March 2023
iji EPA Document No.
LhI r> r\r\T\r\ AA/-
822r23006
PUBLIC COMMENT DRAFT
APPENDIX: Toxicity Assessment and Proposed
Maximum Contaminant Level Goal for Perfluorooctanoic
Acid (PFOA) in Drinking Water
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PUBLIC COMMENT DRAFT
APPENDIX: Toxicity Assessment and Proposed Maximum Contaminant
Level Goal for Perfluorooctanoic Acid (PFOA)
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 822P23006
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 i
• Contents ii
• Figures vii
• Tables xv
• Acronyms and Abbreviations xxix
• Appendix A. Systematic Review Protocol for Updated PFOA Toxicity
Assessment A-l
A. 1 Overview of Background Information and Systematic Review Protocol A-2
A. 1.1 Summary of Chemical Identity and Occurrence Information A-2
A. 1.2 Problem Formulation A-3
A. 1.3 Overall Objective and Specific Aims A-5
A. 1.4 Populations, Exposures, Comparators, and Outcomes (PECO) Criteria A-6
A. 1.5 Literature Search A-9
A. 1.6 Literature Screening Process to Target Dose-Response Studies and PK
Models A-22
A. 1.7 Study Quality Evaluation Overview A-54
A. 1.8 Data Extraction for Epidemiological Studies A-102
A. 1.9 Data Extraction for Animal Toxicological Studies A-l 11
A. 1.10 Evidence Synthesis and Integration A-l 16
A. 1.11 Dose-Response Assessment: Selecting Studies and Quantitative Analysis ... A-
121
A. 1.12 Candidate Toxicity Value Derivation and Selection A-124
A.2 Meta-Analysis Table A-127
A.3 Studies Identified After Assessment Literature Cut-Off Date A-133
• 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-3
B. 1.4 Dermal Exposure B-3
B. 1.5 Developmental Exposure B-4
B.l.6 Bioavailability B-4
B.2 Distribution B-5
B.2.1 Protein Binding B-5
B.2.2 Subcellular Distribution B-8
B.2.3 Tissue Distribution B-8
B.2.4 Distribution during Reproduction and Development B-21
B.2.5 Volume of Distribution Data B-47
ii
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
B.3 Metabolism B-54
B.4 Excretion B-54
B.4.1 Urinary and Fecal Excretion B-54
B.4.2 Physiological and Mechanistic Factors Impacting Excretion B-61
B.4.3 Maternal Elimination Through Lactation and Fetal Partitioning B-67
B.4.4 Other Routes of Elimination B-69
B.4.5 Half-life Data B-70
• Appendix C. Non-priority Health System 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-15
C.1.3 Mechanistic Evidence Synthesis C-28
C. 1.4 Evidence Integration C-29
C.2 Endocrine C-18
C.2.1 Human Evidence Study Quality Evaluation and Synthesis C-18
C.2.2 Animal Evidence Study Quality Evaluation and Synthesis C-26
C.2.3 Mechanistic Evidence C-35
C.2.4 Evidence Integration C-36
C.3 Metabolic/Systemic C-42
C.3.1 Human Evidence Study Quality Evaluation and Synthesis C-42
C.3.2 Animal Evidence Study Quality Evaluation and Synthesis C-60
C.3.3 Mechanistic Evidence C-72
C.3.4 Evidence Integration C-73
C.4 Nervous C-80
C.4.1 Human Evidence Study Quality Evaluation and Synthesis C-80
C.4.2 Animal Evidence Study Quality Evaluation and Synthesis C-88
C.4.3 Mechanistic Evidence C-91
C.4.4 Evidence Integration C-92
C.5 Renal C-101
C.5.1 Human Evidence Study Quality Evaluation and Synthesis C-101
C.5.2 Animal Evidence Study Quality Evaluation and Synthesis C-106
C.5.3 Mechanistic Evidence C-l 15
C.5.4 Evidence Integration C-115
C.6 Hematological C-122
C.6.1 Human Evidence Study Quality Evaluation and Synthesis C-122
C.6.2 Animal Evidence Study Quality Evaluation and Synthesis C-125
C.6.3 Mechanistic Evidence C-128
C.6.4 Evidence Integration C-129
C.7 Respiratory C-132
C.7.1 Human Evidence Study Quality Evaluation and Synthesis C-132
iii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.7.2 Animal Evidence Study Quality Evaluation and Synthesis C-134
C.7.3 Mechanistic Evidence C-138
C.7.4 Evidence Integration C-138
C.8 Musculoskeletal C-142
C.8.1 Human Evidence Study Quality Evaluation and Synthesis C-142
C.8.2 Animal Evidence Study Quality Evaluation and Synthesis C-145
C.8.3 Mechanistic Evidence C-146
C.8.4 Evidence Integration C-147
C.9 Gastrointestinal C-151
C.9.1 Human Evidence Study Quality Evaluation and Synthesis C-151
C.9.2 Animal Evidence Study Quality Evaluation and Synthesis C-153
C.9.3 Mechanistic Evidence C-156
C. 9.4 Evi dence Integrati on C-156
C.10 Dental C-160
C.10.1 Human Evidence Study Quality Evaluation and Synthesis C-160
C.10.2 Animal Evidence Study Quality Evaluation and Synthesis C-162
C.10.3 Mechanistic Evidence C-162
C.10.4 Evidence Integration C-162
C.ll Ocular C-164
C. 11.1 Human Evidence Study Quality Evaluation and Synthesis C-164
C.11.2 Animal Evidence Study Quality Evaluation and Synthesis C-165
C.11.3 Mechanistic Evidence C-166
C.l 1.4 Evidence Integration C-167
C.12 Dermal C-169
C.12.1 Human Evidence Study Quality Evaluation and Synthesis C-169
C.12.2 Animal Evidence Study Quality Evaluation and Synthesis C-170
C.12.3 Mechanistic Evidence C-171
C.12.4 Evidence Integration C-172
• Appendix D. Detailed Information from Epidemiology Studies D-l
D. 1 Developmental D-2
D.2 Reproductive D-56
D.2.1 Male D-56
D.2.2 Female D-64
D.3 Hepatic D-74
D.4 Immune D-88
D.5 Cardiovascular D-l 29
D.5.1 Cardiovascular Endpoints D-129
D.5.2 Serum Lipids D-148
D.6 Endocrine D-179
D.7 Metabolic/Systemic D-187
D. 8 Nervous D-205
iv
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
D.9 Renal D-231
D. 10 Hematological D-243
D.ll Respiratory D-245
D.12 Musculoskeletal D-248
D.13 Gastrointestinal D-253
D.14 Ocular D-256
D.15 Dermal D-256
D.16 Cancer D-257
• Appendix E. Benchmark Dose Modeling E-269
E.l Epidemiology Studies E-269
E. 1.1 Modelling results for Immunotoxicity E-269
E. 1.2 Modeling Results for Decreased Birthweight E-291
E. 1.3 Modeling Results for Increased Cholesterol E-297
E. 1.4 Modeling Results for Liver Toxicity E-308
E. 1.5 Modeling Results for Cancer E-322
E.2 Toxicology Studies E-328
E.2.1 Butenhoff, 2012, 2919192 E-328
11.2.2 Dewitt, 2008, 1290826 11-331
11.2.3 Lau, 2006, 1276159 11-334
11.2.4 Li, 2018, 5084746 11-347
11.2.5 Loveless, 2008, 988599 11-350
E.2.6 Macon, 2011, 1276151 Error! Bookmark not defined.
11.2.7 NTP, 2020, 7330145 11-355
11.2.8 Song, 2018, 5079725 11-378
11.2.9 Wolf, 2007, 1332672 11-389
• Appendix F. Pharmacokinetic Modeling F-2
F.l Animal Pharmacokinetic Model F-2
F.l.l Comparison of Fits to Training Datasets Used in Wambaugh et al. (2013,
2850932) F-2
F.l .2 Visual Inspection of Test Datasets not Used for Initial Fitting F-5
F.l.3 Consideration of Hinderliter et al. (2006, 3749132) in the Animal Model ...F-9
F.2 Human Model Validation F-l 1
• 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 PFOA Sources G-3
G.3.1 Dietary Sources G-4
G.3.2 Consumer Product Uses G-9
v
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
G. 3.3 Indoor Dust G-10
G.3.4 Ambient Air G-10
G.3.5 Other Exposure Considerations G-l 1
G.4 Recommended RSC G-ll
vi
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figures
Figure A-l. Overview of Study Quality Evaluation Approach A-54
Figure A-2. Possible Domain Scores for Study Quality Evaluation A-56
Figure A-3. Overall Study Confidence Classifications A-56
Figure B-l. Localization of Transport Proteins B-62
Figure C-l. Summary of Study Evaluation for Epidemiology Studies of PFOA and Male
Reproductive Effects C-4
Figure C-2. Summary of Study Evaluation for Epidemiology Studies of PFOA and Female
Reproductive Effects C-9
Figure C-3. Summary of Study Evaluation for Epidemiology Studies of PFOA and Female
Reproductive Effects (Continued) C-10
Figure C-4. Summary of Study Evaluation for Epidemiology Studies of PFOA and Female
Reproductive Effects (Continued) C-l 1
Figure C-5. Summary of Study Evaluation for Toxicology Studies of PFOA and
Reproductive Effects C-l6
Figure C-6. Sperm Counts in Rodents Following Exposure to PFOA (logarithmic scale) C-19
Figure C-l. Percent Change in Male Reproductive Hormone Levels Relative to Controls in
Rodents Following Exposure to PFOA C-22
Figure C-8. Percent Change in Female Reproductive Hormone Levels Relative to Controls
in Rodents Following Exposure to PFOA C-24
Figure C-9. Summary of Mechanistic Studies of PFOA and Reproductive Effects C-29
Figure C-10. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Endocrine Effects C-21
Figure C-l 1. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Endocrine Effects (Continued) C-22
Figure C-12. Summary of Study Evaluation for Toxicology Studies of PFOA and
Endocrine Effects C-26
Figure C-13. Percent Change in Endocrine Organ Weights Relative to Controls in Rodents
Following Exposure to PFOAa C-28
Figure C-14. Percent Change in Thyroid and Thyroid-Related Hormone Levels of Male and
Female Rats Exposed to PFOA for 28 Days as Reported by NTP (2019,
5400977 )ab C-30
Figure C-15. Endocrine Organ Histopathology in Rodents Following Exposure to PFOA
(logarithmic scale) C-3 5
vii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure C-16. Summary of Mechanistic Studies of PFOA and Endocrine Effects C-36
Figure C-17. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Metabolic Effects C-45
Figure C-18. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Metabolic Effects (Continued) C-46
Figure C-19. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Metabolic Effects (Continued) C-47
Figure C-20. Summary of Study Evaluation for Toxicology Studies of PFOA and
Metabolic Effects C-61
Figure C-21. Summary of Study Evaluation for Toxicology Studies of PFOA and Systemic
Effects C-64
Figure C-22. Effects on Survival in Rodents Following Exposure to PFOA (logarithmic
scale) C-64
Figure C-23. Effects on Body Weight in Rodents Following Exposure to PFOA
(logarithmic scale) C-68
Figure C-24. Effects on Body Weight in Rodents Following Developmental Exposure to
PFOA (logarithmic scale) C-70
Figure C-25. Effects on Food Consumption in Rodents Following Exposure to PFOA C-71
Figure C-26. Summary of Mechanistic Studies of PFOA and Metabolic Effects C-72
Figure C-27. Summary of Mechanistic Studies of PFOA and Systemic Effects C-73
Figure C-28. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Neurological Effects C-82
Figure C-29. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Neurological Effects (Continued) C-83
Figure C-30 Summary of Study Evaluation for Toxicology Studies of PFOA and Nervous
Effects C-89
Figure C-31. Summary of Mechanistic Studies of PFOA and Nervous Effects C-92
Figure C-32. Summary of Study Evaluation for Epidemiology Studies of PFOA and Renal
Effects C-103
Figure C-33. Summary of Study Evaluation for Toxicology Studies of PFOA and Renal
Effects C-107
Figure C-34. Absolute Kidney Weights in Rodents Following Exposure to PFOA
(logarithmic scale) C-108
Figure C-35. Percent Change in Relative Kidney Weights of Male Rats Following
Exposure to PFOA C-l 10
viii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure C-36. Percent Change in Relative Kidney Weights of Female Rodents Following
Exposure to PFOA C-l 12
Figure C-37. Summary of Mechanistic Studies of PFOA and Renal Effects C-l 15
Figure C-38. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Hematological Effects C-124
Figure C-39. Summary of Study Evaluation for Toxicology Studies of PFOA and
Hematological Effects C-126
Figure C-40. Hematological Effects in Male and Female Sprague Dawley Rats Dosed with
PFOA for 28 Days as Reported by NTP (2019, 5400977) C-127
Figure C-41. Summary of Mechanistic Studies of PFOA and Hematological Effects C-129
Figure C-42. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Respiratory Effects C-133
Figure C-43. Summary of Study Evaluation for Toxicology Studies of PFOA and
Respiratory Effects C-135
Figure C-44. Incidence of Nonneoplastic Nasal Lesions in Male and Female Sprague-
Dawley Rats Following 28-day Oral Exposure to PFOA, as Reported by NTP
(2019, 5400977) C-l37
Figure C-45. Summary of Mechanistic Studies of PFOA and Respiratory Effects C-138
Figure C-46. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Musculoskeletal Effects C-144
Figure C-47. Summary of Study Evaluation for Toxicology Studies of PFOA and
Musculoskeletal Effects C-146
Figure C-48. Summary of Mechanistic Studies of PFOA and Musculoskeletal Effects C-147
Figure C-49. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Gastrointestinal Effects C-l53
Figure C-50. Summary of Study Evaluation for Toxicology Studies of PFOA and
Gastrointestinal Effects C-l54
Figure C-51. Gastrointestinal Effects in Rodents and Non-Human Primates Following
Exposure to PFOA (logarithmic scale) C-155
Figure C-52. Summary of Mechanistic Studies of PFOA and Gastrointestinal Effects C-156
Figure C-53. Summary of Study Evaluation for Epidemiology Studies of PFOA and Dental
Effects C-161
Figure C-54. Summary of Study Evaluation for Epidemiology Studies of PFOA and Ocular
Effects C-l 65
ix
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Figure C-55. Summary of Study Evaluation for Toxicology Studies of PFOA and Ocular
Effects C-166
Figure C-56. Summary of Mechanistic Studies of PFOA and Ocular Effects C-167
Figure C-57. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Dermal Effects C-170
Figure C-58. Summary of Study Evaluation for Toxicology Studies of PFOA and Dermal
Effects C-171
Figure C-59. Summary of Mechanistic Studies of PFOA and Dermal Effects C-172
Figure D-l. Overall Levels of Total Cholesterol in Adults from Epidemiology Studies
Following Exposure to PFOA D-l49
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-272
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-274
Figure E-3. Regression coefficients and 95% CIs between the log of the RCC ORs and
serum PFOA concentrations using data from Shearer et al. (2021, 7161466):
adjusted (gist) and unadjusted (vwls) for OR dependence E-326
Figure E-4. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 1 Model for Leydig Cell Adenomas in the Testes in Male Sprague-
Dawley Crl:COBS@CD(SD)BR Rats Following Exposure to PFOA with BMR
4% Extra Risk {Butenhoff, 2012, 2919192} E-331
Figure E-5. Plot of Mean Response by Dose with Fitted Curve for the Polynomial Degree 4
Model for Serum Sheep Red Blood Cells-specific IgM Antibody Titers in
Female C57BL/6N Mice (Study I) Following Exposure to PFOA {Dewitt,
2008, 1290826} 11-333
Figure E-6. Plot of Mean Response by Dose with Fitted Curve for the Selected Polynomial
Degree 2 Model for Time to Eye Opening in Fi Male and Female CD-I Mice
Following Exposure to PFOA {Lau, 2006, 1276159} E-341
Figure E-7. Plot of Mean Response by Dose with Fitted Curve for the Selected Hill Model
for Time to Eye Opening using Cavg,pup,iact in Fi Male and Female CD-I Mice
Following Exposure to PFOA {Lau, 2006, 1276159} E-343
Figure E-8. Plot of Mean Response by Dose with Fitted Curve for the Selected Power
Model for Time to Eye Opening using Cmax,pup,gest in Fi Male and Female CD-I
Mice Following Exposure to PFOA {Lau, 2006, 1276159} E-345
x
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-9. Plot of Mean Response by Dose with Fitted Curve for the Selected Hill Model
for Time to Eye Opening using Cmax,pup,iact in Fi Male and Female CD-I Mice
Following Exposure to PFOA {Lau, 2006, 1276159} E-347
Figure E-10. Plot of Incidence Rate by Dose with Fitted Curve for the Selected
Dichotomous Hill Model for Focal Necrosis in Male Crl:CD-l(ICR)BR Mice
Following Exposure to PFOA {Loveless, 2008, 988599} E-351
Figure E-l 1. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Probit
Model for Individual Cell Necrosis in Male Crl:CD-l(ICR)BR Mice Following
Exposure to PFOA {Loveless, 2008, 988599} E-353
Figure E-12. Plot of Mean Response by Dose with Fitted Curve for the Selected
Exponential 3 Model for IgM Serum Titer in Male Crl:CD-l(ICR)BR Mice
Following Exposure to PFOA {Loveless, 2008, 988599} E-355
Figure E-13. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 3 Model for Hepatocyte Single Cell Death in Fi Male Sprague-Dawley
Rats Following Postweaning Exposure to PFOA {NTP, 2020, 7330145} E-357
Figure E-14. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Gamma
Model for Hepatocyte Single Cell Death in Fi Male Sprague-Dawley Rats
Following Perinatal and Postweaning Exposure to PFOA {NTP, 2020,
7330145} E-358
Figure E-l 5. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 4 Model for Hepatocyte Single Cell Death in Fi Male Sprague-Dawley
Rats Following Exposure to PFOA (Pooled) {NTP, 2020, 7330145} E-360
Figure E-16. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Log-
Logistic Model for Necrosis in the Liver in Fi Male Sprague-Dawley Rats
Following Postweaning Exposure to PFOA {NTP, 2020, 7330145} E-362
Figure E-17. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 1 Model for Necrosis in the Liver in Fi Male Sprague-Dawley Rats
Following Perinatal and Postweaning Exposure to PFOA {NTP, 2020,
7330145} E-363
Figure E-18. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Log-
Logistic Model for Necrosis in the Liver in Fi Male Sprague-Dawley Rats
Following Exposure to PFOA (Pooled) {NTP, 2020, 7330145} E-365
Figure E-19. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 3 Model for Hepatocellular Adenomas in Fi Male Sprague-Dawley Rats
Following Postweaning Exposure to PFOA {NTP, 2020, 7330145} E-367
Figure E-20. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 2 Model for Hepatocellular Adenomas in Fi Male Sprague-Dawley Rats
Following Perinatal and Postweaning Exposure to PFOA {NTP, 2020,
7330145} E-368
xi
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-21. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 3 Model for Hepatocellular Adenomas in Fi Male Sprague-Dawley Rats
Following Exposure to PFOA (Pooled) {NTP, 2020, 7330145} E-369
Figure E-22. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 3 Model for Hepatocellular Adenoma or Carcinoma in Fi Male
Sprague-Dawley Rats Following Postweaning Exposure to PFOA {NTP, 2020,
7330145} E-371
Figure E-23. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 2 Model for Hepatocellular Adenoma or Carcinoma in Fi Male
Sprague-Dawley Rats Following Perinatal and Postweaning Exposure to PFOA
{NTP, 2020, 7330145} E-373
Figure E-24. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 2 Model for Hepatocellular Adenoma or Carcinoma in Fi Male
Sprague-Dawley Rats Following Exposure to PFOA (Pooled) {NTP, 2020,
7330145} E-374
Figure E-25. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 2 Model for Pancreatic Acinar Cell Adenoma in Fi Male Sprague-
Dawley Rats Following Postweaning Exposure to PFOA {NTP, 2020,
7330145} E-376
Figure E-26. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 1 Model Pancreatic Acinar Cell Adenoma in Fi Male Sprague-Dawley
Rats Following Perinatal and Postweaning Exposure to PFOA {NTP, 2020,
7330145} 11-377
Figure E-27. Plot of Mean Response by Dose with Fitted Curve for the Selected
Exponential 2 Model for Offspring Survival using Cavg,pup,gest in Fi Male and
Female CD-I Mice Following Exposure to PFOA {Song, 2018, 5079725} E-381
Figure E-28. Plot of Mean Response by Dose with Fitted Curve for the Selected
Polynomial Degree 3 Model for Offspring Survival using Cavg,pup,iact in Fi Male
and Female CD-I Mice Following Exposure to PFOA {Song, 2018, 5079725} .E-383
Figure E-29. Plot of Mean Response by Dose with Fitted Curve for the Selected
Polynomial Degree 3 Model for Offspring Survival using Cavg,pup,gest,iact in Fi
Male and Female CD-I Mice Following Exposure to PFOA {Song, 2018,
5079725} E-385
Figure E-30. Plot of Mean Response by Dose with Fitted Curve for the Selected
Polynomial Degree 2 Model for Offspring Survival using Cmax,pup,gest in Fi Male
and Female CD-I Mice Following Exposure to PFOA {Song, 2018, 5079725} .E-387
Figure E-31. Plot of Mean Response by Dose with Fitted Curve for the Selected
Polynomial Degree 3 Model for Offspring Survival using Cmax,pup,iact in Fi Male
and Female CD-I Mice Following Exposure to PFOA {Song, 2018, 5079725} .E-389
XII
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-32. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Gamma
Model for Dams with Whole Litter Loss in Po Female CD-I Mice Following
Exposure to PFOA (constant variance) {Wolf, 2007, 1332672} E-l
Figure F-l. Experimentally Observed Serum Concentrations {Lou, 2009, 2919359} and
Median Predictions for a Single Oral Dose of 1, 10, or 60 mg/kg PFOA to
Female CD1 Mice F-2
Figure F-2. Experimentally Observed Serum Concentrations {Kemper, 2003, 6302380}
and Median Prediction for a Single IV Dose of 1 mg/kg or an Oral Dose of 0.1,
1, 5, or 25 mg/kg PFOA to Male Sprague-Dawley Rats F-3
Figure F-3. Experimentally Observed Serum Concentrations {Kemper, 2003, 6302380}
and Median Prediction for a Single IV Dose of 1 mg/kg or a Single Oral Dose
of 0.1, 1, 5, or 15 mg/kg PFOA to Female Sprague-Dawley Ratsa F-3
Figure F-4. Model prediction summary for PFOA training data F-4
Figure F-5. PFOA Sensitivity Coefficients of the Adult Model and Developmental Model F-5
Figure F-6. Experimentally Observed Serum Concentrations {Dzierlenga, 2020, 5916078}
and Median Predictions for a Single IV Dose of 6 mg/kg or a Single Oral Dose
of 6, 12, or 45 mg/kg PFOA to Male Sprague-Dawley Rats F-6
Figure F-7. Experimentally Observed Serum Concentrations {Dzierlenga, 2020, 5916078}
and Median Predictions for a Single IV Dose of 40 mg/kg or a Single Oral Dose
of 40, 80, or 320 mg/kg PFOA to Female Sprague-Dawley Ratsa'b F-6
Figure F-8. Experimentally Observed Serum Concentrations and Median Predictions for a
Single IV Dose of 1 mg/kg or an Oral Gavage Dose of 1 mg/kg PFOA {Kim,
2016, 3749289} or an IV Dose of 20 mg/kg PFOA {Kudo, 2002, 2990271} to
Male Sprague-Dawley Rats F-7
Figure F-9. Experimentally Observed Serum Concentrations and Median Predictions for a
Single IV Dose of 1 mg/kg or an Oral Gavage Dose of 1 mg/kg PFOA {Kim,
2016, 3749289} or an IV Dose of 20 mg/kg PFOA {Kudo, 2002, 2990271} to
Female Sprague-Dawley Ratsa F-7
Figure F-10. Observed and Predicted PFOA Plasma Concentration in Female Sprague-
Dawley Rats following Perinatal, Lactational, and Post-weaning Exposure
during Study 1 of NTP (2020, 7330145)a b F-8
Figure F-l 1. Observed and Predicted PFOA Plasma Concentrations in Male Sprague-
Dawley Rats following Perinatal, Lactational, and Post-weaning Exposure
during Study 2 of NTP (2020, 7330145 )ab F-8
Figure F-12. Model Prediction Summary for PFOA Test Data F-9
Figure F-13. Model Prediction Summary for PFOA Data from Hinderliter et al. (2006,
3749132) F-10
Figure F-14. Model Comparison F-l 1
xiii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure F-15. Predicted Child Serum Levels Compared to Reported Values F-12
Figure F-16. Comparison of Predicted and Observed Child Serum Levels F-13
Figure F-17. Sensitivity Coefficients F-14
Figure F-18. 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-15
Figure F-19. Predicted Child Serum Levels Compared to Reported Values with Constant
Volume of Distribution and Variable Exposure Based on Drinking Water Intake. F-15
Figure G-l. Application of the Exposure Decision Tree {U.S. EPA, 2000, 19428} for
PFOA G-l 2
xiv
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Tables
Table A-l. Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for a
Systematic Review on the Health Effects from Exposure to PFOA and PFOS A-7
Table A-2. Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for
Absorption, Distribution, Metabolism, and/or Excretion (ADME) Studies A-8
Table A-3. Populations, Exposures, Comparators, and Outcomes (PECO) Criteria for
Mechanistic Studies A-9
Table A-4. Search String for April 2019 Database Searches A-10
Table A-5. Search String for September 2020 and February 2022 Database Searches A-12
Table A-6. Key Epidemiological Studies of Priority Health Outcomes Identified from the
2016 PFOA Health Effects Support Document A-15
Table A-7. Key Animal Toxicological Studies Identified from the 2016 PFOA Health
Effects Support Document A-19
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-23
Table A-9. DistillerSR Form for Title/Abstract Screening A-25
Table A-10. SWIFT-Active Form for Title/Abstract Screening A-26
Table A-l 1. Supplemental Tags for Title/Abstract and Full-Text Screening A-26
Table A-12. Mechanistic Study Categories Considered as Supplemental A-27
Table A-13. DistillerSR Form for Full-Text Screening A-29
Table A-14. Health Effect Categories Considered for Epidemiological Studies A-32
Table A-l5. litstream Forms for ADME Screening and Light Data Extraction A-36
Table A-16. litstream Forms for Mechanistic Screening and Light Data Extraction A-46
Table A-17. Study Quality Evaluation Considerations for Participant Selection A-58
Table A-18. Study Quality Evaluation Considerations for Exposure Measurement A-60
Table A-19. Study Quality Evaluation Considerations for PFAS-Specific Exposure
Measurement A-63
Table A-20. Study Quality Evaluation Considerations for Outcome Ascertainment A-65
Table A-21. Study Quality Evaluation Considerations for Confounding A-67
Table A-22. Study Quality Evaluation Considerations for Analysis A-70
Table A-23. Study Quality Evaluation Considerations for Selective Reporting A-72
xv
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table A-24. Study Quality Evaluation Considerations for Study Sensitivity A-73
Table A-25. Study Quality Evaluation Considerations for Overall Study Confidence -
Epidemiological Studies A-74
Table A-26. Study Evaluation Considerations for Reporting Quality A-76
Table A-27. Study Quality Evaluation Considerations for Selection and Performance -
Allocation A-79
Table A-28. Study Quality Evaluation Considerations for Selection and Performance -
Observational Bias/Blinding A-81
Table A-29. Study Quality Evaluation Considerations for Confounding/Variable Control A-85
Table A-30. Study Quality Evaluation Considerations for Selective Reporting and Attrition
- Reporting and Attrition Bias A-87
Table A-31. Study Quality Evaluation Considerations for Exposure Methods Sensitivity -
Chemical Administration and Characterization A-89
Table A-32. Study Quality Evaluation Considerations for Exposure Methods Sensitivity -
Exposure Timing, Frequency, and Duration A-92
Table A-33. Study Quality Evaluation Considerations for Outcome Measures and Results
Display - Endpoint Sensitivity and Specificity A-94
Table A-34. Study Quality Evaluation Considerations for Outcome Measures and Results
Display - Results Presentation A-97
Table A-35. Study Quality Evaluation Considerations for Overall Study Confidence -
Animal Toxicological Studies A-99
Table A-36. DistillerSR Form Fields for Data Extraction of Epidemiological Studies A-102
Table A-37. Epidemiological Study Design Definitions A-109
Table A-38. HAWC Form Fields for Data Extraction of Animal Toxicological Studies A-l 11
Table A-39. Framework for Strength-of-Evidence Judgments for Epidemiological Studiesa A-l 17
Table A-40. Framework for Strength-of-Evidence Judgments for Animal Toxicological
Studiesa A-l 18
Table A-41. Evidence Integration Judgments for Characterizing Potential Human Health
Effects in the Evidence Integration51 A-l 19
Table A-42. Epidemiologic Meta-Analysis Studies Identified From Literature Review A-127
Table A-43. Toxicologic Meta-Analysis Studies Identified From Literature Review A-132
Table A-44. Studies Identified After Updated Literature Review (Published or Identified
After February 2022) A-133
xvi
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-l. Cellular Accumulation and Retention Relative to Lipophilicity and
Phospholipidicity as Reported by Sanchez Garcia et al. (2018, 4234856) B-l
Table B-2. PFOA Parameters from Toxicokinetic Studies Informing Bioavailability in
Sprague-Dawley Rats B-4
Table B-3. Dissociation Constants of Binding Between PFOA and Albumin as Reported by
Han et al. (2003, 5081471) B-5
Table B-4. Tissue Distribution of PFOA in Wistar Rats After Exposure via Gavage for
28 Days as Reported by Ylinen et al. (1990, 5085631) B-12
Table B-5. Distribution of PFOA in Male Sprague-Dawley Rats After a Single Oral Gavage
Dosea as Reported by Kemper et al. (2003, 6302380) B-14
Table B-6. Distribution of PFOA in Female Sprague-Dawley Rats After a Single Oral
Gavage Dosea as Reported by Kemper et al. (2003, 6302380) B-15
Table B-7. Distribution of PFOA in Male C57BL/6 Mice Following Exposure to 14C-
PFOA for 1, 3, or 5 days in Feeda as Reported by Bogdanska et al. (2020,
6315801) B-l 8
Table B-8. PFOA Concentrations in Wild-type and PPARa-null Male Mice Exposed to
PFOA by Gavage for Four Weeksa as Reported by Minata et al. (2010,
1937251) B-20
Table B-9. PFOA Concentrations in Human Cord Blood, Maternal Blood, and
Transplacental Transfer Ratios (RCM) B-26
Table B-10. PFOA Concentrations in Human Maternal Blood, Cord Blood, Placenta and
Amniotic Fluid Across Studies B-30
Table B-l 1. Summary of Studies Evaluating PFOA concentrations in Maternal Serum,
Breast Milk, and Infant Serum B-35
Table B-12. Percent Change in PFOA Ratios in Maternal Serum to Breast Milk and Breast
Milk to Infant Serum by Infant Age in Humans as Reported by Mondal et al.
(2014, 2850916) B-37
Table B-13. Percent Change in PFOA Serum Concentration by Exclusive, Mixed or No
Breastfeeding Per Month in Humans as Reported by Mogensen et al. (2015,
3859839) B-3 7
Table B-14. Maternal Plasma PFOA Levels in Sprague-Dawley Rats During Gestation and
Lactation51 as Reported by Hinderliter et al. (2005, 1332671) B-38
Table B-15. Placenta, Amniotic Fluid, and Embryo/Fetus PFOA Concentrations in
Sprague-Dawley Ratsa as Reported by Hinderliter et al. (2005, 1332671) B-39
Table B-16. Fetus/Pup PFOA Concentration in Sprague-Dawley Rats During Gestation and
Lactation51 as Reported by Hinderliter et al. (2005, 1332671) B-39
xvii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-17. Maternal Milk PFOA Concentration in Sprague-Dawley Rat During Lactation51
as Reported by Hinderliter et al. (2005, 1332671) B-39
Table B-18. Plasma PFOA Concentrations in Postweaning Sprague-Dawley Ratsa as
Reported by Han (2003, 9978263) B-40
Table B-19. Plasma PFOA Concentrations in Male Sprague-Dawley Rats at 2 and 24 hours
after Oral Gavage as Reported by Hinderliter et al. (2006, 3749132) B-41
Table B-20. Plasma PFOA Concentrations in Female Sprague-Dawley Rats at 2 and 24
hours after Oral Gavage as Reported by Hinderliter et al. (2006, 3749132) B-41
Table B-21. Select Fluids and Tissues PFOA Concentrations in CD-I Mice During
Gestation and Lactation51 as Reported by Fenton et al. (2009, 194799) B-43
Table B-22. Serum, Liver, and Brain PFOA Concentration in Female CD-I Mouse Pups
After GD 10-17 Exposure51 as Reported by Macon et al. (2011, 1276151) B-44
Table B-23. Serum PFOA Concentrations in Female CD-I Mouse Pups After GD 10-17
Exposure as Reported by Macon et al. (2011, 1276151) B-45
Table B-24. Serum PFOA Concentration in CD-I Mice Over Three Generations51 as
Reported by White et al. (2011, 1276150) B-46
Table B-25. Maternal Serum, Amniotic Fluid, and Whole Embryo PFOA Concentrations in
CD-I Mice Exposed During Gestation Day 1.5-17.5 as Reported by Blake et al.
(2020, 6305864) B-46
Table B-26. Summary of PFOA Volume of Distribution Values Assigned in Human
Studies B-48
Table B-27. PFOA Volume of Distribution in Serum of FVB/NJcl Mice as Reported by
Fujii et al. (2015, 2816710) B-49
Table B-28. Summary of PFOA Volume of Distribution Calculations in Rats B-51
Table B-29. Urine PFOA Concentrations in Male and Female Sprague-Dawley Rats, 24-
Hours After Oral Gavage51 as Reported by Hinderliter et al. (2006, 3749132) B-57
Table B-30.Cumulative Percent 14C-PFOA Excreted in Urine and Feces by Male and
Female CD Rats51 as Reported by Hundley et al. (2006, 3749054) B-58
Table B-31. Percentage of Dose Excreted in Urine and Feces of Male and Female Sprague-
Dawley Rats exposed to 14C-PFOA via Oral Gavage as Reported by Kemper
(2003, 6302380) B-58
Table B-32. Cumulative Percent 14C-PFOA Excreted in Urine and Feces in Mouse,
Hamster, and Rabbit51 as Reported by Hundley et al. (2006, 3749054) B-60
Table B-33. Kinetic Parameters of Perfluorinated Carboxylate Transport by OAT1, OAT3,
and OATPlal as Reported by Weaver et al. (2010, 2010072) B-64
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-34. 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-70
Table B-35. Summary of PFOA Concentration in Blood and Urine in Relation to Half-life
values in Humans B-76
Table B-36. Summary of Human PFOA Half-Life Values B-78
Table B-37. PK Parameters in Male Sprague-Dawley Rats Following Administration of
PFOA as Reported by Kemper et al. (2003, 6302380) B-81
Table B-38. PK Parameters in Female Sprague-Dawley Rats Following Administration of
PFOA as Reported by Kemper et al. (2003, 6302380) B-82
Table B-39. PK Parameters in Male and Female Sprague-Dawley Rats Following Oral and
IV Administration of PFOA as Reported by Kim et al. (2016, 3749289) B-83
Table B-40. Summary of Animal PFOA Half-life Values Identified in the Literature
Review B-84
Table C-l. Evidence Profile Table for PFOA Reproductive Effects in Males C-2
Table C-2. Evidence Profile Table for PFOA Reproductive Effects in Females C-l 1
Table C-3. Associations Between PFOA Exposure and Thyroid and Thyroid-Related
Hormone Effects in Rodents and Non-human Primates C-31
Table C-4. Associations Between PFOA Exposure and Adrenocortical Hormone Effects in
Rodents C-33
Table C-5. Evidence Profile Table for PFOA Endocrine Effects C-38
Table C-6. Evidence Profile Table for PFOA Systemic and Metabolic Effects C-75
Table C-l. Evidence Profile Table for PFOA Nervous System Effects C-94
Table C-8. Incidences of Nonneoplastic Lesions in the Kidneys of Female Sprague-Dawley
Rats as Reported by NTP (2020, 7330145) C-l 13
Table C-9. Evidence Profile Table for PFOA Renal Effects C-l 17
Table C-10. Evidence Profile Table for PFOA Hematological Effects C-130
Table C-l 1. Incidences of Non-Neoplastic Pulmonary Lesions in Male Rats as Reported by
Butenhoff et al. (2012, 2919192) C-l36
Table C-12. Evidence Profile Table for PFOA Respiratory Effects C-140
Table C-13. Evidence Profile Table for PFOA Musculoskeletal Effects C-149
Table C-14. Evidence Profile Table for PFOA Gastrointestinal Effects C-158
Table C-15. Evidence profile table for PFOA Dental Effects C-163
Table C-16. Evidence Profile Table for PFOA Ocular Effects C-168
XIX
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-17. Evidence Profile Table for PFOA Dermal Effects C-173
Table D-l. Associations Between PFOA Exposure and Developmental Effects in Recent
Epidemiological Studies D-3
Table D-2. Associations Between PFOA Exposure and Male Reproductive Effects in
Recent Epidemiologic Studies D-56
Table D-3. Associations between PFOA Exposure and Female Reproductive Health Effects
in Female Children and Adolescents D-64
Table D-4. Associations between PFOA Exposure and Female Reproductive Health Effects
in Pregnant Women D-68
Table D-5. Associations between PFOA Exposure and Female Reproductive Health Effects
in Non-Pregnant Adult Women D-71
Table D-6. Associations Between PFOA Exposure and Hepatic Effects in Epidemiology
Studies D-74
Table D-7. Associations between PFOA Exposure and Vaccine Response in Recent
Epidemiological Studies D-88
Table D-8. Associations between PFOA Exposure and Infectious Disease in Recent
Epidemiological Studies D-100
Table D-9. Associations Between PFOA Exposure and Asthma in Recent Epidemiologic
Studies D-l 08
Table D-10. Associations Between PFOA Exposure and Allergies in Recent Epidemiologic
Studies D-l 17
Table D-l 1. Associations Between PFOA Exposure and Eczema in Recent Epidemiologic
Studies D-l 22
Table D-12. Associations Between PFOA Exposure and Autoimmune Health Effects in
Recent Epidemiologic Studies D-l25
Table D-13. Associations Between PFOA Exposure and Cardiovascular Effects in Recent
Epidemiological Studies D-129
Table D-14. Associations Between PFOA Exposure and Serum Lipid Effects in Recent
Epidemiologic Studies D-149
Table D-l5. Associations Between PFOA Exposure and Endocrine Effects in Recent
Epidemiologic Studies D-l79
Table D-16. Associations Between PFOA Exposure and Metabolic Effects in Recent
Epidemiologic Studies D-l87
Table D-17. Associations Between PFOA Exposure and Neurological Effects in Recent
Epidemiologic Studies D-205
xx
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table D-18. Associations Between PFOA Exposure and Renal Effects in Recent
Epidemiologic Studies D-231
Table D-19. Associations Between PFOA Exposure and Hematological Effects in Recent
Epidemiologic Studies D-243
Table D-20. Associations Between PFOA Exposure and Respiratory Effects in Recent
Epidemiologic Studies D-245
Table D-21. Associations Between PFOA Exposure and Musculoskeletal Health Effects in
Recent Epidemiologic Studies D-248
Table D-22. Associations Between PFOA Exposure and Gastrointestinal Health Effects in
Recent Epidemiologic Studies D-253
Table D-23. Associations Between PFOA Exposure and Dental Health Effects in Recent
Epidemiologic Studies D-255
Table D-24. Associations Between PFOA Exposure and Ocular Effects in Recent
Epidemiologic Studies D-256
Table D-25. Associations Between PFOA Exposure and Dermal Health Effects in Recent
Epidemiologic Studies D-256
Table D-26. Associations Between PFOA Exposure and Cancer in Recent Epidemiologic
Studies D-257
Table E-l. Results specific to the slope from the linear analyses of PFOA 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-270
Table E-2. BMDs and BMDLs for effect of PFOA 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-274
Table E-3. Results of the linear analyses of PFOA 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-275
Table E-4. BMDs and BMDLs for effect of PFOA measured perinatally and anti-tetanus
antibody concentrations at age five years {Budtz-j0rgensen, 2018, 508363 l}....E-277
Table E-5. BMDs and BMDLs for effect of serum PFOA in children on anti-tetanus
antibody concentrations using a BMR of V2 SD change in logio (tetanus
antibodies concentration) and a BMR of 1 SD change in logio (tetanus
antibodies concentration) Timmerman et al. (2021, 9416315) E-278
Table E-6. BMDLs for effect of PFOA on anti-tetanus antibody concentrations using a
BMR of1 - SD E-279
xxi
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-7. Results specific to the slope from the linear analyses of PFOA 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-280
Table E-8. BMDs and BMDLs for effect of PFOA 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-282
Table E-9. Results of the analyses of PFOA 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-283
Table E-10. BMDs and BMDLs for effect of PFOA measured perinatally and anti-
diphtheria antibody concentrations at age five years {Budtz-j0rgensen, 2018,
5083631} E-285
Table E-l 1. BMDs and BMDLs for effect of PFOA 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
(diphtheria antibodies concentration) E-286
Table E-12. BMDLs for effect of PFOA on anti-diphtheria antibody concentrations using a
BMR of - SD E-287
Table E-13. Dose-Response Modeling Data for Decreased HiB Antibody Levels in
Abraham et al. (2020, 6506041) E-288
Table E-14. Summary of Benchmark Dose Modeling Results for Decreased HiB Antibody
Levels in Abraham et al. (2020, 6506041) E-289
Table E-l 5. BMDs and BMDLs for effect of PFOA on decreased birth weight, by using
percentage (8.27%) of live births falling below the public health definition of
low birth weight, or alternative study-specific tail probability E-296
Table E-16. BMDs and BMDLs for effect of PFOA 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-297
Table E-17. NHANES mean and standard deviation of TC (mg/dL) and mean PFOA
(ng/ml.) I >299
Table E-l 8. BMDs and BMDLs for effect of PFOA on increased cholesterol in Dong et al.
(2019, 5080195) E-300
Table E-19. NHANES mean and standard deviation of ln(TC) (ln(mg/dL)) and mean
ln(PFOA) (ln(ng/ml.)) E-301
Table E-20. BMD and BMDL for effect of PFOA on increased cholesterol in Steenland et
al. (2009, 1291109) E-301
xxii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-21. Regression Results for Serum Total Cholesterol by Deciles of serum PFOA
from Steenland et al. (2009, 1291109) E-302
Table E-22. Summary of Benchmark Dose Modeling Results for Increased Mean Serum
Total Cholesterol in Steenland et al. (2009, 1291109) E-303
Table E-23. Odds ratios for elevated serum TC by quartiles of serum PFOA from
Steenland et al. (2009, 1291109) E-304
Table E-24. Summary of Benchmark Dose Modeling Results for Elevated Total
Cholesterol in Steenland et al. (2009, 1291109) E-305
Table E-25. Adjusted Mean Differences in Serum Total Cholesterol by Quartiles of Serum
PFOA (ng/ml.) from Lin et al. (2019, 1291109) E-306
Table E-26. Summary of Benchmark Dose Modeling Results for Increase Mean Serum
Total Cholesterol from Lin et al. (2019, 5187597) E-307
Table E-27. BMDLs for effect of PFOA on serum total cholesterol using a BMR of 5% E-308
Table E-28. Odds Ratios for Elevated ALT by Decile of PFOA serum concentrations
(ng/ml.) from Gallo et al. (2012, 1276142) E-309
Table E-29. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et
al. (2012, 1276142) Using the Unadjusted Mean PFOA Serum Concentration ...E-311
Table E-30. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et
al. (2012, 1276142) Using the Adjusted, No Intercept Mean PFOA Serum
Concentration E-312
Table E-31. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et
al. (2012, 1276142) Using the Unadjusted Median PFOA Serum Concentration E-313
Table E-32. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et
al. (2012, 1276142) Using the Adjusted, No Intercept Median PFOA Serum
Concentration E-314
Table E-33. NHANES mean and standard deviation of ln(ALT) (In IU/L) and mean PFOA
(In ng/mL) E-315
Table E-34. Prevalence of elevated ALT E-316
Table E-3 5. BMD and BMDL for effect of PFOA (ng/mL) on increased ALT in Gallo et al.
(2012, 1276142) E-317
Table E-36. BMD and BMDL for effect of PFOA (ng/mL) on increased ALT in Nian et al.
(2019, 5080307) E-318
Table E-37. Linear Regression results for In (ALT) by quintiles of serum PFOA
Concentration in Darrow et al. (2016, 3749173) E-318
Table E-38. BMD and BMDL for effect of PFOA (ng/mL) on increased ALT in Darrow et
al. (2016, 3749173) E-319
xxiii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-39. Dose-Response Modeling Data for Increased Mean ALT Concentration in
Darrow et al. (2016, 3749173) E-319
Table E-40. Summary of Benchmark Dose Modeling Results for Increased Mean ALT
concentrations in Darrow et al. (2016, 3749173) E-321
Table E-41. BMDLs for effect of PFOA on serum ALT using a BMR of 5% E-322
Table E-42. ORs for the association between PFOA serum concentrations and RCC in
Shearer et al. (2021, 7161466) and data used for CSF calculations E-324
Table E-43. Internal CSF calculations for Shearer et al. (2021, 7161466) and Vieira et al.
(2013, 2919154) studies E-325
Table E-44. ORs for the association between PFOA serum concentrations and RCC in
Vieira et al. (2013, 2919154) and data used for CSF calculations E-327
Table E-45.CSF calculations pooling dose-response for Shearer et al. (2021, 7161466) and
Vieira et al. (2013, 2919154) studies E-328
Table E-46. Dose-Response Modeling Data for Leydig Cell Adenomas in the Testes in
Male Sprague-Dawley Crl:COBS@CD(SD)BR Rats Following Exposure to
PFOA {Butenhoff, 2012, 2919192} E-329
Table E-47. Summary of Benchmark Dose Modeling Results for Leydig Cell Adenomas in
the Testes in Male Sprague-Dawley Crl:COBS@CD(SD)BR Rats Following
Exposure to PFOA {Butenhoff, 2012, 2919192} E-330
Table E-48. Dose-Response Modeling Data for Serum Sheep Red Blood Cells-specific IgM
Antibody Titers in Female C57BL/6N Mice (Study I) Following Exposure to
PFOA {Dewitt, 2008, 1290826} 11-332
Table E-49. Summary of Benchmark Dose Modeling Results for Serum Sheep Red Blood
Cells-specific IgM Antibody Titers in Female C57BL/6N Mice (Study I)
Following Exposure to PFOA (nonconstant variance) {Dewitt, 2008, 1290826} E-332
Table E-50. Dose-Response Modeling Data for Serum Sheep Red Blood Cells-specific IgM
Antibody Titers in Female C57BL/6N Mice (Study II) Following Exposure to
PFOA {Dewitt, 2008, 1290826} 11-333
Table E-52. Summary of Benchmark Dose Modeling Results for Serum Sheep Red Blood
Cells-specific IgM Antibody Titers in Female C57BL/6N Mice (Study II)
Following Exposure to PFOA (nonconstant variance) {Dewitt, 2008, 1290826}E-334
Table E-53. Dose-Response Modeling Data for Prenatal Loss in Po Female CD-I Mice
Following Exposure to PFOA {Lau, 2006, 1276159} E-335
Table E-54. Summary of Benchmark Dose Modeling Results for Prenatal Loss using
Cavg,dam,gest in Po Female CD-I Mice Following Exposure to PFOA (nonconstant
variance) {Lau, 2006, 1276159} E-335
xxiv
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-55. Summary of Benchmark Dose Modeling Results for Prenatal Loss using
Cmax,dam in Po Female CD-I Mice Following Exposure to PFOA (nonconstant
variance) {Lau, 2006, 1276159} E-336
Table E-56. Dose-Response Modeling Data for Fetal Body Weight in Fi Male and Female
CD-I Mice Following Exposure to PFOA {Lau, 2006, 1276159} E-337
Table E-57. Summary of Benchmark Dose Modeling Results for Fetal Body Weight
Change in Fi Male and Female CD-I Mice Following Exposure to PFOA
(constant variance) {Lau, 2006, 1276159} E-338
Table E-58. Dose-Response Modeling Data for Time to Eye Opening in Fi Male and
Female CD-I Mice Following Exposure to PFOA {Lau, 2006, 1276159} E-339
Table E-59. Summary of Benchmark Dose Modeling Results for Time to Eye Opening
using Cavg,puP,gest in Fi Male and Female CD-I mice Following Exposure to
PFOA (nonconstant variance) {Lau, 2006, 1276159} E-340
Table E-60. Summary of Benchmark Dose Modeling Results for Time to Eye Opening
using Cavg,pup,iact in Fi Male and Female CD-I mice Following Exposure to
PFOA (nonconstant variance) {Lau, 2006, 1276159} E-342
Table E-61. Summary of Benchmark Dose Modeling Results for Time to Eye Opening
using Cmax,puP,gest in Fi Male and Female CD-I mice Following Exposure to
PFOA (nonconstant variance) {Lau, 2006, 1276159} E-344
Table E-62. Summary of Benchmark Dose Modeling Results for Time to Eye Opening
using Cmax,puP,iact in Fi Male and Female CD-I mice Following Exposure to
PFOA (nonconstant variance) {Lau, 2006, 1276159} E-346
Table E-63. Dose-Response Modeling Data for Fetal Body Weight in Fi Male and Female
Kunming Mice Following Exposure to PFOA {Li, 2018, 5084746} E-347
Table E-64. Summary of Benchmark Dose Modeling Results for Fetal Body Weight in Fi
Male and Female Kunming Mice Following Exposure to PFOA (constant
variance) {Li, 2018, 5084746} I>349
Table E-65. Dose-Response Modeling Data for Focal Necrosis in Male Crl:CD-l(ICR)BR
Mice Following Exposure to PFOA {Loveless, 2008, 988599} E-350
Table E-66. Summary of Benchmark Dose Modeling Results for Focal Necrosis in Male
Crl:CD-l(ICR)BR Mice Following Exposure to PFOA {Loveless, 2008,
988599} E-350
Table E-67. Dose-Response Modeling Data for Individual Cell Necrosis in Male Crl:CD-
1(ICR)BR Mice Following Exposure to PFOA {Loveless, 2008, 988599} E-352
Table E-68. Summary of Benchmark Dose Modeling Results for Individual Cell Necrosis
in Male Crl:CD-l(ICR)BR Mice Following Exposure to PFOA {Loveless,
2008, 988599} E-352
XXV
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-69. Dose-Response Modeling Data for IgM Serum Titer in Male Crl:CD-
1(ICR)BR Mice Following Exposure to PFOA {Loveless, 2008, 988599} E-353
Table E-70. Summary of Benchmark Dose Modeling Results for IgM Serum Titer in Male
Crl:CD-l(ICR)BR Mice Following Exposure to PFOA (constant variance)
{Loveless, 2008, 988599} E-354
Table E-71. Dose-Response Modeling Data for Hepatocyte Single Cell Death in Fi Male
Sprague-Dawley Rats Following Exposure to PFOA {NTP, 2020, 7330145} E-355
Table E-72. Summary of Benchmark Dose Modeling Results for Hepatocyte Single Cell
Death in Fi Male Sprague-Dawley Rats Following Postweaning Exposure to
PFOA {NTP, 2020, 7330145} E-356
Table E-73. Summary of Benchmark Dose Modeling Results for Hepatocyte Single Cell
Death in Fi Male Sprague-Dawley Rats Following Perinatal and Postweaning
Exposure to PFOA {NTP, 2020, 7330145} 11-357
Table E-74. Summary of Benchmark Dose Modeling Results for Hepatocyte Single Cell
Death in Fi Male Sprague-Dawley Rats Following Exposure to PFOA (Pooled)
{NTP, 2020, 7330145} 11-359
Table E-75. Dose-Response Modeling Data for Necrosis in Fi Male Sprague-Dawley Rats
Following Exposure to PFOA {NTP, 2020, 7330145} E-360
Table E-76. Summary of Benchmark Dose Modeling Results for Necrosis in the Liver in Fi
Male Sprague-Dawley Rats Following Postweaning Exposure to PFOA {NTP,
2020, 7330145} 11-361
Table E-77. Summary of Benchmark Dose Modeling Results for Necrosis in the Liver in Fi
Male Sprague-Dawley Rats Following Perinatal and Postweaning Exposure to
PFOA {NTP, 2020, 7330145} 11-362
Table E-78. Summary of Benchmark Dose Modeling Results for Necrosis in the Liver in Fi
Male Sprague-Dawley Rats Following Exposure to PFOA (Pooled) {NTP,
2020, 7330145} 11-364
Table E-79. Dose-Response Modeling Data for Hepatocellular Adenomas in Fi Male
Sprague-Dawley Rats Following Exposure to PFOA {NTP, 2020, 7330145} E-365
Table E-80. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenomas
in Fi Male Sprague-Dawley Rats Following Postweaning Exposure to PFOA
{NTP, 2020, 7330145} 11-366
Table E-81. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenomas
in Fi Male Sprague-Dawley Rats Following Perinatal and Postweaning
Exposure to PFOA {NTP, 2020, 7330145} 11-367
Table E-82. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenomas
in Fi Male Sprague-Dawley Rats Following Exposure to PFOA (Pooled) {NTP,
2020, 7330145} 11-369
xxvi
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-83. Dose-Response Modeling Data for Hepatocellular Adenoma or Carcinoma in
Fi Male Sprague-Dawley Rats Following Exposure to PFOA {NTP, 2020,
7330145} E-370
Table E-84. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenoma
or Carcinoma in Fi Male Sprague-Dawley Rats Following Postweaning
Exposure to PFOA {NTP, 2020, 7330145} E-371
Table E-85. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenoma
or Carcinoma in Fi Male Sprague-Dawley Rats Following Perinatal and
Postweaning Exposure to PFOA {NTP, 2020, 7330145} E-372
Table E-86. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenoma
or Carcinoma in Fi Male Sprague-Dawley Rats Following Exposure to PFOA
(Pooled) {NTP, 2020, 7330145} E-373
Table E-87. Dose-Response Modeling Data for Pancreatic Acinar Cell Adenoma in Fi
Male Sprague-Dawley Rats Following Exposure to PFOA {NTP, 2020,
7330145} E-374
Table E-88. Summary of Benchmark Dose Modeling Results for Pancreatic Acinar Cell
Adenoma in Fi Male Sprague-Dawley Rats Following Postweaning Exposure
to PFOA {NTP, 2020, 7330145} 11-375
Table E-89. Summary of Benchmark Dose Modeling Results for Pancreatic Acinar Cell
Adenoma in Fi Male Sprague-Dawley Rats Following Perinatal and
Postweaning Exposure to PFOA {NTP, 2020, 7330145} E-376
Table E-90. Summary of Benchmark Dose Modeling Results for Pancreatic Acinar Cell
Adenoma in Fi Male Sprague-Dawley Rats Following Exposure to PFOA
(Pooled) {NTP, 2020, 7330145} 11-378
Table E-91. Dose-Response Modeling Data for Offspring Survival in Fi Male and Female
Kunming Mice Following Exposure to PFOA {Song, 2018, 5079725} E-378
Table E-92. Summary of Benchmark Dose Modeling Results for Offspring Survival using
Cavg,puP,gest in Fi Male and Female Kunming Mice Following Exposure to PFOA
(constant variance) {Song, 2018, 5079725} E-380
Table E-93. Summary of Benchmark Dose Modeling Results for Offspring Survival using
Cavg,pup,iact in Fi Male and Female Kunming Mice Following Exposure to PFOA
(constant variance) {Song, 2018, 5079725} E-382
Table E-94. Summary of Benchmark Dose Modeling Results for Offspring Survival using
Cavg,puP,gest,iact in Fi Male and Female Kunming Mice Following Exposure to
PFOA (constant variance) {Song, 2018, 5079725} E-384
Table E-95. Summary of Benchmark Dose Modeling Results for Offspring Survival using
Cmax,puP,gest, in Fi Male and Female Kunming Mice Following Exposure to
PFOA (constant variance) {Song, 2018, 5079725} E-386
xxvii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-96. Summary of Benchmark Dose Modeling Results for Offspring Survival using
Cmax,pup,iact, in Fi Male and Female Kunming Mice Following Exposure to
PFOA (constant variance) {Song, 2018, 5079725} E-388
Table E-97. Dose-Response Modeling Data for Pup Body Weight Change in Fi Male and
Female CD-I Mice Following Exposure to PFOA {Wolf, 2007, 1332672} E-389
Table E-98. Summary of Benchmark Dose Modeling Results for Pup Body Weight Change
in Fi Male and Female CD-I Mice Following Exposure to PFOA (constant
variance) {Wolf, 2007, 1332672} E-390
Table E-99. Dose-Response Modeling Data for Time to Eye Opening Fi Male and Female
CD-I Mice Following Exposure to PFOA {Wolf, 2007, 1332672} E-391
Table E-100. Summary of Benchmark Dose Modeling Results for Time to Eye Opening for
Cavg,puP,gest in Fi Male and Female CD-I Mice Following Exposure to PFOA
(constant variance) {Wolf, 2007, 1332672} E-391
Table E-101. Summary of Benchmark Dose Modeling Results for Time to Eye Opening for
Cavg,pup,iact in Fi Male and Female CD-I Mice Following Exposure to PFOA
(constant variance) {Wolf, 2007, 1332672} E-392
Table E-102. Summary of Benchmark Dose Modeling Results for Time to Eye Opening for
Cavg,puP,gest,iact in Fi Male and Female CD-I Mice Following Exposure to PFOA
(constant variance) {Wolf, 2007, 1332672} E-392
Table E-103. Summary of Benchmark Dose Modeling Results for Time to Eye Opening for
Cmax,puP,gest in Fi Male and Female CD-I Mice Following Exposure to PFOA
(constant variance) {Wolf, 2007, 1332672} E-393
Table E-104. Summary of Benchmark Dose Modeling Results for Time to Eye Opening
(Cmax,puP,iact) in Fi Male and Female CD-I Mice Following Exposure to PFOA
(constant variance) {Wolf, 2007, 1332672} E-393
Table E-105. Dose-Response Modeling Data for Dams with Whole Litter Loss in Po
Female CD-I Mice Following Exposure to PFOA {Wolf, 2007, 1332672} E-394
Table E-106. Summary of Benchmark Dose Modeling Results for Dams with Whole Litter
Loss in Po Female CD-I Mice Following Exposure to PFOA {Wolf, 2007,
1332672} E-395
Table G-l. Summary of EPA national freshwater fish tissue monitoring results for PFOA G-7
xxviii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Acronyms and Abbreviations
17-OHP
17-hydroxyprogesterone
BDI
ABC
ATP-binding cassette
transporter
BDI-II
aBMD
areal bone mineral
density
BMC
ACD
anterior chamber depth
BMD
ACE
America's Children and
the Environment
BMDL
ACTH
adrenocorticotropic
hormone
BMDLo.5si
ADHD
attention deficit
hyperactivity disorder
ADME
absorption, distribution,
metabolism, and
excretion
AGD
anogenital distance
BMDLisd
AIC
Akaike information
criterion
AMH
anti-Mullerian hormone
ANOVA
analysis of variance
APFO
ammonium
perfluorooctanoate
BMDL4
apoB
apolipoprotein B
aPPT
activated partial
thromboplastin time
ASD
autism spectrum disorder
ASQ
Ages and Stages
Questionnaire
BMDLs
AT SDR
Agency for Toxic
Substances and Disease
Registry
BMDL10
AUC
area under the curve
AUMC
area under the first
moment curve
P
regression coefficients
BMDS
BBB
blood-brain barrier
BCRP
BD
breast cancer resistance
protein
bolus dose
BMI
BMR
BSID-II
Beck Depression
Inventory
Beck Depression
Inventory-II
bone mineral content
benchmark dose
lower limit of benchmark
dose
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
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
lower bound on the dose
level corresponding to the
95% lower confidence
limit for a 4% change in
the response
lower bound on the dose
level corresponding to the
95% lower confidence
limit for a 5% response
level
lower bound on the dose
level corresponding to the
95% lower confidence
limit for a 10% change
Benchmark Dose
Software
body mass index
benchmark response
Bayley Scales of Infant
Development
xxix
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
BUN
blood urea nitrogen
BW
body weight
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,total
area under the curve in
gestation/lactation added
to the area under the
curve from diet (post-
weaning) divided by two
years
C7,avg
average concentration
over final week of study
CalEPA
California Environmental
Protection Agency
CAR
constitutive androstane
receptor
C-F
carbon-fluorine
CH
congenital
hypothyroidism
CHARGE
Childhood Autism Risk
from Genetics and
Environment
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
CL
post-dosing clearance
CLr
renal clearance
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
COPD
chronic obstructive
pulmonary disease
CSF
cancer slope factor
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
E
embryonic day
EFSA
European Food Safety
Authority
eGFR
estimated glomerular
filtration rate
eNT
equilibrative nucleoside
transporter
EPA
U.S. Environmental
Protection Agency
ES3
estrone-3-sulfate
Fi
first generation
f2
second generation
FDA
U.S. Food and Drug
Administration
FEV1
forced expiratory volume
in one second
FR
folate receptor
XXX
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
FSH
follicle stimulating
hormone
IC50
median inhibiting
concentration
FT3
free triiodothyronine
ID
intellectual disability
FTI
free thyroxine index
INUENDO
Biopersistent
FTOH
fluorotelomer alcohols
Organochlorines in Diet
FVC
forced vital capacity
and Human Fertility
FXR
farnesoid X receptor
GD
gestation day
i.p.
