimizing the sum of squared differences between the curves
generated by the re-expressed P and the reported p. Doing so results in a re-expressed P
coefficient of-11.0 g (95% CI: -17.6, -4.4) per ng/mL.
Typically, for continuous data, the preferred definition of the BMR is to have a basis for what
constitutes a minimal level of change in the endpoint that is biologically significant. For birth
weight, there is no accepted percent change that is considered adverse. However, there is a
clinical measure for what constitutes an adverse response. Babies born weighing less than 2500 g
are considered to have low birth weight, and further, low birth weight is associated with a wide
range of health conditions throughout life {Hack, 1995, 8632216; Reyes, 2005, 1065677; Tian ,
2019, 8632212}. Given this clinical cut-off for adversity and that 8.27% of all live births in the
US in 2018 fell below this cut-off, the hybrid approach can be used to define the BMR. The
hybrid approach harmonizes the definition of the BMR for continuous data with that for
E-19
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
dichotomous data, and therefore is an advantageous approach9. Essentially, the hybrid approach
involves the estimation of the dose that increases the percentile of responses falling below (or
above) some cut-off for adversity in the tail of the response distribution. Application of the
hybrid approach requires the selection of an extra risk value for BMD estimation. In the case of
birth weight, an extra risk of 5% is selected given that this level of response is typically used
when modeling developmental responses from animal toxicology studies, and that low
birthweight confers increased risk for adverse health effects throughout life, thus supporting a
BMR lower than the standard BMR of 10% extra risk.
Therefore, given a background response and a BMR = 5% extra risk, the BMD would be the
dose that results in 12.86% of the responses falling below the 2500 g cut-off value:
Extra Risk(ER) = (P(d) - P(0)) / (1 - P(0))
P(d) = ER( 1 - P(0)) + P(0) = 0.05(1 - 0.0827) + 0.0827 = 0.1286
Based on the mean birth weight for all birth in the U.S. in 2018 of 3261.6 g with a standard
deviation of 590.7 g, EPA calculated the mean response that would be associated with the
12.86th percentile of the distribution falling below 2500 g. In this case, the mean birth weight
would be 3169.2g. Given the median exposure of 2.6 ng/mL from ACE Biomonitoring on
Perfluorochemicals as x , the mean birth weight in the U.S. as y and the re-expressed P as m
term, the intercept b can be estimated as:
b =y - nix = 3261.6 g - (-11.0 S©"1) 2.6^ = 3290.3 g (3)
The BMD was calculated by rearranging the equation y = mx + b and solving for x, using
3290.3 g for the b term and -11.0 for the m term. Doing so results in a value of 11.0 ng/mL:
nq „
x = (y -b)/m = (3169.2 g - 3290.3 g)/(-11.0 #(^t)_1) = 11.0 ng/mL
mL
To calculate the BMDL, the method is essentially the same except that the lower limit (LL) on
the P coefficient (Pll= -17.6) is used for the m term. However, Chu et al. (2020, 6315711)
reported a two-sided 95% confidence interval for the P coefficient, meaning that the lower limit
of that confidence interval corresponds to a 97.5% one-sided lower limit. The BMDL is defined
as the 95% lower limit of the BMD (i.e., corresponds to a two-sided 90% confidence interval), so
the corresponding lower limit on the P coefficient needs to be calculated before calculating the
BMDL. First, the standard error of the P coefficient can be calculated as:
9 While the explicit application of the hybrid approach is not commonly used in IRIS dose/concentration/exposure-
response analyses, the more commonly used SD-definition of the BMR for continuous data is simply one specific
application of the hybrid approach. The SD-definition of the BMR assumes that the cut-off for adversity is the 1.4th
percentile of a normally distributed response and that shifting the mean of that distribution by one standard deviation
approximates an extra risk of 10%.
E-20
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
SE =
Upper Limit — Lower Limit 4.4 g(~[) 1 ( 17-6
3.92
3.92
Then the corresponding 95% one-sided lower bound on the P coefficient can be calculated as:
Sagiv et al. (2018, 4238410) reported a P coefficient of-17.9 g (95% CI: -40.9, 5.1) per IQR
increase in PFOS (ng/mL), corresponding to a P coefficient of-1.1 g (95%CI: -2.6, 0.3) per
ng/mL increase, for the association between birth weight and maternal PFOS serum
concentrations (collected during 5 weeks to 19 weeks of pregnancy with a median of 9 weeks) in
a United States cohort. The intercept b is 3264.5 g based on the P coefficient of-1.1 g per
ng/mL. A BMD of 85.2 ng/mL is calculated from Sagiv et al. (2018, 4238410) using the same
approach as above with the same values for the mean birth weight in the United States.
To calculate the BMDL, the same procedure as above is used to calculate the corresponding 95%
one-sided lower limit for the P coefficient from the lower limit on the 95% two-sided confidence
interval of-2.6 g per ng/mL. Using the corresponding lower limit (-2.3 g per ng/mL), a BMDL
of 41.0 ng/mL is calculated.
E.1.2.3 Starling et al. {2017, 3858473)
Starling et al. (2017, 3858473) reported a P coefficient of-13.8 g (95%CI: -53.8, 26.3) per
ln(ng/mL) for the association between birth weight and maternal PFOS serum concentrations
(collected during 20 to 34 weeks of pregnancy with a median of 27 weeks) in a United States
cohort. Given the reported study-specific median (2.4 ng/mL) and the 25th and 75th percentiles
(1.5, 3.7 ng/mL) of the exposure, EPA estimated the mean (0.88) and standard deviation (0.67)
of the log normally distributed exposure. The re-expressed P coefficient is -5.5 g (95%CI: -21.4,
10.5) per ng/mL and the intercept b is 3275.9 g. The 95% one-sided lower limit for the re-
expressed P coefficient is -18.9 g per ng/mL. The values of the BMD and BMDL are
19.4 ng/mL and 5.7 ng/mL, respectively.
E.l.2.4 Wikstrom et al. (2020, 6311677)
Wikstrom et al. (2020, 6311677) reported a P coefficient of-46.0 g (95%CI: -88.0, -3.0) per
ln(ng/mL) for the association between birth weight and maternal PFOS serum concentrations
(collected during 9 weeks to 10 weeks of pregnancy with a median of 10 weeks) in a Swedish
cohort. Given the reported study-specific median (5.4 ng/mL) and the 25th and 75th percentiles
(4.0, 7.6 ng/mL) of the exposure, EPA estimated the mean (1.68) and standard deviation (0.48)
of the log normally distributed exposure. The re-expressed P coefficient is -8.4 g (95%CI: -16.0,
-0.5) per ng/mL and the intercept b is 3,283.4 g. The 95% one-sided lower limit for the re-
95% one - sided LL = j3 - 1.645(SE(j3)) = -11 g^r)'1 - 1.645 (3.37 g^r)'1)
mL V mL '
Using this value for the m term results in a BMDL value of 7.3 ng/mL maternal serum
concentration.
E.1.2.2 Sagiv et al. (2018, 4238410)
E-21
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
expressed P coefficient is -14.8 g per ng/mL. The values of the BMD and BMDL are
13.7 ng/mL and 7.7 ng/mL, respectively.
E. 1.2.5 Darrow et al. (2013, 2850966)
Darrow et al. (2013, 2850966) reported a P coefficient of-49.0 g (95%CI: -90.0, -8.0) per
ln(ng/mL) for the association between birth weight and maternal PFOS serum concentrations in a
United States cohort. Given the reported study-specific median (13.9 ng/mL) and the 25th and
75th percentiles (9.5, 19.7 ng/mL) of the exposure, EPA estimated the mean (2.63) and standard
deviation (0.54) of the log normally distributed exposure. The re-expressed P coefficient is
-3.4 g (95%CI: -6.3, -0.6) per ng/mL and the intercept b is 3270.5 g. The 95% one-sided lower
limit for the re-expressed P coefficient is -5.8 g per ng/mL. The values of the BMD and BMDL
are 29.6 ng/mL and 17.4 ng/mL, respectively.
E. 1.2.6 Yao et al. (2021, 9960202)
Yao et al. (2021, 9960202) reported a P coefficient of-32.3 g (95%CI: -116.2, 51.6) per
ln(ng/mL) for the association between birth weight and maternal PFOS serum concentrations
(collected within 3 days of delivery) in a China cohort. Given the cohort-specific median
(4.6 ng/mL) and the 25th and 75th percentiles (3.2, 5.9 ng/mL) of the exposure reported in Han
et al. (2018, 5080230), EPA estimated the mean (1.52) and standard deviation (0.45) of the log
normally distributed exposure. The re-expressed P coefficient is -6.9 g (95%CI: -25.0, 11.1) per
ng/mL and the intercept b is 3279.7 g. The 95% one-sided lower limit for the re-expressed P
coefficient is -22.1 g per ng/mL. The values of the BMD and BMDL are 15.9 ng/mL and
5.0 ng/mL, respectively.
E. 1.2.7 Summary of Modeling Results for Decreased Birthweight
For all of the above calculations, EPA used the exact percentage (8.27%) of live births in the US
in 2018 that fell below the cut-off of 2500 g as the tail probability to represent the probability of
extreme ("adverse") response at zero dose (P(0)). However, this exact percentage of 8.27% was
calculated without accounting for the existence of background PFOS exposure in the U.S.
population (i.e., 8.27% is not the tail probability of extreme response at zero dose). Thus, EPA
considers an alternative control-group response distribution (N(jj.c,
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Perfluorochemicals as background exposure ( x), the tail probabilities using this alternative
approach was study-specific and ranged from 9.05% to 9.78%. As such, the results from this
alternative approach, presented under the column of "Alternative Tail Probability" in Table E-13,
are very similar to the main results, presented under the column of "Exact Percentage" in Table
E-13, when background exposure was not accounted for while estimating the tail probability.
Table E-13 presents the BMDs and BMDLs for all studies considered for POD derivation, with
and without accounting for background exposure while estimating the percentage of the
population falling below the cut-off value. The BMDLs across the studies ranged from
5.0 ng/mL to 57.6 ng/mL.
E-23
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-13. BMDs and BMDLs in ng/mL for effect of PFOS on decreased birth weight, by using the exact percentage (8.27%)
of live births falling below the public health definition of low birth weight, or alternative study-specific tail
Study
Exposure
Median
Exposure
Distribution
Reported p
(95% CI)
Re-expressed p
(95% CI)
Intercept
SE(P)
Pll
Exact Percentage
(P(0) = 8.27%)
Alternative Tail Probability3
(25th - 75th
percentiles)
(H, <*)
units
g/(ng/mL)
b
BMD
BMDL
P(0)
BMD
BMDL
Chu et al. (2020,
6315711)
7.2 (4.4-11.9)
(1.97,0.75)
-83.3
(-133.2, -33.4)
g/ln(ng/mL)
-11.0 (-17.6,-4.4)
3290.3
3.37
-16.5
11.0
7.3
9.05%
12.8
8.5
Sagiv etal. (2018,
4238410)
25.7(18.9-34.9)
(3.25,0.45)
-17.9 (-40.9, 5.1)
g/IQR (ng/mL)
-1.1 (-2.6,0.3)
3264.5
0.73
-2.3
85.2
41.0
9.78%
119.8
57.6
Starling et al.
(2017, 3858473)
2.4(1.5-3.7)
(0.88, 0.67)
-13.8 (-53.8, 26.3)
g/ln(ng/mL)
-5.5 (-21.4, 10.5)
3275.9
8.14
-18.9
19.4
5.7
9.45%
25.0
7.3
Wikstrom et al.
(2019,6311677)
5.4 (4.0-7.6)
(1.68,0.48)
-46.0 (-88.0, -3.0)
g/ln(ng/mL)
-8.4 (-16.0, -0.5)
3283.4
3.94
-14.8
13.7
7.7
9.24%
16.7
9.4
Darrow et al.
(2013,2850966)
13.9(9.5-19.7)
(2.63, 0.54)
-49.0 (-90.0, -8.0)
g/ln(ng/mL)
-3.4 (-6.3, -0.6)
3270.5
1.46
-5.8
29.6
17.4
9.60%
40.0
23.3
Yao etal. (2021,
9960202)
4.6(3.2-5.9)
(1.52,0.45)
-32.3 (-116.2, 51.6)
g/ln(ng/mL)
-6.9 (-25.0, 11.1)
3279.7
9.22
-22.1
15.9
5.0
9.34%
19.9
6.3
Notes: SE = standard error.
a The alternative study-specific tail probability of live births falling below the public health definition of low birth weight based on Normal distribution with intercept b as mean
and standard deviation of 590.7 based on the US population.
E-24
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
ACE Biomonitoring on Perfluorochemicals also provides the median blood serum levels of
PFOS among women ages 16 to 49 in 1999-2000 (23.8 ng/mL), in 2009-2010 (5.7 ng/m) and in
2013-2014 (3.0 ng/mL). EPA performed a sensitivity analysis by estimating BMD and BMDL
using these values as background exposures. The results for Wikstrom et al. (2020, 6311677),
presented in Table E-14, demonstrate the robustness of EPA's approaches with alternative
assumptions on background exposures.
Table E-14. BMDs and BMDLs for effect of PFOS on decreased birth weight by
background exposure, using the exact percentage of the population (8.27%) of live births
falling below the public health definition of low birth weight, or alternative tail probability
Exact percentage
(P(0) = 8.27%)
Alternative Tail Probability1"
aiuuy
Exposure"
b
BMD
BMDL
P(0)
BMD
BMDL
(ng/mL)
(ng/mL)
(ng/mL)
(ng/mL)
Wikstrom
2.6
3283.4
13.7
7.7
9.24%
16.7
9.4
et al. (2020,
3.0
3286.7
14.1
7.9
9.14%
16.8
9.5
6311677)
5.7
3309.2
16.8
9.4
8.53%
17.6
9.9
23.8
3460.4
34.9
19.6
5.20%
24.1
13.6
Notes:
a Assumptions on background exposure for the estimation of intercept using Equation (3).
b The tail probability of live births falling below the public health definition of low birth weight based on Normal distribution.
For decreased birth weight associated with PFOS exposure, the POD selected from the
available epidemiologic literature is 7.7 ng/mL maternal serum concentration, based on
birth weight data from Wikstrom et al. (2020, 6311677). Of the six individual studies, Sagiv
et al. (2018, 4238410) and Wikstrom et al. (2020, 6311677) assessed maternal PFOS serum
concentrations primarily or exclusively in the first trimester, minimizing concerns surrounding
bias due to pregnancy-related hemodynamic effects. Therefore, the PODs from these two studies
were considered further for POD selection. The POD from Wikstrom et al. (2020, 6311677) was
ultimately selected as it was the lowest POD from these two studies.
E.1.3 Modeling Results for Increased Cholesterol
This updated review indicated that there was as association between increases in PFOS and
increases in total cholesterol (TC) in adults. Three medium confidence studies were considered
for POD derivation {Dong, 2019, 5080195; Lin, 2019, 5187597; Steenland, 2009, 1291109}.
These candidate studies offer a variety of PFOS exposure measures across various populations.
Dong et al. (2019, 5080195) investigated an NHANES population (2003-2014), while Steenland
et al. (2009, 1291109) investigated effects in a high-exposure community (the C8 Health Project
study population). Lin et al (2019, 5187597) collected data from prediabetic adults from the
Diabetes Prevention Program (DPP) and DPP Outcomes Study at baseline (1996-1999).
E. 1.3.1 Dong et a I. (2019, 5080195)
Using data from NHANES (2003-2014) on 8,948 adults, Dong et al. (2019, 5080195) calculated
a BMD for PFOS and TC using a hybrid model {Crump, 1995, 2258}. The cut-off for adverse
response (i.e., elevated TC) was set at the upper 5th percentile of TC values in the lowest PFOS
E-25
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
exposure group (the actual TC value at this cutoff point was not provided), and the BMR was
defined as a 10% increase in the number of people with TC values above this level. Using this
method, Dong et al. (2019, 5080195) reported a BMDio and BMDLio of 44.2 ng/mL and
24.1 ng/mL, respectively. Key variables or other results such as the cut-off point used to define
elevated TC or model fit parameters were not provided.
Although the hybrid approach has several advantages {Crump, 1995, 2258}, few details were
provided in Dong et al. (2019, 5080195) on several important aspects of this approach or on
other key issues, including the definition of the unexposed reference group, the distribution of
PFOS or TC values in this group, model fit (e.g., the fit of linear vs. non-linear models), the
impact of potential confounders, or the potential role of reverse causality.
EPA re-analyzed the data using the regression models from the Dong et al. (2019; 5080195)
study, together with updated NHANES data, applied to a modified hybrid model to develop
BMD and BMDL estimates for various time periods and assumptions. The BMD values for a
BMR of 5% ranged from 15.84 ng/mL for the period 1999-2018, excluding adults taking
cholesterol medications, up to 36.20 ng/mL for the period 2017-2018, for all adults. The BMDL
values for a BMR of 5% ranged from 9.34 ng/mL for the period 1999-2018, excluding adults
taking cholesterol medications, up to 21.35 ng/mL for the period 2017-2018, for all adults. The
BMD values for a BMR of 10% ranged from 35.79 ng/mL for the period 1999-2018, excluding
adults taking cholesterol medications, up to 55.71 ng/mL for the period 2017-2018, for all
adults. The BMDL values for a BMR of 10% ranged from 21.11 ng/mL for the period 1999-
2018, excluding adults taking cholesterol medications, up to 32.86 ng/mL for the period 2017-
2018, for all adults.
An important caveat is that these calculations assume that Dong's regression model is still
applicable, or at least a good approximation, for all the time periods, for all adults and for adults
taking cholesterol medications, and for the recently updated NHANES data.
Dong et al. (2019, 5080195) reported a regression coefficient P, which we also call /??, of
0.4 mg/dL TC per ng/mL PFOS (95% CI: 0.06, 0.6). From correspondence with the author, EPA
obtained an updated estimated coefficient of 0.35 (95% CI: 0.06, 0.64) mg/dL TC per ng/mL
PFOS, which EPA used for these analyses. The regression model applies to all adults 20 to
80 years old and was adjusted for age, gender, race, poverty income ratio, body mass index,
waist circumference, physical activity level, diabetes status, smoking status, and number of
alcoholic drinks per day. Using a normal approximation, the standard error of the regression
coefficient is estimated as
These analyses were for the periods 1999-2008, 2003-2014, 2003-2018, and 2017-2018,
assuming that regression model coefficient developed for the period 2003-2014 in the Dong et
al. (2019, 5080195) study can be applied to the alternative NHANES periods. These analyses
used the NHANES-recommended reference method data for TC. EPA used the NHANES PFOS
data for each NHANES period including data adjustments to stored biospecimen data collected
in 1999-2000 and 2013-2014 that were publicly released in April 2022. Alternative analyses
SE =
Upper Limit — Lower Limit 0.64 — 0.06
3.92
3.92
E-26
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
were for all adults ages 20 and over, and for adults ages 20 and over that reported not taking
prescribed cholesterol medications. NHANES survey weights were applied.
EPA estimated the distribution of TC assuming a normal distribution and also estimated the
mean PFOS. The means and standard deviations for each group are shown in the following table:
Table E-15. NHANES mean and standard deviation of TC (mg/dL) and mean PFOS
(ng/mL)
Time Period
1999-
2018
1999-
2018
2003-
2014
2003-
2014
2003-
2018
2003-
2018
2017-
2018
2017-2018
Taking prescribed cholesterol
No
No
No
No
medication?
Mean TC (y)
196.17
197.89
196.36
198.01
194.86
196.96
189.01
192.12
Standard Deviation TC (S)
41.99
41.47
41.84
41.39
41.80
41.28
40.57
39.67
Mean PFOS (x)
13.73
13.73
15.64
15.64
13.21
13.21
6.13
6.13
For the BMD analyses, the response of interest is having elevated serum cholesterol, defined as
greater than or equal to 240 mg/dL. The baseline probability of such a response is P(0), estimated
as 11.5%, for adults aged 20 and older in 2015-2018, as reported by the CDC Health, United
States, 2019 Data Finder {NCHS, 2019, 10369680}.