intraperitoneal
GM
geometric mean
IQ
intelligence quotient
GSD
geometric standard
IQR
interquartile range
deviation
IRIS
Integrated Risk
Hb
hemoglobin
Information System
HDL
high-density-lipoprotein
IUFD
intrauterine fetal death
HED
human equivalent dose
IV
intravenous
HEK293
human embryonic kidney
IVD
in vitro digestion method
cells
Kd
disassociation constant
HERO
Health and
Environmental Research
Kmem/w
membrane/water partition
coefficients
Online
Koc
organic carbon-water
HESD
health effects support
partitioning coefficient
document
LD
lactation day
HFD
high fat diets
LDL
low-density lipoprotein
HHRA
human health risk
assessment
L-FABP
liver fatty acid binding
protein
HOMA-B
Homeostatic Model
LFD
low fat diets
Assessment of Beta-Cell
LH
luteinizing hormone
Function
LIFE
Longitudinal
HOMA-IR
Homeostatic Model
Assessment for Insulin
Resistance
Investigation of Fertility
and the Environment
Study
HOME
Health Outcome
Measures of the
LOAEL
lowest-observed-adverse-
effect level
Environment
LOD
limit of detection
HPA
hypothalamic-pituitary-
adrenal
LOQ
limit of quantification
MCDI
MacArthur
HPLC/MS
high-performance liquid
chromatography mass
spectrometry
Communicative
Development Inventories
for Infants
HUMIS
Norwegian Human Milk
Study
MCLG
Maximum Contaminant
Level Goal
IBD
inflammatory bowel
disease
MDH
Minnesota Department of
Health
xxxi
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
MDI
Mental Development
OATs
Index
OATPs
MDR1
p-glycoprotein
MeSH
medical subject headings
OCC
Mg/kg-day
milligrams per kilogram
per day
OCISS
MLR
mixed linear regression
OECD
MOA
mode of action
MoBA
Norwegian Mother,
Father, and Child Cohort
OR
Study
ORD
M/P
milk/plasma
MRL
minimum reporting level
Po
mRNA
messenger ribonucleic
acid
PBET
MRP
multi-drug resistance-
associated protein
PBPK
MP AH
2-(N-methyl-PFOSA)
PCBs
acetate
PECO
MS
multiple sclerosis
NCI
National Cancer Institute
NEPSY-II
neuropsychological tests
PEF
NHANES
National Health and
PFAS
Examination Survey
PFBA
NICHD
U.S. National Institute of
Child Health and Human
PFBS
Development
PFCA
NJDEP
New Jersey Department
PFDA
of Environmental
PFDoDA
Protection
PFHpA
NMR
nuclear magnetic
PFHxA
resonance
PFHxS
NOAEL
no-ob served-adverse-
PFOA
effect level
PFOS
NOAEC
no observed adverse
effect concentration
PFSA
NPDWR
national primary drinking
water regulation
P
NTCP
sodium-taurocholate
-T ion
cotransporting
polypeptide
PFUnDA
PK
NTP
National Toxicology
Program
organic anion transporters
organic anion
transporting polypeptides
Odense Child Cohort
Ohio Cancer Incidence
Surveillance System
Organisation for
Economic Co-operation
and Development
Odds Ratio
Office of Research and
Development
parental generation
physiologically based
extraction test
physiologically-based
pharmacokinetic
poly chlorinated biphenyls
Populations, Exposures,
Comparator, and
Outcomes
peak expiratory flow rate
per- and polyfluoroalkyl
substances
perfluorobutanoic acid
perfluorobutane sulfonate
perfluorocarboxylates
perfluorodecanoic acid
perfluorododecanoic acid
perfluoroheptanoic acid
perfluorohexanoic acid
perfluorohexane sulfonate
perfluorooctanoic acid
perfluorooctane sulfonic
acid
perfluoroalkanesulfonic
acid
passive anionic
permeability
perfluoroundecanoic acid
pharmacokinetic
xxxii
-------
DRAFT FOR PUBLIC COMMENT
PLCO
Prostate, Lung,
Colorectal, and Ovarian
Screening Trial
PND
postnatal day
PNW
postnatal week
POD
point-of-departure
PODhed
point-of-departure human
equivalent dose
POPUP
Persistent Organic
Pollutants in Uppsala
Primiparas
PPARa
proliferator-activated
receptor alpha
PXR
pregnane X receptor
Qi
quantile 1
Q2
quantile 2
Q3
quantile 3
Q4
quantile 4
QA
quality assurance
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
SAB
Science Advisory Board
SE
standard errors
SERT
serotonin transporter
SES
socioeconomic status
SD
standard deviation
SDQ
Strengths and Difficulties
Questionnaire
SDWA
Safe Drinking Water Act
MARCH 2023
SRBC
serum sheep red blood
cells
SMBCS
Shanghai Minhang Birth
Cohort Study
SWAN
Study of Women's Health
Across the Nation
T3
triiodothyronine
T4
thyroxine
TA
thyroid antibody
TC
total cholesterol
TDS
Total Diet Study
TgAB
thyroblobulin antibodies
TiAb
title-abstract
Tmax
maximum plasma
concentration
TPoAb
thyroid peroxidase
antibody
TRR
total reactive residues
TSH
thyroid stimulating
hormone
TTE
transplacental transfer
efficiencies
TTR
transthyretin
UBM
unified BARGE method
UCMR3
third Unregulated
Contaminant Monitoring
Rule
UF
uncertainty factor
Vi
volume of central
distribution
v2
volume of peripheral
distibution
Vd
volume of distribution
Vdss
volume of distribution at
steady state
VI
visual impairment
VLDL
very low-density
lipoproteins
VMWM
Virtual Morris Water
Maze
WBHGB
whole blood hemoglobin
WCST
Wisconsin Card Sorting
Test
xxxiii
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
WHO
WIAT-II
WVCR
World Health
Organization
Wechsler Individual
Achievement Test-II
West Virginia Cancer
Registry
xxxiv
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix A. Systematic Review Protocol for
Updated PFOA 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 PFOA according to the scope and methods outlined in this protocol builds upon
several other assessments, including the Health Effects Support Document for Perfluorooctanoic
Acid (PFOA) {U.S. EPA, 2016, 3603279} (hereafter referred to as the 2016 PFOA HESD) and
Proposed Approaches to the Derivation of a Draft Maximum Contaminant Level Goal for
Perfluorooctanoic Acid (PFOA) (CASRN 335-67-1) in Drinking Water {U.S. EPA, 2021,
10428559}, 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 reviews and dose-
response analyses for the assessment of PFOA (Draft Toxicity Assessment and Proposed
Maximum Contaminant Level Goal (MCLG) for PFOA) and has been updated in response to
comments from the SAB. It should be noted that PFOA and PFOS underwent some steps of
systematic review (e.g., literature searches) concurrently.
A-l
-------
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 PFOA 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}. 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 this 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.1.1 Summary of Chemical identity and Occurrence
information
This section summarizes more detailed sections on these topics found in Proposed Maximum
Contaminant Level Goal (MCLG) for PFOA (hereafter referred to as the PFOA MCLG main
document) and is provided for context. Please refer to the PFOA MCLG main document for
more detailed information about chemical identity, physical-chemical properties, and occurrence.
A.l.1.1 Chemical Identity
The systematic review described by this protocol applies to all isomers of PFOA, as well as
nonmetal salts of PFOA that would be expected to dissociate in aqueous solutions of pH ranging
from 4 to 9 (e.g., in the human body). PFOA is a perfluorinated aliphatic carboxylic acid. It is a
strong acid that is generally present in solution as the perfluorooctanoate anion. PFOA is water
soluble and mobile in water, with an estimated log organic carbon-water partitioning coefficient
(log Koc) of 2.06. PFOA 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. PFOA has low volatility in ionized form but can adsorb to particles and be
A-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
deposited on the ground and into water bodies. It can be transported long distances in air or
water, as evidenced by detections of PFOA in arctic media and biota including polar bears,
ocean-going birds, and fish found in remote areas {Lindstrom, 2011, 1290802; Smithwick, 2006,
1424802}.
A.1.1.2 Occurrence Summary
Key PFOA occurrence information is summarized below. More detail is provided in Chapter 1 of
the PFOA MCLG main document.
A. 1.1.2.1 Biomonitoring
The U.S. Centers for Disease Control and Prevention (CDC) National Health and Nutrition
Examination Survey (NHANES) has measured blood serum concentrations of several PFAS in
the general U.S. population since 1999. PFOA has been detected in up to 98% of analyzed serum
samples representative of the U.S. general population; however, blood levels of PFOA dropped
60% to 80%) between 1999 and 2014, presumably due to reductions in its commercial usage in
the United States.
A.l.1.2.2 Occurrence in Water
PFOA is one of the dominant PFAS detected in ambient water, along with PFOS {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 nationally representative finished water occurrence
information for PFOA {U.S. EPA, 2017, 9419085; U.S. EPA, 2021, 7487276; U.S. EPA, 2023,
10692764}. UCMR 3 analyzed 36,972 samples from 4,920 PWSs for PFOA. The minimum
reporting level (MRL)1 for PFOA was 0.02 |ig/L. A total of 379 samples from 117 PWSs had
detections of PFOA (i.e., greater than or equal to the MRL). PFOA concentrations for these
detections ranged from 0.02 |ig/L (the MRL) to 0.349 |ig/L (median concentration of 0.03 |ig/L;
90th percentile concentration of 0.07 |ig/L).
A. 1.2 Problem Formulation
EPA performed this updated assessment for PFOA (including all isomers as well as nonmetal
salts of PFOA that would be expected to dissociate in aqueous solutions of pH ranging from 4 to
9 (e.g., in the human body)) to support development of an MCLG and national primary drinking
water regulation (NPDWR) for PFOA (see Chapter 1 of the PFOA main document for more
information). 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 PFOA
(2016 PFOA HESD and 2021 Proposed Approaches to the Derivation of a Draft Maximum
Contaminant Level Goal for Perfluorooctanoic Acid (PFOA) (CASRN 335-67-1) in Drinking
Water) as well ongoing EPA assessments of other PFAS (e.g., perfluorobutanoic acid (PFBA)
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-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
and draft perfluorohexanoic acid (PFHxA), perfluorohexane sulfonate (PFHxS),
perfluorononanoic acid (PFNA), and perfluorodecanoic acid (PFDA) IRIS assessments).
The 2016 PFOA HESD identified several adverse health outcomes associated with PFOA
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 PF OA. EPA's 2021 draft Proposed Approaches to the Derivation of a Draft Maximum
Contaminant Level Goal for Perfluorooctanoic Acid (PFOA) (CASRN 335-67-1) in Drinking
Water {U.S. EPA, 2021, 10428559} evaluated PFOA 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 PFOA.
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
Acid Forms) IRIS Assessments outlines key science issues relevant to PFAS in general {U.S.
EPA, 2020, 8642427}, many of which are relevant to PFOA. 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 PFOA toxicokinetics across species and sexes were
accounted for in the PFOA-specific animal and human pharmacokinetic models (see PFOA
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 PFOA 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 PFOA-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 PFOA, this is less of an issue as there
has been substantial research and publication of both epidemiological and animal toxicological
studies.
A-4
-------
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 PFOA 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 PFOA 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 Perfluorooctanoic Acid (PFOA) (CASRN 335-67-
1) in Drinking Water {U.S. EPA, 2021, 10428559}, the 2016 PFOA HESD {U.S. EPA, 2016,
3603279}, and the 2016 PFOA Drinking Water Health Advisory {U.S. EPA, 2016, 3982042}.
A.1.3.2 Specific Aims
The specific aims of the PFOA MCLG main document, which support the overall objective of
deriving an MCLG for PFOA, are as follows:
• 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
PFOA 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 PFOA (and its associated salts and isomers) as outlined in the PECO criteria.
• Evaluate and document the available mechanistic information (including toxicokinetic
understanding) associated with PFOA 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 on 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—but 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.
A-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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.
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 PFOA 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 PFOA and
describe the rationale.
Determine PFOA's cancer classification using a weight-of-evidence approach.
Characterize the effects associated with PFOA 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 PFOA 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 PFOA 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.
2 Notes: Although this appendix and its accompanying main document pertain to PFOA, the PECO criteria also cover PFOS
because the literature searching and screening were performed concurrently for PFOA and PFOS.
A-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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, in utero, lactation, peripubertal, and adult stages).
In vitro!cell studies or in .v/V/Wmodcling 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 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.
A-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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, orbiomolecules.
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).
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 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.
Outcome 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 vs. intestine, nasal vs. 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).
A-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Note: ADME = absorption, distribution, metabolism, and/or excretion; CAS = Chemical Abstracts Service;
PK = pharmacokinetic.
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.
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: CAS = Chemical Abstracts Service; MOA = mode of action.
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
A-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 PFOA HESD into
the literature inventory. The literature search strategy included searches within core literature
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
(PFOA 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
A-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Database Search String Date Run
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=(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 "Heptadecafhioro-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
A-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Database Search String Date Run
"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]
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-l-octanoic
acid"+"Pentadecafluoro-n-octanoic acid"+"Octanoic acid, pentadecafluoro-
"+"Perfluorocaprylic acid"+"Pentadecafluorooctanoic
acid"+"perfluoroheptanecarboxylic acid"+@TERM+@rn+3 3 5 -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-1-sulphonic acid"[tw] OR "Perfluorooctane
A-12
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Database Search String Date Run
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]
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="1-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
Web of
Science
Batch IDs:
39681, 46144
A-13
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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/;
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 Strateeies.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.gov/data/index.html). In 2020, the NTP website was searched for
PFOA toxicity study final reports that could provide relevant health effects information.
A-14
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• 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:
• Manual review of the reference lists of studies identified as PECO-relevant after full-text
review were reviewed at the title level for potential 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 PFOA Health Effects
Support Document
The 2016 HESD for PFOA contains 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 PFOA. 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 PFOA HESD that addressed one or more of the five main
health outcomes into this updated PFOA assessment, as described below.
Over 140 epidemiological studies were captured in the 2016 PFOA HESD. The 2016 HESD did
not use the epidemiological data quantitatively. For the current 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 62 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 the
2016 PFOA Health Effects Support Document
HERO ID
Reference
Title
Cancer
2850946
Barry et al., 2013
Perfluorooctanoic acid (PFOA) exposures and incident cancers among
adults living near a chemical plant
2851186
Boncfcld-Jorgcnscn
et al., 2014
Breast cancer risk after exposure to perfluorinated compounds in
Danish women: a case-control study nested in the Danish National
Birth Cohort
2150988
Boncfcld-Jorgcnscn
et al., 2011
Perfluorinated compounds are related to breast cancer risk in
Greenlandic Inuit: a case control study
2919344
Eriksen et al., 2009
Perfluorooctanoate and perfluorooctanesulfonate plasma levels and
risk of cancer in the general Danish population
2968084
Hardell et al., 2014
Case-control study on perfluorinated alkyl acids (PFAAs) and the risk
of prostate cancer
A-15
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
2850270
Raleigh et al., 2014
Mortality and cancer incidence in ammonium perfluorooctanoate
production workers
2851015
Steenland et al.,
2015
A cohort incidence study of workers exposed to perfluorooctanoic
acid (PFOA)
2919168
Steenland and
Woskie, 2012
Cohort mortality study of workers exposed to perfluorooctanoic acid
2919154
Vieira et al., 2013
Perfluorooctanoic acid exposure and cancer outcomes in a
contaminated community: a geographic analysis
Cardiovascular
1429922
Costa et al., 2009
Thirty years of medical surveillance in perfluorooctanoic acid
production workers
1290905
Emmett et al., 2006
Community exposure to perfluorooctanoate: Relationships between
serum levels and certain health parameters
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.,
Reductions in serum lipids with a 4-year decline in serum
2013
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
3981585
Maisonet et al., 2015
Prenatal exposures to perfluoroalkyl acids and serum lipids at ages 7
and 15 in females
1291110
Nelson et al., 2010
Exposure to polyfluoroalkyl chemicals and cholesterol, body weight,
and insulin resistance in the general US population
1290836
Olsen and Zobel,
2007
Assessment of lipid, hepatic, and thyroid parameters with serum
perfluorooctanoate (PFOA) concentrations in fluorochemical
production workers
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.
A-16
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
1424954
Olsen et al., 2000
Plasma cholecystokinin and hepatic enzymes, cholesterol and
lipoproteins in ammonium perfluorooctanoate production workers
2850270
Raleigh et al., 2014
Mortality and cancer incidence in ammonium perfluorooctanoate
production workers
1291103
Sakr et al., 2007
Cross-sectional study of lipids and liver enzymes related to a serum
biomarker of exposure (ammonium perfluorooctanoate or APFO) as
part of a general health survey in a cohort of occupationally exposed
workers
1430761
Sakr et al., 2007
Longitudinal study of serum lipids and liver enzymes in workers with
occupational exposure to ammonium perfluorooctanoate
1276141
Savitz et al., 2012
Perfluorooctanoic acid exposure and pregnancy outcome in a highly
exposed community
1424946
Savitz et al., 2012
Relationship of perfluorooctanoic acid exposure to pregnancy
outcome based on birth records in the mid-Ohio Valley
2850928
Starling et al., 2014
Perfluoroalkyl substances and lipid concentrations in plasma during
pregnancy among women in the Norwegian Mother and Child Cohort
Study
2851015
Steenland et al.,
2015
A cohort incidence study of workers exposed to perfluorooctanoic
acid (PFOA)
1291109
Steenland et al.,
2009
Association of perfluorooctanoic acid and perfluorooctane sulfonate
with serum lipids among adults living near a chemical plant
2919168
Steenland and
Woskie, 2012
Cohort mortality study of workers exposed to perfluorooctanoic acid
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
2851142
Winquist and
Steenland, 2014
Modeled PFOA exposure and coronary artery disease, hypertension,
and high cholesterol in community and worker cohorts
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 et al., 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
1290822
Fei et al., 2008
Prenatal exposure to perfluorooctanoate (PFOA) and
perfluorooctanesulfonate (PFOS) and maternally reported
developmental milestones in infancy
A-17
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
2349574
Fei et al., 2008
Fetal growth indicators and perfluorinated chemicals: a study in the
Danish National Birth Cohort
1005775
Fei et al., 2007
Perfluorinated chemicals and fetal growth: A study within 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
2349575
Monroy et al., 2008
Serum levels of perfluoroalkyl compounds in human maternal and
umbilical cord blood samples
1290813
Nolan et al., 2010
Congenital anomalies, labor/delivery complications, maternal risk
factors and their relationship with perfluorooctanoic acid (PFOA)-
contaminated public drinking water
2349576
Nolan et al., 2009
The relationship between birth weight, gestational age and
perfluorooctanoic acid (PFOA)-contaminated public drinking water
1276141
Savitz et al., 2012
Perfluorooctanoic acid exposure and pregnancy outcome in a highly
exposed community
1424946
Savitz et al., 2012
Relationship of perfluorooctanoic Acid exposure to pregnancy
outcome based on birth records in the mid-Ohio Valley
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
1429922
Costa et al., 2009
Thirty years of medical surveillance in perfluorooctanoic acid
production workers
1290905
Emmett et al., 2006
Community exposure to perfluorooctanoate: Relationships between
serum levels and certain health parameters
1276142
Gallo et al., 2012
Serum perfluorooctanoate (PFOA) and perfluorooctane sulfonate
(PFOS) concentrations and liver function biomarkers in a population
with elevated PFOA exposure
1291111
Lin et al., 2010
Investigation of the Associations Between Low-Dose Serum
Perfluorinated Chemicals and Liver Enzymes in US Adults
1290836
Olsen and Zobel,
2007
Assessment of lipid, hepatic, and thyroid parameters with serum
perfluorooctanoate (PFOA) concentrations in fluorochemical
production workers
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.
1424954
Olsen et al., 2000
Plasma cholecystokinin and hepatic enzymes, cholesterol and
lipoproteins in ammonium perfluorooctanoate production workers
A-18
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
1291103
Sakr et al., 2007
Cross-sectional study of lipids and liver enzymes related to a serum
biomarker of exposure (ammonium perfluorooctanoate or APFO) as
part of a general health survey in a cohort of occupationally exposed
workers
1430761
Sakr et al., 2007
Longitudinal study of serum lipids and liver enzymes in workers with
occupational exposure to ammonium perfluorooctanoate
2919168
Steenland and
Woskie, 2012
Cohort mortality study of workers exposed to perfluorooctanoic acid
2851015
Steenland et al.,
2015
A cohort incidence study of workers exposed to perfluorooctanoic
acid (PFOA)
Immune
1429922
Costa et al., 2009
Thirty years of medical surveillance in perfluorooctanoic acid
production workers
1937230
Dong et al., 2013
Serum polyfluoroalkyl concentrations, asthma outcomes, and
immunological markers in a case-control study of Taiwanese children
1290905
Emmett et al., 2006
Community exposure to perfluorooctanoate: Relationships between
serum levels and certain health parameters
1290805
Fei et al., 2010
Prenatal exposure to PFOA and PFOS and risk of hospitalization for
infectious diseases in early childhood
1248827
Grandiean 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 et al., 2014
Influenza vaccine response in adults exposed to perfluorooctanoate
and perfluorooctanesulfonate
1332477
Okada et al., 2012
Prenatal exposure to perfluorinated chemicals and relationship with
allergies and infectious diseases in infants
2851015
Steenland et al.,
2015
A cohort incidence study of workers exposed to perfluorooctanoic
acid (PFOA)
1424977
Wang et al., 2011
The effect of prenatal perfluorinated chemicals exposures on pediatric
atopy
Notes: APFO = ammonium perfluorooctanoate; 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 11 animal toxicological studies were
included and are 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 Animal Toxicological Studies Identified from the 2016 PFOA Health
Effects Support Document
HERO ID
Reference
Title
Cancer
A-19
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
673581
Biegel et al., 2001
Mechanisms of extrahepatic tumor induction by peroxisome proliferators in
male CD rats
2919192
Butenhoff et al., 2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
Cardiovascular
2919192
Butenhoff et al., 2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
988599
Loveless et al., 2008
Evaluation of the immune system in rats and mice administered linear
ammonium perfluorooctanoate
Developmental
1335452
Abbott et al., 2007
Perfluorooctanoic acid induced developmental toxicity in the mouse is
dependent on expression of peroxisome proliferator activated
receptor-alpha
1291063
Butenhoff et al.,
2004
The reproductive toxicology of ammonium perfluorooctanoate
(APFO) in the rat
1276159
Lau et al., 2006
Effects of perfluorooctanoic acid exposure during pregnancy in the
mouse
988599
Loveless et al.,
2008
Evaluation of the immune system in rats and mice administered linear
ammonium perfluorooctanoate
1276151
Macon etal., 2011
Prenatal perfluorooctanoic acid exposure in CD-I mice: Low-dose
developmental effects and internal dosimetry
1332672
Wolf et al., 2007
Developmental toxicity of perfluorooctanoic acid in the CD-I mouse
after cross-foster and restricted gestational exposures
Endocrine
2919192
Butenhoff et al.,
2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
1291063
Butenhoff et al.,
2004
The reproductive toxicology of ammonium perfluorooctanoate
(APFO) in the rat
988599
Loveless et al.,
2008
Evaluation of the immune system in rats and mice administered linear
ammonium perfluorooctanoate
Gastrointestinal
2919192
Butenhoff et al.,
2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
Hepatic
1335452
Abbott et al., 2007
Perfluorooctanoic acid induced developmental toxicity in the mouse is
dependent on expression of peroxisome proliferator activated
receptor-alpha
673581
Biegel et al., 2001
Mechanisms of extrahepatic tumor induction by peroxisome
proliferators in male CD rats
2919192
Butenhoff et al.,
2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
A-20
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
1291063
Butenhoff et al.,
2004
The reproductive toxicology of ammonium perfluorooctanoate
(APFO) in the rat
1276159
Lau et al., 2006
Effects of perfluorooctanoic acid exposure during pregnancy in the
mouse
988599
Loveless et al.,
2008
Evaluation of the immune system in rats and mice administered linear
ammonium perfluorooctanoate
1276151
Macon etal., 2011
Prenatal perfluorooctanoic acid exposure in CD-I mice: low-dose
developmental effects and internal dosimetry
1291118
Perkins et al., 2004
13-week dietary toxicity study of ammonium perfluorooctanoate
(APFO) in male rats
1332672
Wolf etal., 2007
Developmental toxicity of perfluorooctanoic acid in the CD-I mouse
after cross-foster and restricted gestational exposures
Immune
2919192
Butenhoff et al.,
2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
1291063
Butenhoff et al.,
2004
The reproductive toxicology of ammonium perfluorooctanoate
(APFO) in the rat
1290826
Dewitt et al., 2008
Perfluorooctanoic acid-induced immunomodulation in adult
C57BL/6J or C57BL/6N female mice
988599
Loveless et al.,
2008
Evaluation of the immune system in rats and mice administered linear
ammonium perfluorooctanoate
Metabolic
1335452
Abbott et al., 2007
Perfluorooctanoic acid induced developmental toxicity in the mouse is
dependent on expression of peroxisome proliferator activated
receptor-alpha
2919192
Butenhoff et al.,
2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
Nervous
2919192
Butenhoff et al.,
2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
1291063
Butenhoff et al.,
2004
The reproductive toxicology of ammonium perfluorooctanoate
(APFO) in the rat
1276151
Macon etal., 2011
Prenatal perfluorooctanoic acid exposure in CD-I mice: low-dose
developmental effects and internal dosimetry
Renal
2919192
Butenhoff et al.,
2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
1291063
Butenhoff et al.,
2004
The reproductive toxicology of ammonium perfluorooctanoate
(APFO) in the rat
Reproductive
1335452
Abbott et al., 2007
Perfluorooctanoic acid induced developmental toxicity in the mouse is
dependent on expression of peroxisome proliferator activated
receptor-alpha
A-21
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
HERO ID
Reference
Title
673581
Biegel et al., 2001
Mechanisms of extrahepatic tumor induction by peroxisome
proliferators in male CD rats
2919192
Butenhoff et al.,
2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
1291063
Butenhoff et al.,
2004
The reproductive toxicology of ammonium perfluorooctanoate
(APFO) in the rat
1291118
Perkins et al., 2004
13-week dietary toxicity study of ammonium perfluorooctanoate
(APFO) in male rats
Respiratory
2919192
Butenhoff et al.,
2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats'
1291118
Perkins et al., 2004
13-week dietary toxicity study of ammonium perfluorooctanoate
(APFO) in male rats
Systemic
1335452
Abbott et al., 2007
Perfluorooctanoic acid induced developmental toxicity in the mouse is
dependent on expression of peroxisome proliferator activated
receptor-alpha
2919192
Butenhoff et al.,
2012
Chronic dietary toxicity and carcinogenicity study with ammonium
perfluorooctanoate in Sprague-Dawley rats
1291063
Butenhoff et al.,
2004
The reproductive toxicology of ammonium perfluorooctanoate
(APFO) in the rat
1290826
Dewitt et al., 2008
Perfluorooctanoic acid-induced immunomodulation in adult
C57BL/6J or C57BL/6N female mice
1276159
Lau et al., 2006
Effects of perfluorooctanoic acid exposure during pregnancy in the
mouse
1291118
Perkins et al., 2004
13-week dietary toxicity study of ammonium perfluorooctanoate
(APFO) in male rats
1332672
Wolf et al., 2007
Developmental toxicity of perfluorooctanoic acid in the CD-I mouse
after cross-foster and restricted gestational exposures
3981487
Yu et al., 2016
Effects of perfluorooctanoic acid on metabolic profiles in brain and
liver of mouse revealed by a high-throughput targeted metabolomics
approach
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-22
-------
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, in utero, lactation, peripubertal, and adult stages).
In vitro!cell studies or in .v/V/Wmodcling 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.
Note: 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 PFOA 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 contained
(e.g., mechanistic, ADME). Studies that met the PECO criteria were tagged as having relevant
A-23
-------
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, except at the full-text level (described further in
Section A. 1.6.5).
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 material 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 PFOA 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
title/abstract and/or full-text screening underwent further screening against the mechanistic-
specific PECO criteria outlined in Table A-3. Studies that met the mechanistic-specific PECO
A-24
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
criteria were extracted 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 PFOA 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?
• Yes
[Select one]
• Noa
• Tag as potentially relevant supplemental material
• Unclear
If "Yes" to Question #1:
2a What type of evidence?
• Human
[Select all that apply]
• Animal (mammalian models)
• PBPK model
If "Tag as potentially relevant supplemental material" to Question #1:
2b What kind of supplemental
• Mechanistic0
material?b
• Non-mammalian model
[Select all that apply]
• 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-25
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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.
Table A-10. SWIFT-Active Form for Title/Abstract Screening
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
If "Yes" to Question #1:
2a Identify the Type of Evidence
[Select all that apply]
• Human/Epidemiological
• Animal
• Unsure
If "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
Note:
a Erratums and corrections were considered not relevant.
b Refer to the list of supplemental tags in Section A. 1.6.4.2.
A.l.6.4.1 Supplemental Tags
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 ADME/TK
or mechanistic 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
A-26
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Category
Evidence
ADME and Toxicokinetic
Acute/Short-Term Duration
Exposures
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.
Animal studies of less than 28 days (unless the study is a
developmental/reproductive study)
Only One Exposure Group Animal studies with only one exposure group, e.g., control and 1 mg/kg/day PFOA.
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 Exposure characteristic studies include data that are unrelated to toxicological
Health Outcome) endpoints, but which provide information on exposure sources or measurement
properties of the environmental agent (e.g., demonstrate a biomarker of exposure).
Susceptible Populations
(No Health Outcome)
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 With No Original
Data (e.g., Reviews,
Editorials, Commentaries)
Secondary studies (e.g., reviews, editorials, commentaries, assessments) that do not
provide any primary research/results.
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.
Note: C. elegans = Caenorhabditis elegans.
A.l.6.4.2 Mechanistic Study Categories and Keywords
The following categories were considered mechanistic throughout the title/abstract and full-text
screening (Table A-12). 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
Chromosome or DNA structure, function,
repair, or integrity
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
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
A-27
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Category
Examples of Keywords
Metabolism
anabolic or catabolic pathways for lipids, carbohydrates, amino acids,
nucleotides; energy metabolism; biochemical pathways; metabolomics;
lipidomics; enzyme or coenzyme activity or function.
Cell signaling or signal transduction ligand interactions with membrane, cytoplasmic and nuclear receptors
pathway (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 or organelle structure, motility, membrane integrity, cell scaffolding, cytoskeleton, actin, microtubules,
integrity 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 growth, differentiation, proliferation, cell cycle (Gl, S, G2, M), cyclins, CDKs, p53, p27, Rb, E2F stem cell,
or viability progenitor, apoptosis, Annexin V, TUNEL, necrosis, blebbing,
pyknosis, Bax, Bcl-2, hyperplasia, dysplasia
Activation of intrinsic cell defense cytokines, chemokines, caspases, MHC/HLA molecules, pattern
molecules or systems recognition receptors (PRRs), NLR, proteasomes, autophagy
Oxidative stress 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
Hormone function
GnRH, CRF, ADH/vasopressin, FSH, LH, ACTH, GH, TH, TSH, PTH,
Cortisol, epinephrine/norepinephrine, melatonin, oxytocin, estrogen,
testosterone, adiponectin, leptin, insulin, glucagon
Biomarkers of cerebral function
Apoptotic neurodegeneration protein markers, cerebral glucose
metabolism, brain glucose levels
Other (provide details)
Please provide specific details regarding reason for supplemental tag in
the notes section.
Notes: 8-OHdG = 8-hydroxy-2'-deoxyguanosine; ACTH = adrenocorticotropic hormone; ADH = antidiuretic hormone;
AHR = aryl hydrocarbon receptor; Bcl-2 = B-cell lymphoma 2; CAR = constitutive androstane receptor; CDK = cyclin-
dependent kinase; CRF = corticotropin-releasing factor; DAG = diacylglycerol; DNA = deoxyribonucleic acid; ECM =
extracellular matrix; ER = estrogen receptor; EROD = ethoxyresorufin-O-dealkylase; FSH = follicle stimulating hormone;
GH = growth hormone; GTPase = guanosine triphosphate; GnRH = gonadotropin-releasing hormone; LH = luteinizing hormone;
MHC/NHLA = major histocompatibility complex/human leukocyte antigen; microRNA = micro ribonucleic acid;
mTOR = rapamycin; NF-RB = nuclear factor kappa B; NLR = nucleotide-binding oligomerization domain-like receptors; NO =
nitric oxide; PERK = protein kinase R-like endoplasmic reticulum kinase; PTE! = parathyroid hormone; RAR = retinoic acid
receptor; RNA = ribonucleic acid; TE[ = thyroid hormone; TGFp = transforming growth factor beta; TUNEL = terminal
deoxynucleotidyl transferase-mediated dUTP nick end labeling; UPR = unfolded protein response.
A-28
-------
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, health outcome(s), and whether PFOA and/or PFOS 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-29
-------
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-30
-------
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: ALT = alanine transaminase; AST = aspartate aminotransferase; PBPK = physiologically-based pharmacokinetic; PK = pharmacokinetic.
a Refer to list of health outcomes and examples in Appendix A. 1.6.5.1.
b Refer to list of supplemental tags in A. 1.6.4.1.
c Refer to list of mechanistic information in Appendix A. 1.6.4.2.
d Refer to definition of acute/short-term duration exposures in Appendix A. 1.6.6.
A-31
-------
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).
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;
• Markers of development specific to
small for gestational age)
other systems are organ/system-specific
• Preterm birth
(e.g., tests of sensory maturation are
• Sex ratio
under Nervous System).
• Postnatal growth
• 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
Reproductive.
blood (non-reproductive)
Gastrointestinal
• Symptoms of the stomach and
-
intestines (e.g., diarrhea, nausea,
vomiting, abdominal pain and cramps)
Hematologic
• Blood count
• White blood cell counts and globulin
• Red blood cells
are under Immune.
• Blood Hematocrit or hemoglobin
• Serum lipids are under
• Corpuscular volume
Cardiovascular.
• Blood Platelets or reticulocytes
• Serum liver markers are under Hepatic.
• Blood biochemical measures (e.g.,
sodium, calcium, phosphorus)
Hepatic
• Liver enzymes (e.g., ALT; AST; ALP)
• Serum lipids are under
• Liver disease
Cardiovascular.
• Liver-specific serum biochemistry
• Biochemical markers, such as albumin,
(e.g., albumin)
are under Hepatic. Liver tissue
cytokines are under Immune.
• Globulin is under Immune.
• Serum glucose is under Metabolic.
Immune
• Asthma
• Red blood cells are under
• Allergy
Hematological.
• Atopic dermatitis/eczema
• Non-immune measures of pulmonary
• Vaccine response
function are under Respiratory.
• IgE
• Interleukin 6 (IL-6) is considered a
• Autoimmune or infectious disease
Mechanistic outcome.
A-32
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Health Effect Category
Example Health Outcomes
Notes
• Hypersensitivity
• General immune assays (e.g., white
blood cell counts)
• 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
A-33
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Health Effect Category
Example Health Outcomes
Notes
• 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: ALP = alkaline phosphatase; ALT = alanine transaminase; AST = aspartate aminotransferase; FEV1 = forced expiratory
volume in one second; FVC = forced vital capacity; GFR = glomerular filtration rate; HDL = high-density lipoprotein;
LDL = low-density lipoprotein; PBPK = physiologically-based pharmacokinetic; PCOS = polycystic ovary syndrome;
PK = pharmacokinetic; T3 = triiodothyronine; T4 = thyroxine; TSH = thyroid stimulating hormone.
A-34
-------
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-35
-------
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) also 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, etc.). 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 Forms for ADME Screening and Light Data Extraction
Question/Prompt Response Options Suggested Considerations
1 General Questions
1.1 Does the article meet PECO • Yes
criteria? • No
[Select one]
1.2 What PFAS did the study report? • PFOA
[Select all that apply] • PFOS
1.3 Does this study contain multiple
time points, multiple tissues,
and/or multiple doses?
[Select one]
• Yes
• No
• Use ADME-specific PECO statement (see Main PFOA
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.
• 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-36
-------
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-37
-------
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 Mfero/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 Mfero/placental" AND by "lactational transfer",
select "in Mfero/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 Mfero/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 Mfero/placental" AND by "breast milk", select "in
Mfero/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-38
-------
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).
A-39
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt Response Options Suggested Considerations
[Free-text]
Animal Studies
IF the sludv docs not contain ;in animal sludv. skip this section and move on lo Section 4 - Mammalian Cells//;? Vim
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 "metabolism"
absorption, distribution,
• Distribution
is an unlikely selection.
metabolism, and/or excretion.
• Metabolism
[Select all that apply]
• Excretion
3 .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]
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 Mfero/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 Mfero/placental" AND by "lactational transfer",
select "in Mfero/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 Mfero/placental transfer
» Corn oil
»Filtered air
• 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 Mfero/placental" AND by "breast milk", select "in
Mfero/placental" and use the free text field to note that
lactational transfer also occurred via breast milk.
A-40
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
• Olive oil
• Ethanol
• 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: GD 0 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.
A-41
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
3.11 What is the life stage when the
ADME data was collected?
[Select all that apply; Free-text]
• Prenatal
• Weaning
• Adolescent
• Adult
• Elderly
• Prenatal
o Non-human primates: conception to birth
o Rodents: GD 0 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
• Use the free text field to provide additional life stage
notes.
• If there is more than one life stage when ADME data
was collected, add an additional population in another
form.
3 .12 What is the number of animals per
dosing group?
Use the free text field to report the
number of animals per dosing group.
[Free-text]
• Example: Control = 10, low dose = 20, high dose = 20;
All groups = 20.
• Use "Not Reported" if appropriate.
3 .13 Dose Levels
Use the free text field to enter the
numeric dose levels.
[Free-text]
• Example: 0, 450, 900.
3 .14 Dose Units
Use the free text field to report the
dosage units as presented in the
paper.
[Free-text]
• Examples: mg/kg-d; mg/m3; ppm
• Use "Not Reported" if appropriate.
3 .15 Dose Duration
Use the free text field to enter the
details of the dose duration if known.
• Use abbreviations (h, d, wk, mon, y).
A-42
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
[Free-text]
• For reproductive and developmental studies, where
possible instead include abbreviated age descriptions
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.
3 .16 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: 14 or 28 d; 13 wk; 2 y.
• Use "Not Reported" if appropriate.
3.17 Measured Tissues
Use the free text field to enter the
tissues measured in the study (e.g.,
plasma, liver, adipose).
[Free-text]
Mammalian Cells///? Vitro
IF (he study docs not contain an in vitro component, skip (his section and move on lo Section 5 - Notes.
4.1 Population Name
[Free-text]
• Name a population (e.g., Primary Human Hepatic,
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
protein binding? • No
[Select one; Free-text]
• If "Yes" option is selected, use the free text field to list
the binding proteins.
A-43
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
4.5 Does the study present data on
active transport?
[Select one; Free-text]
• Yes
• No
• If "Yes" option is selected, use the free text field to list
the transporters.
4.6 Cell Line Name or Tissue Source
Use the free text field to list the cell
line name or tissue source the cells
were derived from.
[Free-text]
• Examples: A549; liver tissue from adult Sprague
Dawley female rats.
• 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.
• Use abbreviations (h, d, wk, mon, y).
o Examples: 28 d; 13 wk; 2 y.
A-44
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
[Free-text]
• Use "Not Reported" if appropriate.
4.14 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.
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: GD = gestational day; MOA = mode of action; PBPK = physiologically-based pharmacokinetic; PND = postnatal day; ppm = parts per million; QA/QC = quality
assurance/quality control; TK = toxicokinetic.
A-45
-------
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 Forms for Mechanistic 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
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-46
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
• In Mfero/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 Mfero/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 do
outcome system?
• Cardiovascular
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-47
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response 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 the study docs nol 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
• Rabbit
• Guinea pig
• Free text field to list species for "other rodent model"
option.
A-48
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options Suggested Considerations
• Other rodent model
3.3
What is the strain?
- -
[Free-text]
3.4
Identify the Exposure Route
• Inhalation (nose only) • Free text field for "other" option.
[Select one]
• Inhalation (whole head exposure)
• Inhalation (whole body exposure)
• Oral (diet)
• Oral (drinking water)
• Oral (gavage)
• Dermal
• Lactational transfer
• In Mfero/placental transfer
• Other (e.g., intraperitoneal, intramuscular, intravenous,
intranasal)
3.5
What is the exposure vehicle?
• Diet • Free text field for other "other" option.
[Select one]
• Water
• Breast milk
• In Mfero/placental transfer
• Corn oil
• Filtered air
• Olive oil
• Ethanol
• DMSO
• Mineral oil
• Corn oil:acetone
• Other
3.6
What is the life stage when the
• Prenatal • Free text field for life stage notes.
animal was dosed?
• Weaning
[Select one; Free-text]
• Adolescent
• Adult
• Elderly
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]
• Adolescent
• Adult
A-49
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options Suggested Considerations
• 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 system.
• Dermal
• Developmental
• Endocrine
• Gastrointestinal
• Hematologic
• Hepatic
• Immune
• Lymphatic
• Metabolic
• Musculoskeletal/connective tissue
• Nervous
• Ocular
• Renal
• Reproductive
• Respiratory
• Systemic/whole body
• Other
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 remodeling • Free text field for "other" option.
[Select all that apply; Free-text]
• Atherogenesis and clot formation
• Big data, non-targeted analysis
• Cell growth, differentiation, proliferation, or viability
• Cell signaling or signal transduction
A-50
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt Response Options Suggested Considerations
• 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
• Rabbit
• Guinea pig
• Free text field to list "other rodent model" option.
A-51
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
• Other rodent model
4.7 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
system.
• Dermal
• Developmental
• Endocrine
• Gastrointestinal
• Hematologic
• Hepatic
• Immune
• Lymphatic
• Metabolic
• Musculoskeletal/connective tissue
• Nervous
• Ocular
• Renal
• Reproductive
• Respiratory
• Systemic/whole body
• Other
4.8 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
Angiogenic, antiangiogenic, vascular tissue remodeling • Free text field for "other" option.
Atherogenesis and clot formation
Big data, non-targeted analysis
Cell growth, differentiation, proliferation, or viability
Cell signaling or signal transduction
4.9 Mechanistic Pathway
[Select all that apply; Free-text]
A-52
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt Response Options Suggested Considerations
• 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: DMSO = dimethyl sulfoxide; DNA = deoxyribonucleic acid; QA/QA = 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-53
-------
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)
A
ident e
vo revi
T
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 Hkely to have had a notable impact on die result or prevent
reliable interpretation of study findings)
^ Critically Deficient (Serious flaws that make observed effects uninterpretable)
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 mpact on die study results or their interpretation)
L'ninforma trv e (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; (bJEvaluation 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,
adequate, deficient (or "not reported," which carried the same functional interpretation as
A-54
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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(s) 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-55
-------
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 limitation(s) 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-56
-------
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-57
-------
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-58
-------
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 • 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 • 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
sen'ices availability, without consideration of similar
selection forces affecting recruitment of controls.
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).
A-59
-------
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)?
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).
' Exposure misclassification is expected to be minimal.
Adequate • 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.
Deficient • 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-60
-------
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 of
exposure:
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?
What exposure time period is reflected by the
biomarker? If the half-life is short, what is the
correlation between serial measurements of
exposure?
Additional suggested considerations for biomarkers of exposure (should be
evaluated in addition to the general considerations above):
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 method.
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
studies of occupational exposures:
Is exposure based on a comprehensive job history
describing tasks, setting, time period, and use of
specific materials?
Additional suggested considerations for occupational exposures (should be
evaluated in addition to the general considerations above):
• 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.
• 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,
tasks performed, or materials used) - may be Critically
Deficient, depending on severity of this limitation.
A-61
-------
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?
Critically
Deficient
JEM with data indicating it cannot differentiate between
exposure levels over time, area, or between individuals.
Notes: JEM = job exposure matrix; LOD = limit of detection; QA/QC = quality assurance/quality control.
A-62
-------
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. Study Quality Evaluation Considerations for PFAS-Specific Exposure
Measurement
Rating
Criteria
Good • Evidence that exposure was consistently assessed using well-established analytical methods
that directly measure exposure (e.g., measurement of PFAS in blood, serum, or plasma).
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.
Adequate • 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.
Deficient 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.
Critically I Any of the following:
Deficient | • 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-63
-------
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-64
-------
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
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-65
-------
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
Deficient
Critically
Deficient
• 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.
• Outcome definition was not specific or sensitive.
• Uncertainty regarding validity of assessment instrument.
• Invalid/insensitive marker of outcome.
• 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.
A-66
-------
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-67
-------
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.
• If confounders not considered by design or in the analysis
(e.g.. only simple correlations presented).
A-68
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
ADHD = attention deficit hyperactivity disorder.
A-69
-------
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-70
-------
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 I • Results of analyses of effect modification examined without
Deficient 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-71
-------
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)?
Adequate
Deficient
Suggested Considerations
• 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.
• 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-72
-------
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
Suggested Considerations
Is the exposure range/contrast adequate to detect
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?
Adequate • 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.
Deficient • 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-73
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.7.1.9 Overall Confidence
Table A-25. Study Quality Evaluation Considerations for Overall Study Confidence - Epidemiological Studies
Provide judgment 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-74
-------
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-75
-------
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 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 for
on description of recruitment process (e.g., test species, strain, sex, age, exposure
Critical information necessary to ncrform
selection of comparison population, methods, experimental design, endpoint
study evaluation:
population-based random sample evaluations and the presentation of results.
• Species; test article name; levels and duration of
selection, recruitment from sampling
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 one
employees) such that study participants conducted in compliance with the OECD
endpoint of interest
were unlikely to differ from a larger cohort guidelines for Good Laboratory Practice
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 of
administration
study (e.g., initial enrollment, follow-up.
• Experimental design: frequency of exposure.
selection into analysis sample). If rate is
animal age and lifestage during exposure and at
not high, there is appropriate rationale for
endpoint/outcome evaluation
why it is unlikely to be related to exposure
• Endpoint evaluation methods: assays or
(e.g., comparison between participants and
procedures used to measure the
nonparticipants or other available
endpoints/outcomes of interest
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-76
-------
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.
Critically
Deficient
Adequate. All critical information is
reported but some important information is
missing. Specifically, it is unclear what
strain of rats was used.
Deficient • 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.
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.
' 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
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.
A-77
-------
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?
¦ 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).
Notes: For the Reporting Quality 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-78
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A. 1.7.2.2 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."
Adequate • 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-79
-------
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)." 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
• No indication of randomization of groups Not reported (interpreted as Deficient). The
(Interpreted as
or other methods (e.g., normalization) to authors did not indicate randomization or
Deficient)
control for important modifying factors other normalization procedures for balancing
across experimental groups. important variables across groups.
Critically
• Bias in the animal allocations was reported Critically Deficient. There is direct evidence
Deficient
or inferable. 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-80
-------
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 Exami)le 1: Good. Historatholoev:
endpoints/outcomes in a study:
were described (e.g., blinding to conceal Although the study did not indicate blinding.
treatment groups during endpoint blinding during the initial evaluation of
Does the study report blinding or other
evaluation; consensus-based evaluations of tissues for initial or non-targeted evaluations
methods/procedures for reducing observational
histopathology lesions3). is generally not recommended as masked
bias?
evaluation can make the task of separating
treatment-related changes from normal
If not, did the study use a design or approach for
variation more difficult and may result in
which such procedures can be inferred?
subtle lesions being overlooked {Crissman,
2004, 51763}. The study did include a
What is the expected impact of failure to
secondary evaluation by a pathology
implement (or report implementation) of these
working group (PWG) review on coded
methods/procedures on results?
pathology slides which minimized the
potential for observational bias.
Examiile 2: Good. Orsan weiehts. 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. Histopathology measures:
Authors report "lesions were counted by 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 non-targeted
Adequate • Methods for reducing observational bias
(e.g., blinding) can be inferred or were
reported but described incompletely.
A-81
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Did the study implement measures to reduce observational bias?
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
mitiratpd hnspd nn iisp nf measures: Authors did not indicate whether
automated/computer driven systems, investigators were blinded during outcome
standard laboratory kits, relatively simple, assessment. Potential concern for bias was
objective measures (e.g., body or tissue mitigated for these endpoints which were
weight), or screening-level evaluations of measured using automated/computer driven
histopathology. systems, standard laboratory kits, relatively
simple, objective measures (e.g., body or
tissue weight).
Example 2: Not reported (interpreted as
Adeauate). Histonatholoev: Blindins durine
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).
Example 3: Not reported (interpreted as
Adequate). Fetal evaluation for
A-82
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Did the study implement measures to reduce observational bias?
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
• Strong evidence for observational bias that Critically Deficient. Neurobehavior after
Deficient
could have impacted results 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 behavioral testing in treated mice.
There was no mention of blinding of
investigators.
Notes:
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).
A-83
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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-84
-------
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:
Good
• Outside of the exposure of interest. Good. Based on the study report, vehicle
variables that are likely to confound or (deionized water with 2% tween 80) and
Are there differences across the treatment groups
modify results appear to be controlled for husbandly practices were inferred to be the
(e.g., co-exposures, vehicle, diet, palatability.
and consistent across experimental groups, same in controls and treatment groups. The
husbandry, health status, etc.) that could bias the
experimental conditions described provided
results?
no indication of concern for uncontrolled
variables or different practices across
If differences are identified, to what extent are
groups.
they expected to impact the results?
Adequate
• Some concern that variables that were Exami)le 1 (oral): Adeauate. Hormone
likelv to confound or modify results were measurements: Authors did not use a sov-
uncontrolled or inconsistent across groups free diet. Soy-based rodent feeds contain
but are expected to have a minimal impact phytoestrogens that may act as a confounder
on the results. for endocrine-related measures. Since this
study includes relatively high doses (100
and 1500 mg/kg/day) the concern is
minimal.
Example 2 (inhalation): Adequate.
Behavior, immunoloeical responses, and
hormonal chanees: 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-85
-------
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 throughout
results. 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.
Notes: DMSO = dimethyl sulfoxide; ppm = parts per million.
Critically
Deficient
• Confounding variables were presumed to
be uncontrolled or inconsistent across
groups, and are expected to be a primary
driver of the results.
A-86
-------
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.
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
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
A-87
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Did the study report results for all prespecified outcomes and tested animals?
significantly impact the interpretation of were 8 animals used in experiment 1 and 6
the results. 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.
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
• Extensive results omission and/or animal Critically Deficient. None of the animals in
Deficient
attrition are identified and prevents the high and medium dose groups survived
comparisons of results across treatment and there was high mortality (> 75%) in the
groups. low dose group.
A-88
-------
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
individual study level. Relevance and utility of the
routes of exposure are considered in the PECO
• 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 concentrations are
missing or verified with less reliable
methods).
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-89
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Did the study adequately characterize exposure to the chemical of interest and the exposure administration methods?
criteria for study inclusion and during evidence
synthesis.
Critically
Deficient
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.
Deficient • 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.
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.
• Uncertainties in the exposure
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.
A-90
-------
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.
Example 2 (inhalation): Critically
Deficient. Dams were exposed in static
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-91
-------
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
Good
• The duration and frequency of the Example 1: Good. Study uses a standard
endpoints/outcomes in a study:
exposure was sensitive and the exposure OECD short-term (28-day) study design to
included the critical window of sensitivity examine toxicological effects that are
Does the exposure period include the critical
(if known). routinely evaluated in this testing guideline.
window of sensitivity?
Example 2: Good. The experimental design
and exposure period were appropriate for
Was the duration and frequency of exposure
evaluation of potential male reproductive
sensitive for detecting the endpoint of interest?
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.
Adequate
• The duration and frequency of the Adequate. The study does not include the
exposure was sensitive and the exposure full developmental window of exposure
covered most of the critical window of most informative to evaluating potential
sensitivity (if known). effects on androgen-dependent development
of male reproductive organs. Specifically,
the study exposed rats from GD18-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,
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-92
-------
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
exposure is not sensitive and did not
include the majority of the critical window
of sensitivity (if known). These limitations
are expected to bias the results towards the
null.
Deficient. The experimental design is not
considered appropriate for evaluation of
male fertility. Male rats were exposed for
chemical X for 1 week and fertility was
assessed on week 2 of the study. This 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 and
Deficient is expected to strongly bias the results
towards the null. The rationale should
indicate the specific concern(s).
Critically Deficient. The experimental
design is not appropriate for evaluation of
cancer endpoints. Animals were 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).
Note: OECD = Organisation for Economic Co-operation and Development; OPPT = Office of Pollution Prevention and Toxics.
A-93
-------
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 timing of these
Are there concerns regarding the specificity and
evaluations. Study authors used standard
validity of the protocols?
methodology (i.e., commercial kits)
appropriate for use in adult liver tissue
Are there serious concerns regarding the sample
samples.
size (see note)?
Example 2: Good. Orsan weieht. bodv
Are there concerns regarding the timing of the
weiehts. and hormone measures: no
endpoint assessment?
concerns regarding the specificity and
validity of the protocols and measures were
NOTE: Sample size alone is not a reason to
identified. Study authors used standard
conclude an individual study is critically deficient.
methodology for evaluating 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
A-94
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Are the procedures sensitive and specific for evaluating the endpoint(s)/outcome(s) of interest?
pathology 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.
Deficient - 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, 2009, 3987435;
Foley, 2001, 4003913).
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
A-95
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Are the procedures sensitive and specific for evaluating the endpoint(s)/outcome(s) of interest?
sample sizes used to evaluate these
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. lEndDoint 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; OECD = Organisation for Economic Co-operation and Development.
A-96
-------
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
endpoints/outcomes in a study:
Does the level of detail allow for an informed
interpretation of the results?
Good
Good. There are no notable concerns about
the way the results are analyzed or
presented.
Are the data analyzed, compared, or presented in a
way that is inappropriate or misleading?
Adequate
Examiile 1: Adeauate. Reproductive orsan
weiehts. hormone measures: results are
presented graphically; however, the authors
do not clarify whether error bars correspond
to SD or SE.
Example 2: Adequate. Developmental
effects: the study 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.
Example 3: Adequate. Anogenital distance
(AGP): 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 - Example 1: Deficient. Histopathology:
Incidence and severity of individual effects
was unclear, as only scores across multiple,
disparate pathological endpoints were
reported.
A-97
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Are the results presented in a way that makes the data usable and transparent?
Exami)le 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 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 central nervous system
(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
Deficient
- Critically Deficient. Endooint name: The
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-98
-------
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. Study Quality Evaluation Considerations for Overall Study Confidence - Animal Toxicological Studies
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
Confidence
• No notable concerns are identified (e.g.,
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 expected
Confidence 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
report information on the severity of
histological effects for which this is
A-99
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Considering the identified strengths and limitations, what is the overall confidence rating for the endpoint(s)/outcome(s) of interest?
routinely provided. The authors also failed
to describe use of methods to reduce
potential observational bias.
Low • Identified concerns are expected to
Confidence 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).
Example 1: Low Confidence.
Developmental effects: Substantial concerns
were raised regarding quantitative analyses
without addressing potential litter effects.
Other significant limitations included
incomplete data 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 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.
Example 1: Uninformative. Critical
information was not reported. Specifically,
the study authors did not report the duration
of the exposure or the results (qualitative or
quantitative). Given this critical deficiency,
the other domains 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 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
A-100
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Core Question: Considering the identified strengths and limitations, what is the overall confidence rating for the endpoint(s)/outcome(s) of interest?
resulted in an interpretation that the results
were unreliable.
Example 3: Uninfonnative. 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
uninfonnative.
A-101
-------
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 Fields 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
• General population, children
and adolescents <18 years
• Do not select "pregnant women" if pregnant
women are only included as part of a general
population sample.
A-102
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
¦ Occupational
• Pregnant women
• Occupational/general
population, adults
• Other
> When exposure is measured in cord blood and
outcome in children, the study population
would be "children".
Population Summary
[Free-text]
• Briefly describe the study population (e.g.,
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.
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 Section A.l.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
• Only use the name of an official study or
applicable)
cohort. Leave blank if there is no name.
[Free-text]
7
Country (or Countries)
• Use full names such as "United States" (not
[Free-text]
US).
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
"recruitment 2012-2015, outcome assessed at
5-year follow-up".
Exposure Measurement
[Select all that apply]
• Biomonitoring
• Air
A-103
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
• 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 also
matrix.
• Plasma
blood). Only select blood if something more
[Select all that apply]
• Maternal blood
specific is not specified (e.g., cord blood,
• Cord 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
Quantitative Data Extraction (Sub-Forms)
11.1 Health Effect Category
[Select one]
• Cancer
• Cardiovascular
• Dermal
• Developmental
• Endocrine
• Gastrointestinal
• Hematologic
• Hepatic
• Immune
• Metabolic
• Musculoskeletal/Connective
Tissue
• Nervous
• Ocular
• Reproductive, female
> See Appendix A. 1.6.5.1 for what kind of
health outcomes are grouped under which
health effect category. Please create a
separate form for each outcome.
A-104
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
• 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:
o Weight (ln-grams)
o Triglyceride (loglO mg/dL)
• Some outcomes are dichotomous (e.g., high
blood pressure, high cholesterol, etc.),
indicate the outcome definition in
parentheses. For example:
oHigh 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.
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)).
o If median is not available, please extract
other measures of distribution, such as
A-105
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
mean or geometric mean, range, other
percentiles.
• Extract levels for the overall study
population. If only available by subgroups,
specify which subgroup.
Example:
• 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., loglO or 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 ng/mL-1.4 ng/mL) vs. Ql
(0.83 ng/mL)].
• OR [fortertile 2 (0.83 ng/mL-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.
11.11 Rank this Comparison
Group by Exposure
[Free-text]
• For standalone result of unit change, leave
blank.
• 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.
A-106
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
11.12 Effect Estimate Type
[Select one]
Odds Ratio (OR)
Relative Risk Ratio (RR)
Absolute Risk %
Beta Coefficient (b)
Beta Coefficient
(standardized)
Standardized Mortality Ratio
(SMR)
Standardized Incidence 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)
• If the effect estimate is a regression
coefficient (a beta or (3), select from the menu
"Regression Coefficient" rather than "Beta
Coefficient".
• If PFOS/PFOA was the outcome of interest
(e.g., study looked at the impact of a disease
on PFOS/PFOA level), please still extract the
data but make a note under the Results
Comments (11.19).
11.13 Effect Estimate
[Free-text]
• Only report the effect estimate from the
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).
11.14 CILCL: Confidence
Interval - Lower
Confidence Limit
[Free-text]
11.15 CIUCL: Confidence
Internal - Upper
Confidence Limit
A-107
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Question/Prompt Response Options Suggested Considerations
[Free-text]
11.16 SD or SE
• Enter the SD or SE if reported for the effect
[Free-text]
estimate.
• Leave blank if not reported.
11.17 p-value
• Enter the quantitative p-value if available
[Free-text]
(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-textp. 650").
• Enter any relevant p-values, such as
interaction p-values or trend p-values.
• Enter any additional details on the outcome
measurement or definition.
A-108
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Question/Prompt
Response Options
Suggested Considerations
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 are 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: ALT = alanine transaminase; BMt = body mass index; IQR = interquartile range; LOAEL = lowest-observed-adverse-
effect level; LOD = limit of detection; LOQ = limit of quantification; NOAEL = no-observed-adverse-effect level.
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.
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.
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-109
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
A-110
-------
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 performed 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
Name Field
[Free-text]
• Name should be short and simple. For example, '28-Day Oral' '2-Year Drinking Water', '1-Week 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.
1.2
Type Field
[Select one]
• 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
[Free-text]
• Enter the preferred name of the chemical (i.e., PFOA or PFOS).
• Refer to the PECO statement in for a list of synonyms for each chemical.
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.
A-lll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Questions/Prompts and Options
Suggested Considerations
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., Fi Male Sprague
Dawley Rats or P0 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
Po) group. For example, a Po 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.
3.4
Dose-Groups Field
[Free-text]
• Dose groups should be listed lowest to highest (dose group 1 = 0 mg/kg-d).
• 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.
A-112
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Questions/Prompts and Options Suggested Considerations
• 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 • The 'Observation time' text field is included in visualizations and should be filled in; the 'Observation time'
[Free-text] numeric field and 'Observation time units' can be left blank.
• 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).
A-113
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Questions/Prompts and Options
Suggested Considerations
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 arc qualitatively discussed
in the text, but actual data arc not
provided.
• For instances where study authors
specify that only the significant effects
arc described - and certain endpoints
arc then not discussed - assume that no
change occurred in these endpoints.
Create dummv variables for all
• For endpoints for which no quantitative data arc 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 arc creating the dummy
variable (e.g.. "incidence" for histopathology observations, "grams" for body weight)
• Under "Dosc-rcsponsc 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.
• "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.
A-114
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Questions/Prompts and Options
Suggested Considerations
cndpoinls staled to be measured with
the assumption if lliey arc not
discussed thev 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,
unchcck 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 arc grouped together.
• If a single endpoint called "Clinical Observation."" create the dummy variables above using all 0 with nothing
lagged 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 lagged as
significant.
• Whenever using dummy variables instead of actual data, make sure to note in the results comment text box
that the data arc dummy variables using the standard language given in the instructions in H AWC under the
"Results notes" box.
Notes: CAS = Chemical Abstracts Service.
A-115
-------
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 PFOA
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 humans and animals was integrated for each
health outcome. Integrated judgments 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 PFOA
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 effects 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 PFOA 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-116
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table A-39. Framework for Strength-of-Evidence Judgments for Epidemiological Studies3
Strength-of-Evidence
Judgment
Description
Moderate (©©O)
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.
• Multiple studies showing generally consistent findings, including at least one high or
medium 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)
One or more studies reporting an association between exposure and the health outcome,
where 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.
Slight
(©OO)
A-117
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Strength-of-Evidence
Judgment
Description
• 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.
Compelling evidence
of no effect ( )
• 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 adapted from Table 11-3 in the IRIS Handbook.
Table A-40. Framework for Strength-of-Evidence Judgments for Animal Toxicological
Studies"
Strength-of-Evidence
Judgment
Description
Robust (©©©)
Moderate (©©O)
Slight
(©OO)
• 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.
• At least one high- or medium-confidence study with supporting information increasing
the 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).
Scenarios in which there is a signal of a possible effect, but the evidence is conflicting or
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
A-118
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Strength-of-Evidence
Judgment
Description
description in moderate) that increases the overall evidence strength, where conflicting
evidence exists from 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.
Indeterminate
(OOO)
Compelling evidence
of no 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 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
judgment level
Explanation and example scenarios
Evidence • A strong evidence base demonstrating that [chemical] exposure causes [health effect] in
demonstrates humans
• 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
indicates humans, 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
A-119
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence
integration Explanation and example scenarios
judgment level
• 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,
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)
Evidence • An evidence base that suggests that [chemical] exposure may cause [health effect] in humans,
suggests 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
Evidence • This conveys either a lack of information or an inability to interpret the available evidence for
inadequate13 [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
Notes: MOA = mode of action.
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.
Strong evidence •
supports no
effect
A-120
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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
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
A-121
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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
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 judgments indicate stronger evidence of hazard (i.e., evidence
demonstrates and evidence indicates integration judgments). 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,
A-122
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• 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,
• 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 PFOA, 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
A-123
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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
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 PFOA MCLG main document for additional detail).
Concentrations of PFOA 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 PFOA 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
A-124
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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
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 a 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
A-125
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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.
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-126
-------
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 PFOA, 17 meta-analysis studies were identified and summarized below (Table
A-42, Table A-43).
Table A-42. Epidemiologic Meta-Analysis Studies Identified From Literature Review
Reference
Number of
Studies
Countries
Health Outcome
Results/Conclusions3
Johnson et al.
(2014, 2851237)
Canada, Denmark,
Germany, Japan,
South Korea, Taiwan,
United Kingdom,
United States
Developmental
Birthweight:
• Pooled (3 per 1 ng/mL increase in serum or plasma PFOA (9 studies):
-18.9 g (-29.8, -7.9), I2= 38%
Length:
• Pooled (3 per 1 ng/mL increase in serum or plasma PFOA (5 studies):
-0.06 cm (-0.09, -0.02), I2= 0%
Ponderal Index:
• Pooled (3 per 1 ng/mL increase in serum or plasma PFOA (5 studies):
-0.01 g/cm3 (-0.03, 0.01), I2= 63%
Head Circumference:
• Pooled (3 per 1 ng/mL increase in serum or plasma PFOA (4 studies):
-0.03 cm (-0.08, 0.01), I2= 26%
Verneretal. (2015,
3150627)
Canada, Denmark,
Japan, Norway,
Taiwan, United
Kingdom, United
States
Birthweight:
• Pooled (3 per 1 ng/mL increase in PFOA in maternal or cord blood (7
studies): -14.72 g (-8.92, -1.09)
Developmental • 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)
16
Canada, China,
Denmark, Germany,
Greenland, Japan,
Norway, Poland,
South Korea, Taiwan,
Ukraine, United
Birthweight:
• Pooled (3 per 1 ng/mL increase in PFOA (12 studies): -12.8 g (-23.2,
Developmental 2.4), I2 = 53%
• Pooled (3 per 1-ln ng/mL increase in PFOA (9 studies): -27.1 g (-
50.6, -3.6), I2 = 28%
A-127
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Number of
Studies
Countries
Health Outcome
Results/Conclusions3
Kingdom, United
States
Steenland et al.
(2018, 5079861)
24
NR
Developmental
Birthweight:
• Pooled (3 per 1 ng/mL increase in PFOA in maternal or cord blood
(24 studies): -10.5 g (-16.7, -4.4), I2 = 63%
• After inclusion of one additional large study (Savitz, 2012) (25
studies): -1.0 g (-2.4, 0.4)
• Cord blood studies only (9 studies): -13.3 g (-24.7, -1.8), I2 = 47%
• Maternal blood studies only (15 studies): -9.2 g (-15.6, -2.8),
I2 = 66%
• Comparison between early- and late-pregnancy blood sampling
yielded a p-value of 0.02
• Early-pregnancy blood PFOA 9 studies): -3.3 g (-9.6, 3.0), I2= 68%
• Late-pregnancy blood PFOA (17 studies): -17.8 g (-25.0, -10.6),
I2 =29%
Cao et al. (2021,
9959525)
Korea, Spain, Taiwan,
United States
Developmental
LBW:
• Pooled OR for maternal PFOA (6 studies): OR: 0.90 (0.80-1.01),
I2 = 18.4%
Deji et al. (2021,
7564388)
21
Brazil, Canada,
China, Denmark,
Norway, Spain,
United States
Developmental,
Female Reproductive
Miscarriage:
• Pooled OR (6 studies): 0.98 (0.92, 1.05); I2 = 0%, heterogeneity
p = 0.502
PTB:
• Pooled OR (16 studies): 0.98 (0.89, 1.08); I2 = 54.6%, heterogeneity
p = 0.005
Gao et al. (2021,
9959601)
Brazil, Canada,
China, Denmark,
29 Norway, Spain,
Sweden, United
States
Developmental,
Female Reproductive
PTBC:
• (8 studies): inverted U-shaped association, increased risk in middle
exposure range (p-nonlinear trend = 0.030)
• GDM (7 studies), miscarriage (2 studies), preeclampsia (4 studies),
pregnancy-induced hypertension (2 studies), SGA (6 studies),
LBW (2 studies): Associations not statistically significant
Yang et al. (2022,
10176603)
23
Belgium, Canada,
China, Denmark,
Netherlands, Norway,
Slovakia, Spain,
Sweden, United
States
Developmental
PTB:
• Pooled OR (14 studies): 1.22 (0.95, 1.57), I2 = 58.8%
o Significant associations for PFOA in maternal blood sampled in
3rd trimester to delivery (2 studies, pooled OR = 2.25 (1.07,
4.74), I2 = 0%), and for maternal blood sample type overall (13
studies, pooled OR = 1.29 (1.01, 1.66), I2 = 52.6%)
A-128
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Number of
Studies
Countries
Health Outcome
Results/Conclusions"
Miscarriage:
• Pooled OR for PFOA in maternal blood (5 studies): 1.40 (1.15, 1.70),
I2 = 0%
o Pooled OR for PFOA in maternal blood sampled in lst-2nd
trimester (3 studies): 1.50 (1.16, 1.95), I2 = 0%
SGA:
• Pooled OR (11 studies): 1.08 (0.93, 1.27), I2 = 0%
LBW:
» Pooled OR (7 studies): 1.02 (0.80, 1.29), I2 = 0%
Costello et al.
(2022, 10285082b)
25
Asia (NOS), Europe
(NOS), United States
Hepatic
ALT:
• Positive relationship between PFOA and ALT in adults and
adolescents
o Cross-sectional (8 studies) weighted z-score = 6.20, p < 0.001
oLongitudinal (3 studies) weighted z-score = 5.12, p < 0.001
• In children < 12 years of age, associations not statistically significant
GGT:
• Positive relationship between PFOA and GGT in adults
o Cross-sectional (8 studies) weighted z-score = 4.13, p < 0.001
o One longitudinal study reported positive association
• 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 (18
studies)
Serum Lipids (11
studies)
CVD Risk:
• Small overall effect between serum PFOA and CVD risk (16
studies); z = 1.56, p = 0.12,12 = 72.1%
• Inconsistent associations between serum PFOA and coronary heart
disease and stroke
• Consistent associations between serum PFOA and increased serum
TC, LDL, triglyceride levels, and uric acid
• Inconsistent associations between serum PFOA and increased GDM
in pregnant mothers compared to non-pregnant mothers
Metabolic (3 studies)
Kim et al. (2018,
5079795)
China, Korea, Japan,
11 Norway, Taiwan,
United States
Endocrine - Thyroid
Total T3:
• Pooled z-value (7 studies): 0.03 (0.00, 0.06), I2 = 43%
• Free T4 (8 studies), Total T4 (8 studies), TSH (11 studies):
Associations not statistically significant
o Sensitivity analyses removed one outlier for total T4; z
value = -0.06 (-0.08, -0.03), I2 = 47%
A-129
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Number of
Studies
Countries
Health Outcome
Results/Conclusions3
• Subgroup analyses stratified by PFOA levels or pregnancy status:
Associations not statistically significant
Liu et al. (2018,
5079852)
10
Denmark, Faroe
Islands, Greenland,
Norway, Spain,
Sweden, Taiwan,
Ukraine, United
States
Metabolic
Overweight risk:
• Overall effect size for maternal plasma/serum PFOA and childhood
overweight risk (8 studies): 1.25 (1.04, 1.50), I2 = 40.5%
BMI z-score:
• Pooled (3for maternal plasma/serum (9 studies): 0.10 (0.03, 0.17),
I2 = 27.9%
• Significant association between early-life exposure to PFOA and
childhood BMI z-score among studies in Europe (7 studies) but not
studies in North America (3 studies) or Asia (1 study)
• Subgroup of analyses adjusted by maternal parity (7 studies):
beta = 0.13 (0.02, 0.24), I2 = 47.4%
• Subgroup of analyses stratified by sex (4 studies): Associations not
statistically significant for either sex
Zare Jeddi et al.
(2021, 8347183)
7
Canada, China,
Croatia, Italy, United
States
Metabolic
Metabolic syndrome:
• Pooled OR: 1.06 (0.9, 2.34), I2 = 67.6%
Stratakis et al.
(2022, 10176437)
China, Demark, Faroe
Islands, Greenland,
Netherlands, Norway,
21 Spain, Sweden, Metabolic
Taiwan, Ukraine,
United Kingdom,
United States
BMI z-score:
• Inverse association reported between prenatal PFOA exposure and
BMI-z in infancy (3 studies): (3 = -0.02 (-0.08, 0.05), I2 = 70.9%
• BMI-z in childhood (2-9 years) (10 studies): (3 = 0.03 (-0.02, 0.08),
I2 = 55.5%
• Waist circumference in childhood (4 studies): (3 = 0.30 (-0.50, 1.09),
I2 = 85.7%
• 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 = 1.00 (0.75, 1.25), I2 = 76.6%
• Subgroup analyses for differences by region or exposure type not
significant
Bartell and Vieira.
(2021, 7643457)
NR
Cancer
Kidney cancer:
• Concluded sufficient evidence that PFOA is a likely cause of kidney
cancer and testicular cancer in humans
• Average relative increase in cancer risk per 10 ng/mL increase in
serum PFOA:
A-130
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
,, „ Number of
Reference
Studies
Countries
Health Outcome
Results/Conclusions"
oKidney cancer (7 studies): 16% (3%, 30%)
o Testicular cancer (3 studies): 3% (2%, 4%)
Notes'. ADHD = attention deficit-hyperactivity disorder; ALT = alanine aminotransferase; AST = aspartate aminotransferase; BMt = body mass index; CVD = cardiovascular
disease; GDM = gestational diabetes mellitus; GGT = gamma-glutamyl transferase; LBW = low birth weight; LDL = low density lipoproteins; NR = not reported; PTB = preterm
birth; SGA = small for gestational age; TC = total cholesterol.
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-131
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table A-43. Toxicologic Meta-Analysis Studies Identified From Literature Review
Reference
Number of
Studies
Animal Sex and
Model/Species
Health Outcome(s)
Results/Conclusions3
Wang et al., 2021,
7152781
16
Male Rats, Male Mice
Male Reproductive
(including Cancer)
¦ SMD for reproductive toxicity (14 studies): -0.39 (-0.71, -0.07),
p = 0.02,12 = 80%
¦ SMD for serum testosterone levels (6 studies): -0.54 (-0.95, -0.13),
p = 0.01,12 = 71%
> Mean difference for serum estradiol levels (3 studies): 4.75 (2.29,
7.21), p = 0.0002,12 = 91%
¦ SMD for absolute testicular weight (7 studies): -0.20 (-0.33, -0.06),
p = 0.005,12 = 44%
> SMD for absolute epididymis weight (2 studies): -0.01 (-0.02, -
0.01), p< 0.0001,12 =39%
¦ OR for incidence of Leydig cell adenoma (2 studies): 8.47 (2.74,
26.18), p = 0.0002,12 = 0%
> Mean difference for percentage of abnormal sperm (2 studies): 1.48
(0.65, 2.30), p = 0.0004,12 = 87%
> Day of preputial separation, risk of testis atrophy, risk of epididymis
tubular atrophy, sperm motility: Associations not statistically
significant
Notes: SMD = standard mean difference; OR = odds ratio.
a Results reported as effect estimate and 95% confidence interval (CI) unless otherwise stated.
A-132
-------
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 PFOA, 7 studies were identified after the updated
literature review and they are summarized below (Table A-44).
Table A-44. 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 linear PFOA concentrations had adjusted
HRs of 1.47 (95% CI: 1.24, 1.75) (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.
PFOA might be associated with increased risk of
hypertension in women. Possible mixture effects with
hypertension in women. No change.
Feng et al., 2022
(10328872)
Goodrich et al.,
2022 (10369722)
Case-cohort study within the Dongfcng-Tongji cohort,
including incident breast cancer cases (n = 226) and a random
sub-cohort (n = 990). Significant increase in risk per each 1-
ln ng/mL increase in PFOA (HR = 1.35, 95% CI: 1.03, 1.78),
in the highest (> 1.80 ng/mL) uv. lowest quartile
(< 0.84 ng/mL) (HR = 1.69, 95% CI: 1.05, 2.70)), and among
postmenopausal women (HR= 1.34, 95% CI: 1.01, 1.77).
Quantile g-computation analysis observed a 19% increased
incident risk of breast cancer along with each simultaneous
quartile increase in all ln-transformed PFOA concentrations
(HR= 1.19, 95% CI: 1.01, 1.41), with PFOA accounting for
56% of the positive effect.
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). Non-significant increase in risk in in those with high
PFOA exposure (> 85th percentile; > 8.6 ug/L) vs. low
PFOA may be associated with increased risk of breast cancer
in postmenopausal women. Similar findings with {Mancini,
2020, 5381529}. No change.
No change.
A-133
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Major Findings
Assessment Implications
Gui et al., 2022
(10365824)
Jiang et al., 2022
(10328207)
exposure (< 85th percentile; < 8.6 ug/L) (OR = 1.20, 95% CI:
0.49, 2.80).
Meta-analysis of 24 studies, pooled change in birthweight per
1 ng/mL increase in PFOA (unadjusted for gestational
age/unstandardized birth weight): -37.02 g (95% CI: -54.37,
-19.66), I2= 56.5 %. 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.
Meta-analysis of 8 studies across 8 countries, pooled breast
cancer risk for PFOA was 1.32 (95% CI: 1.19, 1.46),
I2 = 98.5%. However, results seem to be influenced by the
largest study {Omoike, 2021, 7021502}.
Supports an association between PFOA and birth weight,
birth length and ponderal index. Similar conclusions as
previous meta-analyses.
No change. Serious methodological limitations warrant
cautious interpretations of results from this publication.
Luo et al., 2022
(10273290)
Prospective study in the Danish National Birth Cohort, 656 No change
children. Prenatal exposure to PFOA was not associated with
facial features (measures of palpebral fissure length, philtrum
groove, and upper-lip thickness) in children at age 5.
Velarde et al., 2022 Case-control study of 150 Filipino women (75 breast cancer
(9956482) cases and 75 controls). PFOA was not statistically
significantly associated with breast cancer risk.
No change.
Wen et al., 2022
(10328873)
Population -based cohort study of 11,747 participants from No change.
1999-2014 NHANES followed up to December 2015. PFOA
was not statistically significantly associated with all-cause,
heart disease or cancer mortality.
A-134
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference Major Findings Assessment Implications
Zhang et al., 2022 Prospective cohort study (the Shanghai Birth Cohort) of 2,395 No change.
(9944433) mother-infant pairs. Prenatal PFOA 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.
Notes: HR = hazard ratio; OR = odds ratio; NHANES = National Health and Nutrition Examination Survey; SWAN = Study of Women's Health Across the Nation.
A-135
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix B. Detailed Toxicokinetics
B.l Absorption
B.l.l Cellular Uptake
Several studies using cell lines transfected with specific transporters or vector controls support
cellular accumulation of PFOA through facilitated transport. Several transporters classically
considered specific to renal or enterohepatic resorption have also been found to be expressed in
tissues relevant to absorption. Specifically, organic anion transporter 2 (OAT2) transcripts have
been identified in several tissues in addition to kidney including the small intestine {Cropp,
2008, 9641964}. OATP1A2 expression has also been identified in intestine {Kullak-Ublick,
1995, 9641965}.
A single study in immortalized intestinal Caco-2 cells found that uptake was fast and saturable,
supporting a carrier-mediated uptake process. The Km for PFOA uptake was calculated to be
8.3 ±1.2 |iM and uptake clearance (Vmax/Km) was 55.0 [xL/mg protein/min. Uptake was found to
be independent of sodium ions, while concentration, temperature, and pH all influenced uptake.
Substrates or inhibitors of organic anion transporting polypeptides (OATPs) significantly
decreased the uptake of PFOA, suggesting that the uptake of PFOA from the apical membranes
of Caco-2 cells was at least partially mediated by OATPs {Kimura, 2017, 3981330}.
Lipid binding may influence PFOA 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 higher levels compared to
azithromycin-dihydrate, 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 pH7.4 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 pH7.4. 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 PFOA
phospholipophilicity influences absorption through the gastrointestinal 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
Phospholipid
Lipid Binding
LogP
Lung Epithelium
Epithelium
Binding
(LogD7.4)
(% AZI)
(CHIIAM7.4)
PFOS
313±101*
26 ±4
39 ±3*
2.33 ±0.11*
5
PFOA
15 ±3
ND
29 ± 1
1.29 ±0.02
4.9
B-l
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 uptake of
PFOA into cells. This is consistent with observations that cells transfected with vector only could
take up PFOA, albeit at lower levels than cells transfected with PFOA-specific transporters
(discussed further in Section B.4.2.1). Ebert et al. (2020, 6505873) determined membrane/water
partition coefficients (Kmem/w) for PFOA 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/[LW/kgmem]) for PFOA was 3.93, similar to the
experimentally determined value of 3.52 ± 0.08. Kmem/w values 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 values. The predicted anionic
permeability (log Pion/[cm/s]) for PFOA ranged from -6.89 to -7.45, 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
Based on animal data, PFOA is well absorbed following oral exposure. Studies on male rats
administered PFOA by gavage using a single dose (11.4 mg/kg, CD rats) or daily doses over
28 days (5 or 20 mg/kg/day, Sprague-Dawley rats) all estimated dose absorption of at least
92.3% {Gibson, 1979, 9641813; Cui, 2010, 2919335}.
Toxicokinetic parameters informing absorption were derived by comparing oral to intravenous
(IV) dosing in two studies conducted in rats {Kim, 2016, 3749289; Dzierlenga, 2019, 5916078}.
In the study by Kim and colleagues, rapid differences in absorption based on sex were observed
for PFOA but not PFOS {Kim, 2016, 3749289}. Male and female Sprague-Dawley rats were
administered 1 mg/kg by either the IV or oral route. Urine and feces were collected daily for both
males and females, and blood was collected at 11 time points on the first day (females) or 3 time
points on the first day and then up to 12 days after exposure (males). The time to reach the
maximum PFOA plasma concentration (Tmax) following oral exposure in females was 1.44 hours
vs. 2.07 days in males. Dzierlenga et al. (2019, 5916078) administered a single bolus IV
(6 mg/kg) or gavage dose (6, 12, or 48 mg/kg) to adult male Sprague Dawley rats and a single
bolus IV (40 mg/kg) or gavage dose (40, 80, or 320 mg/kg) to adult female Sprague Dawley rats.
Blood and urine were collected for up to 8 time points during the first 24 hours and then up to 12
(females) or 50 (males) days post-dosing. Tmax in rats administered these doses via gavage
ranged from 2.33 to 3.22 hours in females and 4.86 to 8.33 hours in males for PFOA. In females,
maximum blood concentration (Cmax) per dose (mM/mmol/kg) decreased with increasing dose
suggesting saturation of absorption kinetics at higher doses. Similar to the Kim et al. (2016,
3749289) study, shorter Tmax values were observed in females compared with males at all doses.
The data from studies of adverse effects on monkeys, rats, and mice receiving PFOA in capsules,
food, or drinking water demonstrate gastrointestinal absorption. In cynomolgus monkeys, steady -
B-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
state serum and urine PFOA levels were reached 4-6 weeks after dosing with capsules
containing 3, 10, or 20 mg/kg PFOA for 6 months {Butenhoff, 2004, 3749227}. Serum PFOA
concentrations in male Crl:CD BR rats fed diets containing 0.06, 0.64, 1.94, or 6.5 mg PFOA/kg
for 90 days were 7.1, 41, 70, and 138 (J,g/mL, respectively {Perkins, 2004, 1291118}. Peak blood
levels of PFOA were attained 1-2 hours following a 25 mg/kg dose to male and female rats
{Kennedy, 2004, 724950}. Studies on same-sex rats found no differentiation in blood or plasma
levels of PFOA when comparing single and repeated daily PFOA dose administrations
{Kennedy, 2004, 724950; Elcombe, 2010, 2850034}.
In rats, marked sex differences in serum and tissue PFOA levels exist following PFOA
administration. Males consistently have much higher levels than females with differences
maintained and becoming more pronounced over time. Female rats show much greater urinary
excretion of PFOA than do male rats with serum half-life values in hours for females compared
with days for males. These differences account for variability in postexposure serum PFOA
concentrations between males and females.
B.1.3 Inhalation Exposure
Data on exposure to PFOA by inhalation remains unchanged since Hinderliter et al. (2006,
135732) measured the serum concentrations of PFOA following single and repeated nose-only
aerosol inhalation exposures of 0, 1, 10, or 25 mg/m3 PFOA in Sprague-Dawley rats, which
found that PFOA plasma concentrations increased proportional to aerosol exposure
concentrations. The male plasma Cmax values were approximately 2-3 times higher than the
female plasma Cmax values. The female Cmax occurred approximately 1 hour after the exposure
period with plasma concentrations then declining. In males, Cmax was observed immediately after
the exposure period ended and persisted for up to 6 hours. These data demonstrate absorption of
PFOA via inhalation and provide evidence of the sex differences consistent with rate of
excretion.
B.1.4 Dermal Exposure
Evidence that PFOA is absorbed following dermal exposure remains unchanged since 2005, with
in vitro percutaneous absorption studies of PFOA through rat and human skin allowing
calculation of permeability coefficients for PFOA in rat skin to be 3.25 x 10 5 cm/hr, and that of
human skin to be 9.49 x 10~7 cm/hr {Fasano, 2005, 3749187}. Previously, O'Malley and Ebbins
(1981, 4471529) utilized mortality as an indicator of dermal uptake across groups of two male
and two female New Zealand white rabbits receiving 0, 100, 1,000, or 2,000 mg/kg PFOA; after
14 daily dermal doses, all of the animals died at the highest dose, 3 of 4 animals died in the mid-
dose group, and no animals died in the low-dose group. Kennedy (1985, 3797585) detected
elevated blood organofluorine levels in male New Zealand white albino rabbits and male and
female Crl:CD rats that were dermally treated with a total of 10 applications of PFOA at doses of
0, 20, 200, or 2,000 mg/kg. Treatment resulted in elevated blood organofluorine levels that
increased in a dose-related manner.
B-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.1.5 Developmental Exposure
The literature contains no studies on PFOA absorption following developmental exposure.
Additional information on PFOA distribution during reproduction and development is found in
Section B.2.4.B.2.4
B.1.6 Bioavailability
The Kim and Dzierlenga studies discussed above also observed very high bioavailability in rats
(Table B-2) {Kim, 2016, 3749289; Dzierlenga, 2019, 5916078}. At a lower dose of 1 mg/kg,
Kim et al. (2016, 3749289) found that Cmax values after oral administration were 85% and 92%
of values obtained after IV administration (bioavailability values were not reported in this study).
In the Dzierlenga et al. (2019, 5916078) study, bioavailability (calculated by dividing the dose-
adjusted gavage area under the curve (AUC) by the IV AUC) was 140% in males administered
6 mg/kg and 182%) in females administered 40 mg/kg. The authors suggested that the high
bioavailability of PFOA may be attributed to increased reabsorption by intestinal transporters by
the oral route.
Table B-2. PFOA Parameters from Toxicokinetic Studies Informing Bioavailability in
Sprague-Dawley Rats
Study
Dose (mg/kg)
Route
Sex
Cmax (ng/mL)
Tmax (hours)3
Kim et al. (2016,
1
Oral
Male
7.55 ±0.51
49.68 ±5.04
3749289)
IV
Male
8.92 ±2.34
NA
1
Oral
Female
5.41 ±0.38
1.44 ±0.096
IV
Female
5.84 ±0.38
NA
Dzierlenga et al.
6
Oral
Male
36.85 ±2.90
4.86 ±0.81
(2019, 5916078)
IV
Male
52.59 ±2.5
NA
40
Oral
Female
240.16 ±24.84
3.22 ±0.32
IV
Female
369.76 ±81.16
NA
Notes: Cmax = maximum serum concentration; IV = intravenous; NA = not applicable; Tmax = time to Cmax.
a Converted published Tmax (days) to Tmax (hours) for Kim et al. (2016, 3749289).
Li et al. (2015, 2851033) examined bioavailability from food sources in female BALB/c mice
and using in vitro methods. In mice, PFOA was mixed with foods of different nutritional
compositions (e.g., meat, seafood, milk, and fruits/vegetables) and fed to mice over a 7-day
period. By comparing PFOA administration via food mixtures to administration in water, relative
bioavailability was assessed by measuring accumulation in liver. PFOA bioavailability relative to
water ranged from 4.30 ± 0.80 to 69.0 ± 11.9% and was negatively correlated with lipid content
(r = 0.76). The authors suggest that sorption by free fatty acids in foods could limit PFOA access
to intestinal transporters. Another possibility is cations in the gastrointestinal tract, such as Ca2+
and mg2+, can complex PFOA promoting partitioning to the lipid phase. Three different in vitro
methods were used to measure bioavailability using these food mixtures including the in vitro
digestion method (IVD) {James, 2011, 6718854}, the unified BARGE method (UBM) {Smith,
2012, 6702349}, and the physiologically based extraction test (PBET) {Tilston, 2011, 5120687}.
Instead of soil, 0.3 g of food was used at sample/solution ratios of 1:97.5 for UBM, 1:100 for
PBET, and 1:150 for IVD. PFOA bioaccessibility varied by the method (8.7-73% for UBM,
9.8-99%) for PBET, and 21-114%> for IVD). As observed in the in vivo study, bioaccessibility
B-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
was negatively correlated with lipid content for the UBM method (r = 0.82) but not for other in
vitro methods (r = 0.11-0.22). The authors suggest that the UBM method can be used to model
bioaccessibility, possibly because this method is associated with higher lipolysis and better
mimics cations in gastrointestinal fluid of UBM. This may lower the potential to form stable
micelles using this method compared to PBET and IVD methods. Together, these findings
suggest PFOA bioavailability is strongly influenced by diet, with high fat diets associated with
reduced absorption, and that an important factor influencing PFOA bioaccessibility is colloidal
stability in intestinal solutions.
B.2 Distribution
B.2.1 Protein Binding
Kerstner-Wood et al. (2003, 4771364) used in vitro methods to evaluate PFOA binding to
protein in plasma from humans, cynomolgus monkeys, and rats. In all species, plasma proteins
were able to bind 97-100% of the PFOA added at concentrations ranging from 1 to 500 ppm. In
humans, serum albumin carried the largest portion of PFOA among the protein components of
human plasma. Serum albumin is a common carrier of hydrophobic materials in the blood
{Fasano, 2005, 1023584} and approximately 60% of the serum protein in humans and rats is
albumin {Harkness, 1983, 9641985; Saladin, 2004, 9642161}.
Han et al. (2003, 5081471) investigated the binding of PFOA to rat and human plasma proteins
in vitro and determined that the primary PFOA binding protein in plasma was serum albumin.
No significant differences in binding to the serum albumin were found between humans and rats.
Calculation of disassociation constants (Kd) for PFOA, conducted using purified rodent and
human serum albumin binding using labeled 19F nuclear magnetic resonance (NMR) and micro-
size exclusion chromatography and the estimated number of binding sites from this study are
presented in Table B-3.
Table B-3. Dissociation Constants of Binding Between PFOA and Albumin as Reported by
Han et al. (2003, 5081471)
Parameter
Method
Rat Serum Albumin
Human Serum Albumin
Kd (mM)
NMRa
0.29 ± 0.10b
ND
Kd (mM)
micro-SEC0
0.36 ± 0.08b
0.38 ±0.04
Number of Binding Sites
micro-SEC0
7.8 ± 1.5
7.2 ± 1.3
Notes: Kd = dissociation constant; micro-SEC = micro-size exclusion chromatography; ND = not determined; NMR = nuclear
magnetic resonance.
a Average of the two Kd values (0.31 ±0.15 and 0.27 ± 0.05 mM) obtained by NMR.
b On the basis of the result of unpaired t-test at 95% confidence interval, the difference of Kd values determined by NMR and
micro-SEC is statistically insignificant.
c Values were obtained from three independent experiments and their standard deviations are shown.
Several studies have examined the interactions between PFOA and human serum albumin {Wu,
2009, 536376; MacManus-Spencer, 2010, 2850334; Qin, 2010, 3858631; Salvalaglio, 2010,
2919252; Weiss, 2009, 534503; Kerstner-Wood, 2003, 4771364; Luebker, 2002, 1291067;
Zhang, 2013, 5081488; Cheng, 2018, 5024207; Gao, 2019, 5387135; Yue, 2016, 3479514}. Wu
et al. (2009, 536376) examined whether PFOA, after absorption, was transported bound to
B-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
albumin by dialyzing PFOA solutions in the presence and absence of human serum albumin. The
authors found that, in the absence of albumin, 98% of the dissolved PFOA crossed the dialysis
membrane into the dialysate within 4 hours. In the presence of albumin, the amount of PFOA
found in the dialysate decreased in direct proportion to the albumin concentration, demonstrating
binding to the protein. No albumin was identified in the dialysate. Circular dichroism
measurements of the albumin/PFOA complex suggested a conformational change in the protein
as a result of the PFOA binding. These conformational changes could interfere with the
functional properties of serum albumin or other serum proteins impacted by surface monolayers
of PFOA. For example, albumin's ability to transport its natural ligands could be decreased by
the presence of PFOA on the protein surface {Wu, 2009, 536376}.
MacManus-Spencer et al. (2010, 2850334) used a variety of approaches to quantify the binding
of PFOA to serum albumin (e.g., surface tension measurements, 19F NMR spectroscopy,
fluorescence spectroscopy) using bovine serum albumin. Taken together, the results from these
analyses suggested the presence of primary and secondary binding sites on albumin. The results
of the fluorescence spectroscopy suggested a conformational change in albumin following
binding of PFOA that moved tryptophan residue 214 from a slightly polar region of the protein
to a less polar region. Qin et al. (2010, 3858631) also used fluorescence spectroscopy quenching
analysis to study PFOA binding to bovine serum albumin and reported that albumin underwent a
conformational change following the binding of PFOA. They also suggested that van der Waals
forces and hydrogen bonds were the dominant intermolecular binding forces. Similar findings
were observed more recently {Chen, 2020, 6324256} for human serum albumin. This study used
infrared spectroscopy to examine PFOA-mediated effects on albumin secondary structure and
found that PFOA binding led to a decrease in the P-sheet and a-helix conformations.
Salvalaglio et al. (2010, 2919252) conducted a modeling study to determine the binding sites of
PFOA on human serum albumin and classify them by their interaction energy using molecular
modeling. They estimated a maximum number of nine PFOA binding sites on human serum
albumin and determined that these site locations were common to the natural binding sites for
fatty acids, thyroxine (T4), Warfarin, indole, and benzodiazepine. The binding site closest to
tryptophan residue 214 had the highest binding affinity.
Beesoon and Martin (2015, 2850292) examined differences in the binding of the linear and
branched chain PFOA isomers to calf serum albumin and human serum proteins. The linear
PFOA isomer bound more strongly to calf serum albumin than the branched chain isomers.
When arranged in order of increasing binding, the order was 4m <3m<5m<6m (iso)< linear.
In the isomer-specific binding to spiked total human serum protein, the linear molecule clearly
had the strongest binding potential with about 7-10% free. The relationship for the other isomers
was 5m > 6m > 4m > 3m (15-30% free).
Weiss et al. (2009, 534503) screened PFOA and 29 other perfluorinated compounds—differing
by carbon chain length (C4-18), fluorination degree, and functional groups—for potential
binding to the serum thyroid hormone transport protein, transthyretin (TTR), using a radioligand-
binding assay. The natural ligand of TTR is T4. Human TTR was incubated overnight with 125I-
labeled T4, unlabeled T4 (reference), and 10-10,000 nmol PFOA as a competitor for the T4
binding sites. The authors concluded that the binding affinity for TTR was highest for the fully
fluorinated compounds tested and those having at least a carbon chain length of 8, characteristics
B-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
that apply to PFOA. PFOA demonstrated a high binding affinity for TTR with 949 nmol, causing
a 50% inhibition of T4 binding to TTR.
Binding to albumin and other serum proteins may affect transfer of PFOA from maternal blood
to the fetus. Gao et al. (2019, 5387135) correlated placental transfer with experimentally
measured Kd to human serum binding proteins, serum albumin, and L-FABP. For PFOA, KdS
were calculated to be 115 ± 16 |iM for albumin, 166 ± 10 |iM for serum binding proteins, and
197 ± 13 |iM for L-FABP. These KdS 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. Since there is effectively a competition between PFOA 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% per 1 g/L albumin). However,
maternal serum albumin concentration was associated with reduced transfer efficiency (decrease
of 2.5% per 1 g/L albumin). Because albumin cannot cross the placental barrier, the authors
speculate that binding of PFOA to maternal serum albumin can reduce the free PFOA available
to move across the barrier through passive diffusion. Similarly, higher fetal albumin levels will
lead to less free PFOA in cord blood, which may facilitate the rate of placental transfer via
passive diffusion.
In contrast to serum proteins, little is known regarding PFOA binding to proteins in the gut. Yue
et al. (2016, 3479514) examined whether PFOA that enters the digestive tract binds to gastric
enzymes, specifically pepsin. Binding to pepsin was examined using fluorescence quenching of
pepsin's intrinsic fluorescent properties. Scatchard analysis was used to estimate a binding
constant of 0.717 x 104 at 298 K. Spectroscopy including ultraviolet-visible absorption, Fourier
transform infrared fluorescence, and circular dichroism indicated that PFOA induces a
conformation change in pepsin associated with decreased a-helical and P-sheet content.
Molecular docking analysis suggested that PFOA interacts with 16 amino acid residues of
pepsin. It is unclear whether PFOA-pepsin interactions impact absorption or distribution from
the gut to other compartments in the body.
PFAS also binds 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 PFOA 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 concentrations (ICsos) for PFOA and PFOS were 9.0 ± 0.7 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-
B-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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.
B.2.2 Subcellular Distribution
Han et al. (2005, 5081570) examined the subcellular distribution of PFOA in the liver and
kidney of male and female rats. Male and female Sprague-Dawley Crl:CD (SD)IGS BR rats
were gavage-dosed with 25 mg/kg [14C] PFOA and necropsied 2 hours after dosing. Blood was
collected and the liver and kidneys were removed. Five subcellular fractions (nuclei and cell
debris, lysosome and mitochondria, microsome, light microsome and ribosome, and membrane-
free cytosol) were obtained by differential centrifugation. In the male liver, the highest
proportion of total reactive residues (TRR) of PFOA was located in the nuclei and cell debris
(40%), followed by membrane-free cytosol (26% TRR), lysosome and mitochondria (-14%
TRR), microsome (-16% TRR), and light microsome and ribosome (-1% TRR). In the female
liver, the highest proportion of TRR of PFOA was found in the membrane-free cytosol (48%),
followed by nuclei and cell debris (-31% TRR), lysosome and mitochondria (—12% TRR),
microsome (-8% TRR), and light microsome and ribosome (-1% TRR). Based on the results,
the authors concluded that subcellular distribution of PFOA in the rat liver was sex-dependent
because the proportion of PFOA in the liver cytosol of female rats was almost twice that of the
male rats. They hypothesized that females might have a greater amount than males of an
unknown liver cytosolic binding protein with an affinity for perfluorinated acids. They also
hypothesized that the unknown protein or protein complex might normally aid in transport of
fatty acids from the liver. In the kidney, the subcellular distribution did not show the sex
difference seen with the liver; however, the protein-bound fraction for the males (42%) was
about twice that for the females (17%).
Zhang et al. (2020, 6316915) examined the subcellular distribution of PFOA in human colorectal
cancer cells (DLD-1), human lung epithelial cells (A549), and human normal liver cells (L-02).
Cells were incubated with 100 or 300 |iM PFOA for 48 hours and mitochondria, nucleus, and
cytosol were isolated and examined for PFOA levels. Accumulation in these subcellular
compartments corresponded to exposure levels with the highest amounts accumulating in cytosol
followed by nuclei and mitochondria. Cytosolic accumulation was more than 100 times greater
than accumulation in the other analyzed subcellular compartments. The PFOA concentration in
cytosol was highest for liver cells and was comparable between colorectal cancer and lung
epithelial cells. The patterns of accumulation (cytosol > nuclei > mitochondria) were also
comparable.
B.2.3 Tissue Distribution
B. 2.3.1 Human Studies
B.2.3.1.1 Distribution in Blood Fractions
Human blood is a major site of PFOA 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 PFOA concentration was 4.8 ng/mL in adults and 3.0 ng/mL in
B-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
children {Kotlarz, 2020, 6833715}. This value was similar to the estimate of 3.8 ng/mL
predicted using a pharmacokinetic model based on drinking water containing 15 ng/L PFOA and
using the average length of residence of 20 years for the participants.
PFOA accumulation in blood impacts distribution to various tissues and organs, but few studies
have examined PFOA partitioning to human blood fractions. Forsthuber et al. (2020, 6311640)
measured the distribution of PFOA 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 PFOA and that
there was no affinity for lipoproteins. The concentration of PFOA in these fractions was below
the limit of detection (LOD).
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 from C6
(mean 0.24) to CI 1 (0.87). The PFOA plasma:whole blood ratio in the Jin et al. (2016, 3859825)
study was lower (1.2 ± 0.43) compared to the mean plasma:whole blood (2.0-2.1) {Ehresman,
2007, 1429928} and serum:whole blood (1.4-2.2) {Karrman, 2006, 2159543; Hanssen, 2013,
3859848} ratios previously reported. In blood samples obtained from three highly exposed
Canadian subjects, the highest levels of PFOA were measured in plasma (0.27 ng/mL) compared
to red blood cells (RBCs, 0.13 ng/mL) and washed RBCs (0.12 ng/mL). The authors suggest that
these values could be used as more accurate conversion factors to convert 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 PFOA concentrations in plasma, serum, and whole blood were 1.90, 1.60 and
0.93 ng/mL, respectively. Similar to other studies, PFOA preferentially accumulated in plasma
relative to other blood fractions and also suggest that the common practice of multiplying by a
factor of 2 to convert the concentrations in whole blood to serum will not provide accurate
estimates for PFOA.
In another study {De Toni, 2020, 6316907} in which blood from healthy low-exposed donors
was exposed to PFOA, platelets were identified as the preferential site of PFOA accumulation.
The concentrations observed among blood cell components were below the limit of
quantification (LOQ) in erythrocytes, 6.2 ± 0.4 pg/106 cells in leukocytes, and
243.9 ± 122.6 pg/106 cells in platelets. The authors also incubated platelets with Merocyanine
540, a fluorescent dye that has been used as a marker of membrane fluidity. Fluorescence
intensity increased in a dose-dependent manner up to, but not beyond, 400 ng/mL. The authors
suggest these findings support an association between PFOA accumulation and increased
membrane fluidity.
B-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.2.3.1.2 Distribution in Tissues
No clinical studies are available that examined tissue distribution in humans following
administration of a controlled dose of PFOA. However, samples collected in biomonitoring and
epidemiological studies provide data showing distribution of PFOA.
Pirali et al. (2009, 757881) measured intrathyroidal PFOA levels (0.4-6.0 ng/g) in thyroid
surgical patients and found no correlation between serum and thyroid PFOA concentrations.
PFOA has been detected in breast milk samples {Tao, 2008, 1290895; Volkel, 2008, 3103448},
cord blood samples {Apelberg, 2007, 1290833; Monroy, 2008, 2349575}, and follicular fluid
samples {Kang, 2020, 6356899} at concentrations above the LOQ. These studies indicate that
PFOA is distributed within the body, including reproductive tissues.
PFOA concentrations above the LOQ were detected in 5 of 6 postmortem liver samples from
males in Catalonia, Spain. In females, only 1 of 6 liver samples was above LOQ of 0.77 ng/g
{Karrman, 2010, 2732071}. 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. PFOA was present in 45% of the samples but could be
quantified in only 20% (median 1.9 ng/g). PFOA accumulated primarily in the bone (60.2 ng/g),
lung (29.2 ng/g), liver (13.6 ng/g), and kidney (2.0 ng/g), with levels below LOD (2.4 ng/g) in
the brain.
Two studies examined accumulation of PFOA in cerebrospinal fluid and serum {Fujii, 2015,
2816710; Wang, 2018, 5080654}. In both studies, PFOA levels in cerebrospinal fluid were two
orders of magnitude lower than in the serum. These results indicate that PFOA does not easily
cross the adult blood-brain barrier (BBB).
B.2.3.2 Animal Studies
Studies of tissue distributions are available for several species including non-human primates,
rats, and mice. Experiments in non-human primates provide evidence of serum and liver
accumulation of PFOA. While only a few studies exist, they document distribution with repeated
measurements over long periods of time and include recovery time after exposure termination.
Mouse studies demonstrate that PFOA primarily distributes to serum, liver, lungs, and kidney;
however, several of these studies detect PFOA in additional organs and tissues. These tissues
include the central nervous system, cardiovascular, gastrointestinal, renal, reproductive,
endocrine, and musculoskeletal systems. Recent studies have also indicated that a moderate
amount of PFOA enters bone and even crosses the barriers into the central nervous system.
Adipose tissue was observed as a site that contained very little amounts of PFOA accumulation.
These data are characterized based on dosing (low, medium, and high), time exposed (acute vs.
chronic), and any sex differences between males and females. Ranges of dose regimens indicate
changes in deposition patterns as animals are exposed to increased concentrations of PFOA,
indicating possible changes in excretion through bile and urine. Several studies corroborate to
show that there are sex-specific deposition patterns, primarily that male animals accumulate
more PFOA in serum and some tissues including liver. Overall, these studies provide a wide
range of deposition data that can illustrate short- and long-term accumulation of PFOA in animal
tissues.
B-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.2.3.2.1 Non-Human Primates
One of the few studies in cynomolgus monkey that measured distribution of PFOA was
performed by Butenhoff et al. (2002, 1276161; 2004, 3749227). The study followed four to six
male monkeys that received PFOA (0, 3, 10, or 20 mg/kg) daily via oral capsule. Serum, urine,
and fecal samples were collected at 2-week intervals and liver samples were collected at
necropsy. Steady-state concentrations of PFOA in serum were 77 ± 39, 86 ± 33, and
158 ± 100 |ig/mL after 6 weeks and 81 ± 40, 99 ± 50, and 156 ± 103 |ig/mL after 6 months for
the 3-, 10-, and 20-mg/kg dose groups, respectively {Butenhoff, 2002, 1276161; Butenhoff,
2004, 3749227}. The mean serum concentration of PFOA in control monkeys was 0.134-
0.203 |ig/mL, Urine PFOA concentrations reached steady state after 4 weeks and were 53 ± 25,
166 ± 83, and 181 ± 100 |ig/mL in the 3,10, and 20-mg/kg dose groups, respectively, for the
duration of the study. Liver PFOA concentrations at necropsy in the 3-mg/kg and 10-mg/kg dose
groups were similar and ranged from 6.29-21.9 jug/g, while concentrations in two monkeys
exposed to 20 mg/kg were 16.0 and 83.3 jug/g. Liver PFOA concentrations in two monkeys
dosed with 10 mg/kg/day at the end of a 13-week recovery period were 0.08 and 0.15 |ig/g
{Butenhoff, 2004, 3749227}.
B.2.3.2.2 Rats
Numerous studies have been performed on PFOA distribution in rats. These studies range from
acute (hours) to chronic (2 years) and include various levels of dosing. Previous studies have
indicated that humans and rats have similar serum albumin binding, suggesting circulation of
PFOA in the body would be similar {Harkness, 1983, 9641985; Saladin, 2004, 9642161}.
In adult male Sprague-Dawley rats, animals were exposed by gavage to PFOA (20 mg/kg/day)
for 1, 3, or 5 days {Martin, 2007, 758419}. While serum data was only presented for 3-day
exposure animals, it is clear that serum levels had a moderate accumulation of 245 ±41 [j,g/mL.
Additionally, liver concentrations were 92 ± 6, 250 ± 32, and 243 ± 23 j_ig/g after 1, 3, and 5
daily doses, respectively. Liver accumulation appeared to reach its peak by day 3 and remained
steady at this level through day 5. While limited serum levels were presented, data indicates that
at day 3, serum and liver levels were in a 1:1 ratio.
Several studies indicate that the major target organs of PFOA accumulation are liver, kidneys,
and lungs with a large amount of PDOA remaining in blood serum. In an earlier study of PFOA,
Ylinen et al. (1990, 5085631) administered male and female Wistar rats doses of 3, 10, and
30 mg/kg/day PFOA via gavage for 28 days. At necropsy, serum, brain, liver, kidney, lung,
spleen, ovary, testis, and adipose tissue were collected (Table B-4).
Interestingly, measurements of PFOA from adipose tissue resulted in no detectable levels at any
dose or timepoint. For the 3 mg/kg/day dose group, male rats exhibited the highest concentration
of PFOA in their serum followed by, liver, kidneys and then lungs with notable accumulation in
testis. In higher doses of 10 and 30 mg/kg/day, male rats had a significant increase in kidney
PFOA concentration. The levels of PFOA in male rat serum were generally lower in the
30 mg/kg/day dose group than in the 10 mg/kg/day dose group, presumably due to increased
urinary elimination in the 30 mg/kg/day group as a result of saturation of PFOA binding sites in
serum. The PFOA tissue levels were otherwise similar for the 10 and 30 mg/kg/day dose groups
of male rats. In comparison, female rats exhibited much lower serum concentrations that the
B-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
males; the female serum PFOA concentrations were approximately 5-27% of the male
concentrations.
Lower PFOA concentrations were also seen in the female rats' solid tissues as liver and kidney
measurements were -10% and 30% of the concentrations detected in males, respectively. In
females, there was a dose-related increase in tissue and serum PFOA concentrations.
Concentrations of PFOA for female rats at the low dose were highest in serum, followed by liver,
lungs, and spleen. At the higher doses of 10 and 30 mg/kg/day, the highest PFOA concentrations
were found in the serum and kidney, a pattern also observed in male rats.
Table B-4. Tissue Distribution of PFOA in Wistar Rats After Exposure via Gavage for
28 Days as Reported by Ylinen et al. (1990, 5085631)
Tissue3
Males
Females
3 mg/kg/day
10 mg/kg/day
30 mg/kg/day
3 mg/kg/day
10 mg/kg/day
30 mg/kg/day
Serum (|ig/mL)
48.60 ± 10.30
87.27 ± 20.09
51.65 ± 11.47
2.40b
12.47 ±4.07
13.92 ±6.06
Liver (|ig/g)
39.90 ±7.25
51.71 ± 11.18
49.77 ± 10.76
1.81 ±0.49
3.45 ± 1.36
6.64 ±2.64
Kidney (|ig/g)
1.55 ±0.71
40.56 ± 14.94
39.81 ± 17.67
0.06 ± 0.02
7.36 ±3.19
12.54 ±8.24
Spleen (|ig/g)
4.75 ± 1.66
7.59 ±3.50
4.10 ± 1.57
0.15 ±0.04
0.38 ±0.17
1.59 ±0.49
Lung (ng/g)
2.95 ±0.54
22.58 ±4.59
23.71 ±5.42
0.24b
0.22 ±0.15
0.75 ±0.26
Brain (|ig/g)
0.398 ±0.144
1.464 ±0.211
0.710 ±0.320
< LOQ°
0.029 ±0.019
0.044 ±0.018
Ovary (|ig/g)
-
-
-
< LOQ
0.41 ±0.27
1.16 ±0.58
Testis (|xg/g)
6.24 ± 2.04
9.35 ±4.02
7.22 ±3.17
-
-
-
Notes: LOQ = limit of quantification.
a Data are presented as mean ± standard deviation (n = 6).
b Data are presented as the mean (n = 3).
cLOQ = 1 ng/mL.
Kawabata et al. (2017, 3858489) measured PFOA in the tissues of male Wistar rats (including
brains) 9 days after administration of a single dose of 50 mg/kg. Serum PFOA concentrations
were 33.3 |ag/mL and liver concentrations were 58.7 jug/g. However, PFOA levels in brain were
below the limit of detection (i.e., <0.8 jug/g). Although levels are low and detection is variable,
these studies do support PFOA accumulation to low levels in brains of adult rats.
PFOA distribution followed a similar pattern in Sprague-Dawley rats administered a single 14C-
PFOA dose via oral gavage to male (Table B-5) and female (Table B-6) rats {Kemper, 2003,
6302380}. Tissues from male rats were collected at 10.5 hours (Tmax) and 171 hours (Tmax/2)
(time to return to 50% maximum plasma concentration) after dosing. Tissues from female rats
were collected at 1.25 hours (Tmax) and 4 hours (Tmax 2) after dosing. Liver, blood, skin, muscle,
bone, gastrointestinal tract, and adipose were the primary tissues for distribution of 14C-PFOA. In
males, the fraction of dose found in the liver increased between Tmax and Tmax/2 but remained
constant or decreased in other tissues. In females, the fraction of the dose present in all tissues
remained constant or decreased between Tmax and Tmax/2. Livenblood ratios for 14C-PFOA at Tmax
in males were approximately 1:1 but increased between Tmax and Tmax/2. In females, the
livenblood ratio was -1.2:1 at the low dose but increased to -1.5 at higher doses. In males, the
PFOA blood concentration was tenfold or higher than the kidney concentration at Tmax and
declined slightly at Tmax 2, In the female tissues at Tmax/2, -30% of the dosed PFOA retained was
B-12
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
present in the liver, blood, kidney, muscle, and skin tissues in decreasing amounts. This study
confirmed sex-specific differences in PFOA distribution and identified accumulation in
reproductive tissues including testes and ovaries.