The selected BMR is an extra risk of either 5% or 10%. The extra risk of high serum cholesterol
is given by the equation
P(d) -P(0)
Extra Risk = — ———
1 - P(0)
where P(d) is the probability of serum cholesterol greater than or equal to 240 mg/dL for a given
PFOS dose d. Thus
P(d) = {1 - P(0)} x Extra Risk + P(0)
P(d) = {1 — 0.115} x Extra Risk + 0.115
P(d) = 0.1593 for 5% extra risk and P(d) = 0.2035 for 10% extra risk.
The mean serum cholesterol y for a PFOS dose x is given by the equation
y = mx + b
where m is the slope, P, (from the Dong regression model) and b is the intercept. The intercept b
is the mean serum cholesterol for an unexposed population. For the U.S. population, the mean
TC is y (tabulated above) and the mean PFOS is x (tabulated above) so the intercept is given by
the equation
b = y — mx
For a given group and dose, the probability of serum cholesterol greater than or equal to
240 mg/dL is
E-27
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
/240 — y\
P(d) = P(TC > 240) = 1 - ( ^ )
where is the normal cumulative distribution function. Thus, the mean serum cholesterol y is
the solution of the last equation, i.e., y = 240 — S x — P(d)} , where -1 is the inverse
of the normal cumulative distribution function.
The BMD is the corresponding dose x such that y = mx + b. Thus
y-b
BMD =
m
For the BMDL, the lower bound of the dose is calculated, so that in the last equation, instead of
m we use the 95th upper limit for P, which is given by
(395 = 95th Upper limit for (3 = (3 + 1.645 x se{(3)
Thus
y-b
BMDL =
(395
Note that P95 is different from the upper bound of the 95% confidence interval, since that
number is the 97.5th percentile. The estimated BMDs and BMDLs are presented in Table E-16:
Table E-16. BMDs and BMDLs for effect of PFOS on increased cholesterol in Dong et al.
(2019, 5080195)
Time Period
1999-
2018
1999-
2018
2003-
2014
2003-
2014
2003-
2018
2003-
2018
2017-
2018
2017-
2018
Taking prescribed
No
No
No
No
cholesterol medication?
BMR=5%
BMD (ng/mL)
19.28
15.84
21.08
17.63
23.07
18.54
36.20
29.86
BMDL (ng/mL)
11.37
9.34
12.44
10.40
13.61
10.93
21.35
17.61
BMR=10%
BMD (ng/mL)
39.48
35.79
41.21
37.54
43.18
38.39
55.71
48.95
BMDL (ng/mL)
23.29
21.11
24.31
22.14
25.47
22.65
32.86
28.87
Given the potential impact of taking cholesterol medication on the true association between
PFOS and increased TC, the results based on the data excluding such possibility is considered
higher confidence. As illustrated in Table E-16 there was a decline over time in PFOS levels
based on NHANES data, suggesting that reliance on distributional data based on the most recent
NHANES cycle available (2017-2018) might be more reflective of recent exposure levels.
However, given the chronic nature of both exposure and increased TC development, a higher
confidence might be the given to estimates based on the largest period available (1999-2018).
E-28
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
For increased cholesterol associated with PFOS exposure, the POD is based on the data
Dong et al. (2019, 5080195) excluding people taking cholesterol medication, the longest
period available, a BMR of 5% and a BMDLs of 9.3 ng/mL.
E.1.3.2 Steenland et al. (2009,1291109)
The above hybrid approach was also applied to Steenland et al. (2009, 1291109) using log-
transformed values. In Table 4, Steenland et al. (2009, 1291109) reported a linear regression
coefficient for change in ln-transformed TC per ln(PFOS): 0.2660 with a standard deviation of
0.00140. The NHANES data used in this approached is summarized in Table E-17 and
BMD/BMDL values are presented in Table E-18.
Table E-17. NHANES mean and standard deviation of ln(TC) (ln(mg/dL)) and mean
ln(PFOS) (ln(ng/mL))
1999-
1999-
2003-
2003-
2003-
2003-
2017-
2017-
2018
2018
2014
2014
2018
2018
2018
2018
Taking prescribed
No
No
No
No
cholesterol medication?
Mean ln(TC) (y)
5.26
5.27
5.26
5.27
5.25
5.26
5.22
5.24
Standard Deviation
0.21
0.21
0.21
0.21
0.21
0.21
0.22
0.21
ln(TC) (S)
Mean ln(PFOS) (x)
2.17
2.17
2.36
2.36
2.14
2.14
1.50
1.50
Table E-18. BMDs and BMDLs for effect of PFOS on increased cholesterol in Steenland
al. (2009,1291109)
1999-
1999-
2003-
2003-
2003-
2003-
2017-
2017-
2018
2018
2014
2014
2018
2018
2018
2018
Taking prescribed
No
No
No
No
cholesterol medication?
BMR=5%
BMD (ng/mL)
14.16
11.58
16.77
13.48
17.21
13.23
26.36
18.88
BMDL (ng/mL)
11.46
9.52
13.39
10.95
13.72
10.77
20.31
14.94
BMR=10%
BMD (ng/mL)
54.05
43.02
63.79
50.20
66.14
49.34
102.98
69.54
BMDL (ng/mL)
39.33
31.88
45.81
36.75
47.36
36.17
71.18
49.59
Mean serum TC
EPA also conducted dose-response modeling using mean serum TC reported across PFOS
deciles from Table 3 in Steenland et al. (2009, 1291109). The associated standard error terms
were found through author correspondence. BMDS 3.3rcl0 was used to fit the dose response
data using all deciles, no viable models were identified. To further investigate, BMDS 3.3rcl0
was used to fit the dose-response data in the lowest five deciles and regression coefficients for
the mean change of ln-transformed serum TC (Table 3 in Steenland et al. (2009, 1291109)),
summarized in Table E-19. BMRs of a change in the mean equal to V2 and 1 SDs from the
control mean were chosen. The BMD modeling results are summarized in Table E-20.
E-29
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-19. Regression Results for Serum Total Cholesterol by Deciles of serum PFOS
from Steenland et al. (2009,1291109)
Decile
Dose
(ng/mL)
N
Regression coefficient3
(SD)
1
6.37
4629
0.00 (0.192)
2
10.60
4629
0.01 (0.192)
3
13.65
4629
0.01 (0.192)
4
16.19
4629
0.03 (0.192)
5
18.79
4629
0.03 (0.192)
Notes:
a Regression coefficient, change in the natural log of total cholesterol
E-30
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-20. Summary of Benchmark Dose Modeling Results for Increase Mean Serum Total Cholesterol in Steenland et al.
(2009,1291109)
Goodness of Fit Scaled Residual
Model3
p-value
AIC
Dose Group
near BMDisd
Dose Group
near
BMDo.ssd
BMDisd
Control Dose (n2/mL)
Group
BMDLisd
(ng/mL)
BMDo.ssd
(ng/mL)
BMDLo.ssd
(ng/mL)
Exponential 3
<0.0001
-10350.92
0.00
-1.16
-1.54 0.76
0.00
25.38
24.66
Exponential 5
-
-
-
-
-
-
-
-
Hill
-
-
-
-
-
-
-
-
Polynomial Degree 3
0.00
-10588.86
-0.78
-0.78
0.00 45.95
33.33
31.36
26.15
Polynomial Degree 2
0.00
-10588.82
-0.71
-
47.85
39.78
-
-
Power
0.00
-10588.89
-0.75
-0.75
0.02 48.56
47.46
32.31
29.22
Linear
0.01
-10589.87
-0.23
-0.23
0.51 74.49
62.75
37.24
31.37
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDisd = dose level corresponding to a change in the mean equal to
one standard deviation from the control mean; BMDLisd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to one
standard deviation from the control mean. BMDo.5sd = dose level corresponding to a change in the mean equal to 0.5 standard deviations from the control mean;
BMDLo.5sd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.5 standard deviation from the control mean.
aNo viable models. No model was selected.
b BMD Computation failed
E-31
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Elevated TC
In additional to modeling the regression coefficients, dichotomous models using BMDS 3.3rcl0
were used to fit the ORs of elevated TC from Steenland et al. (2009, 1291109) as shown in Table
E-21. Sample sizes, mean PFOS concentrations in each quartile and prevalence of elevated TC in
each exposure group were obtained from Dr. Kyle Steenland. A BMR of 10% and 5% extra risk
were both included. The BMD modeling results are summarized in Table E-22.
Table E-21. Odds ratios for elevated serum TC by quartiles of serum PFOS from Steenland
et al. (2009,1291109)
Quartile
Dose
(ng/mL)
N
Incidence
OR
95% CI
1
6.6
11534
1479
1
Ref
2
16.4
11587
1634
1.14
1.05, 1.23
3
23.8
11441
1795
1.28
1.19, 1.39
4
50.55
11400
2158
1.51
1.40, 1.64
E-32
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-22. Summary of Benchmark Dose Modeling Results for Elevated Total Cholesterol in Steenland et al. (2009,1291109)
Goodness of Fit Scaled Residual
Model"
p-value
AIC
Dose Group
near BMD 10
Dose Group
near BMDs
Control Dose
Group
BMD10
(ng/mL)
BMDLio
(ng/mL)
BMDs
(ng/mL)
BMDLs
(ng/mL)
Dichotomous Hill
_b
-
3.56x10-6
-
-
-
-
31.08
26.59
Gamma
0.53
39272.57
-0.28
-0.28
-0.14
63.00
55.89
30.67
27.21
Log-Logistic
0.57
39272.40
-0.24
-0.24
-0.05
63.18
55.91
29.93
26.39
Multistage Degree 3
0.01
39282.00
-0.58
-0.58
-1.57
62.48
0.00
40.96
40.29
Multistage Degree 2
0.53
39272.57
-0.28
-0.28
-0.14
63.00
55.88
30.67
27.20
Multistage Degree 1
0.53
39272.57
-0.28
-0.28
-0.14
63.00
55.89
30.67
27.20
Weibull
0.53
39272.57
-0.28
-0.28
-0.14
63.00
55.89
30.67
27.21
Logistic
0.27
39274.11
-0.42
-0.42
-0.62
62.30
56.70
34.49
31.47
Log-Probit
0.35
39274.11
-0.10
-0.10
0.16
66.02
57.02
29.71
14.27
Probit
0.31
39273.81
-0.40
-0.40
-0.55
62.43
56.61
33.93
30.84
Quantal Linear
0.53
39272.57
-0.28
-0.28
-0.14
63.00
55.89
30.67
27.21
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level corresponding to a 10% response level;
BMDLio = lower bound on the dose level corresponding to the 95% lower confidence limit for a 10% response level; BMD5 = dose level corresponding to a 5% response level;
BMDL5 = lower bound on the dose level corresponding to the 95% lower confidence limit for a 5% response level.
a Selected model in bold.
b BMD Computation failed
E-33
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Given the potential impact of taking cholesterol medication on the true association between
PFOS and increased TC, the results based on the data excluding such possibility is considered
higher confidence. As illustrated in Table E-17 there was a dramatic decline over time in PFOS
levels based on NHANES data, suggesting that reliance on distributional data based on the most
recent NHANES cycle available (2017-2018) might be more reflective or current impacts.
However, given the chronic nature of both exposure and increased TC development, a higher
confidence might be the given to estimates based on the largest period available (1999-2018).
For increased cholesterol associated with PFOS exposure, the POD is based on the data
from Steenland et al. (2009,1291109) excluding people taking cholesterol medication, the
longest period available, a BMR of 5% and a BMDLs of 9.52 ng/mL. A comparison BMDL
of 14.9 ng/mL based on the most recent period available supports the selected POD.
E-34
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.1.3.3 Lin et al. (2019, 5187597)
Lin et al. (2019, 5187597) collected data from prediabetic adults from the Diabetes Prevention
Program (DPP) and DPP Outcomes Study at baseline (1996-1999). This study included 888 pre-
diabetic adults who were recruited from 27 medical centers in the United States. Median PFOS
levels at baseline were comparable to those from NHANES 1999-2000, 27.2 (25th, 75th
percentiles: 18.0 ng/mL, 40.4 ng/mL). The study presented both cross-sectional and prospective
analyses. The cross-sectional analyses evaluated associations between baseline PFAS and
baseline lipid levels. The prospective analysis evaluated whether baseline PFAS levels predicted
higher risk of incident hypercholesterolemia and hypertriglyceridemia, but in the placebo and the
lifestyle intervention groups, separately.
EPA conducted dose-response modeling using mean serum TC reported across PFOS quartiles
from Table S5 in Lin et al. (2019, 5187597). For its POD calculations, EPA used the results from
the cross-sectional analysis because they were presented in a format that was more amendable to
dose-response analysis.
BMDS 3.3rcl0 was used to fit the dose-response data for the adjusted percent difference in lipid
levels (mg/dL) per quartile of baseline plasma PFOS concentrations (ng/mL), summarized in
Table E-23. BMRs of a change in the mean equal to 0.5 SD and 1 SD from the control mean
were used. The BMD modeling results are summarized in Table E-24. However, the PODs
derived from this study are considered lower confidence since they are based on a poorly fir
PFOS association (adjusted mean difference = 2.53 , 95% CI: -0.10, 5.16).
Table E-23. Adjusted Mean Differences in Serum Total Cholesterol by Quartiles of Serum
PFOS (ng/mL) from Lin et al. (2019,1291109)
Dose
N
Adjusted mean difference
Mean TCa b
(ng/mL)
TC (95% CI) (mg/dL)
12.8
212
Ref
0.00 ±35.48
21.7
224
1.13 (-5.50,7.77)
1.13 ±35.33
32.7
230
5.05 (-1.55, 11.66)
5.05 ±35.39
53
222
5.13 (-1.58, 11.86)
5.13 ±35.70
Notes:
a mean ± standard deviation.
b Adjusted mean difference in lipid levels (mg/dL) per quartile of baseline plasma PFOS concentration (ng/mL)
E-35
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-24. Summary of Benchmark Dose Modeling Results for Increase Mean Serum Total Cholesterol Lin et al. (2019,
5187597)
Goodness of Fit Scaled Residual
BMDisd BMDLisd BMDo.ssd BMDLo.ssd
p-value AIC D()SCd^'"('u|) D()S^™U|) Coi*ro1 Dose (ng/mL) (ng/mL) (ng/mL) (ng/mL)
near BMDisd near BMDo.ssd Group
Exponential 3
0.23
8863.69
-0.21
-0.21
-0.60
108.34
61.19
88.53
57.34
Exponential 5
_b
-
-
-
-
-
-
-
-
Hill
-
-
-
-
-
-
-
-
-
Polynomial Degree 3
0.65
8861.12
-0.35
-0.35
-0.21
261.96
86.09
130.98
66.43
Polynomial Degree 2
0.65
8861.12
-0.34
-
-
262.61
100.07
-
-
Power
0.65
8861.12
-0.34
-0.34
-0.21
262.62
58.47
131.31
66.54
Linear
0.65
8861.12
-0.34
-0.34
-0.21
262.62
133.07
131.31
66.54
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDisd = dose level corresponding to a change in the mean equal to
one standard deviation from the control mean; BMDLisd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to one
standard deviation from the control mean. BMDo.ssd = dose level corresponding to a change in the mean equal to 0.5 standard deviations from the control mean;
BMDLo.ssd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.5 standard deviation from the control mean.
a Selected model in bold.
b BMD Computation failed
E-36
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E. 1.3.4 Summary of Modeling Results for Increased Cholesterol
Table E-25 summarizes the PODs resulting from the modeling approaches for increased
cholesterol. The selected and comparison PODs were based on a BMR of 5%, resulting in BMDLs
ranging from 9.3 ng/mL to 66.5 ng/mL with the selected POD of 9.35 also representing the median
of the BMDLs. The comparison POD based on the data from Lin et al. (2019, 5187597) is
considered low confidence because it is based on a poorly fit PFOS regression parameter.
Table E-25. BMDLs for effect of PFOS on serum total cholesterol using a BMR of 5%.
Study name Effect BMDL (ng/mL)
Dong et al. (2019, Exclude those prescribed cholesterol medication, 1999-2018 9.34
5080195)
Steenland et al. (2009, Exclude those prescribed cholesterol medication 9.52
1291109)
Lin et al. (2019, 5187597) Diabetic adults 66.5
E.1.4 Modeling Results for Liver Toxicity
This updated review indicated that PFOS is associated with increases in the liver enzyme ALT (See
Main PFOS Document). Three medium confidence studies were selected as candidates for POD
derivation. One of the largest studies of PFOS and ALT in adults is Gallo et al. (2012, 1276142)
conducted in 47,092 adults from the C8 Study Project (for detailed descriptions of the study and
findings see Main PFOS Document and Appendix D). Two additional studies {Lin, 2010, 1291111;
Nian, 2019, 5080307} were considered by EPA for POD derivation because they reported
significant association in general populations in the U.S and a high exposed population China,
respectively. In an NHANES adult population, Lin et al. (2010, 1291111) observed elevated ALT
levels per log-unit increase in PFOS in the models adjusted for age, gender, and race/ethnicity, but
not in the fully adjusted models or in the models additionally adjusted for PFOA, PFHxS, and
PFNA. While this is a large nationally representative population, several methodological limitations
preclude its use for POD derivation. Limitations include lack of clarity about base of logarithmic
transformation applied to PFOS concentrations in regression models, and the choice to model ALT
as an untransformed variable, a departure from the typically lognormality assumed in most of the
ALT literature.
Nian et al. (2019, 5080307) examined 1,605 adults in Shenyang (one of the largest fluoropolymer
manufacturing centers in China) part of the Isomers of C8 Health Project and observed significant
increases in ln-transformed ALT per each ln-unit increase in PFOS, as well significant increases in
odds ratios of elevated ALT. Median serum PFOS concentrations were 24.22 ng/mL.
E.l.4.1 Gallo et al. (2012, 1276142)
Gallo et al. (2012, 1276142) evaluated the relationship between PFOS and ALT using two general
types of analyses. In the first, subjects were divided into deciles of PFOS exposure, and linear
regression models were used to compare mean ALT levels by each non-reference quantile vs. mean
ALT level in the lowest decile. In the second type of analysis, a logistic regression evaluated ORs
for having an ALT level above a certain cutoff for each non-reference deciles compared to the
E-37
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
lowest (reference) deciles. The cutoff values used to define elevated ALT levels were 45 IU/L for
men and 34 IU/L for women, clinically based value recommended by the International Federation
of Clinical Chemistry and Laboratory Medicine {Schumann, 2002, 10369681}, and were
approximately the 90th percentile of all ALT values in this study.
Elevated ALT
NOAEC/LOAEC method. The results of the logistic regression analysis of elevated ALT across
deciles of PFOS are presented in Table E-26. The mean, median and ranges of PFOS concentrations
in each decile were not provided with the OR results in the publication. EPA obtained these from
author correspondence and they are illustrated in Table E-26. The no observed adverse effect
concentration (NOAEC) is bolded and is the mean PFOS serum concentration in the highest decile
of PFOS that did not show a statistically significant OR of elevated ALT, which in this case is the
2nd decile, compared to the reference category (the lowest decile of PFOS). The NOAEC based on
the elevated ALT data from Gallo et al. (2012, 1276142) is 10.6 ng/mL.
Table E-26. Odds Ratios for Elevated ALT by Decile of PFOS serum concentrations (ng/mL)
from Gallo et al. (2012,1276142). The NOAEC is bolded.
Participants Participants
Decile
Minimum
(ng/mL)
Maximum
(ng/mL)
Median
(ng/mL)
Mean
(ng/mL)
OR
95% CI
without
Elevated
ALT
with
Elevated
ALT
Total
(N)
0
0.25
00
00
6.4
5.751386
1
reference
4,119
427
4,546
1
8.9
12.2
10.7
10.63289
1.09
0.94,1.26
4,264
446
4,710
2
12.3
14.9
13.6
13.60556
1.19
1.03, 1.37
4,113
459
4,572
3
15
17.5
16.3
16.26427
1.26
1.09, 1.45
4,104
500
4,604
4
17.6
20.2
18.9
18.88567
1.40
1.22, 1.62
4,115
545
4,660
5
20.3
23.3
21.7
21.74935
1.39
1.21, 1.60
4,181
571
4,752
6
23.4
27
25.1
25.11534
1.31
1.14, 1.52
4,099
561
4,660
7
27.1
32
29.3
29.38941
1.42
1.23, 1.64
4,071
586
4,657
8
32.1
40.4
35.6
35.76743
1.40
1.21, 1.62
4,068
547
4,615
9
40.5
585.2
49.7
56.12528
1.54
1.33, 1.78
4,124
552
4,676
BMP method. EPA applied BMDS to calculate a BMD. In addition, EPA performed a sensitivity
analysis using the generalized least-squares for trend (gist) method {Greenland, 1992, 5069}, which
assumes a linear relationship between exposure and log-transformed ORs, and accounts for
covariance between estimates. These analyses were performed in STATA vl7.0 {StataCorp. 2021.
10406419}. Through author correspondence EPA obtained the number of participants with and
without elevated ALT for each decile of PFOS (Table E-26).
Applying BMDS v3.3rcl0 using a BMR of 10% and 5% the data for all ten deciles did not result in
any viable models. Applying BMDS v3.3rcl0 to the data for all first five deciles did result in viable
models. The data associated with the first five deciles was also run using a no intercept approach in
which the lowest dose was subtracted out, subsequently referred to as an adjusted dose. The results
of this modeling using both the mean and median doses are summarized in Table E-27, Table E-28,
Table E-29, Table E-30. This modeling approach results in BMD and BMDL values higher than the
maximum dose included in the modeled data set. The BMD and BMDL values were inside the
range of mean exposure values when considering all ten deciles.
E-38
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-27. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et al. (2012,1276142) Using the
Unadjusted Mean PFOS Serum Concentration
Goodness of Fit Scaled Residual
BMDio BMDLio BMDs BMDLs
n-value AIC Dose Group Dose Group Control Dose (ng/mL) (ng/mL) (ng/mL) (ng/mL)
p near BMDio near BMDs Group
3 I I I I I I I I
Dichotomous Hill
_b
-
-
-
-
-
-
-
-
Gamma
0.92
15296.47
-0.11
-0.11
0.16
28.37
25.58
22.69
20.63
Log-Logistic
0.91
15296.50
-0.11
-0.11
0.17
27.68
22.17
22.50
20.19
Weibull
0.98
15294.50
-0.11
-0.11
0.17
27.47
23.26
22.46
20.46
Logistic
0.52
15296.80
0.67
0.67
0.83
43.97
33.33
25.48
19.53
Log-Probit
0.94
15296.44
-0.10
-0.10
0.14
29.51
22.98
22.98
20.39
Probit
0.51
15296.87
0.69
0.69
0.83
45.41
34.13
25.66
19.47
Quantal Linear
0.45
15297.26
0.80
0.80
0.82
54.66
38.95
26.61
18.96
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level corresponding to a 10% response level;
BMDLio = lower bound on the dose level corresponding to the 95% lower confidence limit for a 10% response level; BMD5 = dose level corresponding to a 5% response level;
BMDL5 = lower bound on the dose level corresponding to the 95% lower confidence limit for a 5% response level.
a Selected model in bold.
b BMD Computation failed.
E-39
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-28. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et al. (2012,1276142) Using the
Adjusted, No Intercept Mean PFOS Serum Concentration
Goodness of Fit Scaled Residual
BMDio BMDLio BMDs BMDLs
p-value AIC D()SC5''/I()II1'|) Dose „Co^ro1 (ng/mL) (ng/mL) (ng/mL) (ng/mL)
near BMDio near BMDs Dose Group
3 I I I I I I I I
Dichotomous Hill
_b
-
-
-
-
-
-
-
-
Gamma
0.95
15296.40
-0.09
-0.09
0.12
24.22
18.67
17.44
15.03
Log-Logistic
0.95
15296.41
-0.09
-0.09
0.14
23.67
16.76
17.30
14.58
Weibull
0.94
15296.42
-0.09
-0.09
0.14
23.39
17.63
17.25
14.87
Logistic
0.52
15296.80
0.67
0.67
0.83
41.00
30.25
23.47
17.42
Log-Probit
0.97
15296.36
-0.07
-0.07
0.10
26.47
17.71
17.96
14.79
Probit
0.51
15296.87
0.69
0.69
0.83
42.78
31.38
23.92
17.64
Quantal Linear
0.45
15297.26
0.80
0.80
0.82
54.66
38.95
26.61
18.96
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level corresponding to a 10% response level;
BMDLio = lower bound on the dose level corresponding to the 95% lower confidence limit for a 10% response level; BMD5 = dose level corresponding to a 5% response level;
BMDL5 = lower bound on the dose level corresponding to the 95% lower confidence limit for a 5% response level.
a Selected model in bold.
b BMD Computation failed.
E-40
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-29. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et al. (2012,1276142) Using the
Unadjusted, Median PFOS Serum Concentration
Goodness of Fit Scaled Residual
BMDio BMDLio BMDs BMDLs
p-value AIC „Con'r°l ^/mL) (ng/mL) („g/mL) (ng/mL)
near BMDio near BMDs Dose Group
nmnnc Will —b — — — — — — — —
Dichotomous Hill
_ b
-
-
-
-
-
-
-
-
Gamma
0.93
15296.46
-0.10
-0.10
0.16
28.47
25.68
22.71
20.60
Log-Logistic
0.92
15296.49
-0.10
-0.10
0.17
27.80
22.17
22.53
20.20
Weibull
0.98
15294.49
-0.10
-0.10
0.17
27.60
23.80
22.49
20.44
Logistic
0.59
15296.40
0.59
0.59
0.79
42.06
32.11
24.42
18.86
Log-Probit
0.94
15296.43
-0.10
-0.10
0.14
29.59
22.97
23.01
20.40
Probit
0.58
15296.47
0.61
0.61
0.79
43.34
32.79
24.53
18.75
Quantal Linear
0.52
15296.83
0.72
0.72
0.79
51.43
36.76
25.04
17.89
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level corresponding to a 10% response level;
BMDLio = lower bound on the dose level corresponding to the 95% lower confidence limit for a 10% response level; BMD5 = dose level corresponding to a 5% response level;
BMDL5 = lower bound on the dose level corresponding to the 95% lower confidence limit for a 5% response level.
a Selected model in bold.
b BMD Computation failed.
E-41
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-30. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et al. (2012,1276142) Using the
Adjusted, No Intercept Median PFOS Serum Concentration.
Goodness of Fit Scaled Residual
BMDio BMDLio BMDs BMDLs
p-value AIC „Con'r°l ^/mL) (ng/mL) („g/mL) (ng/mL)
near BMDio near BMDs Dose Group
nmnnc Will —b — — — — — — — —
Dichotomous Hill
_b
-
-
-
-
-
-
-
-
Gamma
0.96
15296.38
-0.08
-0.08
0.12
23.95
18.49
16.91
14.37
Log-Logistic
0.95
15296.40
-0.08
-0.08
0.13
23.44
16.17
16.78
13.96
Weibull
0.95
15296.40
-0.08
-0.08
0.13
23.14
16.75
16.73
14.27
Logistic
0.59
15296.40
0.59
0.59
0.79
38.74
28.66
22.18
16.50
Log-Probit
0.98
15296.34
-0.06
-0.06
0.09
26.43
17.13
17.48
14.18
Probit
0.58
15296.47
0.61
0.61
0.79
40.40
29.72
22.58
16.70
Quantal Linear
0.52
15296.83
0.72
0.72
0.79
51.43
36.75
25.04
17.89
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level corresponding to a 10% response level;
BMDLio = lower bound on the dose level corresponding to the 95% lower confidence limit for a 10% response level; BMD5 = dose level corresponding to a 5% response level;
BMDL5 = lower bound on the dose level corresponding to the 95% lower confidence limit for a 5% response level.
a Selected model in bold.
b BMD Computation failed.
E-42
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Hybrid method. The hybrid method used the regression slope from the linear regression model of
ln-transformed ALT and ln-PFOS concentrations adjusted for age, sex, alcohol consumption,
socioeconomic status, fasting status, race, month of blood sample collection, smoking status,
body mass index, physical activity, and insulin resistance. The reported regression coefficient P,
which is also referred to as m, was 0.02 (95% CI: 0.014, 0.026) of In ALT (IU/L) per In ng/mL
PFOS (Table 2, Gallo et al. (2012, 1276142), model 3).
Using a normal approximation, the standard error of the regression coefficient is estimated as
Upper Limit — Lower Limit 0.026 — 0.014
SE = — ™™ = 0.0025
3.92 3.92
Elevated ALT is a biomarker of acute liver disease. For the following analyses, the adverse
effect level of ALT for liver disease was chosen to be C = 42 IU/L for males and C = 30 IU/L for
females, based on the sex-specific upper reference limits found in Valenti et al (2021,
10369689).
These analyses were for the periods 1999-2018, 2003-2018, and 2017-2018, separately for
males and females ages 18 and over, assuming that the Gallo regression model coefficient
developed for the C8 Health Project data in Ohio starting in 2005 and 2006 can be applied to the
alternative NHANES periods. These analyses used the NHANES-recommended regression
model adjustment to correct the 2017-2018 ALT data to match the earlier laboratory method.
EPA used the NHANES PFOS data for each NHANES period including data adjustments to
stored biospecimen data collected in 1999-2000 and 2013-2014 that were publicly released in
April 2022. NHANES survey weights were applied.
Using the NHANES data for each period and sex, EPA estimated the mean and standard
deviation of In ALT and the estimated mean In PFOS (Table E-31). The unrounded values were
used in the calculations:
Table E-31. NHANES mean and standard deviation of ln(ALT) (In IU/L) and mean PFOS
(In ng/mL)
Time Period
1999-2018
1999-2018
2003-2018
2003-2018
2017-2018
2017-2018
Sex
Male
Female
Male
Female
Male
Female
Mean In ALT (In IU/L) (y)
3.28
2.96
3.28
2.96
3.29
2.96
Standard Deviation In ALT (In
0.46
0.41
0.46
0.41
0.48
0.42
IU/L) (5)
Mean In PFOS (In ng/mL) (x)
2.40
1.96
2.37
1.93
1.74
1.26
For the BMD analyses, the response of interest is elevated ALT, defined as ALT greater than or
equal to an adverse effect threshold C IU/L defined as 42 IU/C for males and 30 IU/L for
females. EPA estimated P(0), the prevalence of population with elevated ALT using two
approaches. First, the empirical estimate of P(0), "P(0) Empirical," was calculated as the
proportion of the population with ALT greater than or equal to C, using the NHANES survey
weights. Second, the lognormal estimate of P(0), "P(0) Lognormal," was calculated assuming
that ALT is lognormally distributed using the equation:
E-43
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
fln(C) — meanCln ALT))
P(0) Lognormal = 1 — j j
where is the normal cumulative distribution function.
The selected BMR is an extra risk of either 5% or 10%. The extra risk of high ALT is given by
the equation
P(d) -P(0)
Extra Risk = — ———
1 - P(0)
where P(d) is the probability of ALT greater than or equal to C (IU/L) for a given PFOS dose d.
Thus
P(d) = {1 - P(0)} x Extra Risk + P(0)
The values of C, P(0) Empirical, P(d) Empirical, P(d) Lognormal for Extra Risk 5% or 10%, and
P(d) Lognormal for Extra Risk 5% or 10% are shown in Table E-32.
Table E-32. Prevalence of elevated ALT
Time Period
1999-2018
1999-2018
2003-2018
2003-2018
2017-2018
2017-2018
Sex
Male
Female
Male
Female
Male
Female
Adverse effect level C (IU/L)
42
30
42
30
42
30
P(0) Empirical
0.14
0.13
0.15
0.13
0.16
0.13
P(d) Empirical, Extra Risk 5%
0.19
0.17
0.19
0.17
0.20
0.17
P(d) Empirical, Extra Risk 10%
0.23
0.21
0.23
0.21
0.24
0.22
P(0) Lognormal
0.16
0.14
0.16
0.14
0.17
0.15
P(d) Lognormal, Extra Risk 5%
0.20
0.18
0.20
0.18
0.22
0.19
P(d) Lognormal, Extra Risk 10%
0.24
0.23
0.24
0.23
0.26
0.23
The mean In ALTy for a In PFOS dose x is given by the equation
y = mx + b
where m is the slope, P, (from the Gallo regression model) and b is the intercept. The intercept b
is the mean In ALT for a population exposed to 1 ng/mL PFOS. For the U.S. population, the
mean In ALT is y (tabulated above) and the mean In PFOS is x (tabulated above) so the intercept
is given by the equation
b = y — mx
For a given group and dose, the probability of ALT greater than or equal to C is
P = P = P(1-
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
where is the normal cumulative distribution function. Thus, the mean In ALT, y, is the
solution of the last equation, i.e., y = In C — S x — P(d)}
where O-1 is the inverse of the normal cumulative distribution function.
The In PFOS benchmark dose (In BMD) is the corresponding dose x such that y = mx + b. Thus
y — b
In BMD =
m
This gives the PFOS BMD as exp(ln BMD).
For the BMDL, the lower bound of the dose is calculated, so that in the last equation, instead of
m we use the 95th upper limit for P, which is given by
(395 = 95th Upper limit for (3 = (3 + 1.645 x se{(3)
Thus
y — b
In BMDL = V—
(395
This gives the PFOS BMDL as exp(ln BMDL) (Table E-33). Note that P95 is different from the
upper bound of the 95% confidence interval, since that number is the 97.5th percentile.
Table E-33. BMD and BMDL for effect of PFOS (ng/mL) on increased ALT in Gallo et al.
(2012,1276142)
Time Period
1999-2018
1999-2018
2003-2018
2003-2018
2017-2018
2017-2018
Sex
Male
Female
Male
Female
Male
Female
BMR=5%, P(0) Empirical
BMD
124.39
95.88
158.42
91.30
67.81
34.50
BMDL
76.39
56.79
92.20
54.25
41.23
21.81
BMR=5%, P(0) Lognormal
BMD
445.63
269.46
441.62
247.28
210.92
124.77
BMDL
211.73
129.65
209.13
120.26
102.08
60.90
BMR=10%, P(0) Empirical
BMD
3964.56
2624.95
4884.94
2380.02
2011.20
948.37
BMDL
1213.59
799.02
1426.49
733.99
618.43
307.81
BMR=10%, P(0) Lognormal
BMD
11660.73
6185.61
11609.86
5444.59
5311.44
2772.69
BMDL
2873.18
1584.68
2848.49
1421.63
1343.43
725.26
For increased ALT associated with PFOS exposure, the POD is based on the data Gallo et
al. (2012,1276142), a BMR of 5% and a BMDLs of 56.79 ng/mL.
E-45
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.l.4.2 Nian et al. (2019, 5080307)
NOAEC/LOAEC method. Significant positive linear trends were observed for branched PFOS
with ORs of elevated ALT across quartiles of exposure (p-value = 0.04). However, categorical
data, which can be used to develop NOAECs, were not available for total PFOS from the peer-
reviewed publication.
Hybrid method. The previously described hybrid method was implemented using data from Nian
et al. (2019, 5080307). The regression model adjusted for age, sex, career, income, education,
drink, smoke, giblet and seafood consumption, exercise, and BMI. The percentage change in ln-
ALT for ln-unit increase in PFOS was 4.1 (95% CI: 0.6, 7.7) (Table 3, Nian et al. (2019,
5080307). The reported regression coefficient P, which is also referred to as m, was calculated
from the reported percent change expressed as (e|3-l)*100, resulting in a slope of 0.04 (95% CI:
0.01, 0.07) In ALT (IU/L) per In ng/mL PFOS. The estimated BMDs and BMDLs are presented
in Table E-34.
For increased ALT associated with PFOS exposure, the POD is based on the data Nian et al.
(2019, 5080307), a BMR of 5% and a BMDLs of 15.1 ng/mL.
Table E-34. BMD and BMDL for effect of PFOS (ng/mL) on increased ALT in Nian et al.
(2019, 5080307), for 5% and 10% Extra Risk
Time Period 1999-2018 1999-2018 2003-2018 2003-2018 2017-2018 2017-2018
Sex
Male
Female
Male
Female
Male
Female
BMR=5%, P(0) Empirical
BMD
36.82
25.93
41.00
24.89
19.58
10.97
BMDL
22.29
15.12
23.49
14.57
11.73
6.84
BMR=5%, P(0) Lognormal
BMD
69.49
43.37
68.30
40.87
34.44
20.81
BMDL
32.30
20.42
31.64
19.46
16.32
9.94
BMR=10%, P(0) Empirical
BMD
206.25
134.66
225.92
126.14
105.81
57.11
BMDL
60.98
39.58
63.63
37.58
31.43
17.93
BMR=10%, P(0) Lognormal
BMD
352.86
206.31
347.61
190.43
171.58
97.41
BMDL
83.44
50.78
81.84
47.80
41.68
24.50
E. 1.4.3 Summary of Modeling Results for Liver Toxicity
Table E-35 summarizes the PODs resulting from the modeling approaches for increased ALT.
The selected PODs were based on a BMR of 5%, resulting in BMDLs ranging from 15.12 ng/mL
to 56.79 ng/mL, with a selected POD of 15.12 ng/mL.
Table E-35. BMDLs for effect of PFOS on serum ALT using a BMR of 5%.
Study name
BMDL (ng/mL)
Galloetal. (2012, 1276142)
56.79
Nian et al. (2019, 5080307)
15.12
E-46
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.2 Toxicology Studies
E.2.1 Butenhoff et ol. (2012,1276144)/Thomford (2002,
5029075)
EPA conducted dose response modeling of the Butenhoff et al. (2012, 1276144)/Thomford
(2002, 502907) study using the BMDS 3.2 program. This study addresses incidence of adenomas
and/or carcinomas in the liver and pancreas in male rats and the liver and thyroid in female rats,
and individual cell necrosis in the liver in female Sprague-Dawley Crl:CD(SD)IGS BR rats.
E.2.1.1 Hepatocellular Adenomas in Males
Increased incidence of hepatocellular adenomas was observed in male rats. Dichotomous models
were used to fit dose-response data. Multistage models were used consistent with the long-
standing practice of EPA to prefer multistage models to fit tumor dose-response data and a BMR
of 10% extra risk was chosen per EPA's Benchmark Dose Technical Guidance {U.S. EPA, 2012,
1239433}. The dose and response data used for the modeling are listed in Table E-36. The AUC
normalized per day (AUCavg) was selected for this model because the AUC accounts for the
accumulation of effects expected to precede the increased incidence of adenomas and/or
carcinomas. BMD analysis was conducting using both the number of animals at the start of the
study and the number of animals alive at the time of first tumor.
Table E-36. Dose-Response Modeling Data for Hepatocellular Adenomas in Male Rats
Following Exposure to PFOS {Butenhoff, 2012,1276144/Thomford, 2002, 5029075}
Administered Dose
(ppm)
Internal Dose
(mg/L)
Number per Group
at Start of Study
Number per Group
at Time of First
Tumor3
Incidence
0
0.0
50
41
0
0.5
1.4
50
42
3
2
5.9
50
47
3
5
14.3
50
44
1
20
57.8
50
43
7
Notes:
a The time of first occurrence of this tumor was day 512 in males.
BMD modeling results for hepatocellular adenomas following exposure to PFOS for the number
of animals at the start of the study and the number of animals alive at the time of first tumor are
summarized in Table E-37 and Figure E-l and Figure E-2. The best fitting model was the
Multistage Degree 4 model based on adequate p-values (greater than 0.1), the BMDLs were
sufficiently close (less than threefold difference) among adequately fitted models, and the
Multistage Degree 4 model had the lowest Akaike information criterion (AIC). The lower bound
on the dose level corresponding to the 95% lower confidence limit for a 10% response level
BMDLio from the selected Multistage Degree 4 model for the number of animals at the start of
the study is 29.3 mg/L and for the number of animals alive at the time of first tumor is 25.6
mg/L. The number of animals alive at the time of first tumor ensures the potency is not
underestimated by mortality of animals prior to tumor occurrence. The relatively small
E-47
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
difference in the two BMDLio values supports using these values and the selected value is based
on the number of animals alive at the time of first tumor, 25.6 mg/L.