B-13
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table B-5. Distribution of PFOA in Male Sprague-Dawley Rats After a Single Oral Gavage Dosea as Reported by Kemper et
al. (2003, 6302380)
Tissue
1 mg/kg
5 mg/kg
25 mg/kg
'All ilt Tmax
'All ilt Tmax/2
'Al ilt Tmax
'All il t Tmax/2
'All ilt Tmax
°Ao il t Tmax/2
Prostate
0.083 ±0.039
0.030 ± 0.002
0.071 ±0.045
0.057 ±0.020
0.067 ±0.018
0.028 ±0.012
Skinb
14.772 ±2.135
6.061 ±0.274
15.565 ±0.899
7.233 ±0.430
13.836 ±0.969
5.419 ±0.237
Bloodb
22.148 ±0.692
8.232 ± 1.218
24.919 ± 1.942
11.140 ±0.624
22.905 ± 1.177
7.904 ± 1.032
Brain
0.071 ±0.018
0.022 ± 0.002
0.051 ±0.021
0.023 ±0.008
0.063 ± 0.007
0.019 ±0.002
Fatb
2.281 ±0.467
0.593 ±0.136
2.815 ±0.225
0.916 ±0.205
2.153 ±0.430
0.628 ±0.110
Heart
0.451 ±0.119
0.195 ±0.024
0.443 ±0.037
0.252 ±0.030
0.461 ±0.053
0.164 ±0.032
Lungs
0.740 ±0.147
0.341 ±0.043
0.593 ±0.376
0.344 ±0.194
0.863 ±0.103
0.303 ±0.057
Spleen
0.086 ±0.011
0.045 ± 0.006
0.096 ±0.017
0.060 ± 0.007
0.106 ±0.015
0.042 ± 0.005
Liver
21.708 ±5.627
32.627 ±3.601
18.750 ±2.434
25.231 ± 1.289
17.528 ±0.900
20.145 ±3.098
Kidney
1.949 ±0.402
1.140 ±0.215
2.170 ±0.354
1.212 ± 0.115
2.293 ±0.286
1.003 ±0.122
G.I. tract
2.930 ±0.929
0.980 ±0.300
2.508 ±0.713
1.052 ±0.202
2.784 ±0.608
0.808 ±0.189
G.I. contents
2.083 ± 0.625
0.239 ±0.025
2.632 ±0.934
0.270 ± 0.028
4.186 ± 1.349
0.210 ±0.084
Thyroid
0.008 ± 0.005
0.004 ± 0.003
0.011 ±0.006
0.004 ± 0.002
0.009 ± 0.002
0.005 ±0.001
Thymus
0.085 ± 0.008
0.051 ±0.018
0.085 ±0.012
0.053 ±0.003
0.120 ±0.025
0.045 ±0.010
Testes
0.755 ± 0.079
0.356 ±0.037
0.693 ±0.180
0.372 ±0.062
0.623 ±0.098
0.224 ±0.031
Adrenals
0.019 ±0.004
0.010 ±0.001
0.022 ± 0.004
0.009 ±0.001
0.026 ± 0.004
0.009 ±0.003
Muscleb
12.025 ± 0.648
4.984 ± 0.745
13.565 ±0.576
6.429 ±0.648
12.855 ±0.841
4.253 ±0.358
Boneb
3.273 ±0.538
1.120 ±0.094
3.047 ±0.544
1.375 ±0.169
3.062 ±0.438
0.906 ±0.100
Total0
85.465 ± 6.426
57.026 ±3.379
88.033 ± 1.420
56.031 ± 1.025
83.937 ±3.680
42.112 ±4.740
Notes: G.I. = gastrointestinal; Tmax = time to reach maximum plasma concentration; Tmax/2 = time to return to 50% maximum plasma concentration.
a Data are presented as mean percent of dose ± standard deviation recovered at Tmax and Tmax/2 in tissues.
b Percent recovery scaled to whole animal assuming the following: skin = 19%, whole blood = 7.4%), fat = 7%o, muscle = 40.4%, bone = 7.3%o of body weight.
c Totals are calculated from individual animal data.
B-14
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table B-6. Distribution of PFOA in Female Sprague-Dawley Rats After a Single Oral Gavage Dosea as Reported by Kemper et
al. (2003, 6302380)
1 mg/kg 5 mg/kg 25 mg/kg
Tissue
% at Tmax % at Tmax/2 % at Tmax % at Tmax/2 % at Tmax % at Tmax/2
Skinb
0.434 ±0.162
0.403 ± 0.096
0.624 ±0.142
0.307 ±0.121
0.380 ±0.166
0.415 ±0.175
Bloodb
5.740 ± 1.507
4.438 ± 1.625
8.089 ± 2.080
5.411 ± 1.466
7.158 ±2.232
6.407 ± 1.406
Brain
0.037 ±0.009
0.047 ± 0.008
0.066 ±0.019
0.045 ±0.010
0.058 ±0.008
0.058 ±0.018
Fatb
0.134 ±0.032
0.164 ±0.079
0.220 ±0.111
0.110 ±0.069
0.147 ±0.053
0.148 ±0.065
Heart
0.198 ±0.079
0.253 ±0.055
0.388 ±0.057
0.236 ±0.051
0.317 ±0.035
0.287 ± 0.069
Lungs
0.454 ±0.148
0.546 ± 0.082
0.827 ±0.102
0.570 ±0.179
0.678 ± 0.067
0.775 ± 0.204
Spleen
0.063 ± 0.027
0.058 ± 0.006
0.101 ±0.021
0.060 ±0.012
0.091 ±0.007
0.070 ± 0.002
Liver
7.060 ± 1.266
6.817 ± 1.537
11.190 ± 2.192
7.176 ±0.982
10.538 ± 1.723
9.080 ±0.895
Kidney
3.288 ±0.948
2.769 ± 0.784
4.293 ±0.771
2.685 ± 0.736
5.867 ±0.946
4.749 ±0.393
G.I. tract
10.699 ± 9.066
8.462 ±6.519
7.142 ±2.594
8.255 ± 8.967
6.923 ± 1.846
3.547 ± 1.306
G.I. contents
21.956 ± 13.48
3.891 ±2.395
2.896 ±2.305
5.601 ±6.165
2.491 ± 1.548
1.121 ± 1.010
Thyroid
0.010 ±0.003
0.016 ±0.021
0.008 ± 0.002
0.006 ± 0.002
0.009 ±0.003
0.007 ± 0.002
Thymus
0.052 ±0.017
0.058 ± 0.024
0.105 ±0.030
0.068 ±0.021
0.091 ±0.032
0.077 ± 0.020
Ovaries
0.047 ±0.019
0.048 ± 0.006
0.071 ±0.012
0.041 ±0.012
0.071 ±0.012
0.070 ±0.012
Adrenals
0.014 ±0.005
0.018 ±0.004
0.026 ± 0.005
0.015 ±0.004
0.031 ±0.005
0.021 ±0.001
Muscleb
0.170 ±0.051
0.258 ±0.089
0.325 ±0.010
0.229 ±0.031
0.441 ±0.116
0.304 ±0.099
Uterus
0.243 ±0.091
0.374 ± 0.247
0.354 ±0.046
0.247 ± 0.068
0.358 ±0.124
0.365 ± 0.029
Boneb
0.101 ±0.017
0.153 ±0.052
0.174 ±0.057
0.142 ±0.078
0.157 ±0.072
0.181 ±0.090
Total0
50.698 ± 16.485
28.772 ± 10.976
36.897 ±3.187
31.201 ± 12.63
35.803 ±2.554
27.680 ±2.569
Notes: G.I. = gastrointestinal; Tmax = time to reach maximum plasma concentration; Tmax/2 = time to return to 50% maximum plasma concentration.
a Data are presented as mean percent of dose ± standard deviation recovered at Tmax and Tmax/2 in tissues.
b Percent recovery scaled to whole animal assuming the following: skin = 19%, whole blood = 7.4%), fat = 7%o, muscle = 40.4%, bone = 7.3%o of body weight.
c Totals are calculated from individual animal data.
B-15
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Sex dependent dose distribution similar to results found in Ylinen et al. (1990, 5085631) have
also been found in several other reports {Kemper, 2003, 6302380; Lau, 2006, 1276159}.
According to Kemper (2003, 6302380), plasma concentration occurred ten times faster and at
much lower levels in females when compared to males. Lau et al. (2006, 1276159) dosed male
and female Sprague-Dawley rats with 10 mg/kg for 20 days and necropsied them 24 hours after
the last dose. Male rats had serum PFOA levels of 111 |ig/mL compared to 0.69 |ig/mL in
female rats, a sex ratio that was in line with the Kemper et al. results.
Kemper (2003, 6302380) observed levels of PFOA accumulation in the kidneys of females that
were consistently elevated compared to males, indicating that excretion of PFOA may play a role
in the sex differences of PFOA distribution. The results suggest females absorb and excrete
PFOA more rapidly than males. This study also confirmed PFOA can accumulate in reproductive
organs (testes) and observed PFOA accumulation in endocrine (thyroid, adrenals) and immune
(thymus) tissues.
Furthermore, at Tmax/2 there was only -1% of the dosed 14C-PFOA in the gastrointestinal tissues
and contents in males, compared to -14% in females. However, samples were collected at 1.25
and 4 hours in females and 10.5 and 171 hours in males (the timing was based on previous
toxicokinetic experiments determining the Tmax and Tmax/2), thus providing more time for
absorption in the males {Kemper, 2003, 6302380}.
Two NTP studies exemplify sex-specific patterns of PFOA accumulation in blood and liver.
PFOA levels were measured in the context of a both a 28-day toxicity study {NTP, 2019,
5400977} and a two-year chronic toxicity study {NTP, 2020, 7330145}. In the 28-day study
{NTP, 2019, 5400977}, male and female Sprague-Dawley rats were administered 0 to
10 mg/kg/day (males) or 0 to 100 mg/kg/day (females) of PFOA by oral gavage. Although females
were administered a 10-fold higher dose of PFOA, males exhibited higher plasma concentrations
than females across all dose groups. The plasma concentrations in males were 50.7 ± 2.2 and
148.6 ± 15.4 |ig/mL at the lowest and highest dose groups respectively. In females, plasma
concentrations were 0.4905 ± 0.072.1 and 23.444 ± 3.247 |ig/mL at the lowest highest dose
groups respectively. When normalized to dose administered (|iM/mmol/kg), males had a 1,000-
fold higher level than females at the lowest dose and a 63-fold higher level at the highest dose.
Males exhibited a decreasing normalized plasma concentration with dose, whereas females
exhibited an increasing normalized plasma concentration with dose. PFOA in liver was only
measured in males, and the liver:plasma ratios were fairly consistent across dose groups, ranging
from 0.87 to 1.17.
In the two-year study {NTP, 2020, 7330145}, Sprague-Dawley rats were exposed to 0, 150, or
300 ppm PFOA during the perinatal periods. During the postweaning period, first generation (Fi)
male rats were provided 0, 150, or 300 ppm and Fi female rats were provided 0, 300, or
1,000 ppm PFOA via feed. Plasma and liver PFOA levels were measured at the 16-week
interval. Plasma and liver PFOA concentrations in males were within 10% of each other
regardless of whether animals were also dosed during the perinatal period. Plasma concentrations
in females showed a similar pattern to the males (e.g., minor differences between perinatal
exposures and liver patterns). Although exposures in females were 2-3 times higher than in
males, PFOA plasma concentrations were much lower compared to males. For example, at the
highest dose in rats exposed during both perinatal and postweaning periods, plasma
concentrations were 223.4 ± 8.4 |ig/mL in males compared to 70.2 ± 6.9 |ig/mL in females. The
B-16
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
liver:plasma ratios were again fairly consistent across dose groups, ranging from 0.73 to 0.88 in
males and from 0.81 to 0.99 in females.
In a repeated inhalation exposure study, Hinderliter et al. (2006, 135732) exposed male and
female rats to 0, 1, 10, or 25 mg/m3 aerosol concentrations of PFOA for 6 hours/day,
5 days/week for 3 weeks. Blood was collected immediately before and after the daily exposure
period 3 days/week. The aerosols had mass median aerodynamic diameters of 1.3-1.9 |im with
geometric standard deviations (GSDs) of 1.5-2.1. PFOA plasma concentrations were
proportional to the inhalation exposure concentrations, and repeated exposures produced little
plasma carryover in females, but significant day-to-day carryover in males. By 3 weeks, males
reached steady-state plasma levels of 8, 21, and 36 jag/m L for the 1,10, and 25 mg/m3 groups,
respectively. In females, the postexposure plasma levels were 1, 2, and 4 jag/m L for the 1,10,
and 25 mg/m3 groups, respectively. When measured immediately before the next daily exposure,
plasma levels had returned to baseline in females, demonstrating clearance within 24 hours of
each daily dose.
B.2.3.2.3 Mice
Measurements of serum PFOA concentrations in mice have differed from results in rat studies.
Lau and colleagues (2006, 1276159) dosed male and female CD-I mice with 20 mg/kg/day of
PFOA for 7 or 17 days and analyzed serum levels. After 7 days, male mice had serum PFOA
levels of 181 jag/m L and females had levels of 178 |ig/mL. After 17 days of treatment, male
mice had serum PFOA levels of 199 |ig/mL and females had levels of 171 |ig/mL {Lau, 2006,
1276159}. Additionally, in a separate experiment performed by Lou et al. (2009, 2919359)
female CD-I mice were dosed with 20 mg/kg/day for 17 days {Lou, 2009, 2919359}. Serum
samples were collected 24 hours after the final dose and analyzed for PFOA. The mean serum
concentration was 130 ± 23 mg/L, which is comparable to the reported value of 171 |ig/mL
reported above by Lau et al. (2006, 1276159). These data suggest that the sex difference
observed by Lau et al. (2006, 1276159) in rats was not seen in the mice under the conditions of
this study.
Lou et al. (2009, 2919359) measured pharmacokinetics of PFOA in mice administered single
doses of 1 and 10 mg/kg to groups of male and female CD-I mice. Plasma, liver, and kidney
tissues were collected at multiple early time points (4, 8, 12, and 24 hours) as well as a dozen
time points between 3 and 80 days. In female mice, peak serum concentrations were measured at
10 and 100 mg/L and declined to 2 mg/L and < 0.2 mg/L after 80 days for the 1 and
10 mg/kg/day doses, respectively. Peak serum concentrations were slightly lower in the males at
~8 and 80 mg/L, but final serum concentrations were higher in the males at -0.5 and 8 mg/L,
respectively. Liver and kidney concentrations also were higher in males than in females for each
of the two doses. These data suggest a longer half-life in males than in females. Additionally,
this group dosed 60 mg/kg to female mice and measured serum levels over the course of 28 days.
Based on their findings, these mice were able to clear a higher dose of PFOA much more quickly
than animals who had received a 1 or 10 mg/kg dose {Lou, 2009, 2919359}. The 60 mg/kg dose
animals were able to return to a 0.4 mg/L serum concentration in about 28 days while the
10 mg/kg and 1 mg/kg groups took 61 days and 70 days to reach 1 mg/L, respectively.
Several studies of short-term distribution of PFOA in mice have been published that vary
between 4 hours and 28 days and demonstrate the range of PFOA tissue distribution. One of the
B-17
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
earliest of these time points was performed by Burkemper et al. (2017, 3858622) who used a
radioisotope injection (18F-PFOA) and measured deposition in 14 different tissues as well as
serum 4 hours later. Despite the observation that radiolabel was associated with -29% of serum
protein, the majority of signal was found in the bone (femur), liver, and lungs. The next highest
levels of radioisotope detection were in the heart, spleen, large intestines, and then kidneys.
These findings were consistent with recent work by Bogdanska et al. (2020, 6315801). Using a
14C-PFOA radioisotope, authors measured low (0.06 mg/kg/day) and high dose (22 mg/kg/day)
PFOA delivered via feed to C57B1/6 mice and collected measurements at 1, 3 and 5 days
postexposure (Table B-7). Similar to previous finding of the Burkemper paper, liver
accumulation was consistently 4-5 times greater than what was found in serum at all doses and
time points. Lung deposition was also found to be at elevated levels and was measured at nearly
half serum concentrations at all doses and time points. In a study by Li et al. (2017, 4238518)
conducted in BALB/c mice after a 28-day exposure, PFOA concentrations in both liver and
serum increased with PFOA dose in mice, with PFOA concentrations being generally higher in
the liver than the serum.
Table B-7. Distribution of PFOA in Male C57BL/6 Mice Following Exposure to 14C-PFOA
for 1, 3, or 5 days in Feed" as Reported by Bogdanska et al. (2020, 6315801)
0.06 mg/kg/day
22 mg/kg/day
Tissue
Dose Duration
Dose Duration
1 Day
3 Days
5 Days
1 Day
3 Days
5 Days
Blood
0.328
1.222
1.645
90
183
192
Liver
1.59
5.229
7.507
281
671
756
Lung
0.179
0.606
0.873
40
96
110
Kidney
0.16
0.556
0.783
42
91
104
Pancreas
0.087
0.258
0.344
22
51
61
Thyroid gland
0.082
0.294
0.421
24
48
57
Skin
0.096
0.337
0.501
25
47
52
Stomach
0.125
0.259
0.345
14
45
48
Thymus
0.089
0.197
0.237
16
34
47
Inguinal fat pad
0.064
0.209
0.273
15
37
40
Whole bone
0.105
0.282
0.452
20
30
40
Small intestine
0.057
0.174
0.269
10
37
36
Large intestine
0.05
0.166
0.204
10
32
32
Testis
0.054
0.156
0.235
12
28
29
Epididymal fat
0.053
0.152
0.153
12
23
24
Muscle
0.032
0.116
0.169
9
19
20
Brain
0.008
0.029
0.024
2
3
4
Spleen
0.022
< LOD
< LOD
< LOD
5
1
Heart
< LOD
< LOD
< LOD
14
15
< LOD
Notes: LOD = limit of detection.
a Data are presented as mean (nmol/g).
B-18
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Interestingly, while Burkemper et al. (2017, 3858622) measured equal levels of kidney and large
intestine depositions at very early time points (4 hours), Bogdanska et al. (2020, 6315801)
registered a far greater amount of PFOA in the kidneys at the slightly later time points or 1, 3,
and 5 days. This may indicate a change in excretion methods over the course of exposure and/or
reflect differential distribution or detection of 18F-PFOA relative to 14C-PFOA. Burkemper et al.
also measured a large uptake of 18F-PFOA in mouse femurs at 4 hours, while Bogdanska et al.
found moderately low levels at later time points. This difference could be due to rapid fluorine
intake of the bone by potential 18F radioisotope artifacts.
Bogdanska et al. (2020, 6315801) also observed accumulation of PFOA in testes of C57BL/6
mice at levels similar to those observed in epididymal fat and in intestines. In BALB/c mice
exposed to PFOA (0.31 to 20 mg/kg/d) for 28 days, PFOA levels in the testes increased with
increasing dose {Zhang, 2014, 2850230}. Further evidence of distribution to reproductive tissues
in male mice comes from the finding that PFOA accumulated in the epididymis of BALB/c mice
in a dose-dependent manner {Lu, 2016, 3981459}.
Accumulation in both small intestine and the colon was observed in CD-I mice administered
between 1 and 20 mg/kg/day for 10 days {Rashid, 2020, 6833711}. Higher levels of PFOA were
measured in the small intestine relative to colon. The mean concentration of PFOA in small
intestine detected was 1.0, 2.3, 4.4, and 6.5 |ig/g in the 1, 5, 10, and 20 mg/kg/day groups,
respectively. Dose-dependent accumulation was also seen in the colon, where mean
concentrations ranged from 211.12 to 1,834.27 ng/g in colon tissue.
Fujii et al. (2015, 2816710) performed IV injections of 0.313 |imol/kg of PFOA on male and
female animals and collected serum and organ samples after 24 hours. Distribution was
calculated as percentage of total recovered dose from serum and organs. The majority of
administered PFOA was retained in the serum and liver of mice and less than 2% of administered
dose was found in kidney and adipose tissue. While a relatively small amount of PFOA was
measured in the brain (0.1%), it is noteworthy that PFOA can cross the BBB in healthy animals.
Similar findings were observed in both the Burkemper et al., (2017, 3858622) and Bogdanska et
al. (2020, 6315801) studies. Levels in female mouse livers were -30% of the levels measured in
male samples. A larger portion of PFOA was not recovered from serum, organ, and excretions of
female mice, indicating that there may be further distribution in organs that were not examined in
this study. Fujii and colleagues (2015, 2816710) examined distribution based on chain length.
They observed that perfluoroalkyl carboxylic acids (PFCAs) with shorter chain length (C6 and
C7) were excreted rapidly through urine, while longer chains (> C8) accumulated in the liver.
Moreover, PFCA with longer chain lengths were found to exhibit increasing affinity for serum
and liver fatty acid binding proteins. The authors suggest that lipophilicity driven by chain length
may account for the distribution patterns of PFCA, which is consistent with the high levels of
PFOA accumulation in serum and liver. These large sequestration volumes of PFOA observed in
the liver seem to be attributable to the liver's large binding capacity in mice.
Studies that examined PFOA distribution for longer time periods also reveal that the liver is a
primary site of PFOA accumulation. Adult male BALB/c mice exposed to PFOA (0.4, 2, and
10 mg/kg/day) via oral gavage for 28 days exhibited dose dependent increase in both serum and
liver {Guo, 2019, 5080372}. At every dose tested, livenserum ratios appeared to stay near 2:1.
Additionally, it was found that the liver consistently absorbed 10% of the total PFOA each
animal was exposed to. In a study with the same 28-day exposure and similar low dose
B-19
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(1.25 mg/kg/day via oral gavage), Zheng and colleagues found that PFOA distributed in the liver
and serum in an -2.5:1 ratio {Zheng, 2017, 4238507}. These findings further corroborate the
previous radioisotope studies that PFOA accumulates primarily within the liver and secondarily
in serum.
One potential method of removal of PFOA from liver is through activation of PPARa. In human
and rodent hepatocytes, PPARa activation induces expression of genes involved in lipid
metabolism and cholesterol homeostasis. PFOS and PFOA structurally resemble fatty acids and
are well-established ligands of PPARa in the rat and mouse liver. As PPARa agonists, PFOS and
PFOA can induce the B-oxidation of fatty acids, induce fatty acid transport across the
mitochondrial membrane, decrease hepatic very low-density lipoprotein-triglyceride and
apolipoprotein B (apoB) production, and promote lipolysis of triglyceride-rich plasma
lipoproteins {Fragki, 2021, 8442211}. In an experiment using male wild-type 129S4/SvlmJ mice
and PPARa-null 129S4/SvJae-PparatmlGonz/J mice, animals were orally administered 0, 12.5,
25, and 50 [j,mol/kg/day PFOA (~0, 5.4, 10.8, and 21.6 mg/kg/day PFOA, respectively) for four
weeks {Minata, 2010, 1937251}. Blood, liver, and bile were collected for determination of
PFOA concentration at the end of 4 weeks (Table B-8). The PFOA concentration in whole blood
and the liver were similar between wild-type and PPARa-null mice and increased in proportion
to dose. In bile, PFOA concentration in wild-type mice increased by a factor of 13.8 from 12.5 to
25 [j,mol/kg and by a factor of 2.8 from 25 to 50 [j,mol/kg; however, in bile of PPARa-null mice,
PFOA concentration increased by a factor of only 3.2 from 12.5 to 25 [j,mol/kg and by a factor of
6.1 from 25 to 50 [j,mol/kg. The liver can transport PFOA from hepatocytes to bile ducts that is
mediated at least partly by PPARa. The lower PFOA levels in bile of PPARa null mice suggest a
role for PPARa in PFOA clearance in the liver {Minata, 2010, 1937251}.
Table B-8. PFOA Concentrations in Wild-type and PPARa-null Male Mice Exposed to
PFOA by Gavage for Four Weeks" as Reported by Minata et al. (2010,1937251)
Dose
Whole Blood
Bile
Liver
(nmol/kg)
Wild-type
PPARa-null
Wild-type
PPARa-null
Wild-type
PPARa-null
0
ND
ND
ND
ND
ND
ND
12.5
20.6 ± 2.4a
19.3 ±2.2
56.8 ±26.9
19.6 ±2.2
181.2 ±6.3
172.3 ±8.9
25
46.9 ±3.2
36.4 ±2.7
784 ± 137.6
62.9 ± 16.7
198.8 ± 15.4
218.3 ± 14.5
50
64.2 ±6.5
71.2 ±8.0
2174 ±322.4
383 ± 109.9
211.6 ± 13.3
239.7 ±25.0
Notes: ND = not detected; PPARa-null = peroxisome proliferator-activated receptor alpha-null 129S4/SvJae-PparatmlGonz/J
mice; Wild-type = 129S4/SvlmJ mice.
aData are presented as mean ± standard deviation ((ig/mL).
B.2.3.3 Tissue Transporters
As described earlier, protein transporters from a number of families play a role in the tissue
uptake of orally ingested PFOA. The transporters are located at the interface between serum and
a variety of tissues (e.g., liver, kidneys, lungs, heart, brain, testes, ovaries, placenta, and uterus)
{Klaassen, 2010, 9641804}. The liver is an important uptake site for PFOA. OATPs and MRPs,
at least one OAT, and the sodium-taurocholate cotransporting polypeptide (NTCP)—a hepatic
bile uptake transporter—have been identified at the boundary of the liver at the portal blood
and/or the canalicular membranes within the liver {Kim, 2003, 9641809; Kusuhara, 2009,
9641810; Zair, 2007, 9641805}.
B-20
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Transporters responsible for PFOA 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 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, reduced folate carrier 1 (RFC-1), equilibrative nucleoside
transporter (ENT1), folate receptor alpha (FRa), heme carrier protein 1, 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 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.2.4 Distribution during Reproduction and Development
The availability of distribution data from pregnant females plus animal pups and neonates is a
strength of the PFOA pharmacokinetic database because it helps to identify those tissues
receiving the highest concentration of PFOA during development. For this reason, the
B-21
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
information on tissue levels during reproduction and development are presented separately from
those that are representative of other life stages.
B. 2.4.1 Human Studies
Zhang et al. (2013, 3859792) recruited 32 pregnant females (21-39 years) from Tianjin, China,
for a study to examine the distribution of PFOA between maternal blood, cord blood, the
placenta, and amniotic fluid. Samples were collected at time of delivery (35-37 weeks). The
study yielded 31 maternal whole blood samples, 30 cord blood samples, 29 amniotic fluid
samples, and 29 placentas. PFOA was found in all fluids/tissues sampled. Maternal blood
contained variable levels of 10 PFAS: 8 acids and 2 sulfonates. 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. PFOA was transferred to the amniotic fluid to a greater extent than PFOS based
on their relative proportions in the maternal blood and cord blood. Compared with mean PFOA
blood levels in the pregnant females, mean levels of PFOA in the cord blood, placenta, and
amniotic fluid were 47%, 59%, and 1.3%, respectively, of those in the mother's blood. The
correlation coefficients between the maternal PFOA blood levels and placenta, cord blood, and
amniotic fluid levels (0.7-0.9) were statistically significant (p < 0.001).
B.2.4.1.1 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. Chen et al. (2017, 3859806) examined distribution of
PFAS chemicals and their isomers 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 body mass index (BMI) of 20.4 and
gestational age of 38.9 weeks. PFOA isomers examined included nPFOA (linear PFOA), iso-
PFOA, 5m-PFOA, 4m-PFOA, 3m-PFOA, and tb-PFOA; however, the only isomers detected in
maternal serum, cord serum, and/or placenta were linear PFOA, iso-PFOA, and 3m-PFOA.
Linear PFOA contributed approximately 89% of cord serum PFOA and 91% of maternal serum
PFOA. Branched PFOA, including 3m-PFOA and iso-PFOA, contributed approximately 5% and
6%, respectively, of the total PFOA in cord serum and 5% and 5%, respectively of total PFOA in
maternal serum. Notably, the increased proportion of linear isomers was also observed in other
PFAS chemicals including PFOS and PFHxS. Similar findings have been reported in Cai et al.
(2020, 6318671) and Li et al. (2020, 6505874). The ratio of placental :maternal concentrations
(Rpm) for 3m-PFOA was greater than that for linear PFOA, suggesting that 3m-PFOA is
transferred more efficiently than linear PFOA.
Zhang et al. (2013, 3859792) recruited 32 female subjects (mean age of 30.9 years) from a
hospital in Tianjin, China, who reported full-term pregnancies (average gestation period of
B-22
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
40.3 weeks). The authors reported an average of 1.58 ng/g of PFOA in the placenta and
3.35 ng/mL in maternal serum (Table B-9). The Rpm for total PFOA was approximately 0.47,
which is higher than the proportion of total PFOA reported by Chen et al. (2017, 3859806). For
PFOA levels in maternal serum, Zhang et al. (2013, 3859792) reported significantly higher levels
which may have contributed to the increased PFOA accumulation. The fact that participants in
the Zhang et al. (2013, 3859792) study were further along in gestation than participants in the
Chen et al. (2017, 3859806) study may have contributed to their higher maternal PFOA levels.
Mamsen et al. (2019, 5080595) demonstrated that factors such as gestational age can affect
PFOA concentrations in maternal serum and placentas. Using a linear graph of normalized
percentage accumulation as a function of gestational age, the authors found that, for male and
female infant placentas, there was a steady increase in PFOA accumulation during gestation days
50 to 300, with male placentas showing higher levels of than female placentas. Authors
estimated a placenta PFOA accumulation rate of 0.11% per day during gestation.
In summary, the findings from these studies highlight four important points: 1) Linear PFOA is
detected at a higher frequency and at higher concentrations in maternal serum than branched
PFOA isomers; 2) branched and linear PFOA cross the placental barrier and are distributed in
different proportions within the placenta; 3) branched PFOA is more efficiently transferred into
the placenta than linear PFOA; and 4) PFOA concentrations within the placenta increase during
gestation from the first to third trimester.
Several studies have investigated distribution from mother to fetus through analysis of detected
PFAS chemicals in cord blood. 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 the time of delivery,
evaluated the correlation between PFAS levels in maternal serum and matched cord blood.
Maternal serum PFOA levels at 16 weeks of gestation and at time of delivery were 4.8 |ig/L and
3.3 |ig/L, respectively. Authors reported a positive correlation between maternal serum PFOA
concentration at 16 weeks of gestation and cord serum (correlation coefficient = 0.94). Similarly,
the correlation between maternal serum at the time of delivery and cord serum was also positive
(correlation coefficient = 0.88).
Porpora et al. (2013, 2150057) quantified PFOA 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 PFOA concentrations were 2.9 ng/g and 1.6 ng/g,
respectively. A strong positive correlation was observed between maternal and cord serum
concentrations (r = 0.70, p < 0.001). These values suggest a cord to maternal serum ratio of 0.55.
Fromme et al. (2010, 1290877) measured PFOA 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. PFOA was also measured in cord blood and in infant blood at 6 and
19 months after birth. Maternal PFOA serum concentrations ranged from 0.7 to 7.0 |ig/L (38
samples) and cord serum concentrations ranged from 0.5 to 4.2 |ig/L (33 samples). The cord to
maternal serum mean ratio was 0.7.
B-23
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Wang et al. (2019, 5083694) measured the levels of 10 PFAS chemicals in paired maternal and
umbilical cord serum from a prospective birth cohort (n = 369) in Shandong, China. The average
maternal and gestational ages of the participants were 28.4 years and 39.4 weeks, respectively.
PFOA was detected in all maternal and umbilical cord serum samples at a geometric mean of
39.27 ng/mL (range of 1.16-602.79 ng/mL) in maternal serum and 31.83 ng/mL (range 1.52-
291.56 ng/mL) in cord serum. Of the 10 PFAS chemicals measured, PFOA showed the highest
concentration in both maternal and cord serum (r = 0.908). Authors did not explain why PFOA
levels were high. Comparing the studies in Table B-9, geographic location could be a factor in
population exposure to a particular PFAS chemical. In the case of Shandong, China, PFOA
production may be a reason for the high PFOA levels in serum samples. Based on these studies,
cord blood PFOA levels is a biomarker for in utero exposure and provides further evidence that
PFOA readily accumulates in cord blood during gestation.
Study participants from various geographical locations, whether it be Ohio, USA {Kato, 2014,
2851230}, Rome, Italy {Porpora, 2013, 2150057}, Spain {Manzano-Salgado, 2015, 3448674},
France {Cariou, 2015, 3859840}, Faroes Islands, Denmark {Eryasa, 2019, 5412430}, Munich,
Germany {Fromme, 2010, 1290877}, Tianjin Tianjin, China {Zhang, 2013, 3859792}, or
Shandong, China {Wang, 2019, 5083694}, mostly show consistently higher levels of PFOA in
maternal serum versus cord serum regardless of gestational age. However, for studies with
participants of similar gestational ages, the PFOA concentrations in both maternal and cord
serum varied substantially across studies, reflected in RCM ratios that ranged from 0.57 to 1.33
(Table B-9). Factors such as exposure sources, parity, and other maternal demographics can
potentially account for the variations in maternal PFOA concentrations. For example, nulliparous
mothers generally have significantly higher serum PFOA than parous women {Kato, 2014,
2851230}. Conversely, younger women tend to have lower serum PFOA than older women
{Kato, 2014, 2851230}. Therefore, studies with high percentages of young, multiparous women
may report lower levels of PFOA in maternal and cord blood.
To understand the role of the placenta in facilitating the transport of PFAS compounds to the
fetal compartment during gestation, it is important to highlight the transplacental transfer
efficiency (TTE) and the factors that can potentially modulate in utero transport of PFAS. TTE is
a measure of a compound's ability to cross the placenta barrier and is often reported as the ratio
of cord blood to maternal blood concentrations (RCM). A summary of recent studies examining
RCM is presented in Table B-9. The percentages of maternal PFOA that accumulate in cord
blood ranged from 57% to 133% and did not strictly correlate to maternal serum values. This
variability suggests that TTE may differ across populations. For example, Manzano-Salgado et
al. (2015, 3448674) demonstrated that the percentage of maternal PFOA that accumulates in cord
blood tends to increase with maternal age.
Zhang et al. (2013, 3859792) calculated the RCM of 11 PFAS compounds in matched maternal-
cord blood from a population of 32 mothers in Tianjin, China, who delivered their infants at full
term. Authors noted an interesting trend where the highest RCM was reported for
perfluoroheptanoic acid (PFHpA) (C7) and a descending trend of RCM was observed with
increasing chain length from PFHpA (C7) to perfluorodecanoic acid (PFDA) (C10). There was
then an increasing trend in RCM with increasing chain length from PFDA (C10) to
perfluorododecanoic acid (PFDoDA) (C12), creating a "U" shaped curve where the RCM
decreases with increasing chain length until a certain threshold is reached and then the RCM
B-24
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
increases. The authors suggest that this non-linear relationship may be due to differential binding
affinities to maternal serum proteins and that high-affinity PFAS-serum protein interactions may
result in PFAS not being able to cross the placental barrier as efficiently through passive
diffusion. In line with most previous reports {Zhang, 2013, 3859792; Beesoon, 2011, 2050293;
Hanssen, 2010, 2919297; Lee, 2013, 3859850}, but not all {Gutzkow, 2012, 1290878; Kim,
2011, 1424975}, Wang et al. (2019, 5083694) reported that short-chain PFAS were transferred to
cord serum at higher efficiencies than longer-chain PFAS.
Branching also impacts TTE {Zhao, 2017, 5085130} with branched isomers transferring more
efficiently than their linear isomers. The authors observed a U-shaped trend of TTEs with
increasing carbon chain lengths as well as the position of the branching point. TTEs of branched
PFOA isomers (iso-, 5m-, and 4m-PFOA) were 0.71, 0.94, and 2.00, respectively compared to a
TTE of 0.56 for linear isomer (n-PFOA). Thus, higher efficiencies were observed as the
branching point moved closer to the carboxyl moiety of PFOA, which may be due to lower
affinities of branched PFOA isomers for HSA allowing for more efficient transfer to the fetus.
The efficiency of the placenta to modulate the transfer of xenobiotic varies during gestation. To
determine whether RCMs of PFAS in preterm placentas differed from full-term placentas, Li et
al. (2020, 6505874) assessed the RCMs of 32 PFAS chemicals in preterm and full-term
deliveries in the Maoming Birth Cohort in South China. The concentration of PFOA in maternal
blood from preterm subjects (mean =1.2 ng/mL) did not differ significantly from blood levels in
full-term subjects (mean = 1.34 ng/mL). However, the concentration of PFOA in preterm cord
blood (0.70 ng/mL) was significantly lower than full-term cord blood (1.25 ng/mL, p < 0.001).
Interestingly, the proportion of maternal PFOA in cord blood was 33% higher in full-term
pregnancies than in preterm pregnancies. Authors attributed the differences in RCM between
preterm and full-term deliveries to several factors, such as the difference in gestational age
between the two groups. Full-term deliveries have longer gestation periods which means longer
exposure duration. Second, the ability of the placenta to reduce toxin transfer reduces in the later
stages of pregnancy, making it easier for PFAS to diffuse into fetal circulation. Third, most
preterm pregnancies have impaired uteroplacental circulation, potentially reducing the amount of
PFAS entering fetal circulation. Finally, gene expression of RCM transporters varies during the
different stages of gestation, consequently affecting placenta barrier efficiency.
B-25
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-9. PFOA Concentrations in Human Cord Blood, Maternal Blood, and Transplacental Transfer Ratios (RCM)
Number of
Study Country, Cohort Maternal-Infant
Pairs"
Mean Gestational
Age (weeks)b
PFOA
Measurement
Cord Serum
(ng/mL)c
Maternal Serum
(ng/mL)c
Cord:Maternal
Serum Ratios
(RCM)d
Manzano-
Salgado et al.
(2015,
3448674)
53
NR
total PFOA
1.90
2.97
0.746
Sabadell and
Valencia, Spain
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.
Wuhan, China 32 38.9 ±1.6 total PFOA
1.237 ±0.577
1.56 ±0.611
0.808
(2017,
n-PFOA
0.947
1.15
0.842
3859806)
Iso-PFOA
0.067
0.053
1.267
3m-PFOA
0.08
0.06
0.587
Cariou et al.
Toulouse, France 89 NR total PFOA
0.919
1.22
0.78
(2015,
3859840)
Note: Concentrations represent mean values from 100 pairs. Semi-quantified values below LOD were taken into account for mean calculation.
Cai et al.
Maoming birth 424 39.3 ±1.1 total PFOA
0.85 ±0.52
1.21 ± 1.01
0.80
(2020,
cohort, China
6318671)
Note: Ratios were calculated from matched maternal and infant pairs for which all cord blood samples were > limit LOD. PFOA was detected
98.28% of samples PFOA.
Wang et al.
Shandong, China 369 39.4 ±1.3 total PFOA
31.83
39.27
0.83
(2019,
5083694)
Note: PFOA detected in 100% of maternal and cord samples.
Li et al. (2020,
Maoming Birth 86 33.8 ±3.0 total PFOA
0.7
1.2
0.57
6505874)
Cohort, China
(Pre-term births)
Maoming Birth 187 39.5 ±1.1 total PFOA
Cohort, China
1.25
1.34
0.85
(Full term births)
Li et al. (2020,
6506038)
Eryasa et al.
(2019,
5412430)
Note: 273 mother-infant pairs were analyzed, including 86 preterm deliveries and 187 full-term deliveries. Only PFAS quantifiable in > 50%
of maternal and cord sera were included in generating mean concentration values.
Beijing, China 86 39.0± 1.2 total PFOA 4^98 163 L33
Note: PFOA detection rate was 84.62% in maternal serum and 83.76% in cord serum. For PFOA, 86 of 117 matched cord and maternal serum
samples were used to generate RCM.
Faroese Birth
Cohort, Denmark
(cohort 3)
100
39.9 ± 1.3
total PFOA
1.97 (1.42-2.76) 2.33 (1.79-3.29)
0.82
B-26
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Country, Cohort
Number of
Maternal-Infant
Pairs"
Mean Gestational
Age (weeks)b
PFOA
Measurement
Cord Serum
(ng/mL)c
Maternal Serum
(ng/mL)c
Cord:Maternal
Serum Ratios
(RCM)d
51
39.7 ± 1.1
total PFOA
0.81 (0.56-1.26) 1.03 (0.75-1.41)
0.77
Faroese Birth
Cohort, Denmark
(cohort 5)
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).
Pan et al.
Wuhan, China
100
39.4 ± 1.3
total PFOA
1.42
2.19
0.65
(2017,
Note: Maternal blood collected in third trimester (38.4 ±1.6 weeks). PFOA was detected in 100% of maternal and cord samples.
3981900)
Zhao et al.
People's Hospital
63
39.3 ±0.82
n-PFOA
0.551
0.966
0.59
(2017,
of Hong'an
49
39.3 ±0.82
iso-PFOA
0.01
0.014
0.81
5085130)
County, China
36
39.3 ±0.82
5m-PFOA
0.003
0.003
1.7
7
39.3 ±0.82
4m-PFOA
0.001
0.001
2
63
39.3 ±0.82
total-PFOA
0.565
0.984
0.59
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 PFOA
1.1
1.8
0.61
(2011,
and Pregnancy
20
NR
n-PFOA
NR
NR
0.62
2050293)
(CHirP) cohort,
20
NR
Iso-PFOA
NR
NR
0.84
Vancouver,
4
NR
5m-PFOA
NR
NR
0.86
Canada
19
NR
4m-PFOA
NR
NR
0.64
18
NR
3m-PFOA
NR
NR
0.76
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).
Fei et al. (2007, Danish National
50
40.06 ± 1.57
total PFOA
3.7 ±4.7
5.6 ±2.5
0.55
1005775)
Birth Cohort,
maternal blood
obtained in first
trimester
Danish National
50
40.06 ± 1.57
total PFOA
3.7 ±4.7
4.5 ± 1.9
0.68
Birth Cohort,
maternal blood
obtained in second
trimester
B-27
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Number of .
„ , , -w- « . Mean Gestational PFOA
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)d
Note: Authors did not specify if matched maternal and cord blood samples were used to derive ratios.
Hanssen et al.
(2010,
2919297)
Johannesburg, 71 maternal serum, NR total PFOA
South Africa 58 cord blood
Note: Maternal and cord blood samples taken at time of delivery.
1.3
1.3
0.71
Fromme et al.
(2010,
1290877)
Munich, Germany 38 maternal and 33 NR total PFOA
cord serum
Note: Maternal and cord blood samples taken at time of delivery.
1.4
1.9
1.02
Kim et al.
(2011,
1424975)
Seoul and Gumi, 44 mothers, 43 39±1.6 total PFOA
South Korea infants
Note: Median serum concentrations reported. Values in parentheses are 25-75% IQRs.
1.15 (0.95-1.86)
1.46(1.15-1.91)
0.98
Needham et al.
(2011,
1312781)
Faroe Islands 12 NR total PFOA 3.1
Note: Serum concentrations reported as median values, RCMs reported as arithmetic means.
4.2
0.72
Liu et al.
(2011,
2919240)
Jinhu, China 50 (all) NR total PFOA
26 (male infants) NR total PFOA
24 (female infants NR total PFOA
Note: Maternal samples collected in the first weeks after delivery.
1.5
NR
NR
1.655
NR
NR
0.91
0.87
0.95
Midasch et al.
(2007,
1290901)
NR 11 NR total PFOA 3.4
Note: Serum concentrations reported as median values, RCMs reported as arithmetic means.
2.6
1.26
Verneretal. NA NA NA NA NA NA 0.78
(2015, Note: Authors developed a two-compartment, two-generation pharmacokinetic model of prenatal and postnatal exposure to PFOA and PFOS.
3299692) rCMs applied in model were derived from average of studies reported in Aylward et al. (2014, 2920555).
Notes: IQR = Interquartile Range; LOD = Level of Detection; NA = Not Applicable; NR = not reported.
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-28
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.2.4.1.2 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 protein binding capacities in the two matrices.
The authors reported a positive correlation between PFOA in amniotic fluid and maternal blood
(r = 0.621, p < 0.01) and cord blood (r = 0.664, p < 0.01), adding to the evidence that PFOA
levels in amniotic fluid is a potential biomarker for fetal exposure during pregnancy.
Table B-10 presents means or medians and ranges of measured and estimated PFOA
concentrations in maternal blood from recent studies (2013 to present) that also measured fetal
indicators of exposure (cord blood, placenta, and amniotic fluid). These studies demonstrate the
variability of PFOA accumulation in these tissues across geographic regions. Maternal serum
values ranged from 0.02 ng/mL in Rome, Italy {Porpora, 2013, 2150057} to 602.79 ng/mL in
Shandong, China {Wang, 2019, 5083694}. These same studies also showed the greatest range of
PFOA in cord blood (0.17-291.56 ng/mL). Fewer studies measured PFOA in placentas and
amniotic fluid. Placenta values ranged from
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-10. PFOA Concentrations in Human Maternal Blood, Cord Blood, Placenta and
Amniotic Fluid Across Studies
Study (Study Maternal Blood
Location)
Cord Blood
Placenta
Amniotic Fluid
Porpora et al. (2013, Maternal serum
Cord serum
NR
NR
2150057) Mean: 2.9 ng/g
Mean: 1.6 ng/g
(Italy) Median: 2.4 ng/g
Median: 1.6 ng/g
Range: 0.20-9.1 ng/g
Range: 0.17-5.0 ng/g
Zhang et al. (2014, NR
2850251) (Tianiin,
China)
NR
Mean: 1.58 ng/g
Median: 1.41 ng/g
Mean:
0.044 ng/mL
Median:
0.043 ng/Ml
Yang et al. (2016,
Maternal serum
Cord serum
NR
NR
3858535)
Mean: 1.64 ng/mL
Mean: 1.45 ng/mL
(Jiangsu, China)
SD: 1.11 ng/mL
SD: 1.14 ng/mL
Median: 1.24 ng/mL
Median: 1.03 ng/mL
Range: 0.34-5.30 ng/mL
Range: 0.16-6.77 ng/mL
Manzano-Salgado et Maternal plasma
Cord serum
NR
NR
al. (2015, 3448674)
Median: 2.85 ng/mL
Median: 1.90 ng/mL
(Sabadell and
Range: 0.78-11.93 ng/mL
Range: 0.60-
Valencia, Spain)
IQR: 1.87-6.00 ng/mL
10.56 ng/mL
IQR: 1.45-4.70 ng/mL
Maternal serum
Median: 2.97 ng/mL
Range: 0.86-14.54 ng/mL
IQR: 2.26-4.85 ng/mL
Zhang et al. (2013,
Mean: 3.35 ng/mL
1.95 ng/mL
Mean: 1.58 ng/g
Mean:
3859792) (Tianjin,
RSD: 1.03
RSD: 0.71
RSD: 0.54
0.044 ng/mL
China)
Range: 1.17-8.94 ng/mL
Range: 0.70-4.31 ng/mL
Range: 0.45-
RSD: 0.021
3.57 ng/g
Range: < LOQ-
0.145 ng/mL
Cariouetal. (2015,
Maternal serum
Cord serum
NR
NR
3859840)
Mean: 1.22 ng/mL
Mean: 0.919 ng/mL
(Toulouse, France)
Median: 1.045 ng/mL
Median: 0.860 ng/mL
Range: 0.309-7.31 ng/mL Range: 0.311-
7.06 ng/mL
Panetal. (2017,
Maternal Serum T1
Cord serum
NR
NR
3981900)
Mean: 3.15 ng/mL
Mean: 1.42 ng/mL
(Wuhan, China)30
Median: 3.24 ng/mL
Median: 1.41 ng/mL
IQR: 2.44-3.88 ng/mL
IQR: 1.14-1.84 ng/mL
T2 serum
NR
NR
Mean: 2.52 ng/mL
Median: 2.50 ng/mL
IQR: 2.05-3.13 ng/mL
T3 serum
NR
NR
Mean: 2.19 ng/mL
Median: 2.16 ng/mL
IQR: 1.81-2.73 ng/mL
Casertaetal. (2018,
Mean: 1.05 ng/mL
Mean: 0.98 ng/mL
NR
NR
4728855)
SD: 0.35 ng/mL
SD: 0.54 ng/mL
B-30
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study (Study
Location)
Maternal Blood
Cord Blood
Placenta
Amniotic Fluid
(Rome, Italy)
Range: 0.45-1.9 ng/mL
Range: 0.30-2.50 ng/mL
Wangetal. (2019,
Maternal serum
Cord serum
NR
NR
5083694)
GM: 39.27 ng/mL
GM: 31.83 ng/mL
(Shandong, China)
Median: 42.83 ng/mL
Median: 34.67 ng/mL
Range: 1.16-
Range: 1.52-
602.79 ng/mL
291.56 ng/mL
Zhao et al. (2017,
Maternal blood
Cord blood
NR
NR
5085130)
Mean: 0.984 ng/mL
Mean: 0.565 ng/mL
(Hubei, China)
Median: 0.907 ng/mL
Median: 0.535 ng/mL
Range: 0.274-2.72 ng/mL Range:0.126 -
1.44 ng/mL
Brochotetal. (2019,
Group 1 mean (plasma):
Mean: 2.54 ± 1.64 (0.86-
NR
NR
5381552) (INMA
3.26 ± 1.87 (0.39-
10.56) ng/mL
prospective birth
11.93) ng/mL
cohort, Spain)ad
Group 2 mean (plasma):
2.78 ±2.18 (0.20-
31.64) ng/mL
Gao et al. (2019,
Mean: 2.85 ng/mL
Mean: 2.29 ng/mL
NR
NR
5387135)
Median: 2.21 ng/mL
Median: 1.88 ng/mL
(Beijing, China)
Range: < LOD-
Range: 0.03-10.2 ng/mL
25.4 ng/mL
Eryasaetal. (2019,
GM serum: 2.33 ng/mL
Cord serum
NR
NR
5412430)
SD: 0.12 ng/mL
Mean: 1.97 ng/mL
(Faroese Birth
IQR: 1.79-3.29 ng/mL
SD: 0.10 ng/mL
Cohorts, Denmark)b
IQR: 1.42-2.76 ng/mL
(Cohort 3)
Whole cord blood
Mean: 1.08 ng/mL
SD: 0.05 ng/mL
IQR: 0.8-1.45 ng/mL
Eryasaetal. (2019,
Mean: 1.03 ng/mL
Cord serum
NR
NR
5412430)
SD: 0.08 ng/mL
Mean: 0.81 ng/mL
(Faroese Birth
IQR: 0.75-1.41 ng/mL
SD: 0.07 ng/mL
Cohorts, Denmark)b
IQR: 0.56-1.26 ng/mL
(Cohort 5)
Whole cord blood
Mean: 0.41 ng/mL
SD: 0.03 ng/mL
IQR: 0.29-0.59 ng/mL
Cai et al. (2020,
6318671)
(Maoming Birth
Cohort, China)
Maternal serum
Mean: 1.21 ng/mL
SD: 1.01 ng/mL
Median: 0.99 ng/mL
IQR: 0.74-1.37/mL
Cord serum
Mean: 0.85 ng/mL
SD: 0.52 ng/mL
Median: 0.75 ng/mL
IQR: 0.52-1.09 ng/mL
NR
NR
B-31
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study (Study
Location)
Maternal Blood
Cord Blood Placenta
Amniotic Fluid
Li et al. (2020,
Preterm delivery
Preterm delivery NR
NR
6505874)
Mean serum: 1.20 ng/mL
Mean: 0.70 ng/mL
(Maoming Birth
Median: 1.00 ng/mL
Median: 0.57 ng/mL
Cohort, China)
IQR: 0.69-1.47
IQR: 0.43-0.91
Full-term delivery
Full-term delivery
Mean: 1.34
Mean: 1.25 ng/mL
Median: 1.13 ng/mL
Median: 0.99 ng/mL
IQR 0.72-1.74
IQR 0.64-1.49
Li et al. (2020,
6506038)
(Beijing, China)
Mean serum: 3.63 ng/mL
(95% CI: 3.26,4.49)
Median: 3.20 ng/mL
Mean: 4.98 ng/mL (95% NR
CI: 4.41,7.38)
Median: 3.80 ng/mL
NR
Mamsen et al.
Mean: 2.1 ng/g,
NR
Mean: 0.23 ng/g,
NR
(2017, 3858487)
Range: 0.6-8.0 ng/g
Range: 0.04-
(Hospitals in Skelby
0.45 ng/g
and Randers,
Denmark)
Mamsen et al.
Tl serum
NR
Mean: 0.28 ng/g
NR
(2019, 5080595)
Mean: 2.04 ng/mL
SD: 0.09 ng/g
(Denmark)3
SD: 1.63 ng/mL
Median: 0.27 ng/g
Median: 1.51 ng/mL
Range: 0.15-
Range: 0.55-7.95 ng/mL
0.45 ng/g
T2 serum
NR
Mean: 0.39 ng/g
NR
Mean: 1.62 ng/mL
SD: 0.26 ng/g
SD: 0.71 ng/mL
Median: 0.26 ng/g
Median: 1.58 ng/mL
Range: 0.19-
Range: 0.72-3.78 ng/mL
0.99 ng/g
T3 serum
NR
Mean: 0.43 ng/g
NR
Mean: 1.62 ng/mL
SD: 0.16 ng/g
SD: 0.85 ng/mL
Median: 0.36 ng/g
Median: 1.36 ng/mL
Range: 0.21-
Range: 0.62-4.62 ng/mL
0.82 ng/g
Hanssen et al.
Plasma
Cord plasma
NR
NR
(2013, 3859848)
Median: 1.61 ng/mL
Median: 1.00 ng/mL
(Norilsk, Russia)6
Mean: 1.50 ng/mL
Mean: 1.26 ng/mL
Range: 0.63-2.48 ng/mL
Range: 0.36-2.32 ng/mL
Whole blood
Cord whole blood
NR
NR
Median: 0.89 ng/mL
Median: 0.49 ng/mL
Mean: 0.89 ng/mL
Mean: 0.58 ng/mL
Range: 0.33-1.40 ng/mL
Range: 0.15-1.12 ng/mL
Kato et al. (2014,
Maternal Serum at
Cord serum at delivery
2851230)
16 weeks
Median:3.10 |ig/L
(Ohio, USA)f
Median: 4.80|ig/L
Maternal serum at
delivery
Median:3.30 [ig/L
Notes: GM = Geometric mean; LOD = limit of detection; LOQ = limit of quantification; IQR = Interquartile range; NR = Not
reported; SD = Standard deviations; Tl= first trimester; T2 = Second trimester; T3 = Third trimester.
aFor studies that quantified PFOA at different trimesters, first trimester (Tl), second trimester (T2) and third trimester (T3).
B-32
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
bEryasa et al. (2019, 5412430) sampled participants from 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.
c Pan et al. (2017, 3981900) measured PFOA in maternal serum at first, second and third trimester and measured cord blood only
at the time of full-term delivery.
dBrochot et al. (2019, 5381552) collected samples from women in 2 cohorts: Group 1 consist of 52 mother-child pairs that had
available samples of maternal blood and cord serum PFAS during pregnancy. Group 2 consist of 355 mothers who provided
maternal blood during pregnancy. Cord blood was not collected for the Group 2 cohort.
eHanssen et al. (2013, 3859848) measured PFOA in whole blood and plasma from mothers and their infants at the time of
delivery.
fKato et al. (2014,2851230) measured PFOA 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.2.4.1.3 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
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. PFOA was
detected in placenta, fetal liver, extremities, heart, intestines, lungs, connective tissues, spinal
cord, and ribs at different concentrations. Notably, PFOA levels were highest in the placenta and
lung. Mean concentrations of PFOA in maternal serum, placenta, and fetal organs were reported
as 1.9 (0.6-4.1), 0.2 (0.0-0.4), and 0.1 (0-0.3) ng/g, respectively. Mean concentrations of PFOS
in maternal serum, placenta, and fetal organs were reported as 8.2 (2.5-16.7), 1.0 (0.3-2.6), and
0.3 (0-0.7) ng/g, respectively. The concentrations of PFOS in all three matrices were
significantly higher than 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
PFOA decreased from maternal serum to fetal tissues as follows: maternal serum > placenta >
fetal tissues. The relative concentration of PFOA in the placenta was 11% of the concentrations
found in maternal plasma and were further reduced to 7% in fetal tissues. In general, a positive
trend was observed between fetal tissue-maternal serum ratio and gestational age. Although the
gestational age reported in this study is short (37-68 days post conception), the results suggest
that PFOA is retained in several fetal organs and may potentially continue to accumulate across
gestation.
To determine whether PFOA 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, central
nervous system (CNS), and adipose) were collected from 38 first trimester pregnancies, 18
second trimester pregnancies, and 22 third trimester pregnancies. Fetal tissue:maternal serum
ratios of PFAS were calculated by dividing the fetal tissue concentration by the maternal serum
concentration. In general, fetal tissue:maternal ratios of PFOA in fetal tissue increased from first
B-33
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
trimester to third trimester except for the liver and heart which showed the highest
tissue:maternal serum ratios in the second trimester compared with the third trimester. The fetal
tissue:maternal serum ratio of PFOA was highest in adipose tissue during the second trimester
than in any other tissue across gestation.
Interestingly, PFOA concentration in the liver was also highest in the second trimester compared
with the first and third trimesters. 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 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 different
geographical locations of the participants. The first trimester participants were from Denmark
and the second and third trimester participants came from Sweden.
B.2.4.1.4 Partitioning to Infants
Four studies shown in Table B-l 1 analyzed PFOA levels in maternal serum and levels in breast
milk and/or infant blood. Maternal and infant serum PFOA levels were an order of magnitude
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 PFOA concentrations were lower in
breastfeeding mothers (18.32 ng/mL) vs. non-breastfeeding mothers (19.26 ng/mL). Conversely,
breastfed infants had higher GM serum PFOA (48.55 g/mL) than infants who were never
breastfed (21.74 ng/mL).
Cariou et al. (2015, 3859840) reported that PFOA levels in breastmilk were approximately 30-
fold lower relative to maternal serum and the ratio between breastmilk and maternal serum
PFOA was 0.038±0.013 (n= 10). 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 inter-
individual 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 PFOA, although it did reach significance for PFHxS. Only PFOA
exhibited a significant correlation between the total duration of breastfeeding and serum PFOA
levels after adjustment (0.87 (0.80-0.94), p = 0.0007).
B-34
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mogensen et al. (2015, 3859839) relied on maternal PFOA 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-l 1, the
increases in infant blood PFOA 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 PFOA concentrations in 2-4-month old
infants and maternal PFOA 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. Of note, the authors observed that variation in maternal PFOA concentrations
explained 53% of the infant concentration variation, whereas only 13% of the variation in infant
PFUnDA was explained by maternal variation. Also, the PFOA infant:maternal serum ratio was
higher than ratios for other PFAAs (2.8 (0.43-5.7)).
Table B-ll. Summary of Studies Evaluating PFOA concentrations in Maternal Serum,
Breast Milk, and Infant Serum
Study
Subjects
Maternal Blood
Breastmilk
Infant Blood
Mondal et al. A subcohort of the C8
(2014, Science Panel Study (exposed
2850916) 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
mean: 18.69 ng/mL
95% CI: 17.13,20.28
Breastfed
GM: 18.32 ng/mL
95% CI: 16.36, 20.50
Not breastfed
GM: 19.26 ng/mL
95% CI: 16.80, 22.08
NR
Infant serum
Breastfed & not
breastfed
mean: 36.14 ng/mL
95% CI: 24.87, 52.52
Breastfed
GM: 48.55g/mL
95% CI: 31.17, 75.61
Not breastfed
GM: 21.74ng/mL
95% CI: 11.21,42.17
B-35
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Subjects
Maternal Blood
Breastmilk
Infant Blood
Cariou et al. Female volunteers
(2015, hospitalized
3859840) 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.