Table E-37. Summary of Benchmark Dose Modeling Results for Data for Hepatocellular
Adenomas in Male Rats Following Exposure to PFOS {Butenhoff, 2012,
1276144/Thomford, 2002, 5029075}
Goodness of
Fit
Scaled Residual
Model3
P-
value
AIC
Dose Group Control
near BMD Dose Group
BMDio
(mg/L)
BMDLi
0
(mg/L)
Basis for Model
Selection
Multistage 0.260 105.2 0.004 -1.35 56.6 29.3
Degree 4
Multistage 0.254 105.2 0.017 -1.34 56.3 29.1
Degree 3
Animals at
the start of Multistage 0.235 105.4 0.065 -1.32 55.9 28.5
the study Degree 2
Multistage 0.192 105.7 0.204 -1.19 54.5 27.6
Degree 1
EPA selected the
Multistage Degree 4
model. All multistage
models had adequate fit
(p-values greater than
0.1), the BMDLs were
sufficiently close (less
than threefold
difference), and the
Multistage Degree 4
model had the lowest
AIC.
Multistage 0.281 100.9
Degree 4
Multistage 0.275 101.0
Animals Dc8rcc 3
alive at the Multistage 0.252 101.2
time of Degree 2
first tumor
Multistage 0.196 101.6
Degree 1
0.005
0.018
0.071
0.238
-1.31
-1.31
-1.29
-1.16
54.2
53.2
51.4
46.8
25.6
25.4
24.9
23.7
EPA selected the
Multistage Degree 4
model. All multistage
models had adequate fit
(p-values greater than
0.1), the BMDLs were
sufficiently close (less
than threefold
difference), and the
Multistage Degree 4
model had the lowest
AIC.
Notes: AIC = Akaike information criterion; BMD :
corresponding to a 10% response level; BMDLio:
limit for a 10% response level.
a Selected model in bold.
benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level
lower bound on the dose level corresponding to the 95% lower confidence
E-48
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1
0.9
0.8
0.7
8 °-6
° 0.5
(u
ce 0.4
Estimated Probability
Response at BMD
— — Linear Extrapolation
O Data
BMD
BMDL
Figure E-l. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 4 Model for Hepatocellular Adenomas in Male Rats Following Exposure to PFOS,
for Number of Animals Per Group at Start of Study {Butenhoff, 2012,1276144/Thomford,
2002, 5029075}
l
0.9
0.8
0.7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
accounts for the accumulation of effects expected to precede the increased incidence of
adenomas and/or carcinomas. BMD analysis was conducting using both the number of animals at
the start of the study and the number of animals alive at the time of first tumor.
Table E-38. Dose-Response Modeling Data for Incidence of Islet Cell Carcinomas in Male
Rats Following Exposure to PFOS {Butenhoff, 2012,1276144/Thomford, 2002, 5029075}
Administered Dose
(ppm)
Internal Dose
(mg/L)
Number per
Group at Start of
Study
Number per
Group at Time of
First Tumor11
Incidence
0
0.0
50
38
1
0.5
1.4
50
41
2
2
5.9
50
44
2
5
14.3
50
44
5
20
57.8
50
40
5
Notes:
a The time of first occurrence of this tumor was day 542 in males.
The BMD modeling results for incidence of islet cell carcinomas following exposure to PFOS
for the number of animals at the start of the study and the number of animals alive at the time of
first tumor are summarized in Table E-39 and Figure E-3 and Figure E-4. The best fitting model
was the Multistage Degree 1 model based on adequate p-values (greater than 0.1), the BMDLs
were sufficiently close (less than threefold difference) among adequately fitted models, and the
higher degree Multistage models estimated parameters at the zero boundary and reduced to the
Multistage Degree 1 model. The BMDLio from the selected Multistage Degree 1 model for the
number of animals at the start of the study is 29.7 mg/L and for the number of animals alive at
the time of first tumor is 26.1 mg/L. The number of animals alive at the time of first tumor
ensures the potency is not underestimated by mortality of animals prior to tumor occurrence. The
relatively small difference in the two BMDLio values supports using these values and the
selected value is based on the number of animals alive at the time of first tumor, 26. lmg/L.
Table E-39. Summary of Benchmark Dose Modeling Results for Incidence of Islet Cell
Carcinomas in Male Rats Following Exposure to PFOS {Butenhoff, 2012,
1276144/Thomford, 2002, 5029075}
Model"
Goodness of
Fit
Scaled Residual
BMDio
BMDLi
0
(mg/L)
Basis for Model
P-
value
AIC
Dose Group
near BMD
Control
Dose Group
(mg/L)
Selection
Multistage
0.526
114.5
-0.434
-0.633
67.6
29.7
EPA selected the
Degree 4
Multistage
0.526
114.5
-0.434
-0.633
67.6
29.7
Multistage Degree 1
model. All multistage
models had adequate fit
Animals at
Degree 3
(p-values greater than
the start of
the study
Multistage
Degree 2
0.526
114.5
-0.434
-0.633
67.6
29.7
0.1), the BMDLs were
sufficiently close (less
than threefold
Multistage
Degree 1
0.526
114.5
-0.434
-0.633
67.6
29.7
difference), and higher
degree models reduced
E-50
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Goodness of
Fit
Scaled Residual
Model3
P-
value
AIC
Dose Group Control
near BMD Dose Group
BMDio
(mg/L)
BMDLi
0
(mg/L)
Basis for Model
Selection
to the Multistage
Degree 1 model.
Multistage
Degree 4
Multistage
Animals Degree 3
alive at the
time of Multistage
first tumor Degree 2
0.554 111.2
0.554 111.2
0.554 111.2
Multistage 0.554 111.2
Degree 1
-0.417
-0.417
-0.417
-0.417
-0.590
-0.590
-0.590
-0.590
58.5
58.5
58.5
58.5
26.1
26.1
26.1
26.1
EPA selected the
Multistage Degree 1
model. All multistage
models had adequate fit
(p-values greater than
0.1), the BMDLs were
sufficiently close (less
than threefold
difference), and higher
degree models reduced
to the Multistage
Degree 1 model.
Notes: AIC = Akaike information criterion; BMD :
corresponding to a 10% response level; BMDLio:
limit for a 10% response level.
a Selected model in bold.
benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level
lower bound on the dose level corresponding to the 95% lower confidence
1
0.9
0.8
0.7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1
0.9
0.8
0.7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The BMD modeling results for combined incidence of islet cell adenomas and carcinomas
following exposure to PFOS for the number of animals at the start of the study and the number of
animals alive at the time of first tumor are summarized in Table E-41 and Figure E-5 and Figure
E-6. The best fitting model was the Multistage Degree 1 model based on adequate p-values
(greater than 0.1), the benchmark dose lower limits (BMDLs) were sufficiently close (less than
threefold difference) among adequately fitted models, and the higher degree Multistage models
estimated parameters at the zero boundary and reduced to the Multistage Degree 1 model. The
BMDLio from the selected Multistage Degree 1 model for the number of animals at the start of
the study is 25.1mg/L and for the number of animals alive at the time of first tumor is 21.7mg/L.
The number of animals alive at the time of first tumor ensures the potency is not underestimated
by mortality of animals prior to tumor occurrence. The relatively small difference in the two
BMDLio values supports using these values and the selected value is based on the number of
animals alive at the time of first tumor, 21.7mg/L. The combined islet cell adenomas and
carcinomas in males were not considered further because the response of the islet cell adenomas
and carcinomas in the high dose group was not statistically different from the control group,
though the trend of response across dose groups was statistically significant.
Table E-41. Summary of Benchmark Dose Modeling Results for Combined Incidence of
Islet Cell Adenomas and Carcinomas in Male Rats Following Exposure to PFOS
{Butenhoff, 2012,1276144/Thomford, 2002, 5029075}
Model"
Goodness of Fit
Scaled Residual
p-value AIC
Dose Group
near BMD
Control Dose
Group
BMDio
(mg/L)
BMDLio
(mg/L)
Basis for Model Selection
Multistage
0.909
197.34
-0.191
-0.214
63.8
25.1
EPA selected the
Degree 4
Multistage Degree 1
model. All multistage
Multistage
0.909
197.34
-0.191
-0.214
63.8
25.1
models had adequate fit
Degree 3
(p-values greater than 0.1),
Multistage
0.909
197.34
-0.191
-0.214
63.8
25.1
the BMDLs were
Degree 2
sufficiently close (less
than threefold difference),
Multistage
0.909
197.34
-0.191
-0.214
63.8
25.1
and higher degree models
Degree 1
reduced to the Multistage
Degree 1 model.
Multistage
0.938
190.0
-0.162
-0.130
53.6
21.7
EPA selected the
Degree 4
Multistage Degree 1
model. All multistage
Multistage
0.938
190.0
-0.162
-0.130
53.6
21.7
models had adequate fit
Degree 3
(p-values greater than 0.1),
Multistage
0.938
190.0
-0.162
-0.130
53.6
21.7
the BMDLs were
Degree 2
sufficiently close (less
than threefold difference),
Multistage
0.938
190.0
-0.162
-0.130
53.6
21.7
and higher degree models
Degree 1
reduced to the Multistage
Degree 1 model.
Notes: AIC = Akaike information criterion; BMD :
corresponding to a 10% response level; BMDLio:
limit for a 10% response level.
a Selected model in bold.
benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level
lower bound on the dose level corresponding to the 95% lower confidence
E-53
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1
0.9
0.8
0.7
8 °-6
° 0.5
(u
ce 0.4
Estimated Probability
Response at BMD
— — Linear Extrapolation
O Data
BMD
BMDL
10
20
30
Dose
40
50
Figure E-5. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 1 Model for Combined Incidence of Islet Cell Adenomas and Carcinomas in Male
Rats Following Exposure to PFOS, for Number of Animals Per Group at Start of Study
{Butenhoff, 2012,1276144/Thomford, 2002, 5029075}
l
0.9
0.8
0.7
Estimated Probability
Response at BMD
— — Linear Extrapolation
O Data
BMD
BMDL
10
20
30
Dose
40
50
Figure E-6. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 1 Model for Combined Incidence of Islet Cell Adenomas and Carcinomas in Male
Rats Following Exposure to PFOS, for Number of Animals Per Group at Time of First
Tumor {Butenhoff, 2012,1276144/Thomford, 2002, 5029075}
E.2.1.4 Hepatocellular Adenomas in Females
Increased incidence of hepatocellular adenomas was observed in female rats. Dichotomous
models were used to fit dose-response data. A BMR of 10% extra risk per EPA's Benchmark
Dose Technical Guidance {U.S. EPA, 2012, 1239433}. The doses and response data used for the
E-54
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
modeling are listed in Table E-42. The AUCavg was selected for this model because the AUC
accounts for the accumulation of effects expected to precede the increased incidence of
adenomas and/or carcinomas. BMD analysis was conducting using both the number of animals at
the start of the study and the number of animals alive at the time of first tumor.
Table E-42. Dose-Response Modeling Data for Hepatocellular Adenomas in Female Rats
Following Exposure to PFOS {Butenhoff, 2012,1276144/Thomford, 2002, 5029075}
Administered Dose
(ppm)
Internal Dose
(mg/L)
Number per
Group at Start of
Study
Number per
Group at Time of
First Tumor3
Incidence
0
0.0
50
28
0
0.5
1.6
50
26
1
2
6.6
49
15
1
5
16.1
50
28
1
20
65.2
50
31
5
Notes:
a The time of first occurrence of this tumor was day 653 in females.
The BMD modeling results for hepatocellular adenomas following exposure to PFOS for the
number of animals at the start of the study and the number of animals alive at the time of first
tumor are summarized in Table E-43and Figure E-7 and Figure E-8. The best fitting model was
the Multistage Degree 1 model based on adequate p-values (greater than 0.1), the BMDLs) were
sufficiently close (less than threefold difference) among adequately fitted models, and the
Multistage Degree 1 model had the lowest AIC. The BMDLio from the selected Multistage
Degree 1 model for the number of animals at the start of the study is 37.2 mg/L and for the
number of animals alive at the time of first tumor is 21.8 mg/L. The number of animals alive at
the time of first tumor ensures the potency is not underestimated by mortality of animals prior to
tumor occurrence. The relatively small difference in the two BMDLio values supports using
these values and the selected value is based on the number of animals alive at the time of first
tumor, 21.8 mg/L.
Table E-43. Summary of Benchmark Dose Modeling Results for Data for Hepatocellular
Adenomas in Female Rats Following Exposure to PFOS {Butenhoff, 2012,
1276144/Thomford, 2002, 5029075}
Goodness of
Fit
Scaled Residual
Model3
P-
value
AIC
Dose
Group
near BMD
Control
Dose
Group
BMDio
(mg/L)
BMDLi
(mg/L)
Basis for Model
Selection
Multistage
Degree 4b
Animals at ,, . . ,
the start of Multlsta§e
the study Degree 3
Multistage
Degree 2
0.601 69.2
0.598 69.3
0.586 69.3
0.00105
0.00722
0.02918
-0.668
-0.665
-0.655
68.3
69.0
70.5
37.4
37.4
37.3
EPA selected the
Multistage Degree 1
model. All multistage
models had adequate fit
(p-values greater than
0.1), the BMDLs were
E-55
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Goodness of
Fit
Scaled Residual
Model3
P-
value
AIC
Dose
Group
near BMD
Control
Dose
Group
BMDio
(mg/L)
BMDLi
0
(mg/L)
Basis for Model
Selection
Multistage
0.761
67.3
0.08232
-0.608
73.0
37.2
sufficiently close (less
Degree 1
than threefold difference),
the Multistage Degree 1
model had the lowest AIC.
Multistage
0.449
59.8
0.0024
-0.719
46.7
21.8
EPA selected the
Degree 4
Multistage Degree 1
model. All multistage
Multistage
0.447
59.8
0.0094
-0.713
45.4
21.8
models had adequate fit
Degree 3
(p-values greater than
Animals
alive at the
time of
first tumor
0.1), the BMDLs were
Multistage
Degree 2°
0.654
57.8
0.0228
-0.701
43.9
21.8
sufficiently close (less
than threefold difference),
and the Multistage Degree
1 model had the lowest
Multistage
0.654
57.8
0.0228
-0.701
43.9
21.8
AIC (the Degree 2 model
Degree 1
estimated parameters at
the zero boundary and
reduced to the Degree 1
model).
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level
corresponding to a 10% response level; BMDLio = lower bound on the dose level corresponding to the 95% lower confidence
limit for a 10% response level.
a Selected model in bold.
b Degree 3 and 4 models estimated parameters at the zero boundary and reduced to the Multistage Degree 2 model.
c Degree 2 model estimated parameters at the zero boundary and reduced to the Multistage Degree 1 model.
Estimated Probability
Response at BMD
— — Linear Extrapolation
O Data
BMD
BMDL
Figure E-7. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 4 Model for Hepatocellular Adenomas in Female Rats Following Exposure to
E-56
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOS, for Number of Animals Per Group at Start of Study {Butenhoff, 2012,
1276144/Thomford, 2002, 5029075}
l
0.9
0.8
0.7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
a The time of first occurrence of this tumor was day 653 in females.
The BMD modeling results for hepatocellular adenomas following exposure to PFOS for the
number of animals at the start of the study and the number of animals alive at the time of first
tumor are summarized in Table E-45 and Figure E-9 and Figure E-10. The best fitting model was
the Multistage Degree 1 model based on adequate p-values (greater than 0.1), the BMDLs were
sufficiently close (less than threefold difference) among adequately fitted models, and the
Multistage Degree 1 model had the lowest AIC. The BMDLio from the selected Multistage
Degree 1 model for the number of animals at the start of the study is 32.7 mg/L and for the
number of animals alive at the time of first tumor is 19.8 mg/L. The number of animals alive at
the time of first tumor ensures the potency is not underestimated by mortality of animals prior to
tumor occurrence. The relatively small difference in the two BMDLio values supports using
these values and the selected value is based on the number of animals alive at the time of first
tumor, 19.8 mg/L.
Table E-45. Summary of Benchmark Dose Modeling Results for Data for Hepatocellular
Adenomas and Carcinomas in Female Rats Following Exposure to PFOS {Butenhoff, 2012,
1276144/Thomford, 2002, 5029075}
Goodness of
Fit
Scaled Residual
Model"
P-
value
AIC
Dose
Group
near BMD
Control
Dose
Group
BMDio
(mg/L)
BMDLi
(mg/L)
Basis for Model
Selection
61.8
33.2
61.2
33.2
60.6
33.0
60.3
32.7
Multistage
Degree 4
Multistage
Animals at De§ree 3
the start of Multistage
the study Degree 2
Multistage
Degree 1
0.600 73.4
0.597 73.4
0.581 73.5
0.723 71.6
0.0021
0.0081
0.0331
0.1462
-0.668
-0.667
-0.663
-0.565
EPA selected the
Multistage Degree 1
model. All multistage
models had adequate fit
(p-values greater than
0.1), the BMDLs were
sufficiently close (less
than threefold difference),
the Multistage Degree 1
model had the lowest AIC.
Multistage
0.466
63.8
0.0029
-0.716
47.5
20.0
Degree 4
Multistage
0.461
63.8
0.0109
-0.711
45.2
20.0
Animals
Degree 3
alive at the
Multistage
0.449
63.8
0.0415
-0.694
41.7
19.9
time of
Degree 2b
first tumor
Multistage
Degree 1
0.643
61.8
-0.613
-0.630
37.2
19.8
EPA selected the
Multistage Degree 1
model. All multistage
models had adequate fit
(p-values greater than
0.1), the BMDLs were
sufficiently close (less
than threefold difference),
and the Multistage Degree
1 model had the lowest
AIC.
benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level
lower bound on the dose level corresponding to the 95% lower confidence
Notes: AIC = Akaike information criterion; BMD
corresponding to a 10% response level; BMDLio
limit for a 10% response level.
a Selected model in bold.
b Degree 2 model estimated parameters at the zero boundary and reduced to the Multistage Degree 1 model.
E-58
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1
0.9
0.8
0.7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
BMR of 10% extra risk was chosen per EPA's Benchmark Dose Technical Guidance {U.S. EPA,
2012, 1239433}. The doses and response data used for the modeling are listed in Table E-46.
The average concentration over the final week of study Ciast7,avg, was selected for this model
rather than alternate metrics such as Cmax because the average blood concentration is expected to
better correlate with an accumulation of individual cell necrosis in the liver.
Table E-46. Dose-Response Modeling Data for Individual Cell Necrosis in the Liver in
Female Sprague-Dawley Crl:CD(SD)IGS BR Rats Following Exposure to PFOS
{Butenhoff, 2012,1276144/Thomford, 2002, 5029075}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Incidence
0
0
50
3
0.029
1.8
50
4
0.120
7.4
50
4
0.299
18.0
50
5
1.251
72.5
50
9
BMD modeling results for individual cell necrosis in the liver are summarized in Table E-47 and
Figure E-l 1. The Log-Logistic model was selected based on adequate p-values (greater than 0.1)
and had the lowest AIC mong adequately fitting with BMD/BMDL ratios less than 3. The
BMDLio from the selected Log-Logistic model is 27.0 mg/L.
Table E-47. Summary of Benchmark Dose Modeling Results for Individual Cell Necrosis in
the Liver in Female Sprague-Dawley Crl:CD(SD)IGS BR Rats Following Exposure to
PFOS {Butenhoff, 2012,1276144/Thomford, 2002, 5029075}
Goodness of Fit
Scaled Residual
BMDio
BMDLio
Basis for Model
Model3
p-value
AIC
Dose Group
Control Dose
(mg/L)
(mg/L)
Selection
near BMD
Group
Dichotomous
0.947
164.2
0.003
-0.201
57.1
9.4
EPA selected the
Hill
Log-Logistic
Gamma
0.990
162.2
-0.024
-0.239
59.2
29.0
model. All models
Log-Logistic
0.990
162.2
-0.017
-0.226
58.5
27.0
had adequate fit (p-
Multistage
Degree 4
0.990
162.2
-0.024
-0.239
59.2
29.0
values greater than
0.1). The
Dichotomous Hill
Multistage
Degree 3
0.990
162.2
-0.024
-0.239
59.2
29.0
and Log-Probit
were the only
Multistage
0.990
162.2
-0.024
-0.239
59.2
29.0
models that did not
Degree 2
have BMD/BMDL
Multistage
0.990
162.2
-0.024
-0.239
59.2
29.0
ratio <3. Of the
Degree 1
remaining models,
Weibull
0.990
162.2
-0.024
-0.239
59.2
29.0
the Log-Logistic
Logistic
0.981
162.3
-0.040
-0.334
64.2
41.8
model had the
Log-Probit
0.938
164.2
0.022
-0.208
57.0
0.6
lowest AIC.
Probit
0.983
162.3
-0.041
-0.322
63.5
39.9
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level
corresponding to a 10% response level; BMDLio = lower bound on the dose level corresponding to the 95% lower confidence
limit for a 10% response level.
E-60
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
a Selected model in bold.
1
0.9
0.8
Frequentist Log-Logistic Model with BMR of 10% Extra Risk for
the BMD and 0.95 Lower Confidence Limit for the BMDL
10
20
30
Dose
40
50
Estimated Probability
^^Response at BMD
O Data
BMD
BMDL
Figure E-ll. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Log-Logistic
Model for Individual Cell Necrosis in the Liver in Female Sprague-Dawley Crl:CD(SD)IGS
BR Rats Following Exposure to PFOS {Butenhoff, 2012,1276144/Thomford, 2002,
5029075}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
Increased incidence of individual cell necrosis in the liver was observed in male Sprague-Dawley
Crl:CD(SD)IGS BR rats. Dichotomous models were used to fit dose-response data. A BMR of
10% extra risk was chosen per EPA's Benchmark Dose Technical Guidance {U.S. EPA, 2012,
1239433}. The doses and response data used for the modeling are listed in Table E-48. The Czavg
was selected for this model rather than alternate metrics such as Cmax because the average blood
concentration is expected to better correlate with an accumulation of individual cell necrosis in
the liver.
Table E-48. Dose-Response Modeling Data for Individual Cell Necrosis in the Liver in Male
Sprague-Dawley Crl:CD(SD)IGS BR Rats Following Exposure to PFOS {Butenhoff, 2012,
1276144/Thomford, 2002, 5029075}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Incidence
0
0
50
3
0.024
1.1
50
2
0.098
4.5
50
6
0.242
11.0
50
4
0.984
44.8
50
10
E-61
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
BMD modeling results for individual cell necrosis in the liver are summarized in Table E-49 and
Figure E-12. The best fitting model was the Log-Logistic model based on adequate p-values
(greater than 0.1), the BMDLs were sufficiently close (less than threefold difference) among
adequately fitted models, and the Gamma model had the lowest AIC. The BMDLio from the
selected Log-Logistic model is 14.2 mg/L.
Table E-49. Summary of Benchmark Dose Modeling Results for Individual Cell Necrosis in
the Liver in Male Sprague-Dawley Crl:CD(SD)IGS BR Rats Following Exposure to PFOS
{Butenhoff, 2012,1276144/Thomford, 2002, 5029075}
Goodness of Fit
Scaled Residual
Model3
p-value AIC
Dose Group Control Dose
near BMD Group
BMDio
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Dichotomous
0.369
162.02
0.003
-0.201
25.3
1.9
Hill
Gamma
0.283
162.59
-0.004
-0.397
40.4
14.9
Log-Logistic
0.563
160.04
-0.007
-0.004
28.4
14.2
Multistage
0.560
160.05
-0.016
-0.028
29.2
15.6
Degree 4
Multistage
0.560
160.05
-0.016
-0.028
29.2
15.6
Degree 3
Multistage
0.560
160.05
-0.016
-0.028
29.2
15.6
Degree 2
Multistage
0.560
160.05
-0.016
-0.028
29.2
15.6
Degree 1
Weibull
0.560
160.05
-0.016
-0.028
29.2
15.6
Logistic
0.533
160.19
-0.030
-0.184
34.7
24.3
Log-Probit
0.386
161.98
-0.667
0.180
23.8
6.5
Probit
0.536
160.17
-0.031
-0.165
34.0
23.0
EPA selected the
Log-Logistic. All
models had
adequate fit (p-
values greater than
0.1). The
Dichotomous Hill
and Log-Probit
were the only
models that did not
have BMD/BMDL
ratio <3. Of the
remaining models,
the Log-Logistic
model had the
lowest AIC.
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL
corresponding to a 10% response level; BMDLio = lower bound on the dose
limit for a 10% response level.
a Selected model in bold.
= benchmark dose lower limit; BMDio = dose level
level corresponding to the 95% lower confidence
E-62
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Frequentist Log-Logistic Model with BMR of 10% Extra Risk for
the BMD and 0.95 Lower Confidence Limit for the BMDL
^—Estimated Probability
^—Response at BMD
O Data
BMD
BMDL
Figure E-12. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Log-Logistic
Model for Individual Cell Necrosis in the Liver in Male Sprague-Dawley Crl:CD(SD)IGS
BR Rats Following Exposure to PFOS {Butenhoff, 2012,1276144/Thomford, 2002,
5029075}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.2 Lee et al. (2015, 2851075)
EPA conducted dose response modeling of the Lee et al. (2015, 2851075) study using the BMDS
3.2 program. This study addresses fetal body weight in Fi male and female CD-I mice and the
number of dead fetuses in Po female CD-I mice.
E. 2.2.1 Fetal Body Weight
Decreased mean response of fetal body weight was observed in Fi male and female CD-I mice.
Continuous models were used to fit dose-response data. BMR of a change in the mean equal to
0.5 standard deviations from the control mean and 5% change were chosen. The doses and
response data used for the modeling are listed in Table E-50. The Cavg.Pup.gest was selected for this
model rather than alternate metrics such as Cmax because the average blood concentration is
expected to better correlate with an accumulation of effect resulting in decreased fetal body
weight.
Table E-50. Dose-Response Modeling Data for Fetal Body Weight in Fi Male and Female
CD-I Mice Following Exposure to PFOS {Lee, 2015, 2851075}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Mean Response (g)a
0
0
10
1.7 ±0.2
0.5
0.9
10
1.5 ±0.1
2
3.5
10
1.3 ±0.1
l
0.9
0.8
0.7
0 5 10 15 20 25 30 35 40
Dose
E-63
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Administered Dose
Internal Dose
(mg/kg/day)
(mg/L)
Number per Group
Mean Response (g)a
8
14.0
10
1.1 ± 0.2
Notes:
aData are presented as mean ± standard deviation.
BMD modeling results for fetal body weight are summarized in and Table E-51. No models
provided an adequate fit, therefore a NOAEL approach was taken for this endpoint.
E-64
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-51. Summary of Benchmark Dose Modeling Results for Fetal Body Weight in Fi Male and Female CD-I Mice
Following Exposure to PFOS (constant variance) {Lee, 2015, 2851075}
Goodness of
Fit
Scaled Residual
Model
p- AIC Dose Group Dose Group Control Dose
value near BMDo.ssd nearBMDs Group
BMDo.ssd
(mg/L)
BMDLo.ssd
(mg/L)
BMDs
(mg/L)
BMDLs
(mg/L)
Basis for Model
Selection
Exponential 2 0.002
Exponential 3 0.002
Exponential 4 0.360
Exponential 5 0.360
Hill
Polynomial
Degree 3
Polynomial
Degree 2
Power
Linear
0.685
0.001
0.001
0.001
0.001
-19.3
-19.3
-29.4
-29.4
-30.1
-17.5
-17.5
-17.5
-17.5
-0.7
-0.7
0.4
0.4
0.1
-2.4
-2.4
-2.4
-2.4
-0.7
-0.7
-0.7
-0.7
0.1
-0.6
-0.6
-0.6
-0.6
2.2
2.2
0.4
0.4
0.1
2.4
2.4
2.4
2.4
2.0
2.0
0.4
0.4
0.3
2.5
2.5
2.5
2.5
1.5
1.5
0.2
0.2
0.2
1.9
1.9
1.9
1.9
1.8
1.8
0.5
0.5
0.3
2.2
2.2
2.2
2.2
1.4
1.4
0.3
0.3
0.2
1.8
1.8
1.8
1.8
No models had
adequate fit for
the constant or
non-constant
variance (p-
values were less
than 0.05).
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.ssd = dose level corresponding to a change in the mean equal to
0.5 standard deviations from the control mean; BMDLo.ssd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.5
standard deviations from the control mean; BMD5 = dose level corresponding to a 5% change in the mean from the control mean; BMDL5 = lower bound on the dose level
corresponding to the 95% lower confidence limit for a 5% change in the mean from the control mean.
E-65
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.2.2.2 Number of Dead Fetuses
Increased mean response of fetal body weight was observed in Po female CD-I mice. Continuous
models were used to fit dose-response data. A BMR of a change in the mean equal to 0.5
standard deviations from the control mean was chosen. The doses and response data used for the
modeling are listed in Table E-52. The average concentration normalized per day during
gestation (Cavg,dam,gest) and maximum maternal concentration during gestation (Cmax,dam) were
both considered and shown below because fetal death could be a result of exposure during a
sensitive window of development where a Cmax metric is a more appropriate dose metric or an
accumulation of exposure where an average concentration metric is more appropriate. The
Cavg,dam,gest was selected for this model.
Table E-52. Dose-Response Modeling Data for Number of Dead Fetuses in Po Female CD-I
Mice Following Exposure to PFOS {Lee, 2015, 2851075}
Administered Dose
(mg/kg/day)
Internal Dose
AUCavg,dam,gest
(mg/L)
Cmax,dam
(mg/L)
Number per Group
Mean Response
(incidence)
0
0
0
10
0.6 ±0.3
0.5
2.1
9.2
10
1.6 ±0.5
2
8.5
37.0
10
4.8 ±0.5
8
34.1
147.8
10
7.6 ±1.1
Notes:
aData are presented as mean ± standard deviation.
The BMD modeling results for fetal body weight for Cavg,dam,gest and Cmax,dam are summarized in
Table E-53 and Table E-54, respectively. No models provided an adequate fit, therefore a
LOAEL approach was taken for this endpoint using the Cavg,dam,gest internal dose metric.
Table E-53. Summary of Benchmark Dose Modeling Results for Number of Dead Fetuses
for Cavg,dam,gest in Po Female CD-I Mice Following Exposure to PFOS (nonconstant
variance) {Lee, 2015, 2851075}
Model
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control Dose
near BMD Group
BMDo.ssd
(mg/L)
BMDLo ssd Basis for Model
(mg/L) Selection
Exponential 2
< 0.000
148.1
4.2
-3.1
8.2
6.5
No models had
1
adequate fit (p-
Exponential 3
< 0.000
148.1
4.2
-3.1
8.2
6.5
values were less
1
than 0.1).
Exponential 4
0.045
76.8
0.5
0.5
0.2
0.2
Exponential 5
a
74.8
-4.1 x e~3
-4.1 x e"3
0.4
0.2
Hill
a
74.8
-4.1 x e~3
-4.1 x e~3
0.5
0.3
Polynomial
< 0.000
120.7
-1.1
-1.1
0.4
0.3
Degree 3
1
Polynomial
< 0.000
120.7
-1.1
-1.1
0.4
0.3
Degree 2
1
Power
< 0.000
1
120.7
-1.1
-1.1
0.4
0.3
E-66
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Goodness of Fit Scaled Residual
BMDo.ssd BMDLo.ssd Basis for Model
MOdCl p-value AIC D°Se ^°"P ControlDose (mfi/L, (mfi/L, Selection
r near BMD Group
Linear < 0.000 120.7 -LI -LI 04 03
1
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.ssd = dose
level corresponding to a change in the mean equal to 0.5 standard deviations from the control mean; BMDLo.ssd = lower bound
on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.5 standard deviations
from the control mean.
a Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
Table E-54. Summary of Benchmark Dose Modeling Results for Number of Dead Fetuses
for Cmax,dam in Po Female CD-I Mice Following Exposure to PFOS (nonconstant variance)
{Lee, 2015, 2851075}
Goodness of Fit
Scaled Residual
BMDo.ssd
(mg/L)
BMDLo.ssd
(mg/L)
Basis for Model
Selection
Model
p-value
AIC
Dose Group
near BMD
Control Dose
Group
Exponential 2
Exponential 3
<0.0001
<0.0001
148.1
148.1
4.2
4.2
-3.1
-3.1
35.3
35.3
28.1
28.1
No models had
adequate fit (p-
values were less
Exponential 4
0.045
76.8
0.5
0.5
1.0
0.7
than 0.1).
Exponential 5
a
74.8
-4.1 x e~3
-4.1 x e~3
1.9
1.0
Hill
a
74.8
-4.1 x e~3
-4.1 x e~3
2.3
1.3
Polynomial
Degree 3
<0.0001
120.7
-1.1
-1.1
1.9
1.2
Polynomial
Degree 2
<0.0001
120.7
-1.1
-1.1
1.9
1.2
Power
<0.0001
120.7
-1.1
-1.1
1.9
1.2
Linear
<0.0001
120.7
-1.1
-1.1
1.9
1.2
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.ssd = dose
level corresponding to a change in the mean equal to 0.5 standard deviations from the control mean; BMDLo.ssd = lower bound
on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.5 standard deviations
from the control mean.
a Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
E.2.3 Luebker et al. (2005, 757857)
EPA conducted dose response modeling of the Luebker et al. (2005, 757857) study using the
BMDS 3.2 program. This study addresses pup body weight relative to the litter at LD 5 in Fi
male and female Sprague-Dawley rats.
E.2.3.1 Pup Body Weight Relative to Litter at LD 5
Decreased mean response of pup body weight relative to the litter at LD 5 was observed in Fi
male and female Sprague-Dawley rats. Continuous models were used to fit dose-response data.
A BMR of a 5% change from the control mean was selected and a BMR of a 0.5 standard
deviation change from the mean is provided for comparison purposes. The doses and response
data used for the modeling are listed in Table E-55. The Cavg,pup,gest was selected for this model
E-67
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
rather than alternate metrics such as Cmax because the average concentration normalized per day
during gestation is expected to better correlate with an accumulation of effect resulting in
decreased pup body weight.
Table E-55. Dose-Response Modeling Data for Pup Body Weight Relative to the Litter
(LD5) in Fi Male and Female Sprague-Dawley Rats Following Exposure to PFOS
{Luebker, 2005, 757857}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Mean Response (g)a
0
0
17
9.8 ± 2.1b
0.4
15.4
17
8.6 ± 1.9
0.8
30.8
17
8.5 ±2.8
1
38.5
17
8.1 ± 2.5
1.2
46.1
17
7.5 ±2.7
1.6
61.5
17
7.2 ±2.7
2
76.9
17
7.3 ±7.3
Notes:
a Data are presented as mean ± standard deviation.
b Standard deviations were calculated from standard errors.
The dose response data for the highest dose group was removed prior to modeling as the variance
surrounding the mean response for this group was large. The BMD modeling results for pup
body weight relative to the litter at LD 5 are summarized in Table E-56 and Figure E-13. The
Polynomial Degree 6 model was selected as it had the lowest AIC among the viable models. The
BMDLs from the selected Polynomial Degree 6 model is 10.1 mg/L.
E-68
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-56. Summary of Benchmark Dose Modeling Results for Pup Body Weight Relative to the Litter (LD5) in Fi Male and
Female Sprague-Dawley Rats Following Exposure to PFOS (nonconstant variance) {Luebker, 2005, 757857}
Goodness of Fit Scaled Residual
BMDo.ssd BMDLo.ssd BMDs BMDLs Basis for Model
p" AIC DoseGrouP Dose Group Control (mg/L) (mg/L) (mg/L) (mg/L) Selection
value near BMDo s near BMDs Dose Group
Exponential 2
0.069
602.1
-0.73
-0.7
0.38
18.9
11.7
10.8
7.2
EPA selected the
Exponential 3
0.199
599.9
-0.69
-0.3
1.45
43.8
20.4
30.7
12.7
Polynomial
Degree 6 model.
Exponential 4
0.069
602.1
-0.73
-0.7
0.38
18.9
11.7
10.8
7.2
All models had
Exponential 5
0.199
599.9
-0.69
-0.3
1.45
43.8
20.4
30.7
12.7
adequate fit (p-
Hill
0.142
601.3
-0.36
-0.2
1.33
41.4
19.1
28.2
16.4
values greater
than 0.1), and the
Polynomial
0.808
594.2
0.01
-0.7
0.93
33.1
17.9
16.6
10.1
Polynomial
Degree 6
Degree 6 model
Polynomial
0.736
594.6
-0.03
-0.7
0.96
34.1
18.1
17.4
10.2
was selected as it
Degree 5
had the lowest
Polynomial
0.640
595.3
-0.29
-0.7
0.99
35.0
18.3
18.7
10.4
AIC among the
Degree 4
viable models.
Polynomial
0.381
598.1
-0.28
-0.8
1.02
35.7
18.4
20.6
10.6
Degree 3
Polynomial
0.266
599.1
-0.23
-0.9
1.06
35.6
17.9
22.6
10.6
Degree 2
Power
0.245
599.3
-0.36
-0.2
1.34
41.8
18.8
28.5
11.3
Linear
0.133
600.3
-0.81
-0.8
0.34
19.7
13.3
11.3
8.2
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.ssd = dose level corresponding to a change in the mean equal to
0.5 standard deviations from the control mean; BMDLo.ssd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.5
standard deviations from the control mean; BMD5 = dose level corresponding to a 5% change in the mean from the control mean; BMDL5 = lower bound on the dose level
corresponding to the 95% lower confidence limit for a 5% change in the mean from the control mean.
a Selected model in bold
E-69
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
12
Estimated Probability
^^Response at BMD
O Data
BMD
BMDL
10
20
30 40
Dose
50
60
70
Figure E-13. Plot of Mean Response by Dose with Fitted Curve for the Selected Polynomial
Degree 6 Model for Pup Body Weight Relative to the Litter at LD5 in Fi Male and Female
Sprague-Dawley Rats Following Exposure to PFOS {Luebker, 2005, 757857}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.4 NTP (2019, 5400978)
EPA conducted dose response modeling of the NTP (2019, 5400978) study using the BMDS 3.2
program. This study addresses extramedullar hematopoiesis in the spleen in male and female
Sprague-Dawley rats.
E.2.4.1 Extramedullary Hematopoiesis in the Spleen in Male
Sprague-Dawley Rats
Increased incidence of extramedullar hematopoiesis in the spleen was observed in male
Sprague-Dawley rats. Dichotomous models were used to fit dose-response data. A BMR of 10%
extra risk was chosen per EPA's Benchmark Dose Technical Guidance {U.S. EPA, 2012,
1239433}. The doses and response data used for the modeling are listed in Table E-57. The Ciastz
avg was selected for this model rather than alternate metrics such as Cmax because the average
blood concentration is expected to better correlate with an accumulation of extramedullar
hematopoiesis in the spleen.
Table E-57. Dose-Response Modeling Data for Extramedullary Hematopoiesis in Male
Sprague-Dawley Rats Following Exposure to PFOS {NTP, 2019, 5400978}
Administered Dose
Internal Dose
(mg/kg/day)
(mg/L)
Number per Group
Incidence
0
0
10
1
0.312
10.2
10
1
E-70
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Incidence
0.625
20.4
10
2
1.25
40.8
10
7
2.5
81.6
10
8
5
162.7
10
10
The BMD modeling results for extramedullary hematopoiesis in the spleen are summarized in
Table E-58 and Figure E-14. The best fitting model was the Logistic model based on adequate p-
values (greater than 0.1), the BMDLs were sufficiently close (less than threefold difference)
among adequately fitted models, and the Logistic model had the lowest AIC. The BMDLio from
the selected Logistic model is 9.6 mg/L.