Mean: 1.22 ng/mL
Median: 1.045 ng/mL
Range: 0.309-
7.31 ng/mL
Mean: 0.041 ng/mL
Median: < LOQ
LOQ = 0.050 ng/mL
Range: < LOD-
0.308 ng/mL
NR
Mogensen et 80 singleton children in
al. (2016, Faroese birth cohort born
3859839f 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. A
piece-wise linear model was
used to estimate the age
dependence of the PFOA
Concentration.
NR
NR
Median at birth:
2.0 ng/mL (IQR
1.7,2.7)
Median at 11 months:
8.2 ng/mL (IQR 6.1,
10.9)
Median at 18 months:
6.1 ng/mL (IQR 5.1,
10)
Median at 60 months:
3.8 ng/mL (IQR 3.1,
4.9)
Gyllenhamm Primaparae mother/child pairs
ar et al. (201, in 1996-1999 recruited in
8, 4778766) 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: 2.8 ng/g
SD: 0.96 ng/g
Median: 2.7 ng/g
Range: 1.2-6.7 ng/g
NR
Infant serum
Mean: 7.7 ng/g
SD: 3.7 ng/g
Median: 7.2 ng/g
Range: 1.3-20 ng/g
Notes: CI = confidence interval; GM = geometric mean; IQR = interquartile range; LOD = limit of detection; LOQ = limit of
quantification; 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).
Mondal et al. (2014, 2850916) also examined the change in maternal and infant PFOA levels
with duration of breastfeeding (Table B-12). Maternal serum concentrations decreased with
each month of breastfeeding (-3%; 95% CI: -5%, -2%) with the greatest decrease observed
after 12 months of breastfeeding (-41%). Correspondingly, the infant PFOA serum
concentrations increased by 6% (95% CI: 1%, 10%) with each month of breastfeeding, lower
than the estimate of 30% per month in Swedish infants found by Gyllenhammar et al. (2018,
4778766). Increases were modest in the first 6 months (13%) but increased to 141% after 12
months of breastfeeding. Using mixed linear model regression (Table B-13), Mogensen et al.
(2015, 3859839) calculated that, during months with exclusive breastfeeding, significant
B-36
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
increases in the PFOA concentrations in infant serum were estimated (27.8% and 31.2% per
month at 18 and 60 months, respectively). These levels were higher than the continuous (per
month) 6% estimated increases in the Mondal study, respectively. Increases were less striking for
months with partial breastfeeding and small or none for months without breastfeeding.
Altogether, these findings support breastfeeding as the primary source of infant PFOA
accumulation and that distribution to the infant correlates with the length of breastfeeding.
Table B-12. Percent Change in PFOA Ratios in Maternal Serum to Breast Milk and Breast
Milk to Infant Serum by Infant Age in Humans as Reported by Mondal et al. (2014,
2850916)
PFOA {Mondal, 2014,
2850916}
Maternal Serum: Breast Milk
Breastmilk: Infant Serum
Infant Age
Percent Change
95% CI
Percent Change
95% CI
< 6 months
-5%
("18, 8)
13%
(-46, 139)
7-12 months
-29%
(-41, -13)
82%
(-23, -334)
>12 months
-41%
(-57, -17)
141%
(4, 460)
Continuous (per month)
-3%
("5, "2)
6%
(1, 10)
Notes: CI = confidence interval.
Table B-13. Percent Change in PFOA Serum Concentration by Exclusive, Mixed or No
Breastfeeding Per Month in Humans as Reported by Mogensen et al. (2015, 3859839)
Mixed Model up to 18 Months
Mixed Model up to 60 Months
Variable Percent
Change
95% CI
p-value
Percent 95% CI
Change
p-value
Exclusive 27.8
(23.6,32.1)
<0.0001
31.2 (28.0,34.5)
<0.0001
Partial 3.9
(0.5,7.3)
0.0252
0.1 (-1.6, 1.9)
0.8951
None 0.7
("1.1,2.5)
0.4528
-1.3 (-1.5,-1.0)
<0.0001
Notes: CI = confidence interval.
The contributions of placental transfer, breastfeeding, and ingestion of PFAA-contaminated
drinking water to early life PFOA levels in children were analyzed {Gyllenhammar, 2019,
5919402}. This study measured PFOA 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 PFOA for these exposure modes. PFOA
concentrations increased 10% per unit (ng/g serum) of increase in the maternal serum level at
delivery. The association was strongest in 4-year old children. Duration of breastfeeding only
correlated with 4-year old children but not older children in the MLR model (partial R2 = 0.05
for children in this age group). PFOA increased 1.2% per month of cumulative drinking water
exposure. The authors suggested that, in addition to exposure in utero and through lactation,
drinking water with low-to-moderate PFOA contamination is an important source of exposure
for children.
B-37
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.2.4.2 Animal Studies
B. 2.4.2.1 Rats
PFOA levels during gestation and lactation were studied by Hinderliter et al. (2005, 1332671)
(publication of data reported by Mylchreest (2003, 9642031)). Time-mated female Sprague-
Dawley rats were dosed with 0, 3, 10, or 30 mg/kg/day of PFOA during days 4-10, 4-15, and 4-
21 of gestation, or from GD 4 to LD 21. Maternal blood samples were collected at 2 hours ± 30
minutes (mins) post-dose on a daily basis. Plasma, milk, amniotic fluid, and tissue homogenate
(placenta, embryo, and fetus) supernatants were analyzed for PFOA concentrations by high-
performance liquid chromatography mass spectrometry (HPLC/MS). Maternal PFOA plasma
levels during gestation and lactation are presented in Table B-14. Maternal plasma levels at
2 hours post-dosing (approximately the time of peak blood levels following a gavage dose) were
fairly similar during the course of the study with mean levels of 11.2, 26.8, and 66.6 |ag/m L in
the 3, 10, and 30 mg/kg/day groups, respectively; PFOA levels in the control group were below
the LOQ (0.05 |ig/mL). The stability of the maternal plasma PFOA concentrations day-to-day
indicates that PFOA was not accumulating in plasma, despite repeated exposures. This is
possibly because the female rats were able to completely excrete the PFOA dose (3, 10, or
30 mg/kg/day) within 24 hours, before the next dose was administered via gavage.
Table B-14. Maternal Plasma PFOA Levels in Sprague-Dawley Rats During Gestation and
Lactation" as Reported by Hinderliter et al. (2005,1332671)
Exposure Period
Sample Time
Dose
3 mg/kg/day
10 mg/kg/day
30 mg/kg/day
GD4-GD 10
GD 10 plasma
8.53 ± 1.06
23.32 ±2.15
70.49 ±8.94
GD4-GD 15
GD 15 plasma
15.92 ± 12.96
29.40 ± 14.19
79.55 ±3.11
GD4-GD 21
GD 21 plasma
14.04 ±2.27
34.20 ±6.68
76.36 ± 14.76
GD4-LD 3
LD 3 plasma
11.01 ± 2.11
22.47 ± 2.74
54.39 ± 17.86
GD4-LD 7
LD 7 plasma
10.09 ±2.90
25.83 ±2.07
66.91 ± 11.82
GD4-LD 14
LD 14 plasma
9.69 ±0.92
23.79 ±2,81
54.65 ± 11.63
GD4-LD 21
LD 21 plasma
9.04 ± 1.01
28.84 ±5.15
64.13 ± 1.45
NA
Average plasma
11.19 ± 2.76
26.84 ±4.21
66.64 ± 9.80
Notes: GD = gestation day; LD = lactation day; NA = not applicable.
aData are presented as mean ± standard deviation ((ig/mL).
PFOA levels in the placenta, amniotic fluid, and embryo/fetus are presented in Table B-15. The
levels of PFOA in the placenta on GD 21 were approximately twice the levels observed on GD
15, and the levels of PFOA in the amniotic fluid were approximately four times higher on GD 21
than on GD 15. The concentration of PFOA in the embryo/fetus was highest in the GD 10
embryo and lowest in the GD 15 embryo; PFOA levels in the GD 21 fetus were intermediate.
Fetal and pup PFOA plasma levels during gestation and lactation are presented in Table B-16,
and PFOA levels in maternal milk during lactation are provided in Table B-17. The
concentrations of PFOA in the plasma of the GD21 fetus (5.88, 14.48, and 33.11 |ig/mL,
respectively, in the 3, 10, and 30 mg/kg/day groups) were approximately half the levels observed
in the maternal plasma (Table B-14). Pup plasma levels decreased between birth and LD 7
B-38
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(Table B-16) and were, thereafter, similar to the levels observed in the milk (Table B 17). The
pups were not separated by sex. The concentrations of PFOA in maternal milk also were fairly
similar throughout lactation (means of 1.1, 2.8, and 6.2 |ig/ml in the 3, 10, and 30 mg/kg/day
groups, respectively) and were approximately one-tenth of the PFOA levels in the maternal
plasma.
Table B-15. Placenta, Amniotic Fluid, and Embryo/Fetus PFOA Concentrations in
Sprague-Dawley Rats" as Reported by Hinderliter et al. (2005,1332671)
Exposure Period
Tissue
Dose
3 mg/kg/day
10 mg/kg/day
30 mg/kg/day
GD 4-GD 10
GD 10—embryo
1.40 ±0.30
3.33 ±0.81
12.49 ±3.50
GD 4-GD 15
GD 15—placenta
2.22 ± 1.79
5.10 ± 1.70
13.22 ± 1.03
GD 15—amniotic fluid
0.60 ±0.69
0.70 ±0.15
1.70 ±0.91
GD 15—embryo
0.24 ±0.19
0.53 ±0.18
1.24 ±0.22
GD 4-GD 21
GD 21—placenta
3.55 ±0.57
9.37 ± 1.76
24.37 ±4.13
GD 21—amniotic fluid
1.50 ±0.32
3.76 ±0.81
8.13 ±0.86
GD 21—fetus
1.27 ±0.26
2.61 ±0.37
8.77 ±2.36
Notes: GD = gestation day.
aData are presented as mean ± standard deviation ((ig/mL). Samples were
dosing.
pooled by litter and were collected 2 hours post-
Table B-16. Fetus/Pup PFOA Concentration in Sprague-Dawley Rats During Gestation
and Lactation";
as Reported by Hinderliter et al. (2005,1332671)
Exposure Period
Tissue
Dose
3 mg/kg/day
10 mg/kg/day
30 mg/kg/day
GD 4-GD 21
GD 21—fetal plasma
5.88 ±0.69
14.48 ± 1.51
33.11 ±4.64
GD 4-LD 3
LD 3—pup plasma
2.89 ±0.70
5.94 ± 1.44
11.96 ± 1.66
GD 4-LD 7
LD 7—pup plasma
0.65 ± 0.20
2.77 ±0.58
4.92 ± 1.28
GD 4-LD 14
LD 14—pup plasma
0.77 ±0.10
2.22 ±0.38
4.91 ± 1.12
GD 4-LD 21
LD 21—pup plasma
1.28 ±0.72
3.25 ±0.52
7.36 ±2.17
Notes: GD = gestation day; LD = lactation day.
aData are presented as mean ± standard deviation ((ig/mL). Samples were
dosing.
pooled by litter and were collected 2 hours post-
Table B-17. Maternal Milk PFOA Concentration in Sprague-Dawley Rat During
Lactation" as Reported by Hinderliter et al. (2005,1332671)
Exposure Period
Sample Time
Dose
3 mg/kg/day
10 mg/kg/day
30 mg/kg/day
GD 4-LD 3
LD 3-milk
1.07 ±0.26
2.03 ±0.33
4.97 ± 1.20
GD 4-LD 7
LD 7-milk
0.94 ±0.22
2.74 ±0.91
5.76 ±1.26
GD 4-LD 14
LD 14-milk
1.15 ±0.06
3.45 ± 1.18
6.45 ± 1.38
GD 4-LD 21
LD 21-milk
1.13 ±0.08
3.07 ±0.51
7.48 ± 1.63
NA
Average milk
1.07 ±0.09
2.82 ±0.60
6.16 ± 1.06
Notes: GD = gestation day; LD = lactation day; NA
= not applicable.
B-39
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
aData are presented as mean ± standard deviation ((ig/mL). Samples were from 5 dams/group/time point and were collected
2 hours post-dosing.
PFOA accumulation in young rats is impacted by both sex and age. Han (2003, 9978263)
administered groups of 4-8-week-old Sprague-Dawley rats (10 per sex per age) a single dose of
10 mg/kg/day PFOA by oral gavage. Blood samples were collected 24 hours after dosing and the
plasma concentration of PFOA was measured by HPLC-MS. In the 5- and 6-week-old female
rats, the plasma PFOA concentrations were about two-fold lower than in the 4-week-old rats
(Table B-18). However, in the 5-week-old males, the concentration of plasma PFOA was about
fivefold higher than in the 4-week-old group, suggesting a developmental change in excretion
rate. PFOA plasma concentrations were 35-65-fold higher in males than in females at every age
except at 4 weeks. Thus, it appears that maturation of the transport features responsible for the
sex difference in PFOA elimination occurs between the ages of 4 and 5 weeks in the rat.
Table B-18. Plasma PFOA Concentrations in Postweaning Sprague-Dawley Rats" as
Reported by Han (2003, 9978263)
Age (weeks)
Males
Females
4
7.32 ± 1.01
2.68 ±0.64
5
39.24 ±3.89
1.13 ±0.46
6
43.19 ±3.79
1.18 ±0.52
7
37.12 ±4.07
0.57 ±0.29
8
38.55 ±5.44
0.81 ±0.27
Notes:
aData are presented as mean ± standard deviation ((ig/mL).
Hinderliter et al. (2006, 3749132) continued the investigation of the relationship between age
and plasma PFOA in male and female Sprague-Dawley rats. Immature rats at 3, 4, and 5 weeks
of age were administered PFOA via oral gavage at a single dose of 10 or 30 mg/kg. Rats were
not fasted prior to dosing. Two hours after dosing, five rats per sex per age group and dose group
were sacrificed and blood samples were collected. The remaining five rats per sex per age and
dose group were placed in metabolism cages for 24-hour urine collection. These rats were
sacrificed at 24 hours and blood samples were collected.
In the male rats, plasma PFOA concentrations for either the 10- or 30-mg/kg dosage groups did
not differ significantly by sample time (at 2 and 24 hours) or by animal age (3, 4, and weeks),
except at 2 hours for the 5-week old group (p < 0.01), which showed the lowest PFOA level
(Table B-19). PFOA plasma concentrations following a 30-mg/kg dose were 2-3 times higher
than those following a 10-mg/kg dose. These data do not demonstrate a difference between the 5-
week old rats and the younger 3- and 4-week old groups at 24 hours after dosing, and thus do not
support the observations from the Han study (2003, 9978263).
B-40
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-19. Plasma PFOA Concentrations in Male Sprague-Dawley Rats at 2 and 24 hours
after Oral Gavage as Reported by Hinderliter et al. (2006, 3749132)
Plasma PFOA (jig/mL)
Age (weeks)
Dose (mg/kg)
2 Hours Post-Dose
24 Hours Post-Dose
Mean
SD
Mean
SD
3
10
41.87
4.01
34.22
7.89
4
10
39.92
4.45
42.94
5.33
5
10
26.32*
6.89
40.60
3.69
3
30
120.65
12.78
74.16
18.23
4
30
117.40
18.10
100.81
13.18
5
30
65.66*
15.53
113.86
23.36
Notes: SD = standard deviation.
* Statistically significantly different by sample time and animal age (p < 0.01).
In the female rats, plasma PFOA concentrations were significantly lower in the 5-week old group
than in the 3- or 4-week old groups at the 24-hour time period for both doses and for the 30-
mg/kg dose group at 2 hours (Table B-20). Plasma PFOA concentrations following a 30-mg/kg
dose were approximately one and one half to four times higher than those observed following a
10-mg/kg dose.
At 24 hours post-dose, plasma PFOA levels in the female rats were significantly lower than the
plasma PFOA levels in male rats, especially at 5 weeks of age. The data for the 5-week old
female rats compared to the 3- and 4-week old groups at 24 hours are consistent with the Han
(2003, 9978263) data in that they demonstrate a decline in plasma levels compared to their
earlier measurements. Thus, the developmental change is one that appears to be unique to the
female rat.
Table B-20. Plasma PFOA Concentrations in Female Sprague-Dawley Rats at 2 and 24
hours after Oral Gavage as Reported by Hinderliter et al. (2006, 3749132)
Age (weeks)
Plasma PFOA (jig/mL)
Dose (mg/kg)
2 Hours Post-Dose
24 Hours Post-Dose
Mean
SD
Mean
SD
3
10
37.87
5.77
13.55b
3.83
4
10
29.88
12.15
18.98b
7.01
5
10
33.23
7.41
1.36^b
0.87
3
30
84.86
10.51
51 43b
13.61
4
30
80.67
14.10
28.01b
9.90
5
30
56.90 a
29.66
3.42%h
1.95
Notes: SD = standard deviation.
a Statistically significantly different from the 3- and 4-week values (p < 0.01).
b Statistically significantly different from 2-hour values (p < 0.01).
The data demonstrate that both dose and sex influence plasma levels. Post-dosing clearance (CL)
is slow for both doses at 2 and 24 hours in males and females at PNW 3 and 4. At 5 weeks,
B-41
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
however, the plasma levels after 24 hours are greater than those at 2 hours in males. In females,
for the high dose at 2 hours, plasma levels are similar to those in males, while at 24 hours they
are only 3% of the value for males. This suggests that uptake from the intestines is similar while
the rate of excretion at 5 weeks and beyond is considerably greater for female rats than males.
They are comparable for PNW 3 and 4.
In a supplemental study to determine the effect of fasting {Hinderliter, 2006, 3749132}, 4-week
old rats, 4 rats per sex, were administered 10 mg/kg PFOA via oral gavage. Animals (two per
sex) were fasted overnight for 12 hours before dosing with PFOA. All the rats were sacrificed at
24 hours post dosing and blood was collected for analysis of PFOA in plasma. Plasma PFOA
concentrations in male rats were 64.95 and 30.00 |ag/mL for the fasted and nonfasted animals,
respectively. Plasma PFOA concentrations in the female rats were 68.16 and 26.54 |ag/mL for
the fasted and nonfasted animals, respectively. Given the consistency in the 4-week old rat
plasma PFOA concentrations, the authors concluded that age-dependent changes in female
PFOA elimination are observable between 3 and 5 weeks of age. PFOA uptake was greater in the
fasted animals than the fed animals, suggesting competition for uptake in the presence of food
components that share common transporters and/or decreased contact of PFOA with the
intestinal epithelium in the presence of dietary materials. This is consistent with the finding that
dietary fat may negatively impact absorption {Li, 2015, 2851033}.
An oral two-generation reproductive toxicity study of PFOA in rats was conducted {Butenhoff,
2004, 1291063}. Five groups of rats (30 sex/group) were administered PFOA by gavage at doses
of 0, 1, 3, 10, or 30 mg/kg/day. At scheduled sacrifice, after completion of the cohabitation
period in Fo male rats and on lactation day (LD) 22 in Fo female rats, blood samples were
collected. Serum analysis for the Fo generation males showed that PFOA was present in all
samples tested, including low levels in controls (0.0344 ± 0.0148 |ig/mL). Levels of PFOA were
similar in the two male dose groups (51.1 ± 9.30 and 45.3 ± 12.6 |ig/mL, respectively, for 10-
and 30-mg/kg/day dose groups). In the Fo female controls, serum PFOA was below LOQ
(0.00528 |ig/mL), Levels of PFOA found in female sera were lower than in males but increased
between the two dose groups; treated females had an average concentration of 0.37 ± 0.0805 and
I.02 ± 0.425 |ig/mL, respectively, for the 10- and 30-mg/kg/day dose groups.
B.2.4.2.2 Mice
Fenton et al. (2009, 194799) orally dosed pregnant CD-I mice (n = 25/group) with 0, 0.1, 1, or
5 mg PFOA/kg on GD 17. On GD 18, five dams/group were sacrificed and trunk blood, urine,
amniotic fluid, and the fourth and fifth mammary glands were collected Additionally, one fetus
from each dam was retained for whole-pup analysis. The remaining dams were allowed to litter
and samples (excluding amniotic fluid) also were collected on postnatal day (PND) 1, 4, 8, and
18. At each time-point, a single pup was euthanized and retained for whole-pup analysis. Blood
from the remaining pups was collected and pooled. Milk was collected from dams on PND 2, 8,
II, and 18 following a 2-hour separation of the pups from the dam.
PFOA levels in mice during gestation and lactation in selected fluids and tissues are summarized
in Table B-21. The concentrations of PFOA in dam serum were approximately twice that
detected in amniotic fluid. Compared to the amniotic fluid, concentrations of PFOA in the
fetuses were increased by 2.3-, 3.1-, and 2.7-fold at 0.1, 1, and 5 mg/kg, respectively. The
highest concentration of PFOA was detected in the serum of nursing dams. In the dams, the
B-42
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOA serum concentrations exhibited a U-shaped response curve over time; the lowest serum
concentrations were observed at the time of peak lactation. Dam mammary tissue and milk
PFOA concentrations showed a U-shaped response which mirrored that found in dam serum. The
concentrations of PFOA in pup serum were significantly higher than PFOA concentrations in
dam serum and appeared to decrease as the time for weaning approached. When pup PFOA
concentrations were calculated with consideration for pup body weight gain, PFOA body burden
increased through the peak of lactation and began to decrease by PND18, showing an inverse U-
shaped response curve. The authors hypothesized that the U-shaped curve was observed for the
lactating dams because of hydro-dilution; essentially, the increases in blood volume and milk
volume at the time of peak lactation led to lower PFOA concentrations during this particular
time.
Table B-21. Select Fluids and Tissues PFOA Concentrations in CD-I Mice During
Gestation and Lactation" as Reported by Fenton et al. (2009,194799)
Dose
i issue
u ay
0.1 mg/kg
1 mg/kg
5 mg/kg
Dam Serumb
GD 18
143 ± 19
1697 ± 203
7897± 663
PND 1
217.5 ±35
1957.0 ±84
9845.6 ± 1478
PND 4
110.0 ± 12
1269.4 ±235
6776.6 ±561
PND 8
46.7 ±21
360.8 ±98
1961.8 ±414
PND 18
123.3 ±41
1035.2 ±305
5156.5 ± 1201
Amniotic Fluidb
GD 18
99.0 ±28
865.3 ± 191
3203.8 ±492
Dam Urineb
GD 18
21.9 ±8.6
104.9 ±69.7
666.7 ± 169
PND 1
7.7 ± 1.7
116.8 ±64
492.3 ± 119
PND 4
8.4 ±6.4
53.5 ± 15
401.5 ± 117
PND 8
0.8 ±0.22
11.6 ±6.2
40.1 ± 17
PND 18
1.8 ±1.1
18.7 ±8.6
91.7 ±49
Mammary Gland0
GD 18
18.9 ± 1.9
307.2 ±30.4
1429± 186
PND 1
27.4 ±6.8
343.8 ±53
1933.5 ± 194
PND 4
9.6± 8.4
239.2 ±53
1461.8 ±267
PND 8
2.4 ±3.8
71.7 ±22
411.8 ± 78
PND 18
17.1 ± 10
239.9 ±76
1372.8 ±240
Milkb
PND 2
32.5 ± 12
716.7 ± 145
1236.6 ± 1370
PND 8
11.6 ± 8.1
77.4 ± 19
245.1 ±26
PND 11
5.4 ± 1.0
42.3 ±9.1
282.5 ± 162
PND 18
43.5 ± 19
251.8 ± 147
909.8 ±308
Whole Pup0
GD 18
136.3 ± 15
1665.8 ±213
6256.5 ±751
PND 1
150.9 ±21
1606.9 ±288
7134.5 ± 1097
PND 4
91.8 ±8.9
1183.2 ±187
5071.4 ±267
PND 8
60.9 ± 16
729.0 ± 92
3118.5 ±424
PND 18
17.5 ±11
251.9 ± 112
1391.5 ±118
Pup Serumb
PND 1
324.7 ±36
3926.8 ±480
16,286.4 ± 1372
B-43
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Dose
Tissue Day
0.1 mg/kg
PND 4 267.6 ± 47
PND 8 260.2 ± 56
PND 18 111.8 ± 30
Notes: GD = gestation day; PND = postnatal day.
aAnimals were exposed to PFOA dose on GDI7
bData are presented as mean ± standard deviation (ng/mL)
cData are presented as mean ± standard deviation (ng/g)
Macon et al. (2011, 1276151) gavage-dosed CD-I mice with 0, 0.3, 1.0, or 3.0 mg PFOA/kg
from GD 1 to GD 17 or with 0, 0.01, 0.1, or 1.0 mg PFOA/kg from GD 10 to GD 17. As shown
in Table B-22, at the lowest dose, PFOA concentrations in the serum peaked at or before PND 7,
but peaked around PND 14 for the two higher doses. Calculated blood burdens, which take into
account the increasing blood volumes and body weights for females, showed an inverted U-
shaped curve peaking at PND 14 for all doses. In the liver, PFOA concentrations decreased over
time with the highest concentration observed at PND 7. Lower concentrations of PFOA were
detected in the brain of the offspring on PND 7 and PND 14. As shown in Table B-23, after
exposure to low doses of PFOA from GD 10 to GD 17, serum PFOA concentration in the female
offspring declined from PND 1 through the end of the experiment. Calculated blood burden
showed a gradual increase from PND 1 to PND 14, followed by a decline through PND 21.
Table B-22. Serum, Liver, and Brain PFOA Concentration in Female CD-I Mouse Pups
After GD 10-17 Exposure" as Reported by Macon et al. (2011,1276151)
Tissue
Day
Dose
0.3 mg/kg
1.0 mg/kg
3.0 mg/kg
Serum3
PND 7
4980 ±218
11026 ±915
20700 ± 3900
PND 14
4535 ± 920
16950 ± 3606
26525 ± 2446
PND 21
1194 ±394
377 ± 607
8343 ±1078
PND 28
630 ± 162
1247 ± 208
4883 ±1378
PND 42
377± 81
663 ±185
2058 ± 348
PND 63
55 ± 17
176 ± 85
-
PND 84
16 ±5
71 ±8
125
Liverb
PND 7
2078 ± 90
8134 ±740
16700 ± 749
PND 14
972 ± 124
4152 ±483
10290 ±1028
PND 21
1188±182
1939 ±637
2339±1241
PND 28
678 ±130
2007 ± 560
7124±1081
PND 42
342 ± 87
617±145
1145 ±274
PND 63
118 ±22
320 ±113
417±160
PND 84
43 ± 12
55 ± 12
235 ±79
Brainb
PND 7
150 ±26
479 ±41
1594± 162
PND 14
65 ± 12
241 ±20
650 ± 44
PND 21
< LOQ°
31 ±5
133 ±23
1 mg/kg
3020.8 ± 223
2548.2 ± 245
1124.8 ±236
5 mg/kg
11,925.2 ± 1077
9215.8 ±594
5894.3 ±743
B-44
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Tissue
Day
0.3 mg/kg
Dose
1.0 mg/kg
3.0 mg/kg
PND28
PND42
PND63
PND84
< LOQ
< LOQ
< LOQ
< LOQ
< LOQ
< LOQ
< LOQ
< LOQ
62 ±93
< LOQ
< LOQ
< LOQ
Notes: GD = gestation day; LOQ = limit of quantification; PND = postnatal day; - = not measured.
aData are presented as mean ± standard deviation (ng/mL)
bData are presented as mean ± standard deviation (ng/g)
c LOQ: serum full gestation = 10-20 ng/g; liver =35 ng/g; brain =35 ng/g; late gestation serum = 5 ng/mL
Table B-23. Serum PFOA Concentrations in Female CD-I Mouse Pups After GD 10-17
Exposure as Reported by Macon et al. (2011,1276151)
Tissue
Day
Dose
0.01 mg/kg
0.1 mg/kg
1.0 mg/kg
Serum3
PND 1
284.5 ±21.0
2303.5 ± 114.4
16305.5 ±873.5
PND 4
184.1 ± 12.1
-
-
PND 7
150.7 ±20.9
1277.8 ± 122.6
11880.3 ± 1447.6
PND 14
80.2 ± 13.9
645.4 ± 114.2
6083.7 ±662.6
PND 21
16.5 ±2.1
131.7 ±24.5
2025.1 ±281.9
Blood Burden
PND 1
15.2 ± 1.7
114.3 ±5.4
926.0 ± 47.6
(calculated)13
PND 4
20.6 ±0.1
-
-
PND 7
27.3 ±3.8
221.7 ±24.9
1965.9 ±256.7
PND 14
27.0 ±4.6
218.5 ±39.8
2033.6 ±293.5
PND 21
7.9 ± 1.0
66.4 ± 12.8
984.7 ± 142.8
Notes: PND = postnatal day.
aData are presented as mean ± standard deviation (ng/mL).
bBlood burden determined by (body weight x (58.5/1000) x serum x 0.55).
White et al. (2011, 1276150) measured serum PFOA concentrations in three generations of CD-I
mice (Table B-24). Pregnant mice (Fo, n = 10-12 dams/group) were gavage-dosed with 0, 1, or
5 mg PFOA/kg from GD 1 to GD 17. A separate group of pregnant mice (n = 7-10 dams/group)
were gavage-dosed with either 0 or 1 mg PFOA/kg from GD 1 to GD 17 and received drinking
water containing 5 parts per billion (ppb) PFOA beginning on GD 7 and continuing until the end
of the study for their offspring—except during breeding and early gestation—to simulate a
chronic low-dose exposure. Increases in serum PFOA concentrations were observed in the
control + 5 ppb PFOA groups of the Fi and second (F2) generations and in the 1 mg/kg + 5 ppb
PFOA group of the F2 generation. Decreases were observed for the remaining groups.
B-45
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-24. Serum PFOA Concentration in CD-I Mice Over Three Generations" as
Reported by White et al. (2011,1276150)
Dose
Generation/ Day ~
0 mg/kg + 5 ppb
1 mg/kg
1 mg/kg +5 ppb
5 mg/kg
Dams at Weaning
Fo/ PND 22
74.8 ± 11.3
6658.0 ±650.5
4772.0 ± 282.4
26980.0 ± 1288.2
Fi/~PND 91
86.9 ± 14.5
9.3 ±2.6
173.3 ±36.4
18.7 ±5.2
Offspring
Fi/PND 22
21.3 ±2.1
2443.8 ±256.4
2743.8 ± 129.7
10045 ± 1125.6
Fi/PND 42
48.9 ±4.7
609.5 ± 72.2
558.0 ±55.8
1581.0 ±245.1
Fi/PND 63
66.2 ±4.1
210.7 ±21.9
187.0 ±24.1
760.3 ± 188.3
F2/PND 22
26.6 ±2.4
4.6 ± 1.2
28.5 ±3.7
7.8 ± 1.9
F2/PND 42
57.4 ±2.9
0.4 ±0.0
72.8 ±5.8
0.4 ±0.0
F2/PND 63
68.5 ±9.4
1.1 ±0.5
69.2 ±4.3
1.2 ±0.5
Notes: Fo = parent generation; Fi = offspring generation 1; F2 = offspring generation 2; PND = postnatal day.
Data are presented as mean ± standard deviation (ng/mL)
To examine the effect of PFOA on the embryo-placenta unit, Blake et al. (2020, 6305864)
exposed CD-I mice to PFOA at 0, 0.1, or 5 mg/kg-day from embryonic day (E) 1.5 to 11.5 or
17.5 via oral gavage. PFOA levels in the maternal serum, amniotic fluid, and whole embryo are
presented in Table B-25. The mean concentration of PFOA in whole embryo is approximately 7
times higher on E 17.5 than E 11.5 for both the 1- and 5-mg/kg/day dose groups. At E 11.5, the
levels of PFOA in maternal serum is approximately 5.5 times the levels observed in the amniotic
fluid for the 1-mg/kg/day group and 13 times the levels observed in the 5-mg/kg/day group.
Dosimetry for amniotic fluid was not reported for the mice examined at E 17.5.
Table B-25. Maternal Serum, Amniotic Fluid, and Whole Embryo PFOA Concentrations in
CD-I Mice Exposed During Gestation Day 1.5-17.5 as Reported by Blake et al. (2020,
6305864)
Biological Matrix
Gestational Age
Dose
1 mg/kg/day
5 mg/kg/day
Maternal serum3
E 11.5
25.4 ±3.7
117.3 ±20.6
E 17.5
18.7 ±3.2
95.1 ± 14.1
Amniotic fluid3
E 11.5
4.6 ±2.8
8.8 ±2.7
E 17.5
NR
NR
Whole embryob
E 11.5
0.80 ±0.10
2.34 ±0.27
E 17.5
5.78 ±0.71
16.4 ± 1.75
Notes: E = embryonic day; NR = not reported; SD = standard deviation.
3 Data are presented as mean ± standard deviation (ng/mL).
bData are presented as mean ± standard deviation (ng/g).
Transfer of PFAS via lactation does not appear to correlate with lipophilicity {Fujii, 2020,
6512379}. Lactating FVB/NJcl mice were given a single IV dose of PFOA and other PFCAs
chemicals with chain lengths from C8 to C13 on PND 8-PND 13. Maternal blood and milk were
B-46
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
collected from the dam 24 h after administration. The milk/plasma (M/P) concentration ratio for
PFOA was 0.32. Ratios exhibited a U-shaped curve with increasing chain length: 0.30 for C9,
0.17 for C10, 0.21 for CI 1, 0.32 for C12, and 0.49 for C13. While the M/P concentration ratio
did not correlate to lipophilicity of PFCAs, the estimated relative daily intake increased with
chain length: 4.16 for PFOA (C8), 8.98 for C9, 9.35 for C10, 9.51 for Cll, 10.20 for C12, and
10.49 for C13. These findings suggest that the amount transferred from mothers to pup during
lactation may also relate to chain length-dependent clearance.
B.2.5 Volume of Distribution Data
B. 2.5.1 Human Studies
Several researchers have attempted to characterize PFOA exposure and intake in humans through
PK modeling {Lorber, 2011, 2914150; Thompson, 2010, 2919278}. As an integral part of model
validation, the parameter for the volume of distribution (Vd) of PFOA within the body was
calibrated from available data. In the models discussed main document (See PFOA Main
Document), Vd was defined as the total amount of PFOA in the body divided by the blood or
serum concentration.
Two groups of researchers defined a Vd of 170 mL/kg body weight for humans for use in a
simple, single compartment, first-order PK model {Lorber, 2011, 2914150; Thompson, 2010,
2919278}. The models developed by these groups were designed to estimate intakes of PFOA by
young children and adults {Lorber, 2011, 2914150} 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 two contaminated U.S. communities
and assumes that most PFOA intake is from contaminated drinking water. Thus, in using the
models to derive an intake from contaminated water, the Vd was calibrated so that model
prediction of elevated blood levels of PFOA matched those seen in residents.
The assignment of Vd values used in several modelling studies is shown in Table B-26. The
value of 170 mL/kg is frequently used when considering both males and females. Mondal et al.
(2014, 2850916) assigned a value 198 mL/kg for breastfeeding females. Shin et al. (2011,
2572313) assigned values by sex (181 mL/kg for males and 198 mL/kg for females). Gomis et al.
(2017, 3981280) used a higher Vd of 200 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 fractions.
B-47
-------
DRAFT FOR PUBLIC COMMENT
Table B-26. Summary of PFOA Volume of Distribution Values Assigned in Human Studies
MARCH 2023
Study Population
Sex
Compartment
Vd
AUC or Mean/Median
Concentration Measured in
Compartment
Steady State
Considerations
Mondal et al. Adult,
Females
Maternal serum
198 mL/kg
GM Breastfeeding
NR
(2014, 2850916) breastfeeding
: 18.32 ng/mL (95% CI: 16.36,
20.50)
GM Non-breastfeeding:
19.26 (16.80, 22.08)
Zhang et al. Adult
Males and females
Whole blood
170 mL/kg
Mean: 2.71; GM: 2.47
Steady state assumed
(2015,2857764) Adult, pregnant
Females
Whole blood
170mL/kg
Mean: 3.36; GM: 3.09
Steady state not assumed
due to variable PFAS
levels during pregnancy
Worley et al. > 12 years
(2017, 3859800)
Males and females
Blood (2016)
170 mL/kg
bodyweight
Mean: 11.7 jig/L (95 CI: 8.7-
14.6)
NR
>12 years
Males and females
Blood (2010)
170 mL/kg
bodyweight
Mean: 16.3 (95 CI: 13.2-19.6)
NR
Fu et al. (2016, Adult,
Males and females
Serum
170 mL/kg
Mean: 1052 ng/mL
NR
3859819) occupational
Median: 427 ng/mL
Zhang et al. Adults
Males and females
Serum and whole
170 mL/kg
Mean: 3.1 ng/mL
NR
(2013, 3859849)
blood
Shin et al. Adult,
(2011,2572313) nonoccupational
Males
Serum
181 mL/kg
Median predicted: 13.7 ppb;
observed 23.5 ppb (updated
values in Erratum) {Shin, 2013,
5082426}
NR
Adult,
nonoccupational
Females
Serum
198 mL/kg
Median predicted: 13.7 ppb;
observed 23.5 ppb (updated
values in Erratum) {Shin, 2013,
5082426}
NR
Gomis et al. Human and
Males and females
Serum
200 mL/kg
Reports an average of human
Authors note that due to
(2017,3981280) animals
and animal Vd values
declining values in U.S.
and Australian
populations, steady state
was not achieved in the
past decade.
Notes: AUC = area under curve; CI = confidence interval; GM =
geometric mean; NR =
not reported; Vd = volume of distribution.
B-48
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.2.5.2 Animal Studies
In Fujii et al. (2015, 2816710), PFOA distribution in male and female FVB/NJcl mice (8-
10 weeks of age) administered by IV (0.31 |imol/kg) or gavage (3.13 |imol/kg) was determined
using a two-compartment model. Serum PFOA concentrations varied linearly by dose regardless
of route. The Vd after IV injection was calculated as dose/C(0). As shown in Table B-27, the Vd
of PFOA was low in mice after IV injection and exhibited no differences between sexes. The low
serum Vd was consistent with the high percentage (32.3%) of administered dose calculated for
serum. The measured percentage of administered dose was higher in the liver (47.4%) although
Vd for this compartment was not calculated.
In this study, the authors examined PFCAs with chain lengths between 6 and 14 and observed
that Vd increased as a function of chain length in both males and females. The authors suggested
that this may be linked to the lipophilicity of PFCAs and their increasing affinity for serum and
liver fatty acid binding proteins. For PFOA, Vd corresponded to the volume of extracellular
water. Interestingly, Vd values corresponded to different compartments based on chain length,
specifically the total volume of blood for C7 and the volume body water for CI 1 and C12).
Table B-27. PFOA Volume of Distribution in Serum of FVB/NJcl Mice as Reported by
Fujii et al. (2015, 2816710)
Route
Dose (jimol/kg)
Sex
Vd 1 kg-1 a
AUC nmol 1-1 hour (0 to 24 hours)a
IV
0.313
Male
0.18 ±0.04
42.2 ±9.9
IV
0.313
Female
0.15 ±0.04
49.5 ± 11.9
Oralb
3.13
Male
NR
348 ± 76
Oralb
3.13
Female
NR
495 ± 64
Notes: AUC = area under curve; NR = not reported.
a Vd and AUC reported as means ± standard deviation.
b Steady state achieved 8 days after initial dose (oral).
Two recent studies {Kim, 2016, 3749289; Dzierlenga, 2019, 5916078} measured toxicokinetic
parameters in rats, including Vd. 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 (Table B-28). Similar to the Fujii et al. (2015, 2816710) study in mice, Vd values were
similar in males and females. While organ specific Vd values were not determined, the liver and
kidney exhibited partition coefficients greater than 1 in males (2.31 ± 0.38 for liver and
1.18 ± 0.47 for kidney). While the partition coefficients in females for the kidney (1.23 ± 0.39)
were similar to males, they were significantly lower in the livers of females (0.81 ± 0.36)
compared with males. Partition coefficients were similar in males and females for the heart, lung,
and spleen. Although Vd values were not significantly different between males and females, the
differential partition coefficients in liver and kidney may relate to the higher Vd values calculated
for females compared to males.
Dzierlenga et al. (2019, 5916078) calculated the apparent volume of central (Vi) and peripheral
(V2) distribution in rats. In this study, the plasma concentration-time profiles were best described
using one-compartment models in males and a two-compartment model in females. As detailed
in Table B-28, males and females were administered different doses that were higher than those
used in the Kim et al. (2016, 3749289) study. Females were administered 40-320 mg/kg
compared to 6-48mg/kg in males. Several observations were apparent for Vd in males. Vd values
B-49
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
were substantially lower in the peripheral compartment compared to the central compartment,
and VdS were substantially lower in the peripheral compartment after IV administration relative
to oral administration. VdS were similar after oral dosing at 6 and 12 mg/kg (159 ± 12 and
154 ± 11 mL/kg, respectively) and only increased at the highest dose of 48 mg/kg
(202 ±18 mL/kg). In contrast to males after IV dosing, female Vd values were similar in central
and peripheral compartments (108 ± 24 and 98.7 ± 39.8 mL/kg, respectively) although the dose
in females of 40 mg/kg was substantially higher than the 6 mg/kg dose in males.
In females, both peripheral and central VdS were calculated after oral dosing at all doses.
Peripheral Vd values were dramatically lower than central Vd values at all doses by the oral route
(Table B-28). These trends are consistent with the observations that peak tissue levels were
reached readily in both males and females. However, while tissue levels in males were steady
over the course of several days, tissue levels in females dropped quickly (in the span of hours),
which likely reflects the shorter half-life in females.
In a third study {Iwabuchi, 2017, 3859701}, PFOA 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 blood, serum, and several tissues. The whole
blood Vd (0.42 kg tissue volume/kg body weight (BW)) was almost threefold higher than the
serum Vd. Organ Vd values were highest in the brain (9.0 kg tissue volume/kg BW) and spleen
(2.3 kg tissue volume/kg BW). VdS were 1-2 orders of magnitude lower in the heart, kidney, and
liver (0.91, 0.27, and 0.083 kg tissue volume/kg BW, respectively). An interesting observation
from this analysis is that, for PFOA, 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.
A single study examined Vd in primates. Butenhoff et al. (2004, 3749227) calculated a Vd from
noncompartmental PK analysis of data from cynomolgus monkeys. Three males and three
females were administered a single IV dose of 10 mg/kg, and serum PFOA concentrations were
measured in samples collected up to 123 days postdosing. The Vd of PFOA at steady state (Vdss)
was similar for both sexes at 181 ± 12 mL/kg for males and 198 ± 69 mL/kg for females.
B-50
-------
DRAFT FOR PUBLIC COMMENT
Table B-28. Summary of PFOA Volume of Distribution Calculations in Rats
MARCH 2023
Concentration
Study
Metnoa oi v d
Route
Dose
Strain
Age
Sex
Vd
Compartment
Measured in
Cmax
V_- lUV'U.llll'.lvr.l.l
Compartment3
Kim et al.
Dose x
Oral
1 mg/kg
Sprague-
8-12 weeks
Males
106.4 ±8.9
Blood plasma
AUC:
7.55 ± 0.51 ng/mL
(2016,
AUMC/(AUC0-
Dawley
0 mL/kg
24.81 ± 1.41 ng
3749289)
GO) 2
day/mL
Females
153.83 ±9.
Blood plasma
AUC:
5.41 ± 0.38 ng/mL
19 mL/kg
1.39 ± 0.06 ng
day/mL
IV
1 mg/kg
Sprague-
8-12 weeks
Males
Blood plasma
AUC:
8.92 ± 2.34 ng/mL
Dawley
112.12 ±29
21.10 ± 1.51 ng
.41 mL/kg
day/mL
Females
171.37 ±11
Blood plasma
AUC:
5.84 ± 0.38 ng/mL
.19 mL/kg
1.63 ± 0.09 ng
day/mL
Dzierlenga
Standard
Oral
6 mg/kg
Sprague-
8 weeks
Males
159 ± 12 m
Peripheral
AUC: 39.37 ±2.42
0.089 ± 0.007 mM
et al. (2019,
equations
Dawley
L/kg
mM h
5916078)
{Gabrielsson,
12 mg/k
Sprague-
8 weeks
Males
154 ±11 m
Peripheral
AUC: 69.79 ±3.86
0.185 ±0.013 mM
2000,9642135}
g
Dawley
L/kg
mM h
48 mg/k
Sprague-
8 weeks
Males
202 ±18 m
Peripheral
AUC: 178.4 ± 12.1
0.560 ± 0.048 mM
g
Dawley
L/kg
mM h
40 mg/k
Sprague-
8 weeks
Females
73.6 ±20.6
Central
AUC: 5.217 ±0.507
0.580 ± 0.060 mM
g
Dawley
mL/kg
mM h
5.55 ± 1.62
Peripheral
AUC: 5.217 ±0.507
0.580 ± 0.060 mM
mL/kg
mM h
80 mg/k
Sprague-
8 weeks
Females
130 ± 24 m
Central
AUC: 8.066 ± 0.869
0.961 ± 0.118 mM
g
Dawley
L/kg
mM h
19.9 ± 12.9
Peripheral
AUC: 8.066 ± 0.869
0.961 ± 0.118 mM
mL/kg
mM h
320 mg/
Sprague-
8 weeks
Females
272 ±1990
Central
AUC: 57.00 ± 7.97
2.06 ±0.61 mM
kg
Dawley
mL/kg
mM h
69.9 ± 1849
Peripheral
AUC: 57.00 ± 7.97
2.06 ±0.61 mM
.1 mL/kg
mM h
IV
6 mg/kg
Sprague-
8 weeks
Males
114 ± 5 mL/ Central
AUC: 28.0 ± 1.69
0.127 ± 0.006 mM
Dawley
kg
mM h
B-51
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Method of Vd „ ,
Calculation Route
Concentration
Study
Dose Strain Age Sex
Vd
Compartment
Measured in
Compartment3
Cmax
40 mg/k Sprague- 8 weeks Females
g Dawley
39.2 ± 14.5
mL/kg
108 ± 24 m
L/kg
Peripheral
Central
AUC: 28.0 ± 1.69
mM h
AUC: 2.87 ±0.31
mM h
0.127 ± 0.006 mM
0.893 ± 0.196 mM
98.7 ±39.8
Peripheral
AUC: 2.87 ±0.31
mM h
0.893 ± 0.196 mM
Iwabuchi et
al. (2017,
3859701)
Dose / ke x plasma Oral
concentration
(AUC)
100 |ig/k Wistar 7-9 weeks Males
g, single
dose
9.0 kg
tissue
volume/kg
BW
Brain
160 ng/kg tissue
volume - day
8.77 ng/kg
0.91kg
tissue
volume/kg
BW
Heart
1500 ng/kg tissue
volume - day
108 ng/kg
0.083 kg
tissue
volume/kg
BW
Liver
35000 ng/kg tissue
volume - day
1270 ng/kg
2.3 kg
tissue
volume/kg
BW
Spleen
630 ng/kg tissue
volume - day
49.2 ng/kg
0.27 kg
tissue
volume/kg
BW
Kidney
6600 |ig/kg tissue
volume - day
624 ng/kg
0.42 kg
tissue
volume/kg
BW
Whole blood
4300 |ig/kg tissue
volume - day
265 ng/kg
0.15 kg
tissue
volume/kg
BW
Serum
9200 |ig/kg tissue
volume - day
759 ng/kg
B-52
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Notes: AUC = area under curve; AUMC = area under first moment curve; BW = body weight; Cmax = maximum plasma concentration; ke = elimination rate constant; NR = not
reported.
a Presented as AUC or Mean/Median.
B-53
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.3 Metabolism
PFOA does not appear to be metabolized in mammals. In a recent study, Gannon et al. (2016,
3810188) investigated the metabolism of PFOA in vivo and in vitro using rodent models.
Specifically, male and female mice (Crl:CDl(ICR)) and rats (Sprague-Dawley) were exposed to
a single oral dose of PFOA at 3 mg/kg and 30 mg/kg, respectively. Urine samples collected from
both rodent species were analyzed by high-performance liquid chromatography. The authors
subsequently screened for metabolites using the control-comparison tool, IntelliExtractTM. Only
the anionic form of PFOA was detected. There was almost complete recovery of the dose in the
urine, confirming that PFOA is not metabolized. In addition, normal and heat-inactivated rat
hepatocytes (5 x io6 cells/mL) were exposed to 50 |iM of PFOA in a 3-mL suspension. No
differences in clearance rate were found and no metabolites were detected.
B.4 Excretion
B.4.1 Urinary and Fecal Excretion
B. 4.1.1 Human Studies
The majority of human studies predominantly consider PFOA excretion after oral exposure,
either implicitly or explicitly. The urinary excretion of PFOA in humans is impacted by the
isomeric composition of the mixture present in blood and the sex/age of the individual. The
half-lives of the branched-chain PFOA isomers are shorter than those for the linear
molecule, an indication that renal resorption is less likely with the branched chains. Fewer
studies have examined excretion through the fecal route. Animal studies suggest that sex and
competing PFAS compounds influence fecal excretion.
Several major studies highlight the urinary excretion of PFOA in humans. Zhang et al. (2014,
2851103) derived estimates for PFOA'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 detected PFOA concentration in first-draw morning urine samples by the predicted
urinary volume (1600 mL/day for males and 1200 mL/day for females). PFOA was detected in
the blood samples for all participants but for only 76% of the urine samples from the general
population and 30% for the pregnant females. Total daily PFOA intake was modeled for the
general population and used to estimate a daily urinary excretion rate of 25%, but was higher in
males than in females (31% and 19%, respectively). In contrast to the estimates relating to PFOS,
there was little difference in urine:blood ratio between nonpregnant females age 21-50 and those
age 51-61, although the urine:blood ratio was found to be lower for pregnant females than
nonpregnant females (0.0011 and 0.0029, respectively), suggesting the placenta and cord blood
as possible elimination pathways. There was a direct correlation between the PFOA
concentrations in blood and creatinine adjusted urine (r = 0.348 p = 0.013) for the general
population but not for the pregnant females. When limited to the eight females who had
detectable levels in both blood and urine, there was a significant correlation (r= 0.724,
p = 0.042).
B-54
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Zhang et al. (2013, 3859849) calculated median renal clearance rates of 0.16 mL/kg/day in
young women and 0.19 mL/kg/day in men and older women for total PFOA. In a later study, Fu
et al. (2016, 3859819) determined the renal clearance half-lives of PFOA in 302 occupational
workers (213 male, 89 female) from one of the largest producers of PFAS-related compounds in
China. Paired serum and urine samples were collected. The participants were subdivided based
on their work assignment. Serum PFOS and PFHxS were highest in workers of the sulfonation
department and the serum PFOA levels were highest in workers from the electrochemical
fluorination department.
Serum PFOA concentrations were in the ranges of 2.52-32,000 ng/mL (median 424 ng/mL). The
average concentrations of serum PFOA was significantly higher in males (1,215 ± 2,936) ng/mL
than in females (659 ± 743) ng/mL. The median urine concentration for all workers was
4.3 ng/mL (range (LOD - 53.6 ng/mL). The correlation coefficient of PFOA concentrations in
paired serum and urine samples of 0.64 was found to be highly statistically significant,
(p < 0.01), suggesting that urine concentrations could serve as effective bioindicators for PFOA
exposure in occupational settings. Daily renal clearance was calculated for each PFAA as
follows:
Urine PFAA 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/day/kg) followed by PFOS (GM 0.010 mL/day/kg). The high
efficiency of PFOA renal clearance was reflected in the relative abundance of PFOA from 12%
in the serum samples to 42% in the urine samples. Sex did impact daily renal clearance values,
which were significantly lower in males compared to females (p < 0.01).
A single case report study demonstrated fecal excretion of PFOA in humans. Fecal PFOA was
measured in an exposed man before and after treatment with bile sequestering agents {Genuis,
2010, 2583643}. Before treatment, his urine and stool levels of PFOA levels were 3.72 ng/mL
and below detectable limits (0.5 ng/g), respectively. After treatment with cholestyramine, PFOA
measurements in stool increased to 0.96 ng/g in the first weeks after treatment and to 1.19 ng/g
several months later after subsequent treatments with saponins.
Urinary clearance of PFCAs in humans was observed to decrease with increasing alkyl chain
lengths, while biliary clearances increased {Fujii, 2015, 2816710}. In these studies, paired
bile-serum and urine-serum were obtained from the archived samples in the Kyoto
University Human Specimen Bank. Bile samples were taken by nasobiliary drainage,
percutaneous transhepatic biliary drainage or percutaneous transhepatic gallbladder drainage
for 24 hours. Blood samples were taken from the same patients on the same day. Blood and
urine were also collected from healthy volunteers. Human data were analyzed from paired
(bile-serum) archived samples from patients undergoing nasobiliary drainage, percutaneous
transhepatic biliary drainage, or percutaneous transhepatic gallbladder drainage for 24 hours.
Urine-serum pairs were collected from healthy donors. Urinary and biliary clearance was
determined by dividing the cumulative urine or bile excretion in a 24-h period with the
serum concentration. Fecal clearance was calculated using the estimated biliary resorption
rate.
B-55
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The authors estimated that total human clearance was 0.096 mL/kg/day and was 50-100 times
smaller than those clearances estimated in mice after oral gavage dosing. In humans, PFOA
clearance rates via urinary, biliary, and fecal routes were estimated to be 0.044, 2.62, and
0.052 mL/kg/day, respectively. The reabsorption rate of bile excreting PFOA was estimated to be
0.98 (derived by assigning a Vd of 200 mL/kg, a serum half-life of 3.8 years, and the
presumption that that PFOA could only be excreted into the urine and feces via the bile).
Interestingly, PFCAs with chain lengths of C6 and C7 were rapidly excreted into urine, whereas
PFOA and PFCAs with longer chain lengths were deposited mainly in the liver. Thus, chain
length for PFCAs may be a major determinant of bioaccumulation as well as excretion rate and
route. These authors also conducted a toxicokinetics analysis in mice (discussed in the next
section). The ascertained that human urinary clearances for PFCAs were more than 200 times
smaller than those in mice. Fecal clearances in humans were also an order of magnitude lower
than those estimated in mice after oral gavage and IV dosing (ranging from 1.1 to
4.3 mL/day/kg) also differed by one order of magnitude, indicating the other membrane
transporters in the liver may also be involved.
Although no data were identified on urine or fecal excretion of PFOA following inhalation
exposures in humans, the Hinderliter study (2006, 135732) provides evidence of clearance
following single and repeated inhalation exposures in Sprague-Dawley rats. Plasma PFOA
concentrations following a single exposure to 1, 10, or 25 mg/m3 PFOA declined 1 hour after
exposure in females and 6 hours after exposure in males. In females, the elimination of PFOA
was rapid at all exposure levels and, by 12 hours after exposure, their plasma levels had dropped
below the analytical LOQ (0.1 |ig/mL). In males, the plasma elimination was much slower and,
at 24 hours after exposure, the plasma concentrations were approximately 90% of the peak
concentrations at all exposure levels. In the repeated exposure study, male and female rats were
exposed to the same concentrations for 6 hours/day, 5 days/week for 3 weeks. Steady-state
plasma levels were reached in males by 3 weeks, but plasma PFOA levels in females returned to
baseline within 24 hours of each dose.
No data were identified on excretion following dermal exposures. Minimal fecal excretion is
anticipated for the dermal route of exposure although the biliary pathway can be a route for
excretion of material absorbed through the skin, distributed to the liver, and discharged to the
gastrointestinal tract.
B.4.1.2 Animal Studies
Butenhoff et al. (2004, 3749227) studied the fate of PFOA in cynomolgus monkeys in a 6-month
oral exposure study. Groups of four to six male monkeys each were administered PFOA daily via
oral capsule at DRs of 0, 3, 10, and 30/20 mg/kg for 6 months, with urine and fecal samples
collected at 2-week intervals. All dosed groups reached steady-state urine PFOA levels after four
weeks, which were 53 ± 25, 166 ± 83, and 181 ± 100 |ig/mL, respectively. Two monkeys
exposed to 10 mg/kg and three monkeys exposed to 20 mg/kg were monitored for 21 weeks
(recovery period) following dosing. Within two weeks of recovery, urine PFOA concentrations
were < 1% of the value measured during treatment and decreased slowly thereafter. Lower
amounts were excreted in feces. These results are consistent with both renal and biliary excretion
in male monkeys.
B-56
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
There have been a number of studies of excretion in rats because of the sex differences noted in
serum levels. Hinderliter et al. (2006, 3749132) investigated the relationship between age and
urine PFOA concentrations in male and female Sprague-Dawley rats. Immature rats 3, 4, or
5 weeks of age were administered PFOA via oral gavage as a single dose of 10 or 30 mg/kg, and
urine was collected for 24 hours.
Urine PFOA concentrations differed significantly (p < 0.01) with age, dose, and sex. For all
doses and ages, urinary excretion of PFOA was substantially higher in females than in males,
and this difference increased with age, as female excretion increased and male excretion
decreased. In both sexes, urine PFOA was higher (2.5 to 6.5 times) at the 30-mg/kg dose as
compared to the 10-mg/kg dose (Table B-29).
Table B-29. Urine PFOA Concentrations in Male and Female Sprague-Dawley Rats, 24-
Hours After Oral Gavage" as Reported by Hinderliter et al. (2006, 3749132)
Age
(weeks)
Dose
(mg/kg)
Urine PFOA
Male
Female
Mean
SD
Mean
SD
3
10
9.57
4.86
21.17
8.95
4
10
4.53
2.45
23.26
15.27
5
10
4.03
2.36
49.77
24.64
3
30
51.76
28.86
94.89
26.26
4
30
28.70
18.84
104.12
28.97
5
30
15.65
6.24
123.16
51.56
Notes: SD = standard deviation.
Data are presented as mean ± standard deviation (|ig/mL)
Kim and colleagues (2016, 3749289) extended the study of male and female Sprague Dawley
rats to evaluate fecal excretion. They also compared oral and intravenous administration of
PFOA, giving a single 1 mg/mL dose by either pathway. Urine and feces were measured daily
for 12 days in males and females after dosing. Like previous studies, the highest concentrations
were found in urine under all conditions. In males, the levels detected in urine and feces were
very similar from both oral and intravenous exposure. By the oral route, 26.42 ± 2.64 ug was
detected in urine vs. 23.60 ± 9.45 ug in feces. Levels were even more similar in male rats dosed
intravenously (22.47 ± 1.94 ug in urine vs. 21.13 ± 12.31 ug in feces). In contrast, females
excreted much higher levels in urine compared to males and compared to feces. After oral
administration, urine and fecal levels were 124.95 ± 6.38 ug and 24.60 ± 4.18 ug, respectively.
The values measured after intravenous administration were similar to those observed after oral
dosing (131.87 ± 6.82 ug in urine vs. 18.04 ± 1.35 ug in feces). The differences between males
and females in amounts detected in urine and feces translated to significant differences in the
estimated half-life values (1.64 and 1.83 days in males vs. 0.15 and 0.19 days in females by the
oral and intravenous routes).
Other studies comparing urinary and fecal excretion following PFOA administration by gavage
among male Sprague-Dawley rats have found much higher excretion rates from urine than from
feces {Benskin, 2009, 1617974; Cui, 2010, 2919335}. Benskin et al. (2009, 1617974) gave
B-57
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
single doses of 0.5 mg PFOA/kg to each rat and monitored for 38 days, while Cui et al. (2010,
2919335) gave 0, 5, or 20 mg/kg/day over 28 days and monitored for the duration. Among the
single-dose rats, 91-95% of the daily excreted PFOA was eliminated in the urine after the initial
24 hours. On day 3, the mean PFOA concentration in urine and feces were 265 ng/g and 28 ng/g.
The half-life for elimination from plasma in male rats was 13.4 days {Benskin, 2009, 1617974}.
Among the repeated dose rats, a sharp increase in urinary and fecal excretion expressed as
percent of dose/day was observed during week 1 in rats of both dose groups. The excretion rate
leveled off at about 50% for the low-dose animals for the remainder of the 28 days. In the case of
the high-dose animals, the urinary excretion remained level at about 80% for the second and
third weeks and then increased sharply to about 140% at 28 days. The fecal excretion rates
followed an upward trend throughout the 28 days with the terminal percent/day about 25% for
the low-dose group and 40% for the high-dose group.
Studies on male and female CD rats have similar findings to those done in Sprague-Dawley rats;
namely, that females excreted PFOA more efficiently than males, excretion rates increased with
higher dosages, and both sexes excreted more PFOA by urine than by feces. Hundley et al.
(2006, 3749054) examined excretion of PFOA in one male and one female CD rat, giving each a
single dose of 10 mg/kg 14C-PFOA and collecting urine and feces at 12-24 hour intervals for
five days post-dose (Table B-30). Kemper (2003, 6302380) gave either single or repeat doses
ranging from 1-25 mg/kg (Table B-31) and collected urine and feces for 7 or 28 days for females
and males, respectively. Hundley et al. found that the female rat had excreted almost all dosed
14C-PFOA within 48 hours, with urinary excretion accounting for about 2.65 times the amount of
fecal excretion. In the male rat, PFOA was excreted from urine at a similar rate relative to fecal
excretion, but much slower overall; only about 19% had been excreted after 48 hours, and only
34%) after 120 hours. Kemper (2003, 6302380) found that after 28 days, singly dosed male rats
excreted 47-68% of the initial dose; interestingly, while the females consistently excreted more
of the PFOA than males, none of the dose groups were found to eliminate 100% of the 14C-
PFOA after 7 days.
Table B-30.Cumulative Percent 14C-PFOA Excreted in Urine and Feces by Male and
Female CD Rats" as Reported by Hundley et al. (2006, 3749054)
Rat
Hours After Dosing
12
24
48
72
96
120
Male
0.6
8.7
19.2
23.4
30.2
34.3
Female
52.5
96.4
99.8
100.0
100.0
100.0
Notes: 14C-PFOA = 14C-Radiocarbon perfluorooctanoic acid.
a Data are presented in % total dose administered.
Table B-31. Percentage of Dose Excreted in Urine and Feces of Male and Female Sprague-
Dawley Rats exposed to 14C-PFOA via Oral Gavage as Reported by Kemper (2003,
6302380)
Dose and Regimen Sex Urine3 Fecesb
Single Dose 1 mg/kg Male 43.238 ±3.015 14.055 ±4.003
Female 75.872 ±4.066 2.169 ±2.923
B-58
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Dose and Regimen
Sex
Urine3
Fecesb
Single Dose 5 mg/kg
Male
62.201 ±3.656
5.568 ± 1.779
Female
77.867 ±6.034
5.886 ±5.387
Single Dose 25 mg/kg
Male
53.265 ±8.490
12.490 ±4.153
Female
84.381 ± 12.023
1.868 ±2.546
Repeated Dose 1 mg/kg/day
Male
52.430 ±7.959
19.841 ±6.620
Female
68.537 ± 16.631
12.384 ± 15.775
Notes: 14C -PFOA = 14C-Radiocarbon perfluorooctanoic acid; SD = standard deviation.
aData are presented as mean ± standard deviation (|ig/mL)
bData are presented as mean ± standard deviation (|ig/g)
Dose is an important variable that impacts excretion. Rigden et al. (2015, 2519093) exposed
groups of five male Sprague-Dawley rats to doses of 0, 10, 33, and 100 mg/kg/day for 3 days and
maintained them for 3 additional days; overnight urine was collected and body weight was
measured daily. Of greatest interest relative to the limitations on renal resorption, is the dose-
related increase in urine PFOA concentration and urine PFOA concentration per mg creatinine
for the 33- and 100-mg/kg/day groups compared to the 10-mg/kg/day group. The peak in PFOA
excretion normalized to creatinine occurred on day 3 after the cessation of dosing. The
concentration at 33 mg/kg/day was 500 times greater than that at 10 mg/kg/day. At the 100-
mg/kg/day dose, the peak concentration was about 3,200 times greater than for the low dose. The
low-dose excretion was only slightly greater than the controls. The urine results support the renal
resorption hypothesis concept and suggest that there is a threshold limit on resorption that, once
exceeded, dramatically increases PFOA loss in urine. As a consequence, half-life for continuous
low-dose exposures will be longer than for single or short-term high-dose exposures.
Another study {Gao, 2014, 2851191} also compared concentrations in urine and feces of male
and female Wistar rats. A mixture of PFOA/PFNA/PFOS were administered to the rats by
drinking water for 90 days, with each compound at doses of 0, 0.05, 0.5, 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, 0.5, and 5 mg/L. These translated to intake values for
PFOA, PFNA and PFOS of 0.15 and 0.12 mg/kg bw, 1.52 and 1.22 mg/kg bw, and 13.6 and
17.7 mg/kg bw for female and male rats, respectively. At the high dose of 5 mg/L, there were
higher levels of PFOA in urine and feces of males and females. However, and in contrast to that
observed by others, there were far higher levels of PFOS in feces compared to urine for both
males and females. 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 PFNAs.
Hundley et al. (2006, 3749054) examined excretion of PFOA in CD mice, BIO-15.16 hamsters,
and New Zealand White rabbits. One male and one female of each species was given a single
dose of 10-mg/kg 14C-PFOA and housed in metabolism cages. Urine and feces were collected at
12-24 hour intervals for five days post-dose. Additional samples were collected from rabbits at
144 and 168 hours post-dose.
B-59
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Over 120 hours, both mice excreted similar amounts of PFOA, although the male mouse
excreted a greater proportion in feces (3.4% 14C-PFOA in urine and 8.3% 14C-PFOA in feces),
and the female mouse excreted more via urine (6.7% 14C-PFOA in urine and 5.7% 14C-PFOA
in feces). The male hamster excreted far more than the female, although both excreted more via
urine than by feces; the male excreted 90.3% and 8.2% 14C-PFOA in urine and feces,
respectively, and the female hamster excreted 45.3% and 9.3%14C-PFOA. Over 168 hours, both
rabbits excreted most of the original dose, and both predominantly excreted via urine (76.8% and
4.2% 14C-PFOA from the male, and 87.9% and 4.6% 14C-PFOA from the female in urine and
feces, respectively). The cumulative percentages of 14C-PFOA excreted are shown in Table
B-32.
Table B-32. Cumulative Percent 14C-PFOA Excreted in Urine and Feces in Mouse,
Hamster, and Rabbit" as Reported by Hundley et al. (2006, 3749054)
Species
Sex
Hours After Dosing
12
24
48
72
96
120
168
Mouse
Male
0.4
4.1
6.7
8.6
9.1
10.8
-
Female
0.2
4.1
6.5
8.4
9.0
11.0
-
Hamster
Male
67.3
84.5
96.1
97.4
98.2
98.4
-
Female
11.3
24.6
36.4
43.9
50.1
54.0
-
Rabbit
Male
77.8
80.2
80.4
80.4
80.4
80.4
80.4
Female
86.7
90.5
92.0
92.2
92.7
92.9
93.0
Notes: 14C -PFOA = 14C-Radiocarbon perfluorooctanoic acid.
a Data are presented in % of total dose administered
Fujii and colleagues (2015, 2816710) compared elimination in humans and mice exposed to
using a two-compartment model. Toxicokinetics and clearance was investigated in FVB/NJcl
mice exposed by oral gavage and intravenous administration of PFCAs with carbon chain
lengths between C6 and C10. At 24 hours after exposure, urine and feces were collected in
metabolic cages. In mice, the short-chained PFCAs (C6 and C7) were rapidly eliminated in the
urine, whereas long-chain PFCAs (C8 to C14) accumulated in the liver and were excreted slowly
in feces. For PFOA administered IV, urinary clearance was higher in males (13.1 mL/day/kg)
compared to females (9.8 mL/day/kg). PFOA administered by oral gavage was also higher in
males (9.2 mL/day/kg) compared females (6.6 mL/day/kg), but clearance was significantly lower
than rates measured after IV administration.
Fecal clearance of PFOA after IV administration was higher in females (2.0 mL//day/kg)
compared to males (1.1 mL/day/kg). After gavage administration, the opposite was observed
with higher rates observed in males (4.0 mL/day/kg) compared to females (2.4 mL/kg/day). The
feces clearance after 24 hours of gavage administration represents PFOA contained in the bile
and unabsorbed PFCAs that passed through the gut, and this likely accounts for the higher fecal
clearance after gavage dosing. The actual fecal clearances of PFCAs were represented by the
fecal clearances of IV-administrated PFCAs. In contrast to urinary clearance, fecal clearance
rates were still lower than urinary clearance rates by both dosing routes.
B-60
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Interestingly, these authors also estimated urinary and fecal clearance rates in humans, which
were 1-2 orders of magnitude lower than rates estimated in mice. This study illustrates chain
length, sex, and species have dramatic impacts on the rate and route of PFOA excretion.
Studies in animals provide evidence that urine is typically the primary route of excretion but that
sex impacts excretion by both routes, and these sex differences appear to be species-specific.
Limited evidence supports excretion through the fecal route in animals and humans and through
hair in animals. Most studies indicate excretion by the fecal route is substantially lower than that
observed by the urinary route. Excretion through the fecal route appears to be more efficient in
males compared to females and in rodents compared to humans. Also, exposures to mixtures of
PFNAs may also alter the relative amounts of PFOA excreted through the fecal route, quite
possibly due to differential lipophilicity and cellular uptake as well as differential affinities for
transporters associated with chain length and branching. Nevertheless, a comprehensive set of
principles governing resorption by renal, hepatic and enteric routes and how these impact
excretion and retention of PFOA has not been established in either humans or animals.
B.4.2 Physiological and Mechanistic Factors Impacting
Excretion
B. 4.2.1 Renal Resorption
Several studies have been conducted to elucidate the cause of the sex difference in the
elimination of PFOA by rats. Many of the studies have focused on the role of transporters in the
kidney tubules, especially the OATs located in the proximal portion of the descending tubule.
OATs are found in other tissues as well and were discussed earlier for their role in absorption
and distribution. In the kidney, they are responsible for delivery of organic anions (including a
large number of medications) from the serum into the kidney tubule for excretion, as well as
reabsorption of anions from the glomerular filtrate. The transporters are particularly important in
excretion of PFOA because it binds to surfaces of serum proteins (particularly albumin), which
makes much of it unavailable for removal during glomerular filtration. Other transporter families
believed to be involved in renal excretion are the OATPs and the MRPs. However, they have not
been evaluated as extensively as the OATs for their role in renal excretion.
OATs are located on both the basolateral (serum interface) and apical surfaces of the brush
boarder of the proximal tubule inner surface. At the basolateral surface, the OATs transport the
perfluorooctanoate anion from the serum to the tubular cells {Anzai, 2006, 9642039; Cheng,
2008, 758807; Klaassen, 2010, 9641804; Klaassen, 2008, 9642044; Nakagawa, 2007, 2919370;
Nakagawa, 2009, 2919342}. OAT1, 2, and 3 are located on the basolateral membrane surface.
OAT4 and OAT5 are located on the apical surface of the tubular cells, where they reabsorb the
PFOA anions from the glomerular filtrate. Figure B-l diagrams the flow of organic anions such
as the PFOA anion from serum to the glomerular filtrate for excretion and resorption of organic
acids from the glomerular filtrate with transport back to serum. OATs can function for uptake
into the cell across both the basolateral and apical surfaces.
Several MRP transporters also appear to function in the kidney and move organic anions in and
out of cells at both the basolateral surface (e.g., MRP2/4) and the apical surface (e.g., MRP1) as
well as one or more OATPs on each surface {Cheng, 2009, 4116789; Klaassen, 2010, 9641804;
Klaassen, 2008, 9642044; Kusuhara, 2009, 9641810; Launay-Vacher, 2006, 9641802; Yang,
B-61
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
2009, 2919328}. Bidirectional movement of PFOA across both the basolateral and apical
surfaces is driven by concentration gradients and/or active transport. Far more data exist on
PFOA and OATs in the kidneys than on OATPs and MRPs. Abbreviations for individual
transporters on the basolateral and apical surfaces differ across publications. The accepted
convention is to use uppercase letters to refer to human transporters and lowercase letters to refer
to animal transporters. For this report, the data are not reported by species but by transporter
family and the uppercase letters are used.
Proximal Tubule Cells
Excretion Reabsorption
-C3*AsM
r—a Abcal i T*
¦aw« En,1^~
o~ En,2
- —»Oatp4cl
B-62
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Kudo et al. (2002, 2990271) examined the role of sex hormones and OATs on the renal clearance
(CLr) of PFOA. Gonadectomy alone caused an increase in CLR of PFOA in both male and
female rats (14-fold and twofold, respectively). Treatment with testosterone reduced the PFOA
CLr in castrated males and intact females. Conversely, treatment with estradiol increased the
CLr of PFOA in intact male rats, but reduced that of ovariectomized female rats back to normal
values.
Early studies from Kudo et al. (2002, 2990271) and Cheng et al. (2006, 6551310) found that
intact males were found to express less OAT2, more OATPlal, and more OATP3al than their
female counterparts. Castration was found to increase OAT2 and decrease OATPlal.
Ovariectomy increased OAT3 in female rats but did not affect OATPlal, which was already
virtually absence from intact female mice. Treatment with estradiol increased OAT2 in intact
male rats, while 17-P estradiol decreased OATPlal in both castrated and ovariectomized mice
but did not affect OATP3al. Finally, treatment with testosterone increased OAT2 in castrated
rats, while 5a-dihydroxy-testosterone increased both OATPlal and OATP3al in castrated and
ovariectomized mice. Multiple regression analysis of the data suggested that OAT2 and OAT3
are responsible for urinary elimination of PFOA in the rat; however, the possibility of a
resorption process mediated by OATP1 was mentioned as a possible factor in male rat retention
of PFOA. OAT2 and OAT3 are located on the basolateral cell surface. OATP1 is located on the
apical surface of the renal tubule cells {Kudo, 2002, 2990271}.
Based on Hinderliter et al. (2006, 3749132), a developmental change in renal transport occurs in
rats between 3 and 5 weeks of age that allows for expedited excretion of PFOA in females and an
inverse development in males. This was evidenced by changes in measured PFOA in plasma and
urine, such that maturing females experienced decreased plasma PFOA and increased urine
PFOA, while the opposite was seen in males. Taken together with previous information, the
change in female rats seems to involve excretion-promoting OATs {Kudo, 2002, 2990271}
while the change in males seems to involve excretion-reducing OATPs {Cheng, 2006, 6551310}.
Numerous in vitro studies using human embryonic kidney cells (HEK 293) and Chinese hamster
ovary (CHO), time- and concentration-dependent studies as well as competition studies with
known transporters have been utilized to evaluate the role of various transporters in the renal
excretion of PFOA. For example, Yang et al. (2010, 2919288) examined cellular uptake of
PFOA by OATP1A2 in CHO and HEK293 cells transfected with OATP1A2 plasmid DNA or
vector DNA (control). PFOA uptake in OATP1A2-transfected HEK293 cells was no different
than uptake in control cells. Uptake of estrone-3-sulfate (E3S), a known substrate of OATP1A2,
was inhibited -30% in the presence of 100 |iM PFOA (C8). Inhibition varied by PFAS of
different chain lengths (-62% by C9, -70% by C10, -42% by CI 1, and -18% by C12). E3S
uptake was not inhibited by C4-C7.
Other studies observed Michaelis-Menten kinetics in transporter-transfected cells compared to
passive diffusion in control (vector only) cells, and several transporters have been identified as
having PFOA renal transport activity, including OAT1, OAT3, OAT4, OATPlal, and URAT1
{Nakagawa, 2007, 2919370; Nakagawa, 2009, 2919342; Yang, 2009, 2919328; Yang, 2010,
2919288}. Limited data suggest possible roles for OAT2 and OAT1PA2 in uptake of PFOA.
Yang et al. (2009, 2919328) investigated the role of OAT polypeptide lal (OATPlal) in PFOA
uptake. In time-dependent uptake experiments using transfected CHO cells, uptake of PFOA by
B-63
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
OATPlal-transfected cells increased proportionally to time during the first 2 mins of incubation.
Vector-transfected cells had a significant level of uptake of PFOA attributed to nonspecific
passive diffusion. In the concentration-dependent uptake experiments, while saturation levels
were not reached in OATPlal-transfected cells, active PFOA uptake could be derived from the
difference between the uptake of the OATPlal cells and the passive diffusion of the vector-
transfected cells. Based on the results of the uptake and additional inhibition experiments, the
authors suggested that passive diffusion could be an important route of PFOA distribution and
that renal reabsorption in the male rat could be mediated by OATPlal
In vitro studies were supported by in vivo analysis of OATPs gene and protein expression in rat
kidneys {Yang, 2009, 2919328}. OAT polypeptide lal (OATPlal), located on the apical side of
proximal tubule cells and could be the mechanism for renal reabsorption of PFOA in rats. The
level of mRNA of OATPlal in male rat kidney is 5-20-fold higher than in female rat kidney,
OATPlal protein expression is higher in male rat kidneys, and it is regulated by sex hormones.
One of its known substrates is estrone-3-sulfate (E3S). A substantial presence of OATPlal in
male rats would favor resorption of PFOA in the glomerular filtrate and reduce excretion.
Limited evidence exists for a role of OAT and OATP1A2 in PFOA uptake. In transformed HEK
293 cells transfected with OAT 2, prostaglandin F2a uptake by OAT2 was inhibited moderately
by PFOA, 75-85% of control at 10 |imol PFOA, and 65% of control at 100 |imol PFOA
{Nakagawa, 2007, 2919370}. However, in the same study, the authors observed that HEK 293
cells or S2 (cells derived from proximal tubule) transfected with OAT failed to take up
radiolabeled |imol [14C]PFOA. Similarly, Yang et al. (2010, 2919288) observed that PFOA
uptake in OATP1A2-transfected HEK293 cells was no different than uptake in control cells
though they did observe inhibition of E3S uptake. At 100 [j,mol, E3S uptake was inhibited -30%
by PFOA (C8), -62% by PFNA (C9), -70% by PFDA (C10), -42% by PFUnDA (CI 1), and
-18%) by (PFDoDA) C12. E3S uptake was not inhibited by C4-C7 perfluorocarboxylates.
The kinetic response of the OAT1, OAT3, and OATPlal transporters to increasing
concentrations of selected perfluorinated carboxylates also was evaluated by Weaver et al. (2010,
2010072). The change in transport velocity (ng/mg protein/min) with increasing concentrations
of the perfluorinated carboxylate exhibited a Michaelis-Menten-type response. The kinetic data
were analyzed to determine the Km and Vmax, and the results are summarized in Table B-33.
Table B-33. Kinetic Parameters of Perfluorinated Carboxylate Transport by OAT1, OAT3,
and OATPlal as Reported by Weaver et al. (2010, 2010072)
Transporter
PFAS
Km (nmol)
Vmax (nmol/mg protein/min)
OAT1
PFHpA (C7)
50.5 ± 13.9
2.2 ±0.2
PFOA (C8)
43.2 ± 15.5
2.6 ±0.3
OAT3
PFOA (C8)
65.7 ± 12.1
3.8 ±0.5
PFNA (C9)
174.5 ±32.4
8.7 ±0.7
OATPlal
PFOA (C8)
126.4 ±23.9
9.3 ± 1.4
PFNA (C9)
20.5 ±6.8
3.6 ±0.5
PFDA (C10)
28.5 ±5.6
3.8 ±0.3
Notes: Km = Michaelis constant; OAT = Organic Anion Transporter; PFAS = Per- and polyfluoroalkyl substances;
Vmax = maximum rate of transport.
B-64
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The Michaelis-Menten kinetic data (Kmand Vmax (maximum initial rate of an enzyme catalyzed
reaction)) indicate that there are substantial differences in the affinity of the perfluorinated
carboxylate with 8 and 9 carbon chains for OAT3, with PFOA (C8) favored over PFNA (C9).
OAT3 is an export transporter located on the basolateral side of the tubular cells; thus, when
present in a mixture consisting of comparable concentrations of both, renal tubular excretion of
PFOA would tend to decrease excretion of PFNA. For OATPlal, a resorption transporter
located on the apical side of the renal tubular cells, PFNA and PFDA (CIO) have a greater
affinity for the transport protein than PFOA. The kinetic data suggest that the net impact of these
relationships would be to favor excretion of PFOA (C8) over PFNA (C9) and possibly PFDA
(CIO) when all three fluorocarbons are present in the exposure matrix at approximately equal
concentrations. There were minimal kinetic differences between transport of PFHpA (C7) and
PFOA (C8) by OAT1, an export transporter on the basolateral surface of the renal tubular cells.
Sakolish and colleagues developed a 3D microphysiological in vitro model using RPECs
designated as a "kidney tubule chip" of the human proximal tubule {Sakolish, 2020, 6320196}.
The kidney tubule chip results for reabsorption were combined with a physiologically-based
"parallel tube model" {Janku, 1993, 8630776} that was used to model overall renal clearance
kinetics in humans in vivo. When compared to reported in vivo renal clearance (in vivo data were
obtained from Reece et al. (1985, 9642054)) the kidney tubule chip combined with a
physiologically-based kinetic model qualitatively and quantitatively recapitulated in vivo kinetics
in the kidney.
PFOA, used as the positive control in this study, exhibited a low but measurable amount of re-
absorption. The ratio of renal clearance using the combined chip and PBPK model for PFOA was
estimated to be 0.40 [xM at the low dose (0.01 [xM) and 0.32 [xM at the higher dose (1.0 [xM). In
contrast this ratio for creatinine (used as a negative control for resorption) was 0.54 mM and 1.17
mM for doses of 0.1 and 1.0 mM, respectively. The authors suggest the lower than expected
levels of PFOA resorption may be due to one of the following factors: (1) the high degree of
protein binding of PFOA in vivo actually is the primary driver of slow renal clearance as long as
the unbound fraction is < 0.01, with reabsorption contributing to a lesser degree; (2) the lack of a
vascular channel in the tissue chip limits resorption (e.g., tubular secretion is not accounted for);
and (3) basal OAT4 expression in the RPTECs used in the PFOA experiments was relatively low
based on immunohistochemistry observations {Sakolish, 2020, 6320196}.
When considered together, the studies of the transporters suggest that female rats are efficient in
transporting PFOA across the basolateral and apical membranes of the proximal kidney tubules
into the glomerular filtrate, but male rats are not. Males have a higher rate of resorption than
females for the smaller amount they can transport into the glomerular filtrate via OATPlal in the
apical membrane.
Much work remains to be done to explain the sex differences between male and female rats and
to determine whether it is relevant to humans. The broad range of half-lives in human
epidemiology studies suggests a variability in the unbound fraction of PFOA in serum or in
human transport capabilities resulting from genetic variations in structures and consequently in
function. Genetic variations in human OATs and OATPs are described in a review by Zair et al.
(2008, 9641805).
B-65
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
B.4.2.2 Enterohepatic Resorption
In animals, the impact of PFOA on several membrane transporter systems linked to biliary
transport was studied by Maher et al. (2008, 2919367) as part of a more detailed study of PFDA.
A dose of 80 mg/kg by intraperitoneal (IP) injection (propylene glycol: water vehicle) was found
to significantly increase (p < 0.05) the expression of MRP3 and MRP4 in the livers of C57BL/6
mice 2 days after treatment. MRP3 and MRP4 are believed to protect the liver from
accumulation of bile acids, bilirubin, and potentially toxic exogenous substances by promoting
their excretion in bile. There were significant increases in serum bilirubin and bile acids after
PFDA exposure, signifying increased export. Conversely, Western Blot analysis and messenger
ribonucleic acid (mRNA) measurements showed significant decreases (p < 0.05) in the protein
levels for OATPlal, OATPla4, and OATPlb2 following exposure to 40 mg PFOA/kg {Cheng,
2008, 758807}. There was no significant impact on NTCP protein or the serum levels of bile
acids. The OATPs are transporters responsible for the uptake of bile acids and other hydrophobic
substances such as steroid conjugates, ecosinoids, and thyroid hormones into the liver.
These studies, all by the same laboratory, were carried out at high, single-dose exposures, which
limit their value in extrapolating to low- and repeat-dose scenarios. The results suggest a
decrease in the uptake of favored substrates into the liver and an increase in removal of favored
substrates from the liver via bile. Upregulation of MRP3 and MRP4, coupled with decreased
OATp levels, could be beneficial due to increased biliary excretion of bile acids, bilirubin, and
potentially toxic exogenous substances, including PFOA. Based on the results with the more
extensive evaluation of PFDA including mouse strains null for several receptors (PPARa,
constitutive androstane receptor (CAR), pregnane X receptor (PXR), and farnesoid X receptor
(FXR)), the authors concluded that the changes in receptor proteins were primarily linked to
activation of PPARa.
Gastrointestinal elimination of PFOA was reported in a case history of a single human male with
high serum levels of perfluorinated chemicals that was treated with a bile acid sequestrant
(cholestyramine (CSM)) {Genuis, 2010, 2583643}. Before treatment, PFOA was detected in
urine (3.72 ng/mL) but not in stool (LOD = 0.5 ng/g) or sweat samples. After treatment with
CSM for 1 week, his serum PFOA concentration lowered from 5.9 ng/g serum to 4.1 ng/g serum
and stool PFOA levels increased to 0.96 ng/g. This observation suggests that PFOA is excreted
in bile and that enterohepatic resorption via intestinal transporters limits the loss of PFOA via
feces.
Zhao et al. (2017, 3856461) demonstrated that PFOA was a substrate for human OATP1B1,
OATP1B3, and OATP2B1 transporters expressed in liver using in vitro studies of CHO and
HEK-293 cells transfected with transporter cDNA, as well as CHO Flp-In cells expressing
human OATP2B, and compared with wild-type control cells transfected with vector only. Under
these conditions, the three OATPs expressed in human hepatocytes can transport the longer chain
PFOA (C8) and perfluorononanoate (C9), but not the shorter chain perfluoroheptanoate {CI).
The authors suggest that these results may relate to the longer serum elimination half-lives of
these 2 PFCAs.
In summary, relatively few studies have investigated resorption through enterohepatic routes.
The transporters involved in PFOA resorption through these routes may include MRP3 and
MRP4 as well as OATP1A1, OATP1A4, OATP1B1, OATP1B2, OAT2B1, and OAT1B3.
B-66
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Preliminary evidence suggests enterohepatic resorption could limit elimination of PFOA by the
fecal route, including the recent observation that PFOA binds to NTCP, a transporter that
mediates the uptake of conjugated bile acids {Ruggiero, 2021, 9641806}. The extent to which
this pathway operates in vivo and whether enterohepatic resorption plays a substantial role in the
retention of PFOA in humans and animals is still unknown.
B.4.3 Maternal Elimination Through Lactation and Fetal
Partitioning
PFOA 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
BAA, females clearly eliminate PFOA through routes not available to males.
The total daily elimination of PFOA in pregnant females was estimated to be 11.4 ng/day, lower
than the 30.1 ng/day estimated for PFOS {Zhang, 2014, 2850251}. The distribution of PFOA
from maternal serum to the fetus and infants is discussed in detail above (Section. B.2.4). A
study by Zhang et al. (2013, 3859792) exemplifies the routes and amounts of PFOA 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 PFOA was 3.35 ng/mL.
The mean levels in the cord blood, placenta, and amniotic fluid were 58%, 47%, and 1.3%,
respectively, of those in the mother's blood. Thus, pregnant females may eliminate PFOA
through cord blood, placenta, and amniotic fluids. Blood loss during childbirth could be another
source of excretion.
The elimination of PFOA in pregnant women corresponds to an increase in concentrations in the
placenta. Mamsen et al. (2019, 5080595) observed an increase in PFOA accumulation from
gestational age 50 to 300 days, with male placentas showing higher levels of than female
placentas. The authors estimated a placenta PFOA accumulation rate of 0.11% increase per day
during gestation.
Mamsen and colleagues measured placental samples and fetal tissues 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 albeit 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 PFOA were lower than maternal blood. The average concentration of PFOA was
0.17 ng/g in fetal tissues compared to 0.23 ng/g in placenta and 2.1 ng/g in maternal plasma. The
increasing fetal PFOA level with fetal age finding suggest that the rate of elimination of PFAS
from mother to fetus may increase through the gestational period.
The same group {Mamsen, 2019, 5080595} measured PFOA 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, CNS and adipose)
were collected for 38 first trimester pregnancies, 18 second trimester pregnancies and 22 third
trimester pregnancies. PFOA was above LOQ in 100% of maternal serum samples, in 82% of
placenta samples and 70% of fetal organs. In general, the concentrations of PFOA in fetal tissue
increased from first trimester to third trimester except for liver and heart which showed highest
B-67
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
levels in the second trimester compared to the third trimester. Analysis of the placenta: serum
ratio of PFOA revealed a 5.6% higher ratio in male fetuses than in female fetuses (p < 0.05).
These studies support the placenta and fetus as important routes of PFOA elimination in pregnant
women and suggests that the magnitude of elimination may be influenced by the sex of the fetus.
Underscoring the importance of pregnancy as a life-stage when excretion is altered, Zhang et al.
(2015, 2851103) observed that the partitioning ratio of PFOA concentrations between urine and
whole blood in pregnant women (0.0011) was significantly lower (p = 0.017) than the ratios
found in non-pregnant women (0.0028) and may be affected by the increase in blood volume
during pregnancy {Pritchard, 1965, 9641812}.
After birth, women can also eliminate PFOA via lactation. Tao and colleagues (2008, 1290895)
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.
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 PFOA concentrations in breast milk across all samples was 38.5 ng/L (range of
25.1-61.5 ng/L) and the median concentration for all PFAS chemicals measured was 151 ng/L
(range of 105-212 ng/L). Only PFOS concentrations were higher than PFOA with a median
concentration of 47.4 ng/L (36.4-63.8 ng/L).
In this study, pooled breast milk samples were measured to follow the time course of PFOA in
breast milk after birth. Concentrations in breast milk measured 30 days after birth were
significantly higher (ANOVA, p < 0.05) than those measured prior to 7 days after birth. These
findings are contrast with results of other studies. Thomsen et al. (2010, 759807) reported that
breast milk levels of PFOA and PFOS decreased by 7% and 3.1%, respectively, during the first
month after birth. PFOA levels significantly decreased in breast milk over a 4-month lactation
period {Kang, 2016, 3859603}. Demographic factors, maternal diets, sample sizes, the
lactational periods measured may account for these discrepancies.
Lower PFOA levels in the breast milk of multiparous women provides further evidence for
pregnancy and lactation as elimination pathways. Lee and colleagues observed that primiparous
mothers showed higher levels of PFOA in breast milk with a median concentration of 46.0ng/L
compared to 33.4 ng/L for mothers giving birth to more than 1 child (p < 0.05). In another study,
multivariable models estimated that parous women had 40% lower PFOS (95% CI: -56 to
-11%) and 40%) lower PFOA (95% CI: -54, -23%) concentrations compared with nulliparous
women {Jusko, 2016, 3981718}. These authors also measured concentrations in colostrum. The
geometric mean concentration in was 35.3 ng/L for PFOS and 32.8 ng/L for PFOA.
PFOA 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 PFOA concentrations were 1.22, 0.9191 and
0.041 ng/mL in maternal serum, cord serum and breast milk respectively. The observed ratios of
cord and maternal serum for PFOA was 0.78 in this study. However, the ratio between breast
milk and maternal serum was 0.038 ± 0.013 suggesting a low transfer from maternal blood to
breast milk relative to maternal blood to cord blood.
B-68
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Studies in animals support elimination through pregnancy and lactation observed in humans.
Fujii and colleagues (2020, 6512379) used the M/P concentration ratio as a measure of chemical
transferability in FVB/NJcl mice. On PND 8 to PND 13, dams (n = 12) were given a single
administration of PFOA by tail vein injection (3.13 (amol/kg). To facilitate milking, dams were
administered 4.0 U/kg oxytocin and milk was collected from all dams by aspirating with
pulsations using a novel apparatus. After milking, maternal blood was collected to obtain plasma.
Maternal plasma PFOA concentrations were significantly higher than milk (13.78 vs.
4.38 |imol/L, P < 0.05) and the M/P ratios was 0.32. The M/P ratios were similar for PFOA (C8),
PFNA (C9), PFDoDA (C12), and PFTriDA (C13), arguing against a direct relationship with
lipophilicity. Potential roles for binding proteins in breast milk or transporters in breast tissue
have not been investigated.
In summary, partitioning to the placenta, amniotic fluid, fetus, and breast milk represent
important routes of elimination in humans, and may account for some of the sex differences
observed for blood and urinary levels of PFOA by sex and age.
BAA Other Routes of Elimination
Menstruation may be an important factor in the sex-specific differences observed in PFOA
elimination. Zhang et al. (2013, 3859849) estimated a menstrual serum clearance rate
0.029 mL/day/kg. The link between menstruation and PFOA elimination is based on several
observations. First, males and older females have longer PFOA elimination half-lives than young
females (i.e., females of childbearing age) {Zhang, 2013, 3859849}. 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.
Second, several studies examined the 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. 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 PFOA in menstrual blood were not identified. However, for PFOA 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.
Gao et al. (2015, 2851191) examined the possibility that hair could be a potential route of PFAS
elimination. They exposed adult male and female Wistar rats to 0, 0.05, 0.5, and 5 mg/L of
PFOA, PFNA, and PFOS via drinking water for 90 days. The hair samples were cleaned,
sonicated, dried, and alkaline digested to extract PFAAs. PFOA, PFNA, and PFOS were detected
in all the hair samples of treated groups. A dose-dependent increase in hair PFOA concentration
B-69
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
was observed in all exposed animals. The mean hair concentrations of PFOA ranged from 3.31
to 444 ng/g, suggesting that hair may be a potential route for PFOA elimination. Interestingly,
the hair PFOA concentrations for all treatment doses were significantly higher in males than in
females. The sexually dimorphic difference in hair concentrations may be attributed to the sex
differences observed in PFOA elimination rate and the transfer from serum to hair.
Gao et al. (2015, 2851191) also measured the composition of the mixture excreted in in urine,
feces and hair after administration of 0.5 or 0.05 mg/L. As summarized in Table B-34, at the
lower dose of 0.05 mg/mL, PFOA was not detected 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-34. 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: PFNA = perfluorononanoic acid; PFAA = perfluoroalkyl acids; ND = not detected.
aData are presented in % total PFAAs administered. Animals exposed to 0.05 mg/L in Gao et al. {2015,2851191}
Excretion of PFOA through sweat was measured in one study {Genuis, 2013, 2149530}. Sweat
samples were collected during sauna or exercise from 20 human adult subjects. While another
chemical class was readily detected in sweat (polychlorinated biphenyls (PCBs)) no appreciable
levels of PFOA or other PFAS chemicals investigated were detected in sweat despite their
detection in serum. The authors conclude that sweating does not facilitate clearance of PFHxS,
PFOS, or PFOA. In a case report study {Genuis, 2010, 2583643}, excretion through sweat was
also measured in 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 PFOA, were detected in sweat.
Thus far, no single study has conducted a comparative analysis of elimination of PFOA through
all possible routes of excretion. A comprehensive analysis stratified by age and sex would be
necessary to advance the understanding PFOA excretion by all possible routes, and to establish
factors that influence the proportion of PFOA 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
B-70
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
interpreted for excretion (after correction for BW changes). The calculation of PFOA 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:
• The capacity of PFOA to be reabsorbed via renal and enterohepatic routes of excretion
and binding affinities to relevant transporters including OATs, OATPs, MRPs, and
sodium-dependent transporters involved in bile acid transport including NTCP and the
apical sodium-dependent bile acid transporter. Exposures to high levels of PFOA under
acute conditions (e.g., contaminated drinking water) or in occupational settings may result
in saturation of resorption transporters and increased excretion.
• Binding affinity to serum proteins may limit 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.
• Phospholipid lipid binding affinity (phospholipophilicity) can further reduce the unbound
fraction of PFOA as well as uptake into cells. As reported by Sanchez Garcia et al. (2018,
4234856), phospholipophilicity shows the highest correlation to cellular accumulation
data compared to other measures of lipophilicity, raising the possibility that phospholipid
binding affinity could distinguish between high and low accumulating compounds as well
as half-life measures.
• Chain length and branching. The half-lives of the branched-chain PFOA 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.
• 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 PFOA excretion.
• 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.
B. 4.5.2 Human Studies
There have been several studies of half-lives in humans all supporting a long residence time for
serum PFOA with estimates measured in years rather than months or weeks. Using a linear
mixed model, Bartell et al. (2010, 379025) determined an average half-life of 2.3 years based on
a study of the decreases in human serum levels after treatment of drinking water for PFOA
removal was instituted by the Lubeck Public Services District in Washington, West Virginia, and
the Little Hocking Water Association (LHWA) in Ohio.
The results of this assessment showed a 26% decrease in PFOA concentration per year after
adjustment for covariates and a half-life of 2.3 years (confidence interval (CI) = 2.1-2.4). The
only potential confounders determined to be significant were the treatment plant (p = 0.03) and
B-71
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
homegrown vegetable consumption (p < 0.001). This confounder, as well as changes in the
source of drinking water during the study could also have impacted the results.
In another study, the drinking water supply was contaminated with a mixture of perfluorinated
chemicals when a soil-improver mixed with industrial waste was applied upriver to agricultural
lands in Arnsberg, Germany {Brede, 2010, 3859855}. The PFOA levels in the finished drinking
water were measured as 500-640 ng/L in 2006. PFOS and PFHxS also were present. The
estimate for the human half-life was 3.26 years (geometric mean; range 1.03-14.67 years).
Regression analysis of the data also suggested that the elimination rate might have been greater
in younger subjects and older subjects.
Seals et al. (2011, 2919276) determined half-life estimates for 602 residents of Little Hocking,
Ohio, and 971 residents of Lubeck, West Virginia, who were part of the C8 study but had
relocated to a different area of the country. The half-life estimates for Little Hocking ranged
from 2.5-3.0 years (average 2.9 years) and for Lubeck ranged from 5.9-10.3 years (average
8.5 years).
Based on their analysis, the authors suggested that, if their assumptions were correct, a simple
first order elimination model might not be appropriate for PFOA given that the rate of
elimination appeared to be influenced by both concentration and time. There was a difference in
the CL for the two locations even though the range of years elapsed since relocation was the
same for both communities. The authors identified three potential limitations of their analysis:
the cross-sectional design, the assumption that exposure was uniform within a water district, and
a potential bias introduced by exclusion of individuals with serum values <15 ng/mL.
3M {3M, 2000, 8568548; 3M, 2002, 6574114} conducted a half-life study on 26 retired
fluorochemical production workers from their Decatur, Alabama, (n = 24) and Cottage Grove,
Minnesota, (n = 3) plants. The mean serum elimination half-life of PFOA in these workers was
3.8 years (1,378 days, 95% CI: 1,131, 1,624 days) and the median was 3.5 years {Olsen, 2005,
9642064}. No association was reported between the serum elimination half-life and with initial
PFOA concentrations, age, or sex of the retirees, the number of years retired or working at the
production facility, or medication use or health conditions.
Harada et al. (2005, 4564250) studied the relationship between age, sex, and serum PFOA
concentration in residents of Kyoto, Japan. They found that females in the 20-50-year old age
group (all with regular menstrual cycles) had serum PFOA concentrations that were significantly
lower than those in females over age 50 (all post-menopausal). Harada et al. (2005, 4564250)
also estimated the CLr rate of PFOA in humans and found it to be only about 0.001% of the
GFR. There was no significant difference in CLr of PFOA with respect to sex or age group, and
the mean value was 0.03 ± 0.013 mL/day/kg.
Zhang et al. (2013, 3859849) determined half-lives for PFOA 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 CL. The mean half-life for the sum of all PFOA isomers in younger females (n = 12) was
2.1 years (range 0.19-5.2 years) while that for all males and older females (n = 31) was 2.6
(range 0.0059-14 years); the medians were 1.8 and 1.7 years, respectively. The mean values for
the four branched-chain isomers of PFOA were lower than the value for the linear chain,
B-72
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
suggesting that resorption transporters might favor uptake of the linear chain over the branched-
chain isomers. Older females and males have longer half-lives than young females, suggesting
the importance of monthly menstruation as a pathway for excretion {Zhang, 2013, 3859849}.
The rate of serum PFOA 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 (ager range of 15-
50 year) were included in a panel study to estimate PFOA half-lives. Drinking water PFOA
levels were 100 ng/L prior to closure of the waterworks facility and 1.0 ng/L in the unexposed
community. The mean serum levels among the 106 participants 6 months after the end of
exposure was 21.1±14.7 ng/mL. The average decrease in PFOA was 26% of its previous value
each year. The excretion rate constant after the end of exposure was 0.26 (95% CI: 0.24, 0.28)
and was higher in females (0.29) than males (0.25) but this did not reach significance. The mean
half-life was 2.7 years and was also shorter in females (2.4 years) than in males (2.8 years).
There was a high level of inter-individual variation in half-lives.
Fu et al. (2016, 3859819) determined the half-live of PFOA in 302 occupational workers 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 urine concentrations for PFOA among all workers were 4.3 and 1.9 ng/mL,
respectively, whereas in serum, mean and median PFOA were 1052 and 427 ng/mL. The renal
clearance rate for PFOA ranged from 0.00009 to 2.4 mL/kg/day (Geometric mean of
0.067 mg/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 of PFOA was set at 170 mL kg-1 and 230 mL kg-1 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 PFOA estimated using daily clearance rate was 4.1 years (geometric mean value)
and 4.0 years (geometric median value). However, when measured by annual decline rate, the
half-life of PFOA was estimated to be 1.7 years. The GM values of the half-lives of PFOA and
PFOS for men here were 4.7, and 60.9 years (range 0.44-3663 years), respectively, while those
in females were 3.1 and 8.0 years (range 0.76-30475 years). 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.
Worley and colleagues (2017, 3859800) calculated PFOA half-lives in subjects 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
B-73
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
collected blood samples from subjects and followed up with blood and urine measurements in
2016. Biological half-lives were estimated for PFOA using a one-compartment pharmacokinetic
model.
Geometric mean serum PFOA concentrations were significantly higher in subjects (p < 0.0001)
in both 2010 (16.3 ng/L) and 2016 (11.7 ng/L) relative to national averages reported by
NHANES (3.07 ng/L in 2009-2010 and 1.94 ng/L in 2013-2014). Interestingly, the authors
observed a non-significant relationship between PFOA serum and urine concentrations in women
(n = 23, Pearson's r = 0.35) and a significant strong linear relationship in men (n = 22, Pearson's
r = 0.75).
The half-life for PFOS was estimated to be 3.3 years, similar to the 3.9 years estimated for
PFOA. For these calculations, the Vd values were scaled to bodyweight (values of 170 mL/kg
bodyweight for PFOA and 230 mL/kg bodyweight for PFOS were assigned) When the authors
varied the Vd and intake values by 20%, half-life values varied by several months (half-life
estimates for PFOS ranged from 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 PFOA in the airport's
contaminated water were about 1000 times higher than those in the municipal communities
(300 ng/L at airport vs. 0.3 ng/L in municipal water). Specific gravity adjusted urine median
PFOA concentrations were PFOA was 0.031 ng/mL, with a range of 0.010-0.13 ng/mL as
determined from the second to the fifth sampling periods.
The median PFOA concentration in the first serum sample taken from all 26 employees was
9.1 ng/mL and the serum/water ratio was reported as 30. PFOA median concentrations measured
in paired serum and urine samples obtained from the second to the fifth sampling were reported
as 10 ng/mL and 0.031 ng/mL respectively with an average urine/serum ratio of 0.0032. The
significant difference between the serum/water ratio and the urine/serum ratio is suggestive of
the influence of the clearance rate on the overall serum levels (lower the clearance rate and
higher serum levels correlate to longer the half-lives). Similar to Fu and colleagues (2016,
3859819), the half-life of PFOA was estimated as 1.77years.
Half-life estimates in humans rely on measured serum and/or urine concentrations. However,
relatively few studies calculated PFOA half-lives along with measured intake and serum and
urine PFOA concentrations {Xu, 2020, 6781357; Worley, 2017, 3859800; Fu, 2016, 3859819;
Zhang, 2013, 2639569} (Table B-35). PFOA half-life values among these 4 studies varied from
1.7 years in Xu et al. (2020, 6781357) to 4.7 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 females (especially females of
reproductive age) may relate, at least in part, to menstruation as a route of elimination {Zhang,
2013, 3859849}. Second, blood and urine concentrations varied by several orders of magnitude
across these 4 studies. While blood and urine PFOA concentrations varied by two orders of
magnitude across these studies, half-life estimates were similar, ranging from 1.77 to 4.70 years.
This variability in serum and urine concentrations may reflect the role of non-urinary routes of
PFOA excretion; the variability in concentrations may also reflect the difficulty in measuring
B-74
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
renal resorption. Finally, only two studies estimated PFOA intake in subjects {Xu, 2020,
6781357; Worley, 2017, 3859800}. Altogether, there is insufficient data to correlate PFOA
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 PFOA half-life estimates in
humans.
B-75
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-35. Summary of PFOA Concentration in Blood and Urine in Relation to Half-life values in Humans
Study
Number of
Subjects
Age
Primary
Plasma/Serum
Urinary
_ s Exposure Exposure „ ,
Range „ , Concentrations Concentrations
Route
Estimated Half
Life
Considerations
Xu et al.
(2020,
26
19 Males
22-
62 years
Oral, drinking
water
6781357) 7 Females
210 ng/L
(linear)
88 ng/L
(branched)
300 ng/L
Total**
median: 10 ng/mL median: 0.031 ng/mL
(4.1-28 ng/mL) range: 0.010-
13 ng/mL
(not creatinine
adjusted)
1.77 y
• 1 woman was previously
pregnant 2018 during
sampling year
• PFOA also measured in the
private well of one airport
employee living near the
airport (PFOA concentration
in well was lower than the
airport at 0.53 ng/L linear
and <0.3 ng/L branched)
Worley
153 (2010)
2010:
Oral, drinking
NR
2010: GM1
3.9 y
• LOD was 0.01 ng/L,
et al.
63 males
mean 52.0
water
16.3 ng/mL
2016 Creatinine
detection rate 95.6%
(2017,
90 females
2016:
(13.2-19.6 95%
adjusted: mean 0.031
• Clearance rate was not
3859800)
mean 62.6
CI)
ng PFAS/g creatinine
reported
45 (2016)
2016: GM
median 0.024)b
22 males
11.7 ng/mL
23 females
(8.7-14.6, 95%
2016 not adjusted for
CI)
creatinine:
mean 0.027 ng/mL
median 0.022 ng/mL
Fu et al.
302
Males: 19-
Occupational
NR
mean:
mean: 4.3 ng/mL
Male: 4.7 y
• Urinary samples were only
(2016,
213 males
65
1052 ng/mL
median 1.9 ng/mL
Females: 3.1 y
taken from 274 participants
3859819)
89 females
median 41
median
(LOD-53.6 ng/mL)
Overall: 4.1 y
while there were serum
Females:
427 ng/mL,
(not creatinine
samples for every participant
19-50
(2.5-
adjusted)
• For half -life calculation for
median 37
32000 ng/mL).
females, menstrual clearance
was added to renal clearance
•Clearance rate for
PFOA = 0.062 mL/kg-day
Zhang et
86
22-68
Unspecified
NR
mean 3.1 ng/mL
mean 122 ng/g
Young females:
•All participants had paired
al. (2013,
47 males
median 2.3 ng/mL
creatinine
2.1 y
(whole blood/serum and
3859849)
37 females
(0.26-29 ng/mL)
median 23 ng/g
Males and older urine). For young females
creatinine,
females: 2.6 y
menstrual clearance was
B-76
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
. » Number of Age „ „ Plasma/Serum Urinary Estimated Half „
Study „ ® Exposure Exposure „ , T... Considerations
Subjects Range Route Concentrations Concentrations Lite
(3.5-1869 ng/g estimated and added to renal
creatinine) clearance.
• Renal clearance rate for total
PFOA: mean 0.30 mL/day/kg
(young female),
0.77 mL/day/kg (male and
older) female)
Notes: CI = confidence interval; GM = geometric mean; LOD = limit of detection; NR =not reported.
a Measured concentrations in Drinking water at airport before and after mitigation measures. Authors state, "The geometric mean and median value for PFHxS, PFOA, and PFOS
were 14.7 and 11.7,4.1 and 4.0, 32.6 and 21.6 years, respectively, by the daily clearance rates, and they were 3.6,1.7, and 1.9 years estimated by annual decline rates. The half-
lives estimated by the limited clearance route information could be considered as the upper limits for PFAAs, however, the huge difference between two estimated approaches
indicated that there were other important elimination pathways of PFAAs other than renal clearance in human."
bng/g reported in methods but in results reported as (rg/g creatinine.
B-77
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
All human PFOA half-life values identified in the recent literature review are provided in Table
B-36. PFOA half-life values fell within a range from 0.53 years for a branched PFOA in young
females {Zhang, 2013, 3859849} to 22 years in a study of primiparous women in Sweden
{Glynn, 2012, 1578498}. Second, half-life values varied by geographical region. Using a
population model, Gomis et al. (2017, 3981280) derived shorter half-life values for Americans
relative to Australians. Because elimination should be the same at the population level, this
variation may reflect the shorter time frame of biomonitoring data in Australia relative to the
NHANES data set. Third, age and sex difference in PFOA half-lives have not been rigorously
evaluated, though estimates in males are generally longer than those in females {Fu, 2016,
3859819; Gomis, 2017, 3981280; Li, 2017, 4238434} and exhibit an age-related increase
{Genuis, 2014, 2851045; Zhang, 2013, 3859849}. While most studies were conducted in adults
and/or adolescents, at least one study examined PFOA half-lives in a Newborn Screening
Programs {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 NSPs collected in
the state of New York from infants born between 1997 and 2007 were analyzed for PFOA. The
study authors determined the half-life of PFOA using the regression slopes for natural log blood
concentrations vs. the year 2000 and after. The calculated half-life for PFOA was 4.4 years.
Fourth, linear isomers exhibit longer half-lives than branched isomers {Zhang, 2013, 3859849}.
Table B-36. Summary of Human PFOA Half-Life Values
Study
Number of
Subjects
Age Range3
Estimated Half-Life
(years)
Subjects
Bartell et al. (2010,
379025)
200
100 males
100 females
54.5 ± 15
2.3 y
Study of the decreases in human
serum levels after treatment of
drinking water for PFOA removal
was instituted by the Lubeck
Public Services District in
Washington, West Virginia, and
the Little Hocking Water
Association (LHWA) in Ohio.
Source waters for these systems
had become contaminated with
PFAS from the DuPont Works
Plant in Washington, West
Virginia, between 1951 and 2000.
Brede et al. (2010,
3859855)
20 children
22 adult females
23 adult males
Children:
7.4-8.3
Females: 27-
49
Males: 32-71
3.26 y
Subjects exposed to contaminated
drinking water supply s in
Arnsberg, Germany.
3M (2002, 6574114) 9
7 males
2 females
61 (55-64)
4.37 y
(range 1.50 to 14.49 y)
Second interim report with 9
retired fluorochemical production
workers from the 3M Decatur,
Alabama.
Costa et al. (2009,
1429922)
53 males 20-63 5.1 y
(range 2.6-9.7 y)
53 males working in a PFAA
production facility in Italy from
1978 to 2007
Fu et al. (2016, 302 Males: 19-65 based on daily
3859819) 213 males median 41 clearance rate
89 females Male: 4.7 y
Occupationally exposed subjects
working in one of the largest
fluorochemical plants (Henxin
B-78
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Number of
Subjects
Age Range3
Estimated Half-Life
(years)
Subjects
Females: 19-
Females: 3.1 y
Chemical Plant) in Yingcheng,
50
Overall: 4.1 y
Hubei province, China
median 37
based on annual decline
rate
Overall: 1.7 y
Genuis et al. (2014,
53 Father 16-53
Father: 2.61
A family (6 patients) identified to
2851045)
47 Mother
Mother: 2.61
have elevated serum
22 1st male
1st Male child: 2.03
concentrations of PFAAs, likely
child
2nd Female child: 1.85
through
19 2nd female
3rd Male child: 1.80
repeated commercial spraying of
child
4th Male child: 1.59
their home carpets with stain-
17 3rd male
repellents. Patients
child
were treated by intermittent
16 4th male
phlebotomy over a 4-5 year
child 3
period.
Glynn etal. (2012,
1578498)
413 females
19-41
22 y
Primiparous women 3 weeks after
delivery in Uppsala County,
Sweden 1996-2010 (the POPUP
study (Persistent Organic
Pollutants in Uppsala Primiparas).
Gomis et al. (2016,
3749264)
35-60
2.4 y
six occupationally exposed ski
waxers for whom direct and
indirect exposures via inhalation
were characterized.
Gomis et al. (2017, Australia: A
12 + (USA)
Australian men: 2 y
Population based model using
3981280) total of 24-84
<16-> 60
American men: 2.4 y
Australian biomonitoring studies
pools per survey
(Australia)
Australian women:
from 2009-2014 (Toms et al.
containing
1.8 y
2014, 2009) and the National
between 30-100
American women: 2.1 y
Health and Nutrition Survey
individual
(NHANES) from 2003-2011 in
samples.
the USA. A total of 24-84 pools
USA: 2000
per survey were obtained, with
individuals
each pool containing between 30
were sampled
(2007) and up to 100 individual
throughout the
samples (2003, 2009 and 2011)
USA
Study reports intrinsic elimination
half-lives.
Li etal. (2017,
50
15-50
Males: 2.8 y
Subjects in Ronneby, Sweden,
4238434)
Males: 20
Females: 2.4 y
exposed to contaminated water
Females 30
through a municipal water source.
Seals et al. (2011,
602 residents of
<20
2.9 y (Little Hocking)
602 residents of Little Hocking,
2919276)
Little Hocking
20-29
8.5 y (Lubeck)
Ohio, and 971 residents of Lubeck,
OH: 602 Lubeck
30-39
West Virginia, who were part of
WV: 971
40-49
the C8 study but had relocated to a
50-59
different area of the country.
60-69
>70
Splitehoff et al.
240
Newborn
4.4 y
New York State newborn
(2008, 2919368)
infant (1-
screening program blood spot
2 days)
specimens from newborn infants
B-79
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Number of
Subjects
Age Range3
Estimated Half-Life
(years)
Subjects
Worley et al. (2017,
153 (2010)
2010:mean
3.9 y
Residentially exposed population
3859800)
63 males
52.0
from Lawrence, Morgan and
90 females
2016:mean
Limestone Counties, Alabama
62.6
recruited by ATSDR
45 (2016)
22 males
23 females
Xu et al. (2020,
6781357)
26
19 males
7 females
22-62 years
1.77 y
Subjects in Arvidsjaur, Sweden
exposed to contaminated drinking
water occupationally (working at
the airport) and through residential
drinking water
Zhang et al. (2013, 86 22-68 Young females: 2.1 y Healthy volunteers in Shijiazhuang
3859849) 47 males Males and older and Handan, Hebei province,
37 females Females: 2.6 y China, in April-May 2010
n-PFOA young females:
2.3 males and older
females: 2.8
iso-PFOA young
females: 1.4 males and
older females: 2.5
4m-PFOA young
females: 0.64 males and
older females: 1.4
5m-PFOA young
females: 0.53 males and
older females: 1.3
Notes: PFOA = perfluorooctanoic acid; PFAS = perfluorinated alkyl substances; PFAA = Perfluoroalkyl acids,
a Data on age range presented in years (mean ± standard deviation, where applicable).
B.4.5.3 Animal Studies
B.4.53.1 Non-Human Primates
Butenhoff et al. (2004, 3749227) looked at the elimination half-life in monkeys treated for
6 months with 0, 3, 10, and 20 mg/kg/day via capsules. Elimination of PFOA from serum after
cessation of dosing was monitored in recovery monkeys from the 10- and 20-mg/kg dose groups.
For the two monkeys exposed to 10 mg/kg, serum PFOA elimination half-life was 19.5
(r2 = 0.98) days and indicated first-order elimination kinetics. For three monkeys exposed to
20 mg/kg, serum PFOA elimination half-life was 20.8 days (r2= 0.82) and also indicated first-
order elimination kinetics, although dosing was suspended at different time points because of
weight loss.
B.4.5.3.2 Rats
Kemper (2003, 6302380) examined the plasma concentration profile of PFOA following gavage
administration in sexually mature Sprague-Dawley rats. Male and female rats (four per sex per
group) were administered single doses of PFOA by gavage at DRs of 0.1, 1, 5, and 25 mg
PFOA/kg. After dosing, plasma was collected for 22 days in males and 5 days in females.
Plasma concentration vs. time data were then analyzed using noncompartmental PK methods
B-80
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(Table B-37, Table B-38). To further characterize plasma elimination kinetics, animals were
given oral PFOA at a rate of 0.1 mg/kg, and plasma samples were collected until PFOA
concentrations fell below quantitation limits (extended time).