Table E-58. Summary of Benchmark Dose Modeling Results for Extramedullary
Hematopoiesis in Male Sprague-Dawley Rats Following Exposure to PFOS {NTP, 2019,
5400978}
Goodness of Fit
Scaled Residual
BMDio
BMDLio
Basis for Model
Model3
p-value
AIC
Dose Group
Control Dose
(mg/L)
(mg/L)
Selection
near BMD
Group
Dichotomous
0.646
53.0
-0.3
0.2
15.7
7.1
EPA selected the
Hill
Logistic model. All
Gamma
0.594
53.2
-0.3
0.2
13.8
4.6
models had
Log-Logistic
0.646
53.0
-0.3
0.2
15.7
7.1
adequate fit (p-
values greater than
0.1), the BMDLs
Multistage
0.487
53.7
-0.5
0.3
10.9
4.2
Degree 5
were sufficiently
Multistage
0.487
53.7
-0.5
0.3
10.9
4.2
close (less than
Degree 4
threefold
Multistage
0.487
53.7
-0.5
0.3
10.9
4.3
difference), and the
Degree 3
Logistic model had
the lowest AIC.
Multistage
0.487
53.7
-0.5
0.3
10.9
4.3
Degree 2
Multistage
0.475
53.4
-1.0
0.6
5.4
3.7
Degree 1
Weibull
0.549
53.4
-0.4
0.3
12.1
4.4
Logistic
0.558
52.2
-0.6
-0.1
14.0
9.6
Log-Probit
0.676
52.8
-0.4
0.2
16.0
7.5
Probit
0.558
52.3
-0.6
0.0
13.4
9.5
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level
corresponding to a 10% response level; BMDLio = lower bound on the dose level corresponding to the 95% lower confidence
limit for a 10% response level.
a Selected model in bold.
E-71
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-14. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Logistic
Model for Extramedullar Hematopoiesis in the Spleen in Male Sprague-Dawley Rats
Following Exposure to PFOS {NTP, 2019, 5400978}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.4.2 Extramedullary Hematopoiesis in the Spleen in Female
Sprague-Dawley Rats
Increased incidence of extramedullar hematopoiesis in the spleen was observed in female
Sprague-Dawley rats. Dichotomous models were used to fit dose-response data. A BMR of 10%
extra risk was chosen per EPA's Benchmark Dose Technical Guidance {U.S. EPA, 2012,
1239433}. The doses and response data used for the modeling are listed in Table E-59. The C
iast7, avg was selected for this model rather than alternate metrics such as Cmax because the average
blood concentration is expected to better correlate with an accumulation of extramedullar
hematopoiesis in the spleen.
Table E-59. Dose-Response Modeling Data for Extramedullary Hematopoiesis in the
Spleen in Female Sprague-Dawley Rats Following Exposure to PFOS {NTP, 2019, 5400978}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Incidence
0
0
10
2
0.312
10.0
10
3
0.625
20.0
10
3
1.25
40.0
10
8
2.5
80.0
10
10
5
159.6
10
10
E-72
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The BMD modeling results for extramedullary hematopoiesis in the spleen are summarized in
Table E-60 and Figure E-15. The Multistage Degree 1 model was selected based on adequate p-
values (greater than 0.1), the BMDLs were sufficiently close (less than threefold difference)
among adequately fitted models, and the Multistage Degree 1 model had the lowest BMDL. The
BMDLio) from the selected Multistage Degree 1 model is 2.3 mg/L.
Table E-60. Summary of Benchmark Dose Modeling Results for Extramedullary
Hematopoiesis in the Spleen in Female Sprague-Dawley Rats Following Exposure to PFOS
{NTP, 2019, 5400978}
Goodness of Fit
Scaled Residual
Model3
p-value AIC
Dose Group Control Dose
near BMD Group
BMDio
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Dichotomous
0.849
52.8
0.2
-0.5
26.4
9.1
EPA selected the
Hill
Multistage Degree
Gamma
0.966
50.7
0.0
-0.4
21.8
5.7
1 model. All
Log-Logistic
0.956
50.8
0.2
-0.4
25.7
9.1
models had
Multistage
0.989
50.6
-0.2
-0.1
16.1
3.4
adequate fit (p-
Degree 5
values greater than
Multistage
Degree 4
0.981
50.6
-0.2
-0.1
16.5
3.4
0.1), the BMDLs
were sufficiently
close (less than
Multistage
0.959
50.8
-0.3
-0.2
16.5
3.5
threefold
Degree 3
difference), and the
Multistage
0.948
49.2
0.3
0.1
11.5
3.6
Multistage Degree
Degree 2
1 model had the
Multistage
0.448
53.0
0.6
0.6
3.5
2.3
lowest BMDL.
Degree 1
Weibull
0.990
48.7
-0.2
-0.2
18.0
5.0
Logistic
0.877
49.8
0.3
0.5
7.6
5.1
Log-Probit
0.963
50.8
0.1
-0.4
22.5
8.8
Probit
0.888
49.7
0.2
0.5
7.2
5.0
Notes: AIC = Akaike information criterion; BMD =
corresponding to a 10% response level; BMDLio =
limit for a 10% response level.
a Selected model in bold.
: benchmark dose; BMDL = benchmark dose lower limit; BMDio = dose level
lower bound on the dose level corresponding to the 95% lower confidence
E-73
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-15. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 1 Model for Extramedullar Hematopoiesis in the Spleen in Female Sprague-
Dawley Rats Following Exposure to PFOS {NTP, 2019, 5400978}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.5 Zhong et al. (2016, 37488
EPA conducted dose response modeling of the Zhong et al. (2016, 3748828) study using the
BMDS 3.2 program. This study addresses plaque forming cell (PFC) response of splenic cells in
Fi male C57BL/6 mice.
E.2.5.1 Plaque Forming Cell Response of Splenic Cells in Fi Mole
C57BL/6 Mice
Decreased mean response of PFC response of splenic cells was observed in Fi male C57BL/6
mice. Continuous models were used to fit dose-response data. A BMR of a change in the mean
equal to one standard deviation from the control mean was chosen per EPA's Benchmark Dose
Technical Guidance {U.S. EPA, 2012, 1239433}. The doses and response data used for the
modeling are listed in Table E-61. The Cavg.Pup.gest.iact was selected for this model rather than
alternate metrics such as Cmax because the average blood concentration is expected to better
correlate with an accumulation of decreased plaque forming cell response of splenic cells from
across the gestation and lactation lifestages.
Table E-61. Dose-Response Modeling Data for PFC Response of Splenic Cells in Fi male
C57BL/6 mice Following Exposure to PFOS {Zhong, 2016, 3748828}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Mean Response (# cells
per 106 spleen cells)3
0
0.0
12
465.7 ±78.5b
E-74
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Mean Response (# cells
per 106 spleen cells)3
0.1
1.7
12
423.0 ±60.4
1
16.8
12
398.7 ±72.5
5
84.1
12
340.1 ±54.4
Notes:
a Data are presented as mean ± standard deviation,
b Standard deviations were calculated from standard errors.
BMD modeling results for PFC response of splenic cells are summarized in Table E-62 and
Figure E-16. The best fitting model was the Hill model based on adequate p-values (greater than
0.1), the BMDLs were sufficiently close (less than threefold difference) among adequately fitted
models, and the Hill model had the lowest BMDL. The BMDLisd from the selected Hill model
is 1.8 mg/L.
Table E-62. Summary of Benchmark Dose Modeling Results for Plaque Forming Cell
Response of Splenic Cells in Fi Male C57BL/6 Mice Following Exposure to PFOS (constant
variance) {Zhong, 2016, 3748828}
Goodness of Fit
Scaled Residual
Model"
BMDisd BMDLisd
p-value
AIC
Dose Group
near BMD
Control Dose
Group
(mg/L)
(mg/L)
0.181
545.3
0.2
1.4
51.3
34.4
0.181
545.3
0.2
1.4
51.3
34.4
0.174
545.7
0.2
0.9
22.3
6.6
0.174
545.7
0.2
0.9
22.2
6.6
0.190
545.6
0.3
0.8
20.6
1.8
0.161
545.5
0.2
1.4
55.1
38.9
0.161
545.5
0.2
1.4
55.1
38.9
0.161
545.5
0.2
1.4
55.1
38.9
0.161
545.5
0.2
1.4
55.1
38.9
Basis for Model
Selection
Exponential 2
Exponential 3
Exponential 4
Exponential 5
Hill
Polynomial
Degree 3
Polynomial
Degree 2
Power
Linear
EPA selected the
Hill model. All
models had
adequate fit (p-
values greater than
0.1), the BMDLs
were sufficiently
close (less than
threefold
difference), and the
Hill model had the
lowest BMDL.
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDisd = dose
level corresponding to a change in the mean equal to 1 standard deviation from the control mean; BMDLisd = lower bound on
the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 1 standard deviation from the
control mean.
a Selected model in bold.
E-75
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
600
500
200
100
Estimated Probability
^^Response at BMD
O Data
BMD
BMDL
10
20
30
40 50
Dose
60
70
80
Figure E-16. Plot of Mean Response by Dose with Fitted Curve for the Selected Hill Model
for PFC Response of Splenic Cells in Fi Male C57BL/6 Mice Following Exposure to PFOS
{Zhong, 2016, 3748828}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E-76
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix F. Pharmacokinetic Modeling
For the animal pharmacokinetic model, model predictions from Wambaugh et al. (2013,
2850932) were evaluated by comparing each predicted final serum concentration to the serum
value in the supporting animal studies (training data set) and to animal studies published since
the publication of Wambaugh et al. (2013, 2850932) (test data set). The predictions to these two
data sets were generally similar to the experimental values. There were no systematic differences
between the experimental data and the model predictions across species, strain, or sex, and
median model outputs uniformly appeared to be biologically plausible despite the uncertainty
reflected in some of the 95th percentile CIs. The application of the model outputs in the
derivation of a human RfD can be found in the PFOS Main Document.
F-l
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
F.l Comparison of Fits to Training Datasets Used in
Wambaugh et al. (2013, 2850932)
The following figures show comparisons of the model predicted serum concentrations to the data
used for model training. Fits also presented in supplemental material of Wambaugh et al. (2013,
2850932).
lime (days)
Figure F-l. Experimentally Observed Serum Concentrations {Chang, 2012,1289832} and
Median Prediction for a Single Oral Dose of 1 or 20 mg/kg PFOS to Female CD1 Mice
1 mg/kg represented by the squares and solid line; 20 mg/kg represented by the circles and dashed line.
r 101
10°
101
Figure F-2. Experimentally Observed Serum Concentrations {Chang, 2012,1289832} and
Median Prediction for a Single Oral Dose of 1 or 20 mg/kg PFOS to Male CD1 Mice
A) Fits to observed male data using male-specific model. B) Fits to observed male data using female-specific model parameters.
1 mg/kg represented by the squares and solid line; 20 mg/kg represented by the circles and dashed line.
Time (days) Time (days)
F-l
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
£
101 ¦
io-3 -
10"
o
*
y
'-is. A A A
O
CJ\
~ D
~
~
~~
~
\
A 2 mg/kg, IV
~ 2 mg/kg, oral
O 15 rng/kg, oral
10"
10"
10°
Time (days)
101
10*
Figure F-3. Experimentally Observed Serum Concentrations {Chang, 2012, 1289832} and
Median Prediction for a Single IV Dose of 2 mg/kg or a Single Oral Dose of 2 or 15 mg/kg
PFOS to Male Sprague-Dawley Rats
2 mg/kg intravenous (IV) dose represented by the upward triangles and solid line; 2 mg/kg oral dose represented by the squares
and solid line; 15 mg/kg oral dose represented by the circles and dashed line.
dose_range
dose <=5 mg/kg
5 rng/kg < dose <=20 mg/kg
lO0 10l
Literature reported concentration (mg/L)
Figure F-4. Model Prediction Summary for PFOS Training Data
Model predictions on the training data result in a mean squared log error (MSLE) of 0.174. Dashed lines represent +/- one-half
logio.
F-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
We conducted a local, one-at-a-time sensitivity analysis to examine how parameter sensitivity
varied across the adult and developmental models (Figure F-5). For each parameter/dose metric
pair, sensitivity coefficients were calculated to describe the relative change in a dose metric
relative to the proportional change in a parameter value. A sensitivity coefficient of 1 describes
the situation where a 1% increase in a parameter resulted in a 1% increase in the dose metric.
VfccJ
Vcc
¦
Vcc
Qfilc
=
Qfilc
1
¦
free
Fiv2v1
¦In.
free
1
k12
k12
0)
0)
J3
•3
9
KT
—
Cavg pup qest
Cava pup lact
Cavg pup qest lact
Cavg_pup_diet
Cava pup total
ra Ry2v1
5
Cavg
Vfilc
KT
P_milk
¦
Vfilc
halfjife
Tmc
¦
Tmc
ka
ka
rf_m
-0.8 -0.6 -0.4
value
-0.2 0.0
-1.0 -0.5 0.0 0.5 1.0
value
Adult model
Developmental model
Figure F-5. PFOS Sensitivity Coefficients of the Adult Model and Developmental Model
As demonstrated in Figure F-5, the volume of distribution (Vd) represents the most sensitive
parameter for average concentrations in the adult animal. Because of the long half-life and high
degree of plasma protein binding, renal resorption parameters that impact the effective half-life
of PFOS are not as sensitive when compared to PFOA which has a shorter net half-life.
Comparatively, the four one-compartment parameters for the infant (volume of distribution, half-
life, serum:milk partition coefficient, and fetal:maternal ratio) are all sensitive to the
gestational/lactational dose metrics. However, once the pup transitions to the adult model
(Wambaugh model), PFOS transfer during gestation/lactation does not impact the average
concentration during the post-weaning phase (Cavg-PuP-diet). This is because the steady state
concentration for the pup exposed to PFOS in the diet during growth is much larger than the
steady state concentration during the 21 days of lactational exposure.
F-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
F.2 Visual Inspection of Test Datasets not Used for Initial
Fitting
The following figures show a comparison between model predictions and data from more
recently published studies that were not part of the Wambaugh et al. (2013, 2850932)
parameterization.
Time (days)
Figure F-6. mentally Observed Serum Concentrations {Huang, 2019, 7410147} and Median
Predictions for a Single IV Dose of 2 mg/kg or an Oral Dose of 2 or 20 mg/kg PFOS to Male
Sprague-Dawley Rats
2 mg/kg intravenous (IV) dose represented by the squares and solid line; 2 mg/kg oral dose represented by the circles and solid
line; 20 mg/kg oral dose represented by the downward triangles and dashed line.
Time (days) Time (days)
Figure F-7. Experimentally Observed Serum Concentrations {Huang, 2019, 7410147} and
Median Predictions for a Single IV Dose of 2 mg/kg or an Oral Dose of 2 or 20 mg/kg PFOS
to Female Sprague-Dawley Rats
A) Fits to observed female data using female-specific model parameters. B) Fits to observed female data using male-specific
model parameters.
2 mg/kg intravenous (IV) dose represented by the squares and solid line; 2 mg/kg oral dose represented by the circles and solid
line; 20 mg/kg oral dose represented by the downward triangles and dashed line.
F-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E
_ 10°
10"
o ° °
° ~ ~
O
~ O
o
~ 2 mg/kg, IV
O 2 mg/kg, oral
10-
10"
lO0
Time (days)
101
102
Figure F-8. Experimentally Observed Serum Concentrations {Kim, 2016, 3749289} and
Median Prediction for a Single IV Dose of 2 mg/kg or an Oral Dose of 2 mg/kg PFOS to
Male Sprague-Dawley Rats
2 mg/kg intravenous (IV) dose represented by the squares; 2 mg/kg oral dose represented by the circles.
|
£
| 10°
qO
r\
o°o
~
O
B
^°aOCPa
O
<%
i
E
~ 10°
u
c
o
/ o >
o
o
~
£
2
u
M
s
fe 10-'
o
~
2 mg/kg, IV
~ 2 mg/kg, IV
o
2 mg/kg, oral
O 2 mg/kg, oral
10"
10"
10°
Time (days)
101
10";
10"
10°
Time (days)
101
102
Figure F-9.Experimentally Observed Serum Concentrations {Kim, 2016, 3749289} and
Median Prediction for a Single IV Dose of 2 mg/kg an Oral Dose of 2 mg/kg PFOS to
Female Sprague-Dawley Rats
A) Fits to observed female data using female-specific model parameters. B) Fits to observed female data using male-specific
model parameters.
2 mg/kg intravenous (IV) dose represented by the squares; 2 mg/kg oral dose represented by the circles.
F-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
cone mean
Figure F-10. Model Prediction Summary for PFOS Test Data
Model predictions on the adult, single-dose test data result in a mean squared log error (MSLE) of 0.384. Dashed lines
represent +/- one-half logio. Developmental pharmacokinetic summary results not shown as only one study (presented in main
text) is available for comparison.
F.3 Human Model Validation
As mentioned in the main document (see PFOS Main Document), the human model was
implemented in R/MCSim from the original AcslX model {Verner, 2016, 3299692}.
Comparison with model output from the original model shows that, with the original parameters,
the R model exactly replicates the original model (Figure F-l 1). The only difference remaining
was that the start of pregnancy occurs at slightly different times in the two models, but this does
not affect predictions outside of that very narrow time. Validation figures shown in this section
include data for PFOA as well as PFOS. This is because model validation and decisions related
to model structure were made for both chemicals together due to the preference for a similar
model structure for the two chemicals.
F-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Age (yr) Age (yr)
Figure F-ll. Model Comparison
Comparison of the original AcslX model output (red, "Verner" label), the R model output with original model parameters (blue,
"Rep." label), and the R model output with updated parameters (black, "PFAS_TK" label). Note that the red lines are almost
entirely obscured by the blue lines.
The updated parameters result in lower serum concentrations for both the maternal and child. This is mainly due to lower half-
lives selected during the parameter update.
Application of the updated parameters to predictions of serum levels in children showed good
agreement between model predictions and reported values (Figure F-l l;Figure F-12). This
simulation was performed using mean breastmilk consumption estimates rather than the 95th
percentile values from EPA's Exposure Factors Handbook {U.S.EPA, 2011, 786546}. Exposure
in the validation scenario was assumed to be constant relative to body weight and was the same
in the mother and child. This exposure was set such that predicted maternal serum level at
delivery matched the reported value. Unlike the version of the model applied for human
exposure prediction, validation was performed using the age-dependent mean breastmilk
consumption estimates. The main application of the model used the 95th quantile of breastmilk
consumption to provide a health-protective estimate of exposure. Each validation scenario was
customized based on information about the length of breastfeeding typical in that cohort. As a
reminder, the default modeling scenario consisted of 1 year of breastfeeding, with an
instantaneous transition to non-breastfeeding exposure (i.e., with exposure to other PFAS sources
at weaning). One year is more typical of total (exclusive and partial) breastfeeding, as opposed to
exclusive breastfeeding which typically lasts up to around 6 months of age.
For the simulation of the Fromme et al. (2010, 1290877) cohort, information on breastfeeding
status was only available 6 months after birth. At this point 37 of 50 participants were
exclusively breastfed, 6 predominantly breastfed, 6 partially breastfed, and 1 received no breast
milk. As in the analysis by Verner et al. (2015, 3299692), we chose to model this scenario as
exclusive breastfeeding to 6 months of age at which point the constant per bodyweight exposure
starts equivalent to maternal exposure. For the cohort of the MOBA study {Granum, 2013,
1937228}, the average breast-feeding duration was 12.8 months. Because breastfeeding
F-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
parameters were only developed in the model up to 1 year, and the information used to inform
the model only extended to 1 year, the simulation for this scenario used the default 1 year of
breastfeeding. In the Mogensen et al. (2015, 3859839) study, the median length of exclusive
breastfeeding was 4.5 months, and the median length of partial breastfeeding was 4.0 months so
8.5 months was chosen as the breastfeeding duration for simulation of this study.
Age (yr) Age (yr)
Figure F-12. Predicted Child Serum Levels Compared to Reported Values
These values were calculated using the updated parameters with constant Vd and exposure relative to body weight.
Observed PFOA (rig/ml) Observed PFOS (ng/ml)
Figure F-13. Comparison of Predicted and Observed Child Serum Levels
F-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Local, one-at-a-time sensitivity analysis was performed to examine how parameter sensitivity
varied across age and between maternal and child serum (Figure F-13). Sensitivity coefficients
describe the change in a dose metric, in this case serum concentration, relative to the
proportional change in a parameter value, in this case a 1% increase. A sensitivity coefficient of
1 describes the situation where a 1% increase in a parameter resulted in a 1% increase in serum
concentration. Half-life and Vd were sensitive for every dose metric because they govern the
distribution and excretion in all life stages and have a synergistic effect on child levels because
they influence the serum levels in children directly as well as the indirect exposure to the child
early in life through maternal exposure.
For maternal serum at delivery, only the half-life and the Vd influenced the serum concentration.
This was expected as the other parameters evaluated govern distribution of PFOS to the child
and are not in play at this point. For cord blood, we see a similar effect from Vd and half-life as
in the maternal serum, because cord blood levels are based on maternal levels in the model, but
we also see a high sensitivity on the cord blood:maternal serum ratio parameter. This was not
unexpected but emphasizes the importance of this parameter for this endpoint. The 1-year
timepoint occurs at the peak serum concentration associated with the end of breastfeeding.
Consistent with this, we see the parameters that govern lactational transfer of PFOS (i.e.,
breastmilk intake and the milk:maternal serum ratio) have high sensitivity coefficients.
Additionally, sensitivity to Vd is high because that governs the relationship between exposure
and serum levels by accounting for the amount of PFOS distributed to tissues. At the 5-year
timepoint the sensitivity to parameters associated with lactational exposure has decreased. The
sensitivity to Vd is somewhat lower comparted to the value at 1 year, and the sensitivity to half-
life has slightly increased. This reflects the increased importance of excretion relative to the
distribution of incoming PFOS during the time period following lactational exposure.
F-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Maternal at Delivery
Cord Blood
Breastmilk Intake per kg BW
Cord Blood:Maternal Serum Ratio
Milk:Maternal Serum Ratio
Volume of Distribution
Half-life
-2
-2
+
+
Child at 1 yr
Breastmilk Intake per kg BW
Cord Blood:Maternal Serum Ratio
Milk:Maternal Serum Ratio
Volume of Distribution
Half-life
-1 0 1
Child at 5 yr
ti
-2
i-
-2
Figure F-14. Sensitivity Coefficients
Sensitivity coefficients from a local sensitivity analysis of maternal serum at delivery, cord blood at delivery, and child serum at
1 and 5 years old. The child was female. Results for a male child were similar (not shown).
BW = body weight; yr = year.
A model developed by the Minnesota Department of Health (MDH model) {Goeden, 2019,
5080506} was also considered for application to this assessment. This model has a similar model
structure to the chosen model, with single compartments to represent the mother and child and
excretion handled by first-order clearance.
To evaluate the effect of Vd in children, we integrated the Vd scaling in the MDH model into our
model (Figure F-14). The main effect is to reduce the peak serum levels in children that occurs
due to exposure through breastmilk. Based on mean relative error (for PFOA and PFOS
combined), we determined that the model with constant Yd had better performance.
F-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Fromme, 2010
MOBA Cohort
Mogensen, 2015
Fromme, 2010
MOBA Cohort
Mogensen, 2015
Age (yr)
Age (yr)
Figure F-15. Predicted Child Serum Levels Compared to Reported Values with Increased
Volume of Distribution in Children as was Implemented in the Minnesota Department of
Health Model
We also implemented exposure based on drinking water consumption in the modified Verner
model to examine the effect on model predictions and especially on the results of the risk
assessment (Figure F-15). As discussed in the main document (see Main PFOS Document), this
approach was not used for dosimetric extrapolation due primarily to the poor fit to the PFOS
dataset. An MCLG based on constant exposure does not greatly underestimate the risk to
populations with greater water consumption per body weight (e.g., children and lactating
women) because the method for calculating the MCLG from a RfD that assumes constant
exposure accounts for the greater drinking water consumption in these populations.
F-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Fromme, 2010
MOBA Cohort
Mogensen, 2015
Fromme, 2010
MOBA Cohort
Mogensen, 2015
± I
I
Age (yr)
Age (yr)
Figure F-16. Predicted Child Serum Levels Compared to Reported Values with Constant
Volume of Distribution and Variable Exposure Based on Drinking Water Intake
F-12
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix G. Relative Source Contribution
G.l Background
EPA applies a RSC when calculating the MCLG to account for the fraction of an individual's
total exposure allocated to drinking water. EPA emphasizes that the purpose of the RSC is to
ensure that the level of a chemical allowed by a criterion (i.e., PFOS) or multiple criteria, when
combined with other identified sources of exposure (e.g., diet, ambient and indoor air) common
to the population of concern, will not result in exposures that exceed the RfD. In other words, the
RSC is the portion of an exposure for an individual in the general U.S. population estimated to
equal the RfD that is attributed to drinking water ingestion (directly or indirectly in beverages
like coffee tea or soup, as well as from transfer to dietary items prepared with drinking water)
relative to other exposure sources; the remainder of the exposure equal to the RfD is allocated to
other potential exposure sources. The purpose of the RSC is to ensure that the level of a
contaminant (e.g., MCLG value), when combined with other identified sources of exposure
common to the population of concern, will not result in exposures that exceed the RfD {U.S.
EPA, 2000, 19428}. For example, if for a particular chemical, drinking water were to represent
half of total exposure and diet were to represent the other half, then the drinking water
contribution (or RSC) would be 50%. In the case of PFOS, other potential sources include diet,
ambient and indoor air, incidental soil and dust ingestion, consumer products, and others.
The RSC is derived by applying the Exposure Decision Tree approach published in EPA's
Methodology for Deriving Ambient Water Quality Criteria for the Protection of Human Health
{U.S. EPA, 2000, 19428}. The Exposure Decision Tree approach allows flexibility in the RfD
apportionment among sources of exposure. To determine the RSC to be used in the MCLG
calculation, EPA considers whether there are significant known or potential uses/sources other
than drinking water, the adequacy of data or strength of evidence available for each relevant
exposure source and pathway, and whether information on each source is available to
quantitatively characterize exposure. The RSC is developed to reflect the exposure to the general
population or a sensitive population within the general population.
In cases in which there is a lack of sufficient environmental data and/or exposure data, the
Exposure Decision Tree approach results in a recommended RSC of 20%. In the case of MCLG
development, this means that 20% of the exposure equal to the RfD is allocated to drinking water
and the remaining 80% is reserved for other potential sources, such as diet, air, consumer
products, etc. This 20% RSC value can be replaced if sufficient data are available to develop a
scientifically defensible alternative value. If scientific data demonstrating that sources and routes
of exposure other than drinking water are not anticipated for a specific pollutant, the RSC can be
raised as high as 80% based on the available data, allowing the remaining 20% for other
potential sources {U.S. EPA, 2000, 19428}. Applying a lower RSC (e.g., 20%) is a more
conservative approach to public health and results in a lower MCLG. For disproportionately
affected subpopulations, such as the occupationally exposed or site-impacted (e.g., by a
particular source or industry) where there may be higher than average PFAS concentrations in
drinking water, it may be appropriate to apply an RSC greater than 20% if there is sufficient
information to quantitatively characterize sources other than drinking water. This is a less
conservative approach from a public health perspective and would result in a higher MCLG for
those disproportionately affected populations.
G-l
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
G.2 Literature Review
In 2019, EPA's Office of Research and Development (ORD) conducted a broad literature search
to evaluate evidence for pathways of human exposure to PFOA and PFOS. This search was not
date limited and spanned the information collected across the Web of Science, PubMed, and
ToxNet/ToxLine (now ProQuest) databases. An updated literature search was conducted and
captured relevant literature published through March 2021. Literature captured by this search is
housed in EPA's HERO database (https://hero.epa.gov/).
Results of this broad literature search were further distilled to address two questions. First, a
systematic review was conducted to investigate evidence for important PFAS exposure pathways
from indoor environment media including consumer products, household articles, cleaning
products, personal care products, and indoor air and dust {Deluca, 2021, 7277659}. In this study,
literature was identified that reported exposure measures from household media paired with
occupant PFAS concentrations in blood serum. Second, systematic evidence mapping was
conducted for literature reporting measured occurrence of PFAS chemicals in exposure media
{Holder, 2021 in prep., 9419128}. This review focused on real-world occurrences (measured
concentrations) primarily in media commonly related to human exposure (outdoor and indoor
air, indoor dust, drinking water, food, food packaging, articles and products, and soil).
G.2.1 Systematic Review
Deluca and coworkers (2022, 10273296) investigated evidence for important PFAS exposure
pathways from indoor environment media including consumer products, household articles,
cleaning products, personal care products, and indoor air and dust. The authors adapted existing
systematic review methodologies and study evaluation tools to identify and screen exposure
science studies that presented concordant data on PFAS occurrence in indoor media and PFAS
concentrations in blood or serum. Studies included in the systematic review report exposure
measures from household media paired with occupant PFAS concentrations in blood serum,
focusing on PFOS and seven other frequently measured PFAS (PFOA, perfluorobutanoic acid
(PFBA), perfluorobutane sulfonate (PFBS), PFDA, PFHxA, PFHxS, and PFNA). Machine
learning approaches were used during the literature scoping and title/abstract screening to
prioritize exposure pathways of interest by automated tagging and to select studies for inclusion
using an iterative predictive screening model. Title/Abstract screening for the PECO criteria
identified 486 studies for full text screening; only 6 studies fully addressed the protocol
requirements {Wu, 2014, 2533322; Makey, 2017, 3860102; Bryne, 2017, 4165183; Kim, 2019,
5080673; Balk, 2019, 5918617; Poothong, 2019, 5080584}. The extraction of exposure
measurement data and study characteristics from each included study was performed in
DistillerSR software. Exposure intake calculations were used to estimate a percentage of
occupant serum concentrations that could be attributed to indoor exposure pathways other than
drinking water and diet. The included studies were evaluated using an approach modified from
EPA's IRIS Handbook {U.S. EPA, 2022, 2022, 10476098}. Along with providing evidence for
an estimated range of indoor exposure media's contribution to serum PFAS concentrations, this
systematic review highlights the limited availability of concordant measurement data from
indoor exposure media and participant serum.
G-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The Deluca and coworkers review (2022., 7277659) described above focused on indoor
pathways and therefore excluded non-indoor pathways such as drinking or surface water or soil.
Ninety-seven articles fell into this excluded group (i.e., PFOS was measured in sera or a non-
indoor environmental medium). Because the combination of PFOS measured in sera and
drinking water is potentially informative for deriving the RSC, these 97 papers were reviewed
for this effort, though are not described in this appendix.
G.2.2 Evidence Mapping
Holder et al. (2021 in prep., 9419128) investigated evidence for important pathways of exposure
to PFAS chemicals by reviewing literature reporting measured occurrence of PFAS chemicals in
exposure media. The review focused on eight PFAS chemicals (PFOA, PFOS, PFBA, PFBS,
PFDA, PFHxA, PFHxS, and PFNA) and their real-world occurrences primarily in human
matrices and media commonly related to human exposure (outdoor and indoor air, indoor dust,
drinking water, food, food packaging, articles and products, and soil). The initial review
identified 3,622 peer-reviewed papers matching these criteria and published between 2003-2020.
ICF's litstreamTM software was used to conduct title-abstract (TiAb) and full-text screening, and
to extract relevant primary data into a comprehensive evidence database. Parameters of interest
included: sampling dates and locations (focused on locations in the United States, Canada, and
Europe), numbers of collection sites and participants, analytical methods, limits of detection and
detection frequencies, and occurrence statistics.
Detailed data on PFAS occurrence in high-priority household and environmental media from 210
studies were extracted, as well as limited data on human matrices from 422 additional papers.
Studies of PFAS occurrence became numerous after about 2005 and were most abundant for
PFOA and PFOS. Co-measurements for PFAS occurrence in human matrices plus other media,
while relatively infrequent, were typically related to food and drinking water. Most studies found
detectable levels of PFAS, and half or more of the limited studies of indoor air and products
detected PFAS in 50% or more of their samples. Levels of PFOS in these media ranged widely.
Literature search results were categorized into 7 types of exposure pathway categories, including
environmental media, home products/articles/building materials, cleaning products, food
packaging, personal care products, clothing, and specialty products. The environmental media
pathway category included the sub-categories of food, water, air, dust, soil, wastewater, and
landfill.
G.3 Summary of Potential PFOS Sources
PFOS is a synthetic, fully fluorinated, organic compound that is used in many types of consumer
products and is resistant to metabolic and environmental degradation {U.S. EPA, 2016,
3603365}. It has been associated with releases from manufacturing sites, industrial sites,
fire/crash training areas, and industrial and municipal waste sites. PFOS is one of a large group
of perfluoroalkyl substances that are widely used in consumer and industrial products to improve
their resistance to stains, grease, and water. PFOS was a major component of AFFF which were
used to extinguish petroleum-based fires. Most manufacturing of PFOS in the United States was
discontinued voluntarily by its primary manufacturer in 2002 and was completely phased out of
U.S. production in 2016. However, some limited uses of PFOS-related chemicals (i.e., PFOS
replacements such as PFBS) remain for which alternatives are not currently available. Exposure
G-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
to PFOS can occur through food, including fish and shellfish, house dust, air, and contact with
consumer products {U.S. EPA, 2016, 3603365}.
G.3.1 Dietary Sources
Ingestion of food is a potentially significant source of exposure to PFOS and is often claimed to
be the dominant source of exposure for the general population based on early studies that
modeled the relative contributions of various sources among the general populations of North
America and Europe {Fromme, 2009, 1291085; Trudel, 2008, 214241; Vestergren, 2009,
1290815}. The exposure among adults in western countries is typically estimated to be about
1 ng/kg/day, but studies on the dietary exposure among the U.S. population are limited
{Domingo, 2017, 3981385; East, 2021, 9416543}. The dominance of the food ingestion pathway
is attributed to bioaccumulation in food from environmental emissions, relatively large amounts
of foods being consumed, and high GI uptake {Trudel, 2008, 214241}. However, the estimates
are highly uncertain due to limited data availability, relatively low detection frequencies, and
relatively large differences in composition of diets across geographic locations {EFSA, 2020,
6984182; Domingo, 2017, 3981385}.
There is currently no comprehensive, nationwide Total Diet Study (TDS) for PFOS that can be
used to draw conclusions about the occurrence and potential risk of PFOS in the U.S. food
supply for the general population. In 2021, the U.S. Food and Drug Administration (FDA)
released PFAS testing results from their first survey of nationally distributed processed foods,
including several baby foods. Results of the survey showed that 164 of the 167 foods tested had
no detectable levels of the PFAS measured. Three food samples had detectable levels of PFAS:
fish sticks (PFOS (33 parts per trillion (ppt)) and PFNA), canned tuna (PFOS (76 ppt) and
PFDA), and protein powder (PFOS (140 ppt)). In another recent FDA study, PFOS was detected
in one sample (baked cod, 98 ppt) out of 94 food samples collected nationally {FDA, 2021,
9419076}. In a 2019 national survey of produce, meats, dairy and grain products, PFOS was
detected in three of the 179 food samples tested (two samples of tilapia, one sample of turkey)
{FDA, 2019, 9638790; FDA, 2019, 9638792}. PFOS was also detected in produce samples
(collard greens and lettuce) in a 2018 focused study near a PFAS production plant in the
Fayetteville, North Carolina area {FDA, 2018, 9419064}. PFOS was below the lower limit of
quantification (LLOQ; 4 ng/L) in all 30 samples analyzed in a study of domestic and imported
carbonated water and non-carbonated bottled water {FDA, 2016, 9419013}. The sample size in
all of these studies is limited, and thus, the results cannot be used to draw definitive conclusions
about the levels of PFAS in the U.S. food supply more generally {FDA, 2021, 9419076}. In a
2010 study of 31 types of food collected from 5 grocery stores in Texas, PFOS was not detected
in any of the samples {Schecter, 2010, 729962}.
As a component of a scientific evaluation on the risks to human health related to PFAS in food,
the European Food Safety Authority (EFSA) conducted an exposure assessment using
consumption data from the EFSA Comprehensive Food Consumption Database and 69,433
analytical results for 26 PFAS in 1,528 samples of food and beverages obtained from 16
European countries {EFSA, 2020, 6984182}. Samples were collected between the years 2000
and 2016 (74% after 2008), mainly from Norway, Germany, and France. With 92% of the
analytical results below the LOD or LOQ, lower bound dietary exposure estimates were obtained
by assigning zero to values below LOD/LOQ. Median chronic dietary exposures of PFOS for
G-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
children and adults were estimated as 1.02 and 0.58 ng/kg-body weight/day, respectively. The
most important contributors for PFOS were "Fish and other seafood," "Eggs and egg products,"
and "Meat and meat products." It is unclear whether the contribution from food contact material
is reflected in the data. The authors determined diet to be the major source of PFAS exposure for
most of the population but noted that dust ingestion and indoor air inhalation may provide
substantial contributions for some individuals.
The 2020 EFSA report highlighted a recent study of aggregate exposure to PFAS from diet,
house dust, indoor air, and dermal contact among Norwegian adults {Poothong, 2020, 6311690}.
Dietary exposures were estimated for 61 study participants using food diaries and data on
concentrations from an extensive Norwegian database of concentrations in sixty-eight different
food and drinks (including drinking water). For PFOS, dietary intake was by far the greatest
contributor to aggregate exposure (contributing 95% of total estimated PFOS intake), but intake
from ingestion of house dust represented the dominant pathway for some of the top 20% most
highly exposed individuals. The authors reported a significant positive correlation between the
observed and modeled serum concentrations for PFOS (r = 0.29, p < 0.05). The correlation
existed despite the model underestimating serum concentrations of PFOS by a factor of 4, which
was attributed to the long half-life and decreased exposure over recent years. While the authors
did not separately quantify intake from food and drinking water, an earlier article from the same
research group {Papadopoulou, 2017, 3859798} reported measured concentrations in duplicate
diets with median estimated intake of PFOS approximately 150 times higher from solid food
than from liquids.
De Felip et al. (2015, 2850114) investigated correlations of blood concentrations of PFOS with
dietary intake among Italian women. They estimated daily intake of PFOS based on the reported
food consumption frequencies of specific food items and found strongly significant correlations
of blood levels with consumption of beef, pork, and vegetables (p < 0.01), and moderate
correlation with consumption of fish (p < 0.05).
G.3.1.1 Food Contact Materials
Since the 1960s, the FDA has authorized several broad classes of PFAS for use in food contact
substances due to their non-stick and grease, oil, and water-resistant properties. The
authorization of the use of a food contact substance requires that available data and information
demonstrate that there is a reasonable certainty of no harm for that use.
• Non-stick cookware: PFAS may be used as a coating to make cookware non-stick.
• Gaskets, O-Rings, and other parts used in food processing equipment: PFAS may be used
as a resin in forming certain parts used in food processing equipment that require chemical
and physical durability.
• Processing aids: PFAS may be used as processing aids for manufacturing other food
contact polymers to reduce build-up on manufacturing equipment.
• Paper/paperboard food packaging: PFAS may be used as grease-proofing agents in fast-
food wrappers, microwave popcorn bags, take-out paperboard containers, and pet food
bags to prevent oil and grease from foods from leaking through the packaging. {FDA,
2020, 9419078}
G-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Paper products used for food packaging are often treated with PFAS for water and grease
resistance. In previous testing, sandwich wrappers, french-fry boxes, and bakery bags were all
been found to contain PFAS {Schreder, 2018, 9419077}. Older generation PFAS (e.g., PFOA,
PFOS) were manufactured and used in products for decades, and the bulk of the information
available on PFAS toxicity relates to the older compounds. However, because newer-generation
PFAS are more mobile than their predecessors, they migrate more readily into food.
FDA (2020, 9419079) recently prohibited a few PFAS chemicals in food packaging. They
announced in January 2021 that three manufacturers would begin a 3-year phase-out of their
sales of some products containing 6:2 FTOH for use as food contact substances in the U.S.
marketplace. After the phase-out period, they estimated that it could take up to 18 months to
exhaust existing stocks of paper and paperboard products containing these food contact
substances from the market. A fourth manufacturer informed FDA that they have stopped sales
of their short-chain PFAS products to the U.S. market. Maine, Washington, New York, and
Vermont passed restrictions on PFAS in packaging, as have cities like San Francisco and
Berkeley.
Under FDA rules, there are dozens of PFAS chemicals still approved for food contact materials.