Plasma elimination curves were linear with respect to time in male rats at all dose levels. In
males, plasma elimination half-lives were independent of dose level and ranged from
approximately 138 hours to 202 hours. To further characterize plasma elimination kinetics,
particularly in male rats, animals were given oral PFOA at a dose of 0.1 mg/kg, and plasma
samples were collected until PFOA concentrations fell below quantitation limits (2,016 hours in
males). The estimated plasma elimination half-life in this experiment was approximately
277 hours (11.5 days) in male rats.
Plasma elimination curves were biphasic in females at the 5-mg/kg and 25-mg/kg dose levels. In
females, terminal elimination half-lives ranged from approximately 2.8 hours at the lowest dose
to approximately 16 hours at the high dose. The estimated plasma elimination half- life in the
extended time experiment was approximately 3.4 hours in females. Kemper et al. (2003,
6302380) reported half-lives of 6-8 days for male Sprague-Dawley rats (Table B-37) and 3-
16 hours for females (Table B-38).
Table B-37. PK Parameters in Male Sprague-Dawley Rats Following Administration of
PFOA as Reported by Kemper et al. (2003, 6302380)
Dose
Parameter
0.1 mg/kg
1 mg/kg
5 mg/kg
25 mg/kg
1 mg/kg (IV)
0.1 mg/kg
extended
time
Tmax (hr)
10.25
9.00
15.0
7.5
NA
5.5
(6.45)
(3.83)
(10.5)
(6.2)
(7.0)
Cmax (Hg/mL)
0.598
8.431
44.75
160.0
NA
1.08
(0.127)
(1.161)
(6.14)
(12.0)
(0.42)
Lambda z (1/hr)
0.004
0.005
0.0041
0.0046
0.004
0.0026
(0.001)
(0.001)
(0.0007)
(0.0012)
(0.000)
(0.0007)
Ti/2(hr)
201.774
138.343
174.19
157.47
185.584
277.10
(37.489)
(31.972)
(28.92)
(38.39)
(19.558)
(56.62)
AUCinf (hr.^g/mL)
123.224
1194.463
6733.70
25,155.61
1249.817
206.38
(35.476)
(215.578)
(1392.83)
(7276.96)
(113.167)
(59.03)
AUCinf/D
1096.811
1176.009
1221.89
942.65
1123.384
2111.28
(hr. ng/mL/mg/kg)
(310.491)
(206.316)
(250.28)
(284.67)
(100.488)
(586.77)
Clp (mL/kg.hr)
0.962
0.871
0.85
1.13
0.896
0.51
(0.240)
(0.158)
(0.21)
(0.31)
(0.082)
(0.17)
Notes: AUCinf = area under the plasma concentration time curve, extrapolated to infinity; AUCinf/D = AUCinf normalized to
dose; Clp = plasma clearance; Cmax = maximum plasma concentration; Lambda z = terminal elimination constant; T1/2 = terminal
elimination half-life; Tmax = time to Cmax = NA = Not applicable.
Data presented as mean ± (standard deviation)
B-81
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-38. PK Parameters in Female Sprague-Dawley Rats Following Administration of
PFOA as Reported by Kemper et al. (2003, 6302380)
Dose
Parameter
0.1 mg/kg
1 mg/kg
5 mg/kg
25 mg/kg
1 mg/kg (IV)
0.1 mg/kg
Extended
Time
Tmax (hr)
0.56
1.13
1.50
1.25
NA
1.25
(0.31)
(0.63)
(0.58)
(0.87)
(0.50)
Cmax (Hg/mL)
0.67
4.782
20.36
132.6
NA
0.52
(0.07)
(1.149)
(1.58)
(46.0)
(0.08)
Lambda z (1/hr)
0.231
0.213
0.15
0.059
0.250
0.22
(0.066)
(0.053)
(0.02)
(0.037)
(0.047)
(0.07)
Ti/2(hr)
3.206
3.457
4.60
16.22
2.844
3.44
(0.905)
(1.111)
(0.64)
(9.90)
(0.514)
(1.26)
AUCinf (hr.^g/mL)
3.584
39.072
114.90
795.76
33.998
3.34
(0.666)
(10.172)
(11.23)
(187.51)
(7.601)
(0.32)
AUCinf/D
31.721
38.635
20.78
29.54
30.747
34.39
(hr. iig/mL/mg/kg)
(5.880)
(10.093)
(2.01)
(6.92)
(6.759)
(3.29)
Clp (mL/kg.hr)
32.359
27.286
48.48
35.06
34.040
29.30
(6.025)
(7.159)
(4.86)
(.88)
(9.230)
(3.06)
Notes: AUCinf = area under the plasma concentration time curve, extrapolated to infinity; AUCinf/D = AUCinf normalized to
dose; Clp = plasma clearance; Cmax = maximum plasma concentration; Lambda z - terminal elimination constant; T1/2 = terminal
elimination half-life; Tmax = time to Cmax; NA = not applicable.
Data presented as mean ± (standard deviation)
Gibson and Johnson (1979, 9641813) administered a single dose of 14C-PFOA averaging
11.4 mg/kg by gavage to groups of three male 10-week old CD rats. The elimination half-life of
14C from the plasma was 4.8 days.
Toxicokinetic parameters informing half-lives were derived by comparing oral IV dosing in rats
{Kim, 2016, 3749289}. Sprague-Dawley rats were administered 2 mg/kg PFOA 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). Half-life estimates were substantially longer than those
observed by Kemper (2003, 6302380) in Sprague-Dawley rats, as well in CD rats reported by
Gibson and Johnson (1979, 9641813). As shown in Table B-39, Sex differences were also
observed for other TK parameters including Cmax, Tmax, AUC (calculated from time 0 to infinity)
and Vd indicating more rapid clearance of PFOA in females relative to males.
B-82
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-39. PK Parameters in Male and Female Sprague-Dawley Rats Following Oral and
IV Administration of PFOA as Reported by Kim et al. (2016, 3749289)
1 mg/kg
Parameter Oral IV
Male
Female
Male
Female
Tmax (hr)
2.07 ±0.21*
0.06 ± 0.004
8.92 ±2.34
5.84 ±0.38
Cmax (Hg/mL)
7.55 ±0.51
5.41 ±0.38
NA
NA
AUC (ng-day/mL)
24.81 ± 1.41
1.39 ±0.06
21.10 ± 1.51*
1.63 ±0.09
T1/2 (day)
1.83 ±0.47
0.15 ±0.01
1.64 ±0.44*
0.19 ±0.01
vd
106.40 ± 8.90
153.83 ±9.19
112.12 ±29.41
171.37 ± 11.19
Notes: AUC = area under curve; Cmax = maximum plasma concentration; T1/2 = terminal elimination half-life; Tmax = time to
Cmaxj Vd = volume of distribution.
Data presented as mean ± standard deviation.
*p < 0.05 between male and female.
Lou et al. (2009, 2919359) determined values of 21.7 days (95% confidence interval: 19.5-24.1)
for male CD1 mice and 15.6 days (95% confidence interval: 14.7-16.5) for females for use in
their pharmacokinetic model.
Depending on the experimental conditions, half-lives in rats ranged from 0.03 days in the initial
period of high dose (40mg/kg) exposure in females {Dzierlenga, 2019, 5916078} to 13.4 days in
males exposed to a relatively low dose of 0.4mg/kg {Benskin, 2009, 1617974}. Rats exposed by
the IV route exhibited shorter half-lives than rats administered the same dose by the oral gavage
route {Kim, 2015, 2850129, Dzierlenga, 2019, 5916078}. Similar to humans and mice, half-life
estimates were shorter in females rats compared to males rats.
B.4.5.3.3 Mice
Half-life estimates (15.6 to 21.7 days) in the single mouse study {Lou, 2009, 2919359} were
generally longer than those measured in rats.
A summary of animal half-life values identified in animals is shown in Table B-40. Values in
both primates and rodents were much shorter than those estimated in humans as exemplified by
values reported in days rather than in years. Values in cynomolgus monkeys ranged from 13.6 to
41.7 days {Butenhoff, 2004, 3749227}, and were generally longer than those observed in
rodents, but much shorter than values observed in humans. Depending on the experimental
conditions, half-lives in rats ranged from 0.03 days in the initial period of high dose (40 mg/kg)
exposure in females {Dzierlenga, 2019, 5916078} to 13.4 days in males exposed to a relatively
low dose of 0.4mg/kg {Benskin, 2009, 1617974}. Rats exposed by the IV route exhibited shorter
half-lives than rats administered the same dose by the oral gavage route {Kim, 2015, 2850129;
Dzierlenga, 2019, 5916078}. Similar to humans and mice, half-life estimates were shorter in
females rats compared to males rats. In contrast, female half-life values exceeded male values in
cynomolgus monkeys suggesting species-specific factors impacting elimination across sexes.
Similar to results in humans, PFOA isomers exhibited shorter half-lives compared to linear
forms.
B-83
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table B-40. Summary of Animal PFOA Half-life Values Identified in the Literature Review
Study
Species and Strain Exposure Route
Age or
Lifestage
Sex
Dose
Estimated Half Life
Butenhoff et al.
(2004, 3749227)
Monkey, cynomolgus IV
3-4 years
Male
Female
10 mg/kg
10 mg/kg
13.6, 13.7, and 35.3 for 3
males
26.8, 29.3, and 41.7 for 3
females
Lou et al. (2009,
2919359)
Mice, CD-I Oral
70-80 days
Male
Female
1 and 10 mg/kg
1 and 10 mg/kg
21.7
15.6
Benskin et al. (2009,
1617974)
Rat, Sprague-Dawley Oral
Adult (429g)
Male
0.4 mg/kg n-PFOA
(0.5 mg/kg PFOA)
n-PFOA: 13.4
iso-PFOA: 8.11
4m-PFOA: 4.32
5m-PFOA: 3.95
3m-PFOA: 6.26
tb-PFOA: 2.25
5,3/5,4m2-PFOA: 1.79
4,4m2-PFOA: 1.28
B8-PFOA: 9.10
Dzierlenga et al.
(2019, 5916078)
Rat, Sprague-Dawley IV
8 weeks
Male
Female
6 mg/kg - Tl/2 initial phase
6 mg/kg - Tl/2 terminal phase
6 mg/kg - Tl/2 overall
40 mg/kg - Tl/2 initial phase
40 mg/kg - Tl/2 terminal phase
2.8 ± 1.4
10.3 ± 1.2
6.4 ±0.5
0.03 ± 0.02
0.22 ±0.01
Oral
8 weeks
Male
Female
6 mg/kg - Tl/2 overall
12 mg/kg - Tl/2 overall
48 mg/kg - Tl/2 overall
40 mg/kg - Tl/2 initial phase
40 mg/kg - Tl/2 terminal phase
40 mg/kg - Tl/2 overall
80 mg/kg - Tl/2 initial phase
80 mg/kg - Tl/2 terminal phase
12.5 ±0.7
10.8 ±0.5
8.96 ± 0.42 hours
0.11 ±0.02
1.23 ±0.4
0.11 ±0.03
0.16 ±0.02
1.82 ± 1.13
B-84
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Species and Strain Exposure Route
Age or
Lifestage
Sex
Dose
Estimated Half Life
80 mg/kg - Tl/2 overall
0.16 ±0.03
320 mg/kg - Tl/2 initial phase
0.06 ± 1.09
320 mg/kg - Tl/2 terminal phase
0.75 ±0.11
320 mg/kg - Tl/2 overall
0.58 ±4.20
Kemper (2003, Rat, Sprague-Dawley Oral
6302380)
Sexually
mature
Male
0.1 mg/kg
1 mg/kg
5 mg/kg
25 mg/kg
1 mg/kg (IV)
0.1 mg/kg extended
8.4
5.8
7.3
6.6
5.8
11.5
Female
0.1 mg/kg
1 mg/kg
5 mg/kg
25 mg/kg
1 mg/kg (IV)
0.1 mg/kg extended
0.1
0.1
0.2
0.7
0.1
0.1
Kim et al. (2016, Rat, Sprague-Dawley IV
2850129)
8-12 weeks
Male
Female
1 mg/kg
1 mg/kg
1.64 ±0.44
0.19 ±0.01
Oral
8-12 weeks
Male
1 mg/kg
1.83 ±0.47
Female
1 mg/kg
0.15 ±0.01
Kudo et al. (2002, Rat, Wistar IV
9 weeks
Male
48.63 mol/kg body weight
5.68 ±0.99
2990271)
Female
48.63 mol/kg body weight
0.08 ±0.03
Notes: IV = intravenous injection.
a Data presented in mean days ± standard deviation unless otherwise noted.
B-85
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix C. Non-priority Health System
Evidence Synthesis and Integration
C.l Reproductive
EPA identified 64 epidemiological and 16 animal studies that investigated the association
between PFOA and reproductive effects. Of the 22 epidemiological studies addressing male
reproductive endpoints, 2 were classified as high confidence, 15 as medium confidence, 4 as low
confidence, and 1 was considered uninformative (Section C.l.l). Of the 52 epidemiological
studies addressing female reproductive endpoints, 5 were classified as high confidence, 25 as
medium confidence, 20 as low confidence, and 2 were considered uninformative (Section C.l.l).
Of the animal studies, 4 were classified as high confidence, 11 as medium confidence, and 1 was
considered low confidence (Section C.1.2). 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
PFOA 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, 3603279}
reports identified limited evidence of effects of PFOA on reproductive effects in men and boys.
One study {Joensen, 2009, 1405085} of Danish men in the military (n = 105) showed non-
significant inverse associations with serum PFOA and semen volume, sperm concentration,
sperm count, sperm motility, and sperm morphology. Comparing men with combined
PFOA/PFOS serum levels revealed significantly (p < 0.05) less morphologically normal sperm
in those men with higher PFOA/PFOS levels compared to those with low PFOA/PFOS levels.
No associations were observed for serum sex hormones in this study. In healthy young Danish
males Joensen et al. (2013, 2851244) observed no associations with reproductive hormones.
Semen parameters were also assessed in men from the Longitudinal Investigation of Fertility and
the Environment Study (LIFE) cohort {Buck Louis, 2015, 2851189}, and significant associations
were observed for a few morphological parameters, including fewer coiled tails, increased
curvilinear velocity, and a larger acrosome area of the head. One prospective birth cohort study
{Vested, 2013, 2317339} followed offspring for approximately 20 years after mothers provided
a third trimester blood sample. Regarding prenatal PFOA exposure, a significant negative trend
was observed for total sperm count with 34% reductions in total count for each of the highest
two tertiles compared to the lowest PFOA tertile. Additionally, prenatal PFOA exposure was
associated with higher follicle stimulating hormone (FSH) (responsible for stimulating testicular
growth) and luteinizing hormone (LH) (responsible for stimulating testosterone production)
concentrations in these men after 20 years. Three occupational studies {Olsen, 1998, 1290857;
Sakr, 2007, 1291103; Costa, 2009, 1429922} observed minimal evidence of reproductive effects
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
in male employees. A study {Olsen, 1998, 1290857} on male employees (n = 111) at a
Minnesota PFOA production plant (1993-1995) observed non-significant elevated estradiol (E2)
in the highest PFOA exposure group; however, the study authors suggest this may have been
confounded by a high correlation between E2 and BMI. A study {Sakr, 2007, 1291103} of
employees at a DuPont facility in West Virginia observed associations for serum E2 and
testosterone, but they did not address circadian variations in hormone levels and concluded the
biological significance of the result was unclear. No other associations were observed in
occupational studies evaluating males.
For this updated review, 21 studies (22 publications)3 report on the association between PFOA
and male reproductive effects since the 2016 document. 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 {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 a reproductive medical center in Nanjing, China {Pan, 2019,
6315783; Cui, 2020, 6833614}. One pair of studies assessed populations from related cohorts
belonging to the Hokkaido study on the Environment and Children's Health {Itoh, 2016,
3981465; Goudarzi, 2017, 3981462}.
Eleven studies were in children and 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}, and the remainder of the publications were on the
general population. Different study designs were utilized, including four cohort studies {Ernst,
2019, 5080529; Goudarzi, 2017, 3981462; Itoh, 2016, 3981465; Jensen, 2020, 6311643} with
the remainder of the studies following a cross-sectional design. All observational studies
measured PFOA in blood components (i.e., blood, plasma, or serum); however, PFOA in semen
was additionally measured in four studies {Cui, 2020, 6833614; Di Nisio, 2019, 5080655; Pan,
2019, 6315783; Song, 2018, 4220306}. The studies were conducted in different study
populations including populations from Australia, China, Denmark, the Faroe Islands,
Greenland, Italy, Japan, Poland, Taiwan, Ukraine, and the United States. While most studies
evaluated the relationship between exposure to PFOA and sex hormone concentrations, other
male reproductive outcomes investigated included: sex-hormone related steroid hormones (e.g.,
dehydroepiandrosterone (DHEA)), pubertal markers (e.g., voice break), semen analysis, genomic
effects in sperm (e.g., DNA methylation), and anthropometric measurements (e.g., anogenital
distance (AGD), penis length, etc.).
C.l.1.1.2 Study Quality
There are 22 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and male reproductive effects. Study quality evaluations for these 22
studies are shown in Figure C-l.
3 Zhou et al. (2016, 3856472) and Zhou et al. (2017, 3858488) use differing methods to analyze participants from the same
population using the same health outcome.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Of the 22 studies identified since the 2016 assessment, two studies were classified as high
confidence, 15 studies as medium confidence, four studies as low confidence, and one study
{Song, 2018, 4220306} was determined to be uninformative. Publications from the GBCA
{Zhou, 2016, 3856472; Zhou, 2017, 3858488} were considered low confidence because of
concerns of selection bias and confounding. Cases and controls in Zhou et al. (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 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 et al. (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.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Arbuckle et al., 2020, 6356900 -
Cui et al, 2020, 6833614-
Di Nisio et al., 2019, 5080655 -
Ernst et al., 2019, 5080529-
Goudarzi et al., 2017, 3981462 -
Itoh et al., 2016, 3981465-
Jensen et al., 2020, 6311643 -
Kim etal.,2020, 6833596-
Kvist et al., 2012, 2919170-
Leter et al., 2014, 2967406 -
Lewis et al., 2015, 3749030 -
+
+
++
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
+
+
-
-
++
+
+
+
++
+
+
+
+
+
+
+
+
+
+
+
+
+
+
++
+
+
+
+
+
++
+
++
+
+
++
+
+
+
-
+
+
+
+
+
+
++
+
+
+
+
+
+
+
+
-
+
+
+
+
+
+
+
+
+
+
-
Lind et al., 2017, 3858512-
+
+
+
++
Liu etal.,2020, 6569227-
Lopez-Espinosa et al., 2016, 3859832 -
Pan et al., 2019, 6315783-
Petersen et al., 2018, 5080277 -
Song et al., 2018, 4220306-
Tian et al., 2019, 5390052-
Tsai et al., 2015, 2850160-
Wang et al., 2019, 5080598 -
Zhou etal., 2016, 3856472-
Zhou etal., 2017, 3858488-
+
+
+
+
+
+
+
+
+
+
+
+
+
++
++
¦
+
+
+
+
+
+
-
+
+
+
+
+
-*
-
-
+
+
+
++
+
++
+
+
+
-
+
+
+
+
+
+
+
+
++
+
+
++
+
+
+
+
+
+
-
+
+
-
-
+
+
+
+
+
+
-
Legend
3 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-1. Summary of Study Evaluation for Epidemiology Studies of PFOA and Male
Reproductive Effects
Interactive figure and additional study details available on HAWC.
D-4
-------
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 five observed significant effects (Appendix D). A high confidence
study {Jensen, 2020, 6311643 } in boys from to the Odense cohort observed a borderline
significant positive association between prenatal PFOA and FSH at four months (p = 0.06), but
no associations for other serum sex and steroid hormones (i.e., androstenedione, 17-
hydroxyprogesterone (17-OHP), and dehydroepiandrosterone sulfate (DHEAS)). A medium
confidence study {Goudarzi, 2017, 3981462} examined male children from the Sapporo cohort,
in the Hokkaido Study on the Environment and Children's Health and observed a significant
inverse association (p = 0.025) with DHEA in cord blood. Associations were not observed
among other androgenic hormones. Results from an overlapping medium confidence study {Itoh,
2016, 3981465} from the Hokkaido cohort were largely non-significant except for a significant
increase in inhibin B in cord blood. Quartile analyses supported this association, but the trend did
not reach significance (p = 0.063). A medium confidence study {Liu, 2020, 6569227} in male
infants in China observed a significant positive association with progesterone in cord blood.
Kmedium confidence cross-sectional study {Lopez-Espinosa, 2016, 3859832} of boys (6-
9 years) recruited from residents residing near the Mid-Ohio Valley DuPont chemical plant (C8
Health Project) observed a significant inverse association with testosterone, and a significant
inverse trend (p for trend = 0.030) by quartiles of PFOA. In contrast, a cross-sectional study {Di
Nisio, 2019, 5080655} in Italian high school students examined associations between PFOA
levels and possible risk factors for diseases of the male reproductive system and observed
significantly increased semen PFOA levels, testosterone, and LH (p = 0.003) in exposed
individuals compared to unexposed controls. These studies report effects in opposite directions,
however, the significance of this conflicting evidence is not entirely clear as each population had
reached different points in pubertal development. Additionally, Di Nisio et al. (2019, 5080655)
only controlled for age in all analyses, which may result in some residual confounding by SES or
smoking.
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 PFOA
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 significantly
increased semen levels and a reduced number of sperm with normal morphology (p < 0.001) and
a slight increase in semen pH (p = 0.005) in exposed individuals compared to controls.
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,
2017, 3858512} in children from the Odense cohort observed non-significant smaller AGD and
penile width at three months of age with increasing PFOA. Children from the Shanghai-Minhang
Birth Cohort Study {Tian, 2019, 5390052} were evaluated at birth, six months, 12 months of age
for changes in AGD. At six months of age, significant decreases were observed for the second
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
lowest quartile. The effect was consistent in direction for higher quartiles of PFOA exposure but
did not reach significance. At 12 months of age, associations were positive, but none were
significant. 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 in 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 PFOA were significantly correlated (Spearman's r = 0.646, p < 0.01). Total and free
testosterone were inversely associated (p < 0.05) with serum and with semen PFOA levels. E2
and the total testosterone-LH ratio were inversely associated (p < 0.05) with semen PFOA, but
not with serum PFOA levels. Analyses by quartile agreed and showed significant inverse trends
for all outcomes with significant associations in continuous analyses. Analyses stratified by age
showed these associations remained in participants 30 years old or younger but were not
observed in those participants over 30 years of age. A medium confidence cross-sectional study
{Petersen, 2018, 5080277} on Faroese men also observed a decrease in free testosterone with
increasing serum PFOA levels, however, the association was borderline significant (p = 0.05).
The free testosterone-E2 ratio was inversely associated (p = 0.02) with PFOA levels in this
sample. One study {Lewis, 2015, 3749030} analyzed sex hormone concentrations among
NHANES participants, but no clear patterns or significant effects were observed.
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). A medium confidence study {Pan, 2019,
6315783} in men from Nanjing, China observed significant positive associations (p < 0.05) with
sperm concentration, total sperm count, and the sperm DNA fragmentation index (DFI)—a
measure of the percentage of sperm with damaged DNA. In analyses by quartiles, significant
associations were observed for sperm concentration and for the second and fourth quartiles,
however, the trend was not significant. Positive associations were observed for sperm DFI
among the two highest quartiles of exposure, and the trend was significant (p for trend = 0.03). A
significant inverse association (p = 0.03) was observed with progressive motility with a
significant decreasing trend (p for trend = 0.02). Related motility measures, such as sperm
curvilinear velocity and sperm straight-line velocity, did not have significant inverse trends in
continuous analyses, however, an inverse association was observed for the highest quartile of
exposure for each outcome. No other consistent trends for semen parameters were identified
using semen concentrations of PFOA, and no associations were observed with serum PFOA.
One medium confidence study {Kvist, 2012, 2919170} evaluating men from the INUENDO
cohort from Greenland, Poland, or Ukraine, observed a significant positive association (p = 0.05)
with the Y:X chromosome ratio in sperm when pooling data across study countries. This
association was also observed in the Ukraine subset of the cohort but not in other country-
specific analyses. Chromosomal changes were further characterized in another INUENDO study
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
{Leter, 2014, 2967406} using a sperm DNA global methylation assay. Methylation of the LINE-
1 loci was significantly increased (p < 0.05) in men from Ukraine, but no effect was observed in
other INUENDO communities or in the pooled analysis. The LINE-1 loci are a non-transposonic
repetitive satellite DNA sequence generally observed in or adjacent to every centromere and was
used as a surrogate marker of global DNA methylation.
C. 1.1.2 Female
C 1.1.2.1 Introduction
Reproductive health outcomes of interest in females vary by stage of biological maturity and by
pregnancy status. Of interest across the life stages, reproductive hormone levels, such as
prolactin, FSH, LH, testosterone, and E2, are commonly examined as indicators of reproductive
health. Additional reproductive health outcomes of interest include timing of puberty among
children and adolescents; fertility indicators, impacts to menstruation, and occurrence of
menopause among non-pregnant adult females; and gestational hypertension, preeclampsia, and
breastfeeding duration among pregnant females.
The 2016 HESD for PFOA {U.S. EPA, 2016, 3603279} concluded that there was suggestive
evidence of an association with risk of pregnancy-induced hypertension or preeclampsia based
on studies in highly exposed (C8 Health Project) populations {Darrow, 2013, 2850966; Savitz,
2012, 1276141; Savitz, 2012, 1424946; Stein, 2009, 1290816}. There was conflicting evidence
from two studies on altered female pubertal onset, and there were suggestive data from two
studies on reduced fecundity and fertility. Limited suggestive findings on age at menarche or
onset of menopause were hampered by the potential for reverse causation due to PFOA excretion
via menstruation. One study examined female reproductive hormone levels in the C8 Health
Project {Knox, 2011, 1402395} and found no association between PFOA and E2 levels.
For this updated review, 49 studies (53 publications) report on the relationships between PFOA
exposure and female reproductive outcomes.4 Of these, 21 were cohort studies, 20 cross-
sectional studies, and 12 case-control studies. Twenty-one studies were conducted in adults, six
were in children and adolescents, 11 were in both adults and children, and 15 were conducted in
pregnant women. Most studies assessed exposure to PFOA using biomarkers in blood. Others
used amniotic fluid and follicular fluid.
C.1.1.2.2 Study Quality
There are 52 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and female reproductive effects. Study quality evaluations for these
52 studies are shown in Figure C-2, Figure C-3, and Figure C-4.
Among the 52 publications available for review, five were classified as high confidence, 25 as
medium confidence, 20 as low confidence, and two were considered uninformative. Because
menstruation is a primary route of PFOA excretion, reverse causality was a specific concern for
cross-sectional studies that measured blood PFOA and reproductive hormones with known
menstrual fluctuations that failed to report sample collection timing {Heffernan, 2018, 5079713;
4 Singular studies with two associated publications include Avanasi et al. (2016, 3981413) and Avanasi et al. (2016, 3981510);
Dhingra et al. (2016, 3981508) and Dhingra et al. (2017, 3981432); Wang et al. (2019, 5080500) and Wang et al. (2019,
5080598); Zhou et al. (2017, 3858488) and Zhou et al. (2017, 3859799).
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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}. The low confidence studies had deficiencies in participant selection {Zhang, 2018,
5079665; Heffernan, 2018, 5079713}, exposure measurement methods {Avanasi, 2016,
3981413; Avanasi, 2016, 3981510; Campbell, 2016, 3860110}, reliance on self-reporting for
exposure, outcome, or covariate information {Avanasi, 2016, 3981413; Avanasi, 2016, 3981510;
Campbell, 2016, 3860110}, and small sample size {Heffernan, 2018, 5079713; McCoy, 2017,
3858475}. Maekawa et al. (2017, 4238291) was considered uninformative for this assessment
because of lack of information on participant selection and lack of adjustment for key
confounders in the analysis. Lee et al. (2013, 3859850) was also considered uninformative due to
lack of consideration of key confounders in analyses.
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.
Commonly assessed effects were pregnancy-related outcomes (e.g., preeclampsia, gestational
hypertension), menstrual dysfunction (e.g., endometriosis, cycle irregularity), female fertility
indicators, and female reproductive hormone levels (e.g., E2, testosterone, sex hormone binding
globulin (SHGB)). Other female reproductive outcomes discussed in this review include
breastfeeding duration, genital tract infection rate, and female pubertal milestones.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1 1 1 1 1 1 1 1 L
Arbuckle et al„ 2013, 2152344 -
+
+
+
"
+
+
+
~
Avanasi et al., 2016, 3981413-
+
-
+
+
+
+
Avanasi et al., 2016, 3981510 -
+
-
-
+
+
+
+
-
Bach et al., 2015, 3981559-
+
+
4-+
++
+
+
Bach et al„ 2018, 5080557-
"
+
+
4-+
++
+
+
"
Bangma et al., 2020, 6833725 -
+
"
+
"
-
+
Borghese et al., 2020, 6833656 -
+
++
++
+
++
+
+
+
Buck Louis et al., 2016, 3858527 -
+
++
++
+
+
+
+
Campbell et al., 2016, 3860110 -
++
+
+
++
+
+
-
Caserta et al., 2013, 2000966 -
+
+
+
-
+
+
"
"
Caserta et al., 2013, 2001177 -
+
+
+
-
+
"
"
Crawford etal., 2017, 3859813-
+
++
++
+
+
+
+
Dhingra et al., 2016, 3981508 -
+
+
+
+
++
+
+
+
Dhingra etal., 2017, 3981432-
+
+
+
+
+
+
+
Ding et al„ 2020, 6833612-
+
++
+
+
++
+
+
++
Donleyetal., 2019, 5381537-
+
+
++
+
+
+
+
+
Ernstet al., 2019, 5080529-
++
+
+
+
++
+
+
+
Legend
y 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-2. Summary of Study Evaluation for Epidemiology Studies of PFOA and Female
Reproductive Effects
Interactive figure and additional study details available on HAWC.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
.©
,o®
Goudarzi etal., 2017, 3981462-
Heffernan et al„ 2018, 5079713
Huang et al., 2019, 5083564 - +
Huo et al., 2020, 6505752
Itoh etal., 2016, 3981465-
Jensen et al., 2015, 2850253
Jensen et al., 2020, 6311643 -
Kim et al., 2020, 6833596
Lee etal., 2013, 3859850
Lewis etal., 2015, 3749030
Liew et al., 2020, 6387285-
Liu etal., 2020, 6569227-
Lopez-Espinosa etal., 2016, 3859832-
Louis et al., 2012, 1597490-
Lum et al., 2017, 3858516-
Lyngsa et al., 2014, 2850920-
Maekawa et al„ 2017, 4238291
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-3. Summary of Study Evaluation for Epidemiology Studies of PFOA and Female
Reproductive Effects (Continued)
Interactive figure and additional study details available on HAWC.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
ec*°<\e,^6
,0®
Maisonet et al,, 2015, 3859841 -
i
+
|
+
i
+
i
+
i
+
i
+
+
+
McCoy et al., 2017, 3858475-
¦
D
++
-
-
-
-
Mitro et al., 2020, 6833625-
++
++
++
+
+
+
++
Petro et al., 2014, 2850178 -
B
++
++
+
-
-
-
-
Romano etal., 2016, 3981728-
+
+
+
++
-
+
+
Rylander et al., 2020, 6833607 -
+
+
-
+
+
+
+
Starling et al., 2014, 2446669-
+
¦
+
++
+
+
+
Timmermann et al., 2017, 3981439 -
+
+
+
+
++
+
+
+
Tsai etal., 2015, 2850160-
-
+
+
+
+
+
+
+
Vagi et al., 2014, 2718073-
-
+
-
-
+
+
-
-
Wang etal., 2017, 3856459-
+
+
+
+
+
+
+
+
Wang et al., 2019, 5080598 -
+
++
n
+
++
+
+
+
Wikstrom et al„ 2019, 5387145-
+
++
++
+
+
+
+
+
Yao et al., 2019, 5187556-
+
++
B
+
++
+
+
++
Zhang etal.. 2018, 5079665-
-
-
+#
+
+
+
+
-
Zhou et al., 2016, 3856472-
+
+
+
-
-
+
+
¦
Zhouetal., 2017, 3858488-
-
+
+
+
+
+
+
-
Zhou et al„ 2017, 3859799-
+
-
-
+
+
+
+
-
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-4. Summary of Study Evaluation for Epidemiology Studies of PFOA and Female
Reproductive Effects (Continued)
Interactive figure and additional study details available on HAWC.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.l.1.23 Findings from Children and Adolescents
Two high confidence, eight medium confidence, and two low confidence studies assessed
relationships between PFOA 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 {Jensen, 2020, 6311643; Yao, 2019, 5187556} and four medium
confidence {Liu, 2020, 6569227; Itoh, 2016, 3981465; Goudarzi, 2017, 3981462; Wang, 2019,
5080598} studies examined the association between PFOA exposure and female reproductive
hormones in female infants. One medium cross-sectional analysis reported a significant positive
association between cord blood PFOA and cord blood estriol in female infants (beta: 0.29, 95%
CI: 0.02, 0.56) {Wang, 2019, 5080598}. Two high {Jensen, 2020, 6311643; Yao, 2019,
5187556} and three medium confidence studies {Liu, 2020, 6569227; Itoh, 2016, 3981465;
Goudarzi, 2017, 3981462} observed no significant associations between maternal serum or cord
blood PFOA levels and reproductive hormones, such as 17-OHP, DHEA, FSH, and LH {Jensen,
2020, 6311643}, E2, testosterone, or testosterone-to-E2 ratio {Yao, 2019, 5187556}
progesterone {Liu, 2020, 6569227}, prolactin, SHBG, testosterone, DHEA, androstenedione
{Itoh, 2016, 3981465; Goudarzi, 2017, 3981462}.
Three medium confidence studies and one low confidence study examined the effects of PFOA
exposure on female reproductive hormone levels in female adolescents with mixed results. Two
medium confidence studies observed positive associations with E2 in a high exposed population
{Lopez-Espinosa, 2016, 3859832} and testosterone {Maisonet, 2015, 3859841}. As part of the
C8 Health Project, Lopez-Espinosa et al. (2016, 3859832) observed significantly increased E2
levels in serum PFOA quartile 2 compared to quartile 1 (percent difference = 12.6; 95% CI: 3.0,
23.1), but smaller non-significant, positive associations were observed for girls in the two highest
PFOA quartiles. In daughters from the Avon Longitudinal Study of Parents and Children
(ALSPAC), Maisonet et al. (2015, 3859841) reported a positive association for total testosterone
at age 15 when analyzed by maternal serum PFOA tertiles (beta for maternal PFOA tertile 2 vs.
tertile 1: 0.15, 95% CI: -0.02, 0.32; beta for tertile 3 vs. tertile 1: 0.24, 95% CI: 0.05, 0.43).
Maternal serum PFOA was not significantly associated with daughter's SHBG levels. No
associations were observed for follicular stimulating hormone or SHBG in a medium confidence
study {Tsai, 2015, 2850160} or for E2 or testosterone in a low confidence study {Zhou, 2016,
3856472}.
One medium confidence study and one low confidence study reported no evidence of an
association between prenatal PFOA exposure and pubertal milestones in female adolescents.
Breast development, pubic hair development, axillary hair development, and age at menarche
were not associated with maternal blood PFOA during pregnancy in 555 adolescent girls from
the Danish National Birth Cohort (DNBC) {Ernst, 2019, 5080529}. Zhou et al. (2017, 3859799)
reported positive associations between PFOA and risk of hypomenorrhea (OR for PFOA quantile
3 (Q3) vs. quantile 1 (Ql): 2.68, 95% CI: 1.24, 5.78), irregular menstrual cycle (OR for PFOA
quantile 4 (Q4) vs. Ql: 1.99, 95% CI: 1.22, 3.24; OR per log increase PFOA: 1.52, 95% CI: 1.08,
2.15), and long menstrual cycle (OR for PFOA Q4 vs. Ql: 1.95, 95% CI: 1.21, 3.14; OR per log
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
increase PFOA: 1.5 (1.06, 2.1) among female adolescents aged 10-15 years. However, the
analyses were not adjusted for key confounders in this low confidence study.
C.1.1.2.4 Findings from Pregnant Women
Seven studies examined the relationship between PFOA exposure and preeclampsia (Appendix
D). Of these, six observed positive non-significant associations {Huang, 2019, 5083564;
Borghese 2020, 6833656; Rylander, 2020, 6833607; Wikstrom, 2019, 5387145; Avanasi, 2016,
3981510; Avanasi, 2016, 3981413} and one observed a negative non-significant association
{Huo, 2020, 6505752}. Huo et al. (2020, 6505752), a high confidence cohort study of 3,220
pregnant women, observed non-significant decreased odds of preeclampsia in women with
higher serum PFOA levels (OR for women in the 80th percentile or higher for serum PFOA (ln-
ng/mL) vs. women below the 80th percentile = 0.92; 95% CI: 0.5, 1.7; OR per unit increase in
serum PFOA (ln-ng/mL) = 0.89; 95% CI: 0.5, 1.57). All four medium confidence studies
observed, positive non-significant associations between PFOA exposure and preeclampsia, in
cross-sectional {Huang, 2019, 5083564}, case-control {Rylander, 2020, 6833607} and cohort
studies {Wikstrom, 2019, 5387145; Borghese, 2020, 6833656}. One low confidence study re-
analyzed data from a study reviewed in the 2016 HESD, Savitz et al., 2012, and observed non-
significant, positive associations between modeled serum PFOA levels and odds of preeclampsia
{Avanasi, 2016, 3981510; Avanasi, 2016, 3981413}.
One high confidence study {Huo, 2020, 6505752} and two medium confidence studies examined
the relationship between PFOA exposure and gestational hypertension reporting non-significant
mixed effects. Huo et al. (2020, 6505752), a high confidence cohort study of 3,220 pregnant
women, observed non-significant increased odds of gestational hypertension in women with
higher serum PFOA levels. Similarly, Borghese et al. (2020, 6833656) found non-significant
increased odds of gestational hypertension for women in plasma PFOA tertile 3 vs. tertile 1 and
per log2-ng/mL unit increase in plasma PFOA. In contrast, Huang et al. (2019, 5083564) reported
non-significant reduced odds of gestational hypertension with increasing maternal plasma PFOA
levels in both tertile and continuous analyses. When exploring the association between PFOA
exposure and impacts on blood pressure, Borghese et al. (2020, 6833656) found a significant
positive association between first trimester plasma PFOA (ug/L) and systolic blood pressure
(SBP) (beta: 0.82; 95% CI: 0.23, 1.42; p = 0.006) and diastolic blood pressure (DBP) (beta: 0.64;
95%) CI: 0.24, 1.05; p = 0.002). A significant relationship was also observed between continuous
plasma PFOA (ug/L) measured at delivery and SBP (beta: 1.52; 95%> Ci: 0.52, 2.50; p = 0.002)
as well as DBP (beta: 1.11; 95%> CI: 0.44, 1.78; p = 0.001). Results were less consistent when
stratified by infant sex.
Two medium confidence studies assessed the relationship between serum PFOA levels in
pregnancy and breastfeeding duration and both reported significant, inverse associations between
the two {Timmermann, 2017, 3981439; Romano, 2016, 3981728}. Using data from two Faroese
birth cohorts (N = 1,130), one study observed significant, negative associations between
maternal serum PFOA (ng/mL) and both exclusive (regression coefficient per doubling of serum
PFOA (ng/mL): -0.5 months; 95% CI: -0.7, -0.3 months) and total (regression coefficient per
doubling of serum PFOA (ng/mL): -1.3 months; 95% CI: -1.9, -0.7 months) breastfeeding
duration {Timmermann, 2017, 3981439}. These observations were supported by a prospective
birth cohort study which observed a consistent, positive trend between increasing serum PFOA
quartile and relative risk of breastfeeding duration at three and six months postpartum. Relative
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
risk of breastfeeding termination at three months postpartum was significantly increased for
women in serum PFOA quartiles 3 (risk ratio (RR) = 1.63; 95% CI: 1.16, 2.28) and 4
(RR = 1.77; 95% CI: 1.23, 2.54) compared to quartile 1. Relative risk of breastfeeding
termination at six months postpartum was also significantly increased for women in serum
PFOA quartiles 3 (RR = 1.38; 95% CI: 1.06, 1.79) and 4 (RR = 1.41; 95% CI: 1.06, 1.87)
compared to quartile 1.
One high confidence study examined SHBG measured three years postpartum in 812 women
enrolled in the Project Viva birth cohort {Mitro, 2020, 6833625}. The study observed a negative
non-significant association between early pregnancy plasma PFOA and SHBG. These findings
were consistent in analyses stratified by age at pregnancy ( < 35 years vs. > 35 years).
One medium confidence study {Lyngs0, 2014, 2850920} examined the effects of serum PFOA
levels on pre-pregnancy menstruation. The study reported significantly increased odds of long
menstrual cycles for women in the highest PFOA tertile compared to the lowest (OR: 1.8, 95%
CI: 1.0, 3.3) and when analyzing PFOA as a continuous variable (OR: 1.5 (95% CI: 1.0, 2.1).
Significant results persisted when analyses were restricted to nulliparous women.
CI.1.2.5 Findings from the General Adult Population
One high confidence, eight medium confidence, and eleven low confidence studies assessed
relationships between PFOA exposure and female reproductive outcomes in non-pregnant adult
women (Appendix D). Assessed outcomes included various fertility indicators, age at natural
menopause, and reproductive hormone levels.
Five medium confidence studies and eight low confidence studies examined female fertility
indicators and no clear associations or dose-response trends were observed. A cohort study of
501 couples attempting to conceive observed positive significant associations but no trend across
baseline serum PFOA tertiles for day-specific probability of pregnancy or menstrual cycle length
{Lum, 2017, 3858516}. Crawford et al. (2017, 3859813) observed positive association with
cycle-specific time to pregnancy and anti-Mullerian hormone (AMH), a biomarker of ovarian
reserve, and a negative association with day-specific time to pregnancy, but the associations
were non-significant. A low confidence study examining time to pregnancy {Bach, 2018,
5080557} reported a positive association. Another study of AMH examined levels in female
adolescents in the ALSPAC and found a significant positive association between maternal serum
PFOA during pregnancy and AMH concentration (beta: 0.05; 95% CI: 0.01, 0.09). This
association was not significant after missing data imputation {Donley, 2019, 5381537}. A low
confidence study investigated PFOA exposure and premature ovarian insufficiency (POI),
reporting no significant associations {Zhang, 2018, 5079665}, while another low confidence
study found positive associations between the highest PFOA tertile and polycystic ovary
syndrome when compared to the lowest PFOA tertile {Vagi, 2014, 2718073}. Wang et al. (2017,
3856459) observed no associations and no trend in odds of endometriosis-related infertility
across plasma PFOA tertiles. Campbell et al. (2016, 3860110) reported increased odds of
endometriosis only for the third PFOA exposure quartile compared to the lowest PFOA quartile
(OR: 5.45; 95% CI: 1.19, 25.04), while another low confidence study did not observe an
association with endometriosis diagnosis {Louis, 2012, 1597490}. Kim et al. (2020, 6833596)
observed a positive non-significant association between PFOA in follicular fluid and fertilization
rate. Other low confidence studies examining fertility-related outcomes reported non-significant
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
positive associations between PFOA exposure and percent fertilization {McCoy, 2017,
3858475}, minimal correlation with expression of nuclear receptors when examined by fertility
status {Caserta, 2013, 2000966}, and no association between maternal serum PFOA and
infertility {Bach, 2013, 3981559}.
The two studies (3 publications) examined age at natural menopause, and all observed positive
associations. A high confidence study of premenopausal women aged 45-56 in the Study of
Women's Health Across the Nation (SWAN) cohort {Ding, 2020, 6833612} reported a
significantly increased risk of natural menopause for women in the highest exposure tertile
(HR = 1.31; 95% CI: 1.04, 1.65), but no significant association per doubling of serum PFOA. A
medium confidence study (2 publications) {Dhingra, 2016, 3981508; Dhingra, 2017, 3981432}
of women ages 30-65 years in the high exposed Mid-Ohio Valley cohort assessed associations
between both measured and modeled PFOA exposure and self-reported menopause). Menopause
was significantly associated with serum PFOA (p-trend = 0.04), but not modeled PFOA exposure
(p-trend = 0.90) {Dhingra, 2017, 3981432}. However, the findings might be hampered by
reverse causation, likely due to reduced kidney function, as urine is a primary route of PFOA
excretion.
One medium confidence study and four low confidence studies assessed the relationship between
serum PFOA levels and reproductive hormone levels in non-pregnant adult women. In the
medium confidence study, no clear dose-response trends were observed for either FSH or SHBG
across quartiles by age category {Tsai, 2015, 2850160}. While one low confidence study
observed mixed associations between PFOA levels and increased testosterone, with a significant
positive association reported for controls {Heffernan, 2018, 5079713}, another {Zhang, 2018,
5079665} observed no significant associations between PFOA and any female reproductive
hormone outcomes, including E2, prolactin, testosterone, LH, and FSH. Two other low
confidence studies, Lewis et al. (2015, 3749030) and Petro et al. (2014, 2850178), reported no
association for total testosterone or E2, respectively.
C.1.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 4 studies from the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} and 12 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOA HESD that investigated the association between PFOA and reproductive effects. Study
quality evaluations for these 16 studies are shown in Figure C-5.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Biegel et al„ 2001, 673581 -
++
++
L
NR
++
++
•
+
++
'A
++
++
Butenhoffetal., 2004, 1291063-
++
NR
++
++
+
++
++
++
++
Butenhoff et al., 2012, 2919192-
+
++
NR
-
+
++
++
B
++
B
Chen et al., 2017, 3981369-
++
NR
NR
+
++
+
+
+
++
B
Crebelli et al., 2019, 5381564 -
+
+
NR
+
+
+
+
+
~
Hu et al., 2012, 1937235-
+
+
NR
++
++
B
++
++
++
B
Lau et al., 2006, 1276159-
+
+
NR
B
B
++
++
++
B
Li et al., 2018, 5084746-
+
+
NR
++
B
++
B
++
Lu et al., 2016, 3981459-
+
+
NR
fl
++
++
++
B
NTP, 2019, 5400977-
++
++
NR
++
++
++
9
++
++
++
NTP, 2020, 7330145-
++
++
NR
++
++
++
++
++
++
++
Perkins et al., 2004, 1291118 -
B
++
NR
++
++
++
++
B
Song et al., 2018, 5079725-
++
+
NR
++
B
B
B
++
B
Zhang et al., 2014, 2850230-
+
+
NR
++
++
B
++
B
++
Zhang et al., 2020, 6505878 -
+
+
+
B
++
++
++
++
++
B
Zhang et al., 2021, 10176453-
++
+
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-5. Summary of Study Evaluation for Toxicology Studies of PFOA and
Reproductive Effects
Interactive figure and additional study details available on HAWC.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Several animal studies report significant effects on reproductive endpoints following oral
exposure to PFOA; however, the evidence is not consistent across species with effects observed
in mice more frequently than in rats or monkeys. In addition, the effects were observed at dose
levels that have been shown to reduce growth and body weight in several studies, which may
explain the effects observed on reproductive endpoints. Effects observed in male rodents include
reduced fecundity (mice only), decreased epididymal weights, decreased sperm count and
quality, and morphological changes in the testes and epididymides. Female rodents exposed to
PFOA have displayed prolonged diestrus and reduced number and size of corpora lutea
compared to vehicle controls. In addition, alterations in reproductive hormone levels have been
observed in male and female rodents. Oral studies in mice and rats report effects in altered
puberty (delayed vaginal opening in females and altered preputial separation in males).
Developmental studies in mice have reported adverse effects on the weight and histopathology of
the placenta (see Main PFOA Document), and there have been cancers observed in reproductive
organs that are discussed in (see Main PFOA Document).
C. 1.2.1 Reproductive Performance
One standard two-generation reproduction study is available for PFOA that reported no effects
on mating or fertility in rats administered PFOA by gavage for 10 weeks prior to mating with
doses ranging from 1 to 30 mg/kg/day {Butenhoff, 2004, 1291063; York, 2010, 2919279}.
Reproductive endpoints including number of days in cohabitation, fertility index, pregnancy,
implantation, and length of gestation were not affected in either generation. Although Fi pups
exposed to 30 mg/kg/day had decreased birth weight and survival (see Main PFOA Document),
no effects were observed on reproductive performance or fertility in these animals as adults.
Reproductive outcomes in WT mice dosed orally from GD 1-17 with 0.1, 0.3, 0.6, 1, 3, 5, 10
and 20 mg/kg were examined. In the WT mice, the number of implantation sites, number of live
and dead pups per litter and maternal weight were not affected by PFOA. However, the
incidence of full litter resorption was significantly increased at doses of 5 mg/kg/day or higher
{Abbott, 2007, 1335452}. Similarly, the number of pups per litter in CD-I mice exposed to 0.1
and 1 mg/kg PFOA from GD 1.5-17.5 did not significantly differ from control groups {Cope,
2021, 10176465}.
Information on the reproductive performance of mice exposed to PFOA prior to and during
mating is available from two studies. Fecundity was decreased in male BALB/c mice following
exposure to 5 mg/kg/day PFOA by gavage for 28 days when mated to untreated females, shown
by reductions in the numbers of mated females per male mouse and pregnant females per male
mouse {Lu, 2016, 2850390}. The authors did not measure body weight or sperm parameters in
the treated males and did not report if any clinical signs of toxicity were observed, therefore it is
difficult to interpret the toxicological significance of the effect on reproductive performance. In
contrast, Hu et al. (2012, 1937235) administered PFOA (0.02, 0.2, or 2 mg/kg/day) to female
C57BL/6N mice by daily gavage from the day they were paired with untreated males through
weaning of offspring. On average, females were dosed for 12.9 (±7.3) days prior becoming
pregnant. No effects were observed in the number of days to pregnancy or the number of dams
that became pregnant between treated groups and controls {Hu, 2012, 1937235}.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C. 1.2.2 Sperm Parameters
Sperm parameters were quantitatively measured in two studies in rats {NTP, 2019, 5400977;
Butenhoff, 2004, 1291063; York, 2010, 2919279} and two studies in mice {Zhang, 2014,
2850230; Li, 2011, 1294081}. Overall, the findings were not consistent between rats and mice
and therefore do not provide clear evidence of an adverse effect on spermatogenesis.
In a short-term study by NTP, male Sprague Dawley rats were administered 0.625, 1.25, 2.5, 5,
or 10 mg/kg/day PFOA by gavage for 28 days and sperm parameters were evaluated in the
control and three highest dose groups at the end of the treatment period (sample size n = 10)
{NTP, 2019, 5400977}. Cauda epididymal sperm count was significantly decreased (24%) in the
high-dose group compared to controls, but when normalized to sperm count per gram of cauda
epididymis, the difference was no longer statistically significant. No effects were observed on
epididymal sperm motility or testicular spermatid counts. Histopathological examination of the
epididymis revealed hypospermia and exfoliated germ cells in one rat each in the 5 and
10 mg/kg/day groups, though the findings were not significantly different from the control
group. Body weight was significantly reduced in males treated with dose levels >2.5 mg/kg/day
and the highest dose group weighed 19% less than controls at necropsy. This could explain the
reduction in sperm count observed at that dose level. A two-generation reproduction study in
Sprague Dawley rats with doses up to 30 mg/kg/day PFOA found no treatment-related effects on
epididymal sperm count, density, motility, or morphology, as well as testicular spermatid count
or density (sample size n = 28-30) {Butenhoff, 2004, 1291063; York, 2010, 2919279}. The
incidences of hypospermia and exfoliated germs cells in the epididymis were slightly higher for
Po males treated with 30 mg/kg/day vs. controls (2/14 vs. 0/13 for each finding); however, it is
not clear if statistical analyses were performed for those results.
Zhang et al. (2014, 2850230) administered 0.31, 1.25, 5, or 20 mg/kg/day PFOA to adult male
BALB/c mice by gavage for 28 days, but sperm parameters were only evaluated in the control
and 5 mg/kg/day groups (sample size n = 5). At the end of the treatment period, epididymal
sperm count was significantly decreased (32%) in the 5 mg/kg/day group compared to controls.
Sperm motility and progressiveness were also significantly reduced. In addition, the rates of head
and neck teratosperm were significantly increased as was the overall rate of teratosperm.5 Body
weights were not reported in this study, and it is unclear if the mice in the 5 mg/kg/day group
experienced concurrent systemic toxicity.
Li et al. (2011, 1294081) also evaluated sperm parameters in a study designed to examine the
involvement of mouse and human PPARa in male reproductive effects induced by PFOA. Adult
male wild-type, PPARa-humanized, and PPARa-null mice of a 129/Sv background were
administered 1 or 5 mg/kg/day PFOA by daily gavage for 6 weeks. At the end of the treatment
period, body weights did not differ between the control and treated groups. Epididymal sperm
count and motility were unaltered by treatment (sample size n = 8-10); however, the percentage
of sperm abnormalities was significantly increased in both treated groups of wild-type and
5 The text of Zhang et al. (2014,2850230) reports that sperm motility and progressiveness were both significantly reduced and
the overall rate of teratosperm was significantly increased in treated rats, but the results in figures lD(b), (c), and (d) show the
opposite effects. It appears that the figures are mislabeled, and the results were switched. The corresponding author was
contacted for clarification, but no response was received.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
humanized PPARa mice, but not in PPARa-null mice. Therefore, the effects observed in this
particular study are potentially related to PPARa.
The overall evidence is suggestive of an effect of PFOA on spermatogenesis, but there are
several limitations with the dataset that make interpretation difficult. The studies that observed
adverse effects on sperm parameters did not evaluate fertility or fecundity, while the only study
that found an effect on fecundity did not measure sperm parameters or report if overt toxicity
occurred in the males. Furthermore, the studies in mice used relatively small sample sizes (n = 5-
10), while a comprehensive two-generation study in rats with large sample sizes (n = 28-30)
observed no effects on sperm parameters or male fertility {Butenhoff, 2004, 1291063; York,
2010, 2919279}. Epididymal sperm concentration was reduced by 24% in rats treated with
10 mg/kg/day {NTP, 2019, 5400977} and by 32% in mice treated with 5 mg/kg/day {Zhang,
2014, 2850230}; however, the reduction observed in rats was negated when normalized to
weight of the cauda epididymis {NTP, 2019, 5400977}. The study in mice did not normalize
sperm count to organ weight to determine if the effect remained significant. Furthermore, body
weights of rats were significantly reduced at the same dosage that caused reduced sperm
concentration, which could explain the effect on sperm. Body weights were not reported by
Zhang et al. (2014, 2850230) to determine if that was also a confounding factor in mice.
Increased rates of sperm abnormalities were reported in two studies with mice {Zhang, 2014,
2850230; Li, 2011, 1294081}, but not observed in the two-generation study in rats {York, 2010,
2919279}. In summary, it is unclear if the effects on spermatogenesis observed in mice are the
result of direct toxicity to reproductive processes or a reflection of PFOA's effects on body
weight or other systemic effects. Figure C-6 summarizes the effects of PFOA on sperm counts
observed in animal studies.
NTP ¦>!
llUtKIHI
Animal Description
Mim* . RAI Hv id" N-<>>
D,i 1. , \ liu
Hi R H I ill i a si NIC A S5!5 \ \-2 • liu
( IhSD.liiMIR cf N—'M llii
i, -Diult i id" N-I0>
4 Kt pi-miui liu I t1< i K - Sjit rm < mint
illssitEEEIEIsi
T
—
Figure C-6. Sperm Counts in Rodents Following Exposure to PFOA (logarithmic scale)
PFOA 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; Po = parental generation; Fi = first generation; PND = postnatal day; d = day.
C. 1.2.3 Estrous Cyclicity and Ovarian Function
A small number of studies have evaluated estrous cyclicity and effects on corpora lutea
following oral exposure to PFOA, and some significant effects have been observed.
A tendency toward prolonged diestrus was reported in one study with rats {NTP, 2019,
5400977} and in one study with mice {Zhang, 2020, 6505878}. In the study by NTP, adult
female rats were treated for 28 days with doses up to 100 mg/kg/day and estrous cyclicity was
evaluated daily during the last 16 days of treatment. The cycles of treated rats were observed to
be mostly similar to controls; however, rats dosed with 100 mg/kg/day spent around 20% more
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
time in diestrus than controls (62.5% vs. 51.9% of the cycle). Markov analyses indicated that
high-dose females had a higher probability than control animals to transition from a regular cycle
to a cycle with prolonged diestrus (p < 0.001). No effects were observed in the mean estrous
cycle length or the lengths of time spent in other estrous stages. The body weights of females
were not significantly altered by treatment {NTP, 2019, 5400977}.
A two-generation reproduction study in rats {Butenhoff, 2004, 1291063} found no evidence of
extended diestrus in Po or Fi female rats, but the doses were lower than the NTP study and the
authors did not specifically evaluate the proportion of time spent in diestrus. The study authors
observed a significant increase in the number of estrous stages per 21 days in the high-dose
(30 mg/kg/day) Fi females compared to controls (5.4 vs. 4.7 estrous stages/21 days); however,
there were no significant differences observed in the incidences of rats displaying prolonged
diestrus or estrus (defined as > 6 days for each), and no significant changes were observed in the
estrous cycles of females in the P generation. The slight increase observed in number of estrous
stages per 21 days was most likely due to the different stages the rats entered the measurement
period and was probably not related to PFOA treatment.
A study conducted with mice observed significant effects on the estrous cycle at doses much
lower than those causing alterations in the NTP study in rats. Zhang et al. (2020, 6505878)
administered 0.5-5 mg/kg/day PFOA to adult female mice for 28 days by gavage and monitored
daily vaginal cytology throughout the study (sample size n = 8). The number of days spent in
diestrus was significantly increased in females treated with 2 or 5 mg/kg/day, and the authors
noted that the mice in those groups were rarely observed to enter the estrus phase of the cycle
after the second week of exposure to PFOA; however, the durations of estrus and proestrus were
not significantly altered by treatment. Body weight was significantly reduced in the 5 mg/kg/day
group on days 24 and 28 (by 11%) but not significantly affected in the 2 mg/kg/day group.
In the same study, the numbers of corpora lutea were significantly reduced in mice administered
2 or 5 mg/kg/day PFOA for 28 days; however, no effects were observed on the antral follicle
count per ovary (sample size n = 8) {Zhang, 2020, 6505878}. Decreases in the number and size
of corpora lutea were also observed in pregnant mice administered PFOA (2.5, 5 or
10 mg/kg/day) beginning on GD 1 (sample size n = 6) {Chen, 2017, 3981369}. The numbers of
corpora lutea were significantly decreased in the low- and mid-dose groups on GD 7 and in the
mid- and high-dose groups on GD 13. The ratio of corpora lutea to ovarian areas was also
significantly decreased at both time points in a dose-dependent manner. The results of this study
suggest that PFOA treatment can significantly impair ovarian function during pregnancy and the
authors also found evidence of increased oxidative stress and apoptosis in the ovaries of treated
mice. Maternal body weights were not reported in this study.