In 2020, Safer Chemicals Healthy Families and Toxic-Free Future co-published a report where
78 samples of food packaging including take-out containers and deli or bakery paper, among
others, were collected from 20 stores in 12 states {Schreder, 2018, 9419077}. An independent
laboratory tested the samples for fluorine. The utility of measuring fluorine content is limited
because it does not allow for identification and quantification of individual PFAS; however, this
method can be used to determine if a food-packaging material has been treated with PFAS. Over
10% of 78 samples tested contained PFAS. The sample size was not large enough to indicate
how widespread the use of PFAS in food packaging is at this time. However, the study
demonstrated that PFAS in food packaging is still a concern, especially for fiber bowls and trays.
Several other relatively recent studies found PFAS in fast-food packaging collected in the United
States, China, or Europe. The data from the references described below and other publications
likely contributed to the recent regulatory actions of the FDA and a number of states to ban or
restrict the presence of PFAS in food contact materials {Keller and Heckman LLP, 2020,
9419081}. Schaider at al. (2017, 3981864) collected 407 samples of food contact papers,
beverage containers, and paperboard boxes from locations throughout the United States. As was
the case with Schreder & Dickman (2018, 9419077), inorganic fluoride was the analyte for the
initial analysis. 56% of the dessert and bread wrappers were positive for fluoride, 38% of the
sandwich and burger wrappers, and 20% of the paper-board containers. None of the 30 (hot/cold)
paper beverage cups tested positive in contrast to 16% of beverage containers (milk/juice) made
from other materials. Generally, food contact papers had higher fluoride detection frequencies
than food contact paperboard.
An analysis of popcorn bags, snack bags, and sandwich bags purchased in 2018 from
international vendors and grocery stores in the United States found no evidence of PFOS at
concentrations above the LOD (0.63 ng/g paper) {Monge Brenes, 2019, 5080553}. The authors
presented these results as evidence of a reduction in PFOS concentrations in microwave
packaging between 2005 and 2018. In an analysis of microwave popcorn bags from around the
world, Zabaleta et al. (2017, 3981827) reported no measurable concentrations of any PFSA,
including PFOS, in any of the samples. In a second study, Zabaleta et al. (2020, 6505866) looked
G-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
at PFAS in 25 paper- and paperboard packaging materials primarily collected in Spain. Again, no
PFSAs, including PFOS, were found above the level of detection. The packaging materials with
the largest number of detectable analytes was a popcorn bag from China and the inside paper
lining from three individual pet food products, which contained a spectrum of C-3 to CIO
perfluorinated carboxylates.
G.3.1.2 Fish and Shellfish
PFOS has been shown to bioaccumulate and biomagnify with increasing trophic level in a
variety of freshwater ecosystems {Kannan, 2005, 1290874; Martin, 2004, 1291044; Penland,
2020, 6512132; Xu, 2014, 5079760} and saltwater ecosystems {de Vos, 2008, 2919394; Houde,
2006, 1290875; Loi, 2011, 1274155; Powley, 2008, 1332751; Tomy, 2004, 1332758} in North
America, Europe, and Asia. PFOS is often the most abundant PFAS in aquatic organisms, and
this high relative abundance is at least partially explained by the biotransformation of PFOS
precursor chemicals into PFOS {Haukas, 2007, 2158020; Kannan, 2005, 1290874; Kelly, 2009,
1276129; Martin, 2004, 1291044; Tomy, 2004, 1332758}. Higher trophic level organisms have a
greater capacity to metabolize PFOS precursor chemicals, which have been found in lower
concentrations in increasing trophic level {Fang, 2014, 2850900; Kannan, 2005, 1290874;
Martin, 2004, 1291044}.
Global distribution of PFAS chemicals in tissues of aquatic species has been demonstrated in
studies conducted in freshwater and marine environments across every continent, including
remote regions far from direct sources, such as the high arctic, Antarctica, and oceanic islands
{Giesy, 2001, 1290854; Houde, 2006, 1290875}.
EPA collaborates with federal agencies, states, tribes, and other partners to conduct freshwater
fish contamination studies as part of a series of statistically based surveys to produce information
on the condition of U.S. lakes, streams, rivers, and coastal waters. PFOS was detected in nearly
all freshwater fish fillet samples collected during several national studies in rivers and the Great
Lakes (Table G-l).
Table G-l. Summary of EPA national fish tissue monitoring results for PFOS
Reference
Most Commonly
Sampled Species
Site Description Results
U.S. EPA (2010,
Smallmouth bass
162 urban river sites across PFOS was the most
10369692)
Largemouth bass
the United States commonly detected PFAS
Channel catfish
(out of 13 PFAS).
PFOS was detected in 77
percent of samples.
Maximum detected
concentration 127 ng/g.
U.S. EPA (2015,
10369694)
Largemouth bass
Smallmouth bass
Black crappie
White crappie
Walleye/sauger
Yellow perch
White bass
Northern pike
Lake trout
349 urban and nonurban PFOS was the most
river sites across the United commonly detected PFAS
States. (out of 13 PFAS).
PFOS was detected in 99
percent of samples.
Maximum detected
concentration 283 ng/g.
G-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Most Commonly
Sampled Species
Site Description
Results
Brown trout
Rainbow trout
Brook trout
U.S. EPA (2011,
10369695)
Lake trout
Smallmouth bass
Walleye
157 nearshore sites along
the U.S. shoreline of the
Great Lakes
PFOS was the most
commonly detected PFAS
(out of 13 PFAS).
PFOS was detected in 100
percent of samples.
Maximum detected
concentration 80 ng/g;
median 15 ng/g.
U.S. EPA (2016,
10369696)
Freshwater Drum
Longnose Sucker
White Sucker
Lake Whitefish
Northern Pike
Channel Catfish
Burbot
Smallmouth Bass
White Perch
White Bass
Coho Salmon
152 nearshore sites along
the U.S. shoreline of the
Great Lakes
PFOS was the most
commonly detected PFAS
(out of 13 PFAS).
PFOS was detected in 100
percent of samples.
Maximum detected
concentration 64 ng/g;
median 11 parts per billion
(ppb).
Rainbow Trout
Chinook Salmon
Yellow Perch
Brown Trout
Lake Trout
Walleye
Guo et al. (2012, 2919419) measured PFOS in lake trout muscle tissues in Canadian waters of
Lake Superior, Huron, Erie, and Ontario. Average PFOS concentrations correlated with
watershed urbanization, and were 0.85 ng/g, 8.3 ng/g, 27 ng/g, and 46 ng/g wet weight (ww),
respectively. Delinsky et al. (2010, 2587663) measured PFOS in bluegill, black crappie, and
pumpkinseed muscle tissue in 59 lakes in Minnesota, including four lakes in the Minneapolis-St.
Paul metropolitan area. PFOS was detected in muscle tissues of fish collected in 13 of the 59
lakes, and concentrations ranged from 1.08 ng/g ww to 52.4 ng/g ww in lakes where it was
detected. In the four lakes in the Minneapolis-St. Paul metropolitan area, PFOS concentrations in
fish muscle tissues ranged from 4.39 ng/g ww to 47.3 ng/g ww.
Penland et al. (2020, 6512132) measured PFAS concentrations in invertebrates and vertebrates
along the Yadkin - Pee Dee River, in North Carolina and South Carolina in 2015. PFOS was
measured in whole body tissues of snails (6.47 ng/g ww) but was not detected whole body
tissues of in Asian clam, unionid mussels, or crayfish. The highest concentrations in
invertebrates were measured in aquatic insect whole body samples (132.8 ng/g ww) and was
hypothesized to result from dietary uptake of aquatic biofilms. PFOS was measured in muscle
tissue of all 11 sampled fish species and ranged from 11.42 ng/g ww in channel catfish to
37.36 ng/g ww in whitefin shiner. The highest PFOS concentration that Penland et al. (2020,
6512132) measured was 482.9 ng/g ww, from the eggs of a redhorse fish sample.
G-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Houde et al. (2006, 1290875) measured whole body PFOS in six fish species in Charleston
Harbor, South Carolina, and whole body PFOS in zooplankton and five fish species in Sarasota
Bay, Florida. Charleston Harbor was the more developed of the two sites and had higher overall
PFOS concentrations. Average PFOS concentrations in Charleston Harbor ranged from 19 ng/g
in pinfish to 92 ng/g in spot. In Sarasota Bay, PFOS concentrations averaged 0.2 ng/g in
zooplankton, and ranged from 3.1 ng/g in pigfish to 8.8 ng/g in spotted seatrout, suggesting
evidence of trophic biomagnification.
Zafeiraki et al. (2019, 5387058) analyzed about 250 samples of marine fish, farmed fish,
crustaceans, bivalves and European eel, caught in Dutch waters or purchased at Dutch markets
between 2012 and 2018. Of the 16 PFAS that were analyzed, PFOS was generally detected at a
higher frequency and concentration across the tested species. Shrimps and seabass had the
highest average concentrations of PFOS (each over 4 ng/g ww). PFOS was also detected in
mussels, brown crab, eel (100%detection, ranging from 3.3 to 67 ng/g ww) and several farmed
and marine fish species.
Ruffle et al. (2020, 6833737) analyzed marine and freshwater finfish and shellfish from four
regions of the United States and seven countries with significant imports to the United States. A
total of 70 samples were analyzed for 26 PFAS. PFOS represented 80% to 100% of total PFAS
measured in all but one sample. The highest PFOS concentrations (1.2 ng/g ww to 19.1 ng/g ww)
were found in whitefish, walleye, and yellow perch from the Great Lakes region.
In seafood samples collected for the FDA 2021-22 seafood survey, Young et al. (2022,
10601281), analyzed concentrations of 20 PFAS, including PFOS, in 8 of the most highly
consumed seafood products in the U.S. PFOS was detected most frequently (100% of samples;
n=10) and at the highest average concentrations (422.9 ppt) in clams. The study also reported
detections in crab (45.5% of samples; n=l 1; 151.6 ppt average concentration in samples with
detections), tuna (50% of samples; n=10; 86.8 ppt average concentration in samples with
detections), tilapia (20% of samples; n=10; 57.5 ppt average concentration in samples with
detections), and cod (60% of samples; n=10; 62.5 ppt average concentration in samples with
detections). PFOS was not detected above the method detection limits (39 or 45 ppt) in salmon,
shrimp, or pollock.
Based on National Oceanic and Atmospheric Administration (NOAA) National Centers for
Ocean and Coastal Science, National Status and Trends Data, PFOS concentrations (in ww) were
not detected in mussels, oysters, and fish liver samples. However, PFOS was detected in marine
fish fillet samples, up to 75.1 ppb {NOAA, 2017, 9638787}.
PFOS concentrations in aquatic biota tend to be higher in areas with known PFAS
manufacturing, industrial use, and/or application of AFFF, which also tend to be more populated
areas and where recreational and subsistence fishing is more common. Several states have
developed fish consumption advisories for PFOS (e.g., Alabama, Wisconsin, Minnesota,
Michigan).
G.3.2 Consumer Product Uses
An early investigation of consumer exposure to PFOS by Trudel et al. (2008, 214241) used
mechanistic modeling together with information on product-use habits to estimate exposures
G-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
from mill-treated carpets and impregnated clothing. The authors concluded that contact with
consumer products represents less than 1% of total exposure to PFOS, but also pointed out that
because carpets have a relatively long lifetime, the exposure is expected to continue long after
cessation of use of PFOS in carpet treatments. Liu et al. (2014, 2324799) also investigated trends
in PFAS content of household goods between 2007 and 2011. They reported a decrease in the
availability of consumer products that contain PFOS is declining but were still able to find
products that contained PFOS. In an analysis of 52 European products collected between 2014-
2016, Borg and Ivarrson (2017, 9416541) reported that PFSAs were rarely detected in the
samples; PFOS was the only PFSA detected and was only present in one sample, a microwave
popcorn bag. Notably, the authors specifically targeted products that were known or suspected to
contain PFAS in their analyses.
In contrast, Kotthoff et al. (2015, 2850246) reported broad detection of PFOS in a 2010 sampling
effort that collected 115 European consumer products, including carpets, leather, outdoor
materials, cooking materials, and others. PFOS was detected in all but two sample types, often at
the highest median concentration compared to other PFSAs. However, PFSAs were detected at
concentrations often several orders of magnitude lower than perfluorinated carboxylic acids
(PFCAs) and fluorotelomers. The products with the highest concentrations of total PFAS
included ski wax (median concentration of 1.6 |ig/kg), leather products (maximum concentration
of 5.6 |ig/m2), and outdoor materials (median concentration of 9.5 |ig/m2), PFOS was the most
frequently and abundantly detected PFAS in paper-based cooking materials. PFOS has also been
detected in textile samples of outdoor apparel from Europe and Asia {Gremmel, 2016, 3858525;
van der Veen, 2020, 6316195}. PFOS was detected in one-third of the jackets tested by Gremmel
et al. (2016, 3858525) at relatively low concentrations ranging from 0.01 [j,g/m2-0.59 (J,g/m2.
Interestingly, while the concentrations of almost all individual PFAS and total PFAS
concentrations increased when the textiles were subjected to weathering (i.e., increased
ultraviolet light radiation, temperature, and humidity for 300 hours to mimic the average lifespan
of outdoor apparel), PFOS concentrations declined after weathering in the one sample that
exceeded European Commission restrictions on PFOS content of coated materials (1 |ig/m2)
{van der Veen, 2020, 6316195}.
G.33 Indoor Dust
Several studies suggest that PFOS and its precursors in indoor dust may be an important
exposure source for some individuals {Shoeib, 2011, 1082300; Gebbink, 2015, 2850068;
NJDWQI, 2018, 5026035; Poothong, 2020, 6311690}. PFOS is generally a dominant ionic
PFAS constituent in household dust, frequently occurring above detection limits and at relatively
high concentrations in all or most samples {Shoeib, 2011, 1082300; Kim, 2019, 5080673; Wu,
2014, 2533322; Poothong, 2020, 6311690; Makey, 2017, 3860102; Byrne, 2017, 4165183;
Fraser, 2013, 2325338}.
PFOS was measured at the second highest concentrations (geometric mean concentrations
ranging from 29.0 ng/g-34.6 ng/g) and frequencies (ranging from 85%-87% detected) in dust
sampled from Californian households. Similarly, PFOS was found at the highest levels (mean
concentration of 3.06 ng/g) of 15 PFAS measured in dust samples taken from households in
Seoul, Republic of Korea {Kim, 2019, 5080673}. One study of Alaska Natives noted that PFOS
was the predominant compound in dust samples {Byrne, 2017, 4165183}.
G-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
G.3.4 Ambient Air
Air concentrations of PFOS in the atmosphere vary widely across the globe. Areas near
wastewater treatment facilities, waste incinerators, and landfills can be point sources of PFOS to
air {Ahrens, 2011, 2325317}. In an urban area in Albany, NY, perfluorinated acids were
measured in air samples in both the gas and particulate phase in May and July 2006 {Kim, 2007,
1289790}. PFOS in the gas phase had a mean concentration of 1.70 pg/m3 (range: 0.94-3.0
pg/m3) and in the particulate phase had a mean concentration of 0.64 pg/m3 (range: 0.35-1.16
pg/m3). However, at Lake Ontario, concentrations of PFOS in the particulate phase measured in
air samples over the lake were higher {Boulanger, 2005, 1289802}. The mean concentration of
PFOS at Lake Ontario was 6.4 ±3.3 pg/m3; with a range of concentrations from detected to
8.1 pg/m3. In an urban area in Minneapolis, Minnesota, PFOS was measured in both the
particulate and gas phase {MPCA, 2008, 9419086}. PFOS in the particulate phase ranged from
2.1 pg/m3 -7.9 pg/m3 and the gas phase ranged from 1.8 pg/m3 -5.0 pg/m3 across the five
samples.
In Canada, PFOS air concentrations measured in 2009 showed widespread distribution with
remote sites having similar concentrations to urban sites {ECCC, 2018, 9638786}. Using passive
samplers, PFOS concentrations were detected in Toronto, Ontario (8 pg/m3), an agricultural site
in Saskatchewan (5 pg/m3), Whistler, British Columbia (4 pg/m3), and Alert, N Nunavut
(2 pg/m3) {ECCC, 2018, 9638786}.
Other reported concentrations of PFOS in air samples from Sydney, Florida (3.4 pg/m3), Tudor
Hill, Bermuda (6.1 pg/m3), Malin Head, Ireland (3.3 pg/m3), and Hilo, Hawaii (6.6 pg/m3) are
similar to the concentrations reported in Canada {ECCC, 2018, 9638786} and Japan {Sasaki,
2003, 5081390}. The annual geometric mean concentration of PFOS in air samples collected
monthly from 2001-2002 in the town of Oyamazaki and Fukuchiyama City were 5.3 and
0.6 pg/m3, respectively {Sasaki, 2003, 5081390}.
Across Europe, PFOS air concentrations were reported to be variable. In the particulate phase
PFOS concentrations ranged from <1.8 pg/m3-46 pg/m3 {Martin, 2004, 1291044}. Most
locations had low (~1 pg/m3-2 pg/m3) to less than the reported Minimum Detection Limit
(MDL) and included Hazelrigg, United Kingdom, Kjeller Norway, and Mace Head, Ireland
{Barber, 2007, 1049488}. The highest concentrations were reported in Manchester, United
Kingdom. Similarly, high concentrations, 150 pg/m3 for were reported Paris, France {ECCC,
2018, 9638786}.
Even in the Arctic, PFOS, its precursors, and degradation products, have been detected in air
samples in Resolute Bay, Nunavut, Canada, during the summer of 2004 {Stock, 2007, 1289794}.
PFOS in the filter samples were 1-2 orders of magnitude greater than other compounds, with a
mean concentration of 5.9 pg/m3. These concentrations are greater than PFOS concentrations
measured in the particle phase of air samples measured in Zeppelinstasjonen, Svalbard, Norway
{Butt, 2010, 1291056}. PFOS was measured in September and December, 2006 and August and
December, 2007, with mean concentrations of 0.11 pg/m3 (range: 0.03 pg/m3-0.50 pg/m3) and
0.18 pg/m3 (range: 0.02 pg/m3-0.97 pg/m3), respectively.
G-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
G.3.5 Other Possible Exposure Sources
PFOS has also been detected in soils and dust from carpets and upholstered furniture in homes,
offices, and vehicles. Incidental exposure from soils and dust is an important exposure route,
particularly for small children because of their increased level of hand-to-mouth behaviors
compared with adults. Also, the levels in soils and surface waters can affect the concentrations in
local produce, meat/poultry, dairy products, fish, and particulates in the air.
G.4 Recommended RSC
EPA used the Exposure Decision Tree methodology to derive the RSC for this MCLG (Figure
G-l) {U.S. EPA, 2000, 19428}. Findings from studies on populations in the United States, with
supporting evidence from Canada and Western Europe, suggest that diet, particularly fish, is the
major contributor to total PFOS exposure among adults, typically with dust as an important
additional exposure medium, especially for sensitive populations. Additional exposure sources
are consumer products and air (Box 2; Figure G-l). However, adequate data are not available to
describe central tendency and high-end exposures for all relevant exposure sources and pathways
(Box 3; Figure G-l). There is sufficient data on the physical/chemical properties, fate and
transport, and generalized information characterizing the likelihood of exposure to PFOS via
relevant sources (Box 4; Figure G-l). There are significant known or potential sources other than
drinking water (Box 6; Figure G-l), although there is not enough information available for each
pathway, particularly dust, air, consumer products, and food contact materials, to characterize
G-12
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
exposure (Box 8A; Figure G-l). Therefore, an RSC of 20% (0.20) should be used
(Box 8B; Figure G-l).
Figure G-l. Application of the Exposure Decision Tree {U.S. EPA, 2000,19428} for PFOS
Green highlighted boxes indicate selections made at each branch of the Decision Tree.
POD = point of departure; RiD = reference dose; UF = uncertainty factor.
In summary, based on the physical properties, detected levels, and available exposure
information for PFOS, food, and air are potentially significant sources. Following the Exposure
Decision Tree in EPA's 2000 Methodology {U.S. EPA 2000, 19428}, significant potential
sources other than drinking water ingestion exist; however, information is not available to
quantitatively characterize exposure from these different sources. Therefore, EPA recommends
an RSC of 20% (0.20) for PFOS.
G-13
------- |