The overall evidence for adverse effects of PFOA on ovarian function is suggestive but
inconclusive because the effects were mainly observed in mice and in studies with small sample
sizes (n = 6-8). It is likely that prolonged diestrus and reduced corpora lutea observed in mice
were treatment-related effects because they followed a clear dose response, and the effects were
observed at dose levels lower than those causing decrements in body weight (when reported).
Rats also demonstrated a slight increase in the time spent in diestrus, but only at a relatively high
dosage (100 mg/kg/day) {NTP, 2019, 5400977}. Only one study was identified that evaluated
effects on corpora lutea in rats {Staples, 1984, 1332669}, and that study found no difference
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
between the number of corpora lutea in control rats and those treated with 100 mg/kg/day PFOA
from GD 6-GD 15.
Altered Pubertal Timing
Lau et al. (2006, 1276159) reported a slight but significant delay in vaginal opening at 20
mg/kg/day; in contrast, significant accelerations in sexual maturation were observed in males,
with preputial separation occurring 4 days earlier than controls at 1 mg/kg/day and 2-3 days
earlier at 3, 5, and 10 mg/kg/day, whereas preputial separation in the 20 mg/kg/day group was
slightly but significantly delayed compared to controls.
A two-generation study in Sprague-Dawley rats reported significantly delayed sexual maturation
(i.e., vaginal opening and preputial separation) in Fi males and females at 30 mg/kg/day
{Butenhoff, 2004, 1291063}. In a study of direct peripubertal exposure, Yang et al. (2009,
5085085) orally dosed 21-day-old female BALB/c or C57BL/6 mice with 0, 1, 5, or 10
mg/kg/day for 5 days/week for 4 weeks. Vaginal opening was significantly delayed in BALB/c
mice dosed with 1 mg/kg/day and did not occur at all at 5 or 10 mg/kg/day. In C57BL/6 mice,
vaginal opening was delayed at 5 mg/kg/day and did not occur at 10 mg/kg/day.
C. 1.2.4 Reproductive Hormone Levels
C. 1.2.4.1 Moles
Several studies have reported significant alterations in reproductive hormone levels in male
animals following oral exposure to PFOA, but the results are not consistent across species or
study durations. Figure C-7 summarizes the effects of PFOA on reproductive hormone levels
observed in male rodents.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study Name Study Design Observation Tin
Perkins etal., 2004, 1291118 subchronic (I3wk) I3wk
Luteinizing Hormone (LH) Zhang ct al.. 2014.2850230 short-term (28d) 28d
Testosterone
Pcrkinsctal,, 2004, 1291118 subchronic (13wk) I3wk
Zhang ct id.. 2014.2850230 short-term (28d) 28d
Zhang et al., 2014,2850230 short-term (28d) 28d
NTP. 2019,5400977 short-term (28d) 29d
Perkins et al., 2004.1291118 subchronic (13wk) 13wk
PFOA Reproductive Effects - Hormones in Male Rodents
Animal Description Dose (mg/kg/day) | Q Statistically significani # Not stalislically significant M 95% CI |
Ral, Sprague-Dawley Crl:Cd Br
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
NTP {2019, 5400977} administered 0.625-10 mg/kg/day PFOA for 28 days to male rats and
found no significant differences in serum testosterone levels between treated groups and controls
at the end of the treatment period. The high-dose group had serum testosterone levels 22% lower
than controls, but the difference did not attain statistical significance. Likewise, a subchronic
dietary study in rats found no significant treatment-related alterations in serum testosterone, E2,
or LH levels measured after 4, 7, and 13 weeks of exposure with up to 100 ppm PFOA in the diet
(equivalent to 6.5 mg/kg/day) {Perkins, 2004, 1291118}.
Biegel et al. (2001, 673581) measured hormones at 3-month intervals in male rats fed 300 ppm
PFOA for two years (equivalent to 13.6 mg/kg/day), and no apparent treatment-related trends
were observed in serum testosterone, prolactin, LH, or FSH levels. Serum FSH and testosterone
were significantly increased only at 6 months, prolactin decreased significantly at 3 and
6 months, and LH was significantly increased at 6 and 18 months; however, serum E2 levels
were consistently increased at the 1-, 3-, 6-, 9-, and 12-month time points compared to controls.
Serum testosterone was significantly reduced in the male offspring of Kunming mice
administered PFOA (1, 2.5, or 5 mg/kg/day) from GD 1-GD 17 {Song, 2018, 5079725}. On
PND 21, serum testosterone levels were reduced in a dose-dependent fashion in all treated
groups (by 63-71%); however, on PND 70, there was no clear dose-response trend (serum
testosterone was increased by 92% in the low-dose group and decreased in the mid- and high-
dose groups by 74-75%). Zhang et al. (2014, 2850230) administered 0.31, 1.25, 5, or
20 mg/kg/day PFOA to adult male mice for 28 days, and no significant differences were
observed in serum LH levels. Testicular testosterone and progesterone concentrations were both
significantly reduced at dose levels > 1.25 mg/kg/day at the end of the treatment period.
Testicular testosterone was decreased by 34—91 % in a dose-dependent manner, and testicular
progesterone was decreased by 44—55%. In addition, intratesticular cholesterol was significantly
reduced (by 39-44%) at > 5 mg/kg/day.
In the 6-week mechanistic study by Li et al. (2011, 1294081), plasma testosterone levels
measured at the end of treatment were decreased in wild-type mice administered 1 mg/kg/day
(by 37%>), and significantly decreased in wild-type mice administered 5 mg/kg/day (by 57%)
compared to controls. Plasma testosterone was also significantly decreased in low- and high-dose
humanized PPARa mice (by 29% and 31%, respectively). In PPARa-null mice, plasma
testosterone was slightly reduced in a dose-related manner, but statistical significance was not
attained.
Overall, there are no clear treatment-related trends in male reproductive hormone levels across
species and study durations. Serum, plasma, or intratesticular testosterone levels were all
decreased in treated mice {Song, 2018, 5079725; Zhang, 2014, 2850230; Li, 2011, 1294081},
but similar effects on testosterone were not observed in rats after 28 days or longer exposures.
Testosterone in males is pulsatile and can display large random peaks, therefore studies
measuring hormones at various time points over the course of a study are more useful for
determining treatment-related effects than studies that measured concentrations at a single time
point, for example at necropsy. The studies that measured male hormone levels at various times
throughout treatment reported no consistent changes in testosterone {Perkins, 2004, 1291118;
Biegel, 2001, 673581}. Two studies reporting reduced testosterone in mice also observed
adverse effects on sperm concentration and/or quality following exposure to PFOA {Zhang,
2014, 2850230; Li, 2011, 1294081}; however, because of the limited number of studies available
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
and the lack of reproducibility in rats, no firm conclusions can be made about the adversity of
these findings. The 22% decrease in testosterone that was observed in high-dose male rats of the
28-day study by NTP (2019, 5400977) was not large enough to be considered adverse given the
inherent variability in testosterone levels with a male and between males.
Serum E2 levels were consistently increased at multiple time points in one chronic study in male
rats {Biegel, 2001, 673581}; however, the concentrations were very low (in the range of pg/mL),
and it has been shown that estrogen levels are too low to be accurately measured using
radioimmunoassay kits, which was the method used in that study. Therefore, no firm conclusions
can be made about the relevance of those findings as well.
C.l.2.4.2 Females
Figure C-8 summarizes the effects of PFOA on reproductive hormone levels observed female
rodents.
Endpoint
Estradiol
Study Name Study Design Observation Time Animal Description
Perkins et al., 2004, 1291118 subchronic {13wk) 13wk Rat. Sprague-Dawley Crl:Cd Br (n, N=10)
Dose (rng/kg/day)
0
0.06
0.64
1.94
Luteinizing Hormone (LH) Zhang et al., 2014, 2850230 short-term (28d)
Perkins et al.. 2004,1291118 subchronic (13wk) 13wk
Mouse. BALB/c N=0-6)
Rat, Sprague-Dawley Crl:Cd Br (o, N=9-10) 0
0,06
0.64
Testosterone
Zhang etal., 2014, 2850230 short-term (28d)
Song etal., 2018, 5079725 developmental (GD1-17) PND21
Zhang ct al., 2014, 2850230 short-term (28d)
NTP, 2019, 5400977
short-term (28d)
Mouse, BALB/c (-•, N=6)
F1 Mouse. Kunming N=9-10)
F1 Mouse. Kunming {. C, N=5)
Mouse, BALB/c N=6)
Rat. Sprague-Dawley N=10)
0.31
1,25
0.625
1,25
Perkins etal., 2004, 1291118 subchronic (13wk)
Rat, Sprague-Dawley Crl:Cd Br (;?, N=10) 0
PFOA Reproductive Effects - Hormones In Male Rodents
| Q Statistically significant ^ Nol slalislically significant!—| 95% CI |
-150 -100 -50 0 50 100 150 200
Percent control response (%)
Figure C-8. Percent Change in Female Reproductive Hormone Levels Relative to Controls
in Rodents Following Exposure to PFOA
Interactive figure and additional study details available on HAWC.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Po = parental generation; GD = gestation day; d = day.
Only three studies measured female reproductive hormones following oral exposure to PFOA,
and the only effect observed in more than one study was slightly reduced progesterone levels
{Chen, 2017, 3981369; Zhang, 2020, 6505878}.
No significant differences were observed in serum testosterone levels of adult female rats
administered 6.25-100 mg/kg/day PFOA for 28 days {NTP, 2019, 5400977}, but no other
reproductive hormones were measured in that study. A 28-day study in adult female mice
observed significant reductions in several hormone levels following administration of 2 or
5 mg/kg/day, including reduced serum progesterone (17%—21 %), gonadotrophin-releasing
hormone (GnRH) (25%-32%), and LH (18%—21%). Serum E2 was also significantly reduced
(28%) at 5 mg/kg/day {Zhang, 2020, 6505878}. In contrast, when pregnant female mice were
administered PFOA beginning on GD 1 {Chen, 2017, 3981369}, serum E2 was slightly
increased on GD 7 but unaltered on GD 13. Meanwhile, serum progesterone was unaltered on
GD 7 but was significantly reduced on GD 13 at 5 and 10 mg/kg/day (by 16%-22%).
Due to the small dataset and the small percent changes from controls, no firm conclusions can be
made about the effects of PFOA on female reproductive hormones in animals.
C. 1.2.5 Reproductive Organ Weights and Histopathology
C. 1.2.5.1 Moles
Some studies in rats and mice indicate that PFOA exposure can result in changes in the normal
structure of the testes and epididymides; however, the overall body of evidence is inconsistent
with several other studies reporting no histological changes in male reproductive organs.
Absolute weights of the testes were either significantly decreased {NTP, 2020, 7330145; Zhang,
2014, 2850230} or unaltered {NTP, 2019, 5400977; Crebelli, 2019, 5381564; Butenhoff, 2012,
2919192; Butenhoff, 2004, 1291063} in adult male rodents following exposure to PFOA.
Meanwhile, relative weights of the testes were either significantly increased {NTP, 2019,
5400977; NTP, 2020, 7330145; Butenhoff, 2004, 1291063; Biegel, 2001, 673581} or unaltered
{Zhang, 2014, 2850230; Butenhoff, 2012, 2919192}. The decreases observed in absolute
testicular weights in conjunction with unaltered or increased relative weights appear to be
secondary to body weight changes and therefore unrelated to treatment with PFOA.
Several studies observed no histological changes in the testes, including a 28-day study in rats
{NTP, 2019, 5400977}, a 13-week dietary study in rats {Perkins, 2004, 1291118}, a two-
generation reproduction study in rats {Butenhoff, 2004, 1291063}, a 26-week study in monkeys
{Butenhoff, 2002, 1276161}, and a two-year study in rats (See Main PFOA Document for study
design details) {NTP, 2020, 7330145}. However, there is some evidence in mice that suggests
developmental exposure can alter the normal structure of the testes. Song et al. (2018, 5079725)
exposed pregnant mice to 1, 2.5, or 5 mg/kg/day PFOA from GD 1-GD 17 and evaluated
testicular weights and histopathology in the male offspring on PND 21 and PND 70. Absolute
testis weights were significantly increased in the high-dose group on PND 21, but the effect was
not observed on PND 70. There were no significant differences in relative testis weights at either
time point and the change in absolute weight appeared to be related to increased body weights
also observed in the high-dose group. Histopathological examination revealed significant
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
changes in the testes of the 2.5 and 5 mg/kg/day groups on both PND 21 and PND 70. Effects
that were reported quantitatively were decreased numbers of Ley dig cells on PND 21 (by 25%-
27%) and PND 70 (by 17%—25%) and increased intercellular substance areas on PND 21 (by
105%—111%) and PND 70 (by 9%—13%). Other microscopic changes were reported qualitatively
only and included atrophy of the spermatogenic epithelium, reduction in spermatogenic cells,
vacuolization of Sertoli cells and decrease or disappearance of spermatozoa at 5 mg/kg/day. With
increasing dose to the dam, the degree of damage to the testes was noted to increase. From 2.5 to
5 mg/kg/day, the intercellular substance in the testes of offspring became larger and the
interstitial cells gradually decreased. The spermatogenic cells of all levels were arranged in an
irregular pattern; however, vacuolization was not observed on PND 70 indicating some recovery
had occurred since PND 21.
Zhang et al. (2014, 2850230) also reported damage to the testes in adult male mice treated for
28 days, but results were reported qualitatively without incidence data. The findings in rats
treated with 5 or 20 mg/kg/day included atrophy of the seminiferous tubule epithelia, lack of
germ or Sertoli cells between basal membrane and adluminal portions, and detached germ cells
sloughed off into the tubular lumen. In the 6-week mechanistic study in mice by Li et al. (2011,
1294081), histopathological examination of the testes revealed abnormal seminiferous tubules
with vacuoles or lack of germ cells in wild-type and humanized PPARa mice administered
5 mg/kg/day (reported qualitatively without incidence data), but these changes were not observed
in PPARa-null mice. Necrotic cells in testes and significantly reduced weights of the epididymis
and seminal vesicle plus prostate gland were also observed in the 5 mg/kg/day wild-type mice
only.
At the 1-year sacrifice of a chronic dietary study in rats {Butenhoff, 2012, 2919192}, testicular
tubular atrophy with marked aspermatogenesis was observed in in 2/15 (13%) of high-dose
(300 ppm; 14.2 mg/kg/day) males but not in any of the controls (statistical significance not
reported). At the terminal evaluation, there were no significant differences in the incidences of
tubular atrophy, but the incidence of vascular mineralization in the testes was significantly
increased in high-dose males. The incidences of the lesion in the control, 30, and 300 ppm (0,
1.3, and 14.2 mg/kg/day) groups were 0%, 6%, and 18%, respectively. In contrast, a two-year
dietary study conducted by NTP {2020, 7330145} found no treatment-related effects in the testes
of rats fed PFOA at concentrations up to 300 ppm (32 mg/kg/day) for 16 weeks or 80 ppm
(4.6 mg/kg/day) for 2 years (including groups that were also exposed during gestation; see Main
PFOA document for further study design details).
Effects on the epididymis have also been observed following PFOA exposure. Absolute weights
of the epididymis or cauda epididymis were significantly reduced in a few studies {NTP, 2019,
5400977; Lu, 2016, 3981459; Butenhoff, 2004, 1291063}, and relative epididymis weight was
also significantly reduced in one of those studies {Lu, 2016, 3981459}.
In the two-generation reproduction study in rats, absolute weights of several male reproductive
organs were significantly decreased in the high-dose males of the Po (i.e., right and left
epididymis, cauda epididymis, seminal vesicles with and without fluid, and prostate) while the
relative weights of those organs were all significantly increased (except for the prostate)
{Butenhoff, 2004, 1291063; York, 2010, 2919279}. The patterns observed were consistent with
significant decrements in body weights that were also observed in male groups treated with
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
> 1 mg/kg/day, and there were no treatment-related changes observed in histopathology of those
organs.
NTP (2019, 5400977) observed hypospermia and exfoliated germ cells in the epididymis of one
rat each in the 5 and 10 mg/kg/day groups following 28 days of oral exposure, although the
incidences were not statistically different from controls (n = 10 per group evaluated). This
coincided with significantly reduced absolute weights of the left cauda epididymis
(> 5 mg/kg/day) and left epididymis (10 mg/kg/day) as well as reduced epididymal sperm count
(10 mg/kg/day). However, relative epididymal weights were not reported in this study. No
treatment-related effects were observed in the testes, seminal vesicles, or accessory sex glands.
In a 28-day study in mice, absolute weights of the epididymis were reduced in mice treated with
5 or 20 mg/kg/day and relative epididymis weights were also reduced at 20 mg/kg/day {Lu,
2016, 3981459}. Histopathological examination revealed empty spaces in the tubules of cauda
epididymis of mice treated with 5 or 20 mg/kg/day and a lack of normal sperm (reported
qualitatively without incidence data). In addition, the levels of triglycerides in the epididymis
were significantly reduced at 5 and 20 mg/kg/day and the cholesterol content of the epididymis
was significantly reduced at 20 mg/kg/day.
In contrast to the results observed in 28-day studies, chronic studies have reported no treatment-
related changes in the epididymis or accessory sex glands of treated rats or monkeys {NTP,
2020, 7330145; Butenhoff, 2012, 2919192; Butenhoff, 2002, 1276161}.
Overall, the evidence for adverse effects on the male reproductive system is inconsistent for
PFOA. Some studies have reported damage to the testes including atrophy of the seminiferous
tubule epithelia {Song, 2018, 5079725; Zhang, 2014, 2850230; Butenhoff, 2012, 2919192};
however, two comprehensive studies conducted by NTP {2019, 5400977; 2020, 7330145} and a
two-generation reproduction study {Butenhoff, 2004, 1291063} all reported no significant
changes in the histopathology of male reproductive organs and glands. The two-year study by
Butenhoff et al. (2012, 2919192) reported a small but statistically significant increase in the
incidence of vascular mineralization in the testes of high-dose males. The toxicological
significance of that finding is unclear as the study authors did not evaluate any parameters
related to fertility including any hormone levels nor did they see any effects on testes weights. In
addition, this lesion was not observed in another chronic rat study {NTP, 2020, 7330145} or in
any of the shorter duration mouse studies where there were suggestive effects on sperm
parameters and fecundity. When mice were exposed to PFOA in utero, the numbers of Leydig
cells in the testes were decreased and there was evidence of dose-dependent testicular damage on
PND 21 and PND 70 {Song, 2018, 5079725}. Leydig cells are the primary site of testicular
steroidogenesis in males {Huhtaniemi, 1995, 7420539}. BWTs of the pups and growth during
the lactation period were not reported; therefore, it is unclear whether these effects reflect a
specific toxicity to the testes or if they resulted from delayed growth and systemic toxicity. Body
weights were not reduced compared to control on PND 21 or PND 70; therefore, a direct effect
on the developing testes cannot be ruled out.
Reduced epididymal weights were reported in two studies along with reduced epididymal sperm
concentration and/or observations of hypospermia {NTP, 2019, 5400977; Lu, 2016, 3981459}. It
is also unclear if these effects resulted from a specific toxicity to the epididymis or from
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
concurrent systemic toxicity as effects were observed in conjunction with decrements in body
weight {NTP, 2019, 5400977} or body weights were not reported {Lu, 2016, 3981459}.
C. 1.2.5.2 Females
Histopathological changes in the uterus and ovary have been observed following exposure to
PFOA; however, comprehensive studies with chronic exposure durations do not provide
evidence of increased nonneoplastic lesions in female reproductive organs.
Li et al. (2018, 5084746) administered PFOA (1, 5, 10, 20, or 40 mg/kg/day) to pregnant
Kunming mice from GD 1-GD 17 and measured apoptosis in the uterine tissue on GD 18. The
number of apoptotic cells was significantly increased for females dosed with 5 mg/kg/day or
higher in a dose-dependent manner compared to controls, and embryo survival was significantly
decreased at doses > 10 mg/kg/day (see Main PFOA Document). The uterus was examined in
several other studies with no significant changes reported in organ weight or incidences of
nonneoplastic lesions, including a 28-day study in rats {NTP, 2019, 5400977}, a two-generation
reproduction study in rats {Butenhoff, 2004, 1291063} and a 2-year dietary study in rats
{Butenhoff, 2012, 2919192}. No significant differences in uterine weights were observed at the
16-week interim evaluation of the NTP 2-year dietary study in rats (See Main PFOA Document
for study design details){NTP, 2020, 7330145}; however, the terminal evaluation found that
females treated with PFOA had a higher incidence of uterine adenocarcinoma that may have
been related to exposure (see Main PFOA Document). The incidences of nonneoplastic lesions
of the uterus were not significantly increased in any of the PFOA exposure groups {NTP, 2020,
7330145}.
As mentioned above, Chen et al. (2017, 3981369) and Zhang et al. (2020, 6505878) both
observed significant changes in the ovaries of adult female mice administered PFOA, including
reductions in the number of corpora lutea and the ratio of corpora lutea to ovarian areas.
However, the NTP chronic dietary study {NTP, 2020, 7330145} and a two-generation
reproduction study {Butenhoff, 2004, 1291063} both found no treatment-related effects in the
ovaries of treated rats. Butenhoff et al. (2012, 2919192) observed a significant, dose-related
increase in the incidences of ovarian tubular hyperplasia in rats exposed for 2 years to PFOA in
the diet. The incidences of this lesion in the control, 30, and 300 ppm groups were 0%, 14%, and
32%, respectively. The tissues were subjected to a pathology peer review using updated
diagnostic nomenclature and no statistical differences were found between treated groups and
controls {Mann, 2004, 6569580}.
C.1.3 Mechanistic Evidence Synthesis
Mechanistic evidence linking PFOA exposure to adverse reproductive outcomes is discussed in
Sections 3.2.2, 3.2.7, 3.3.3, 3.3.4, and 3.4.3 of the 2016 PFOAHESD {U.S. EPA, 2016,
3603279}. There are 56 studies from recent systematic literature search and review efforts
conducted after publication of the 2016 PFOA HESD that investigated the mechanisms of action
of PFOA that lead to reproductive effects. A summary of these studies is shown in Figure C-9.
Additional mechanistic synthesis will not be conducted since evidence suggests but is not
sufficient to infer that PFOA leads to reproductive effects.
D-4
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Mechanistic Pathway
Animal
Human
In Vitro
Grand T<
Angiogenic, Antiangiogenic, Vascular Tissue Remodeling
1
0
1
2
Big Data, Non-Targeted Analysis
2
0
5
6
Cell Growth, Differentiation, Proliferation, Or Viability
11
0
23
29
Cell Signaling Or Signal Transduction
10
1
24
32
Extracellular Matrix Or Molecules
0
0
3
3
Fatty Acid Synthesis, Metabolism. Storage. Transport, Binding. B-Oxidation
3
0
2
4
Hormone Function
9
1
22
28
Inflammation And Immune Response
2
0
1
2
Oxidative Stress
3
0
6
9
Xenobiotic Metabolism
1
0
6
6
Other
0
0
1
1
Not Applicable/Not Specified/Review Article
1
0
0
1
Grand Total
23
2
41
56
Figure C-9. Summary of Mechanistic Studies of PFOA 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 PFOA exposure and male reproductive
effects based on inverse associations with testosterone in male children and adults, and decreased
AGD in children. Negative effects were observed for some semen characteristics (e.g., semen
motility, DNA fragmentation), but positive associations were also observed (e.g., sperm
concentration). There was inconsistent evidence for the relationship between PFOA exposure
and testosterone in cross-sectional studies {Lopez-Espinosa, 2016, 3859832; Di Nisio, 2019,
5080655} in children and young adults. Inconsistent associations were observed in populations
at different stages of pubertal development, and one positive association was observed in a low
confidence study {Di Nisio, 2019, 5080655}. One medium confidence study {Liu, 2020,
6569227} observed a positive association for progesterone in male infants. Studies in adolescents
did not observe effects on pubertal development, but negative associations were observed for
testicular volume, penis length, penis circumference, and number of sperm with normal
morphology {Di Nisio, 2019, 5080655}. In adults, there was evidence in two studies {Cui, 2020,
6833614; Petersen, 2018, 5080277} of inverse associations between serum PFOA and
testosterone (total and free), and these associations were also observed using semen PFOA.
Inverse associations were also observed for E2 and the total testosterone-LH ratio. For semen
and sperm characteristics in adults, associations were observed for several parameters in analyses
of semen PFOA, including increased sperm concentration and total sperm count, decreased
motility and number of morphologically normal sperm, and increased sperm DNA
fragmentation. Other results for markers of genotoxic effects (e.g., sperm Y:X chromosome
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
ratio, sperm DNA methylation, etc.) in sperm were inconsistent. Overall, these studies provide
additional evidence of potential effects on testosterone levels in adult males.
The animal evidence for an association between PFOA exposure and reproductive toxicity in
males is slight based on several high or medium confidence animal studies; however, the
evidence from animal studies is similarly inconsistent as in epidemiological studies. Despite this,
some studies observed significant alterations in reproductive hormone levels and adverse effects
on sperm parameters. Exposure during development or for short durations in adult rodents has
resulted in changes in the normal structure of the testis and epididymis {Song, 2018, 5079725;
NTP, 2019, 5400977; Lu, 2016, 3981459; Zhang, 2014, 2850230; Li, 2011, 1294081}. Chronic
exposure studies generally found limited histological changes in the testes that included
increased incidence of vascular mineralization {Butenhoff, 2012, 2919192} and Ley dig cell
hyperplasia {Biegel, 2001, 673581}. However, these findings were not observed in another two-
year study by NTP (2020, 7330145). EPA concluded that the observed changes in the testes and
epididymis represent toxicities possibly relevant to humans. In particular, alterations in Ley dig
cell structure or physiology may be driving the reductions in testosterone and effects on sperm
parameters seen in both humans and animals {Zirkin, 2018, 9641879}.
C.1.4.2 Evidence InteQration JudQment
Overall, evidence suggests that PFOA exposure has the potential to cause reproductive effects in
males under relevant exposure circumstances (.
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 PFOA ranging from 1.4 to 34.8 ng/mL. Although there is some
evidence of negative effects of PFOA 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 and in the normal structure of the testis and epididymis in
animal models following exposure to doses as low as 1 mg/kg/day PFOA. However, findings
from animal studies are similarly inconsistent as in epidemiological studies.
C-l
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table C-l. Evidence Profile Table for PFOA Reproductive Effects in Males
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Male reproductive
hormones
1 High confidence study
8 Medium confidence
studies
4 Low confidence studies
Summary and Key
Findings
Results from studies in
children were inconsistent
regarding measures of
testosterone. Significant
increases (1/9) and
significant inverse
associations (1/9) with total
testosterone were observed,
but the remaining studies
reported imprecise results
(6/9). Increases in
estrogenic hormones (i.e.,
estrone, estradiol, and
estriol) were observed in
children (2/4), but only one
result was significant.
Significant increases in LH
(1/9) and progesterone
(1/9), and inverse
associations with androgen
hormones, such as DHEA
and androstenedione (2/9)
were observed in children.
Significant inverse
associations with free
testosterone (2/4) and total
testosterone (1/4) were
observed in adults. Inverse
associations in LH, FSH,
and SHBG were also
observed (1/4).
Factors that Increase
Certainty
• High and medium •
confidence studies •
• Coherence of •
findings between
changes in
androgenic and
estrogenic sex
hormones
• Medium confidence •
studies
Factors that Decrease
Certainty
Low confidence studies
Lmprecision of findings
Potential for residual
confounding by SES
and smoking status
Low confidence study
Evidence Stream
Judgment
©oo
Slight
Evidence for male
reproductive effects is
based on several medium
confidence studies
reporting consistent and
precise associations. More
evidence is available for
associations with male
testosterone levels, male
pubertal development, and
male anthropometric
measurements compared
to other effects. Other
findings were either
inconsistent or are
reported in a limited
number of studies.
Evidence Integration
Summary Judgment
©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
PFOA. Although there is
some evidence of negative
effects of PFOA 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 and in
the normal structure of the
testis and epididymis in
animal models following
exposure to PFOA.
However, findings from
animal studies are similarly
inconsistent as in
epidemiological studies.
Semen parameters The only low confidence
study evaluating
Evidence from Studies of Exposed Humans (Section C.l.l)
C-2
-------
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
4 Medium confidence
studies
1 Low confidence study
adolescents observed
significant inverse
associations with sperm
concentrations and
progressive motility and
increased semen pH with
higher exposure. In four
medium confidence studies
of adults, results were
mixed, with one study
finding evidence of
significantly increased
sperm concentration and
count and significant
inverse associations with
measures of motility and
morphology (1/4). Other
studies reported inverse
associations with semen
parameters (3/4), with one
result for progressive
motility reaching
significance (1/3).
Consistent direction
of effects
Imprecision of most
findings
Potential for residual
confounding by SES
and smoking status
Evidence Integration
Summary Judgment
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
C-3
-------
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
Anthropometric
measurements of male
reproductive organs
1 High confidence study
2 Medium confidence
studies
1 Low confidence study
In children and adolescents,
one medium and one low
confidence study reported
significant effects for
anthropometric
measurements of male
reproductive organs (2/4).
In a medium confidence
study, children from the
Shanghai-Minhang Birth
Cohort study reported
significant inverse
associations with AGD. A
low confidence study
reported smaller AGD in
exposed compared to
unexposed children, and
significant differences in
testicular measurements,
such as smaller testicular
volume and shorter penis
length. A high confidence
study reported inverse
associations with AGD that
did not reach significance.
High and medium
confidence studies
Consistent direction
of effects
Coherence of
findings
Low confidence study
Lmprecision of some
findings
Potential for residual
confounding by SES
and smoking status
Male pubertal
development
1 Medium confidence
study
Findings for changes in
timing of pubertal
development milestones
were non-significant. Voice
break was observed to
occur earlier for those in the
highest exposure tertile, but
the association was not
significant.
Medium confidence
study
Limited number of
studies examining
outcome
Evidence from In Vivo Animal Studies (Section C.1.2)
C-4
-------
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
Organ weights
4 High confidence
studies
6 Medium confidence
studies
Several rodent studies have
shown changes in testis or
epididymis weight
following PFOA exposure
(6/10). However, evidence
is not consistent as one
mouse study (1/10) and
several rat studies (3/10)
show no effect of PFOA on
the weight of male
reproductive organs.
Absolute testis weights
were mostly unchanged in
rats (4/6) and mice (1/3),
although relative testis
weight was increased in rats
(3/5) and unchanged in
mice (2/2). Absolute
epididymis weight was
decreased in two studies
(2/3), with one in mice and
one in rats. The study in
mice also reported
decreased relative
epididymis weight (1/1).
High and medium
confidence studies
Inconsistent direction of ©OO
effects across studies Slight
and species
Changes in body weight Evidence was based on 11
may limit ability to
interpret these
responses
high and medium
confidence studies.
Changes in male
reproductive organs, such
as organ weight or
structural changes, were
observed. However, these
results were inconsistent
among studies. Effects
observed in male rodents
include decreased
epididymal weights,
delayed sexual maturation,
decreased sperm count
and quality, alterations in
reproductive hormone
levels, and morphological
changes in the testes and
epididymides.
Evidence Integration
Summary Judgment
C-5
-------
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
Histopathology
4 High confidence
studies
5 Medium confidence
studies
Several studies in rats and
mice found changes in the
structure of the testes and
epididymides (6/9). In rats,
nonneoplastic changes in
the testes were noted (2/6)
including increased Leydig
cell hyperplasia and
vascular mineralization. A
short-term rat study found a
slight increase in exfoliated
germ cells in the
epididymis. In mice, one
short-term study found
changes in the epididymis
including empty spaces in
the tubules of cauda
epididymis and a lack of
normal sperm. Another
mouse study observed
increased tubular
degeneration and atrophy of
the seminiferous tubules in
the testes. A third mouse
study found decreased
numbers of Leydig cells
and increased intercellular
area in the testes of pups
exposed in utero.
Two chronic rat studies
found no changes in the
testis, epididymis, prostate,
or seminal vesicles.
High and medium
confidence studies
Inconsistent direction of
effects across studies
C-6
-------
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 reproductive
hormones
2 High confidence
studies
3 Medium confidence
studies
Testosterone was decreased •
following PFOA exposure
(2/5), but only in male
mice. In rats, testosterone
was either increased (1/3)
or showed no difference
(2/3). Decreases in
progesterone were observed
in male mice (1/1) and in
prolactin for male rats (1/1).
LH was decreased in one rat
study (1/3). Estradiol was
consistently increased in
one male rat study (1/2). No
changes were observed in
FSH (0/1) in male rats.
High and medium
confidence studies
• Inconsistent direction of
effects among studies
and species
• Limited number of
studies examining
specific outcomes
Sperm parameters
2 High confidence
studies
1 Medium confidence
study
Sperm count was decreased •
following PFOA exposure
in two studies (2/3),
including one study in mice
and one short-term study in
rats. However, a two-
generation reproduction
study in rats found no
effects on sperm count.
Sperm motility was
decreased in one mouse
study (1/3), but not in two
rat studies.
High and medium
confidence studies
• Limited number of
studies evaluating
endpoint
• Inconsistent direction of
effects between species
• Incoherence of findings
between decreased
sperm count and lack of
effects on fertility
Male pubertal
development
1 High confidence study
1 Medium confidence
study
The timing of preputial
separation in males was
altered (2/2). One rat study
found delayed preputial
separation after PFOA
exposure. One mouse study
High and medium
confidence study
Limited number of
studies evaluating
endpoint
C-7
-------
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
found that preputial
separation occurred earlier
at low doses but later at the
highest dose.
Male mating and
fertility
1 High confidence study
One two-generation
reproduction study reported
no effects on mating or
fertility in rats administered
PFOA for 10 weeks prior to
mating (1/1).
High confidence
study
Limited number of
studies evaluating
endpoint
Notes: AGD = anogenital distance; DHEA = dehydroepiandrosterone; FSH = follicle-stimulating hormones; LH = luteinizing hormone; SHBG = sex hormone binding globulin;
SES = socioeconomic status.
C-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C. 1.4.3 Reproductive Effects in Females
There is slight evidence for an association between PFOA exposure and female reproductive
effects in humans based on observed infertility effects across a limited number of
epidemiological studies, observed in populations with high exposure levels and at levels typical
in the general population.
Results for female fertility are mixed. In the 2016 Health Assessment {U.S. EPA, 2016,
3603279}, two studies reported correlations between higher PFOA levels and infertility {Fei,
2009, 1291107; Velez, 2015, 2851037}. Studies published since the 2016 HESD have observed
no clear dose-response trends or directionality for a potential relationship {Lum, 2017, 3858516;
Crawford, 2017, 3859813; Wang, 2017, 3856459; Kim, 2020, 6833596}. However, Kim et al.
(2020, 6833596) did observe some non-significant, positive associations between follicular fluid
PFOA and fertility etiology factors for other gynecologic pathologies, including endometriosis,
polycystic ovarian syndrome (PCOS), genital tract infections, and idiopathic infertility.
There is limited evidence of an inverse association between serum PFOA levels in pregnancy
and breastfeeding duration. Timmermann et al. (2017, 3981439) observed negative associations
between PFOA exposure and exclusive and total breastfeeding duration, while Romano et al.
(2016, 3981728) observed increased relative risk of breastfeeding termination with increasing
PFOA exposure.
Evidence of a relationship between PFOA exposure and the female reproductive milestones of
age at menarche and menopause is mixed. In the 2016 Health Assessment {U.S. EPA, 2016,
3603279}, Kristensen et al. (2013, 2321268) reported a positive association between prenatal
PFOA exposure and later age at menarche, while Christensen et al. (2011, 1290803) reported no
association between the two. Since the 2016 Health Assessment, Ernst et al. (2019, 5080529)
observed a non-significant, negative association between prenatal PFOA exposure and age at
menarche. Other studies have investigated relationships between the menarche as well as
menopause and concurrent PFOA exposure. In the 2016 Health Assessment, Lopez-Espinosa et
al. (2011, 1424973) observed a positive association between concurrent PFOA exposure and age
at menarche. More recently, Ding et al. (2020, 6833612) observed an inverse relationship
between PFOA levels and age at menopause. However, findings from studies concurrently
assessing menstruation events and PFOA levels in blood must be interpreted with caution due to
potential reverse causality, as menstruation is a primary route of PFOA excretion for people who
menstruate.
Since the 2016 PFOA Health Assessment {U.S. EPA, 2016, 3603279}, 11 studies have assessed
relationships between PFOA exposure and various female reproductive hormones, nine of which
studied female infants and adolescents. Most studies did not report significant associations or
consistent trends between PFOA exposure and reproductive hormones including 17-OHP,
DHEA, E2, FSH, SHBG, and testosterone. Medium confidence studies have observed
significant, positive associations between cord blood PFOA and estriol in female infants {Wang,
2019, 5080598}, concurrent PFOA exposure and serum E2 in female adolescents {Lopez-
Espinosa, 2016, 3859832}, and maternal serum PFOA during pregnancy and AMH
concentrations in adolescent daughters {Donley, 2019, 5381537}. There were few studies
assessing relationships between PFOA exposure and female reproductive hormone levels in adult
women (both pregnant and non-pregnant), and those identified did not report consistent evidence
C-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
of relationships between PFOA exposure and these outcomes. Evidence of relationships between
PFOA exposure and human female reproductive hormonal outcomes remains inconsistent.
The recent review observed evidence of an association between PFOA and preeclampsia and
gestational hypertension; there is conflicting evidence on altered puberty onset and limited data
suggesting reduced fertility and fecundity. The associations are inconsistent across reproductive
hormone parameters, and it is difficult to assess the adversity of these alterations.
The animal evidence for an association between PFOA exposure and female reproductive
toxicity is slight based on several high and medium confidence animal studies; however, it is
often unclear if alterations seen in animal studies reflect specific toxicity to the reproductive
system or if they result from concurrent systemic toxicity. Despite this, some studies observed
significant alterations in reproductive hormone levels and ovarian physiology which were not
confounded by alterations in body weight. Specifically, effects of PFOA on the ovary included
altered estrous cyclicity and number of corpora lutea. In female mice, effects on the estrous cycle
(lengthened diestrus phase) were observed at doses that did not significantly reduce body weight
{Zhang, 2020, 6505878}. These results in mice are supported by a study in female rats that
similarly found slightly lengthened diestrus phase, though with a much higher PFOA dose {NTP,
2019, 5400977}. Altered ovarian physiology was also evidenced by two studies (short-term and
gestational) in adult female mice showing reduced numbers of corpora lutea with increasing
PFOA doses {Zhang, 2020, 6505878; Chen, 2017, 3981369} and one study in female rats
(chronic) showing increased tubular hyperplasia of the ovarian stroma {Butenhoff, 2012,
2919192}.
C. 1.4.4 Evidence Integration Judgment
Overall, evidence suggests that PFOA 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 infertility, female reproductive milestones, and female reproductive hormonal
outcomes observed in studies in humans exposed to median PFOA ranging from 3.7 to
30.1 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 alterations in ovarian physiology and hormonal parameters in adult rodents
following exposure to doses as low as 1 mg/kg/day PFOA. However, findings from animal
studies are similarly inconsistent as in epidemiological studies.
C-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-2. Evidence Profile Table for PFOA 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
In 12 studies of female
children and adolescents,
4 studies reported
significant associations.
Positive associations were
reported for estriol in
infants in a medium
confidence study (1/4).
Both E2 and total
testosterone levels had
positive associations
reported in both a medium
and a low confidence
study (2/4). Results from 9
studies of adults, rarely
met significance, though
one low confidence study
reported increased
testosterone relative to
controls, and another low
confidence study reported
increased levels of
prolactin. There were no
significant results for
SHBG.
High and medium •
confidence studies •
Low confidence studies
Lmprecision of most
findings
Potential for selection
bias and residual
confounding by age
and SES
Preeclampsia and
gestational
hypertension
1 High confidence study
5 Medium confidence
studies
3 Low confidence studies
Seven studies examined
preeclampsia in pregnant
women (7/9). None
reported significant
results, though of medium
and low confidence
studies, 6 reported positive
associations (6/6) and two
High and medium
confidence studies
©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.
Low confidence studies
Lmprecision of all
findings
Potential for reverse
causality
eoo
Evidence Suggests
Primary basis:
Human evidence indicted
effects on infertility, female
reproductive milestones,
and female reproductive
hormonal outcomes
observed in studies in
humans exposed to PFOA.
There is considerable
uncertainty in the results
due to inconsistency across
studies and limited number
of studies. Animal evidence
indicated alterations in
ovarian physiology and
hormonal parameters in
adult rodents following
exposure to PFOA.
However, findings from
animal studies are similarly
inconsistent as in
epidemiological studies.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
C-ll
-------
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
reported negative
associations (2/6) for at
least one exposure group
or for continuous analyses.
Of the three studies
examining gestational
hypertension (3/9), two
reported increased odds
but neither reached
significance (2/3).
However, after observing
non-significant increased
odds of gestational
hypertension, one medium
confidence study reported
increased DBP and
significantly increased
SBP.
Female reproductive
milestones
1 High confidence study
3 Medium confidence
studies
1 Low confidence study
Three studies examined
reproductive milestones
related to menstruation,
two in adolescent
populations and one in an
adult population. Two
studies, one low
confidence study in
adolescents (1/2) and one
medium confidence study
in adults (1/1), reported
significant increases in
long menstrual cycles. The
study in adolescents also
reported increased risk of
hypomenorrhea and
irregular menstruation.
High and medium
confidence studies
Consistent direction
of effects
Coherence of
findings
Low confidence study
Potential for residual
confounding by not
identifying
confounders
C-12
-------
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
There were no significant
effects with other pubertal
milestones. Two studies of
medium and high
confidence evaluated age
at natural menopause.
Both observed significant
positive associations,
though only among the
highest exposure group in
the high confidence study.
Fertility indicators
6 Medium confidence
studies
7 Low confidence studies
Examinations of fertility
indicators include
fecundability, fertilization
rate, and measures of
ovarian health, such as
anti-Mullerian hormone
levels or endometriosis.
Thirteen studies evaluated
fertility indicators in non-
pregnant women with
mixed results. Five
reported significant
positive associations
(5/13) with anti- Miillerian
hormone, a marker of
ovarian reserve, in
adolescents (1/5), and
increased odds of
endometriosis (2/5) and
ovarian syndromes (2/5).
Other studies did not
report significant
associations for these
measures and some
Medium confidence
studies
Low confidence studies
Lmprecision of most
findings
Potential for residual
confounding by not
identifying
confounders
C-13
-------
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
observed inverse
associations.
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 medium
confidence study reported
non-significant increases
in anoclitoris distance for
all exposure groups and in
continuous analysis.
Results for anofourchette
distances were non-
significant and mixed. A
high confidence study
observed non-significant
mixed results for both
measures.
Evidence from In Vivo Animal Studies (Section C.1.2)
Organ weights
3 High confidence
studies
3 Medium confidence
studies
Several rodent studies
show a lack of evidence of
changes in female
reproductive organ
weights following PFOA
exposure (5/6). Only one
mouse study found
High and medium
confidence studies
Changes in body
weight may limit
ability to interpret
these responses
©OO
Slight
Evidence is based on!
high and medium
confidence studies.
Changes in female
C-14
-------
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 absolute and
relative gravid uterus
weight following
gestational PFOA
exposure; however,
concurrent decreases in
maternal body weight and
in embryo survival and
body weight make these
results difficult to
interpret. Otherwise, there
were no changes in uterus
weight (4/5) or ovary
weight (2/2) among mouse
or rat studies.
Histopathology
3 High confidence
studies
2 Medium confidence
studies
For non-neoplastic effects
on uterus, one study found
evidence of effects
following PFOA exposure
(1/5). This study in mice
found dose-related
increases in the number of
apoptotic cells in the
uterine tissue of pregnant
mice on GD 18. For non-
neoplastic effects on
ovaries, three rat studies
found no exposure-related
effects on the ovaries
(3/4). However, one rat
study observed a dose-
related increase in ovarian
tubular hyperplasia after 2
years of PFOA exposure.
High and medium
confidence studies
Dose-response
relationship
Inconsistent direction
of effects among
studies
reproductive organs, such
as organ weight or
structural changes, were
observed. However, these
results were inconsistent
among studies. Effects
observed in female rodents
include morphological
changes in the uterus,
delayed sexual maturation,
alterations in reproductive
hormone levels, and
alterations in ovarian
physiology and structure
including effects on the
estrous cycle (prolonged
diestrus), reduced number
and size of corpora lutea
in the ovaries, and
increased tubular
hyperplasia of the ovarian
stroma.
Evidence Integration
Summary Judgment
C-15
-------
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
Female reproductive
hormones
1 High confidence study
2 Medium confidence
studies
Progesterone was slightly •
decreased in female mice
following PFOA exposure
(2/2). No effects on serum
testosterone levels were
reported in a short-term
study in female rats (1/1).
One mouse study found
that estradiol increased in
dams after gestational
PFOA exposure (1/2).
However, another mouse
study in adults (1/2) found
decreases in E2, along
with decreases in LH and
in GnRH.
High and medium •
confidence studies
Limited number of
studies examining
outcome
Inconsistent direction
of effects across
studies
Estrous cyclicity
2 High confidence
studies
1 Medium confidence
study
Exposed rats and mice
spent more time in
diestrus (i.e., prolonged
diestrus) in two studies
(2/3). However, a two-
generation rat study did
not find evidence for
prolonged diestrus. No
changes in estrous cycle
length were noted (3/3).
• High and medium •
confidence studies
•
Inconsistent direction
of effects among
studies
Limited number of
studies examining
outcome
Ovarian function
2 Medium confidence
studies
Decreases in the number
of corpora lutea in the
ovaries were observed in
mice following PFOA
exposure (2/2).
• Medium confidence •
studies
• Consistent direction
of effects
Limited number of
studies examining
outcome
Female pubertal
development
1 High confidence study
Delayed vaginal opening •
was observed in female
rats and mice following •
PFOA exposure (2/2).
High and medium
confidence study
Consistent direction
of effects
Limited number of
studies examining
outcome
C-16
-------
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 Medium confidence
study
Female mating and
fertility
1 High confidence study
1 Medium confidence
study
No effects on female
mating or fertility
parameters were observed
in one- and two-generation
reproduction studies in
rats with PFOA exposure
beginning 10 weeks prior
to mating (2/2).
High and medium
confidence study
Consistent direction
of effects
Limited number of
studies examining
outcome
Notes: E2 = estradiol; SHBG = sex hormone binding globulin; SES = socioeconomic status; DBP = diastolic blood pressure; SBP = systolic blood pressure; GD = gestational day;
LH = luteinizing hormone; GnRH = gonadotropin-releasing hormone.
C-17
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.2 Endocrine
EPA identified 34 epidemiological and 9 animal studies that investigated the association between
PFOA and endocrine effects. Of the epidemiological studies, 4 were classified as high
confidence, 15 as medium confidence, 9 as low confidence, 3 as mixed (1 high/medium, 1
medium/low, and 1 medium!uninformative) confidence, and 3 were considered uninformative
(Section C.2.1). Of the animal studies, 3 were classified as high confidence, and 6 were
considered medium confidence (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
PFOA Document).
C.2.1 Human Evidence Study Quality Evaluation and
Synthesis
C.2.1.1 Introduction
Thyroid disease encompasses conditions such as hypothyroidism and hyperthyroidism, and it is
more common in females than in males. Hypothyroidism is characterized by elevated thyroid
stimulating hormone (TSH) and concurrently low 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, 3982042} and HESD {U.S. EPA, 2016, 3603279}
identified limited evidence of endocrine effects of PFOA for thyroid disease, hypothyroidism,
and hypothyroxinemia. Evidence from occupational cohorts and from general population studies
was mixed. An analysis of an occupational cohort in Minnesota {Olsen, 1998, 1290857} showed
elevated TSH (p = 0.002) levels in a single exposure group (10-30 |ag/mL serum PFOA);
however, this increase was not observed for those with greater exposure ( > 30 |ag/mL serum
PFOA). Pooled occupational analyses, combing the Minnesota cohort with cohorts from
Belgium and Alabama {Olsen, 2003, 1290020; Olsen, 2007, 1290836}, showed a negative
association for free T4, and a positive association was found for T3. Two studies on participants
from the C8 Health Project showed positive associations between estimated PFOA exposure
(cumulative and yearly) and all incident self-reported thyroid disease in women {Winquist, 2014,
2337818}, and thyroid disease in children examining modeled in utero PFOA exposure and
concurrent PFOA serum concentrations {Lopez-Espinosa, 2012, 1291122}. As a result of these
findings, the C8 Science Panel concluded that a probable link exists between PFOA and thyroid
disease {C8 Science Panel, 2012, 1430770}. In general population studies, positive associations
were found with T4 in older adults {Shrestha, 2015, 2851052}, with T3 (free and total) in
females {Wen, 2013, 2850943}, and between prenatal PFOA (cord blood) and T4 concentrations
in thyroid disease-free girls {de Cock, 2014, 2718059}. Other studies did not observe significant
associations in adults and children {Bloom, 2010, 757875; Lin, 2013, 1332458}. Most results in
studies on pregnant women were not significant except for small positive associations with TSH
{Berg, 2015, 2851002}, especially in pregnant women with elevated TPOAb {Webster, 2014,
2850208}.
C-18
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
For this updated review, 32 studies (33 publications)6 report on the association between PFOA
exposure and endocrine effects. Six publications were studies in pregnant women {Aimuzi,
2020, 6512125; 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. One study was a controlled trial {Convertino, 2018, 5080342}, six were
cohort studies {Blake, 2018, 5080657; Crawford, 2017, 3859813; Lebeaux, 2020, 6356361; Liu,
2018, 4238396; Preston, 2018, 4241056; Reardon, 2019, 5412435}, six were cohort and cross-
sectional studies {Dreyer, 2020, 6833676; Itoh, 2019, 5915990; Kim, 2020, 6833758; Kato,
2016, 3981723; Wang, 2014, 2850394; Xiao, 2019, 5918609} two case-control studies {Kim,
2016, 3351917; Predieri, 2015, 3889874}, one case-control and cross-sectional study {Zhang,
2018, 5079665}, and 18 cross-sectional studies {Abraham, 2020, 6506041; Aimuzi, 2019,
5387078; Aimuzi, 2020, 6512125; Byrne, 2018, 5079678; Caron-Beaudoin, 2019, 5097914;
Christensen, 2016, 3350721; 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; Shah-Kulkarni, 2016, 3859821; Tsai, 2017,
3860107; Yang, 2016, 3858535}. All observational studies measured PFOA in blood
components (i.e., blood, plasma, or serum). 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 PFOA and thyroid
hormone concentrations, other endocrine outcomes examined included: thyroid disease, thyroid
antibodies (thyroglobulin antibodies (TgAb) and thyroid peroxidase antibody (TPOAb)), and
thyroid hormone-associated proteins (e.g., thyroglobulin, T4-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 PFOA measurements and thyroid
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 PFOA concentrations
measured during pregnancy and thyroid hormone levels in cord blood or the child's blood
attenuate any concerns for potential reverse causality. Measuring PFOA and thyroid hormone
concentrations concurrently in maternal serum was considered adequate in terms of exposure
assessment timing. Given the long half-life of PFOA (median half-life = 2.7 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 34 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and endocrine effects. Study quality evaluations for these 34 studies
are shown in Figure C-10 and Figure C-l 1.
6 Itoh et al. (2019, 5915990) reports thyroid-related hormone levels in a population overlapping with Kato et al. (2016, 3981723).
C-19
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Of the 34 studies identified since the 2016 assessment, 4 studies were classified as high
confidence, 15 as medium confidence, 9 as low confidence, 3 as mixed (1 high/medium, 1
medium/low, and 1 medium/uninformative) confidence, and 3 were considered uninformative
{Abraham, 2020, 6506041; Kim, 2016, 3351917; Seo, 2018, 4238334}.
C-20
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
ecN^
>0®
Abraham et al., 2020, 6506041 -
Aimuzi et al., 2019, 5387078-
Dreyer et al., 2020, 6833676
Dufouretal., 2018, 4354164
Inoue et al., 2019, 5918599
Itoh et al., 2019, 5915990
Jain et al., 2019, 6315816-
Jain, 2013, 2168068
Kangetal., 2018, 4937567
Kato et al., 2016, 3981723-
KhaMI et al., 2018, 4238547-
1
+
1,
+
1
+
B
i
i
+
•n
+
+
+
++
+
+
+
+
++
+
+
+
+
+
+
+
+
+
+
+
+
+
- -
++
-
+
-
-
+
+
-
+
+
+
+
+
- +
+
-
+
+
-
+
+
+
-
+
+
+
-
+
++
-
+
+
+
+
++
+
++
+
+
++
+
+
+
+
+
-
•
+
++
++ ++
+
+
++
+
++
B
+
+
+
+
+
- +
+
++
++
+
-
+
+
+
+
+
-
-
+
-
+
+
-
+
+
+
+
+
+
+*
++
+
-
+
+
+*
- -
+
+
-
+
+
¦
-
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-10. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Endocrine Effects
Interactive figure and additional study details available on HAWC.
C-21
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
vC®
J L
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
Kim et al., 2016, 3351917-
B
++
B
-
B
-
~
Kim et al., 2020, 6833758-
++
++
++
B
++
+
¦
Lebeaux et al., 2020, 6356361 -
++
++
++
++
++
+
+
Lewis et al., 2015, 3749030-
+
+
+
+
-
+
+
-
Li et al., 2017, 3856460-
-
+
-
-
+
-
+
-
Liu et al., 2018, 4238396-
-
+
++
+
+
+
+
+
Preston et al., 2018, 4241056-
+
+
+
+
+
+
+
+
Reardon et al., 2019, 5412435 -
+
+
-
+
+
+
+
+
Seo et al„ 2018, 4238334-
-
+
-
D
-
-
~
Shah-Kulkami et al„ 2016, 3859821 -
+
+
+
+
+
+
+
+
Tsai et al„ 2017, 3860107-
-
+
-
+
+
+
+
-
Wang etal., 2014, 2850394-
+
+
+
+
++
+
+
+
Xiao et al., 2020, 5918609-
++
++
+
+
++
+
+
++
Yang etal., 2016, 3858535-
+
+
+
+
+
+
+
+
Zhang et al., 2018, 5079665 -
-
-
-
+
+
+
+
-
van den Dungen et al., 2017, 5080340 -
-
+
+
-
+
+
-
-
Figure C-ll. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Endocrine Effects (Continued)
Interactive figure and additional study details available on HAWC.
C-22
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
These differences resulted in high confidence {Lebeaux, 2020, 6356361} and medium
confidence {Dufour, 2018, 4354164; Kato, 2016, 3981723} for infant or child analyses. For
maternal analyses which tend to be cross-sectional in nature, the uncertainty regarding
temporality resulted in medium confidence {Lebeaux, 2020, 6356361}, low confidence {Kato,
2016, 3981723}, or uninformative {Dufour, 2018, 4354164} ratings.
Studies rated as low confidence or uninformative had deficiencies including lack of accounting
for population sampling methods {Lewis, 2015, 3749030}, or residual confounding {Abraham,
2020, 6506041; Convertino, 2018, 5080342; Kim, 2016, 3351917; Predieri, 2015, 3889874}, or
lack of information on allocation of participants to treatment levels {Convertino, 2018,
5080342}, participant recruitment and case definitions {Kim, 2016, 3351917; Predieri, 2015,
3889874} or small sample sizes {Kim, 2016, 3351917; Predieri, 2015, 3889874}.
C.2.1.3 Findings from Children
One high confidence study {Kim, 2020, 6833758} observed no association with subclinical
hypothyroidism in children six years of age. Congenital hypothyroidism (CH) was assessed in
South Korean infants in a very small case-control study {Kim, 2016, 3351917}. PFOA
concentrations were significantly higher in infants with CH compared to controls (means 5.4 and
2.12 ng/mL, respectively, p-value < 0.01) (Appendix D). However, the study was considered
uninformative because of potential key confounding factors were not controlled for in the
analysis, and the small sample size.
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 four observed significant effects (Appendix D). One high confidence
study {Xiao, 2019, 5918609} observed a large positive association between maternal third
trimester PFOA and cord serum TSH. The effect size for TSH was similar after stratification by
infant sex, but no longer significant. Additionally, sex-stratified analyses showed positive
associations between maternal PFOA and measures of T4 (total T4 and free T4 index (FTI)) in
cord blood from female infants. No other significant associations were observed for TSH among
other studies on children. Another high confidence study {Kim, 2020, 6833758} showed positive
associations between serum PFOA concentrations and free T4 levels at age 6. After stratifying
by child sex, the association remained among boys but was not observed in girls. This effect was
also observed in a medium confidence cross-sectional study in newborns {Aimuzi, 2019,
5387078}, which reported significant positive associations with free T4 in cord blood. When
stratified by sex, the effect persisted in male newborns, but was not seen in female newborns.
These three studies report consistent, significant positive associations with T4 in children;
however, the effect was not consistent between boys and girls in different populations. Similarly,
a medium confidence cross-sectional study {Kang, 2018, 4937567} showed a borderline
significant positive association between serum PFOA and free T4 (p = 0.075). Analyses of
children from the Hokkaido Study {Itoh, 2019, 5915990; Kato, 2016, 3981723} did not observe
significant associations with thyroid hormones. The remaining studies that did not observe
significant effects
C-23
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Thyroid antibody (TA) levels were examined in one study {Itoh, 2019, 5915990} which found
significant effects (Appendix D). A medium confidence study on children from the Hokkaido
Study on the Environment and Children's Health {Itoh, 2019, 5915990} showed mixed
associations between maternal PFOA concentrations and thyroglobulin antibody levels. An
inverse association was found for TgAb levels among boys born to TA-negative mothers; no
effects were seen among all boys or boys born to TA-positive mothers. The opposite trend was
seen in girls; a positive association for TgAb levels was observed for girls born to TA- positive
mothers. No effects were observed in all girls or girls born to TA-negative mothers.
C.2.1.4 Findings from Pregnant Women
Thyroid hormone levels were examined in five studies {Aimuzi, 2020, 6512125; Inoue, 2019,
5918599; Itoh, 2019, 5915990; Reardon, 2019, 5412435; Shah-Kulkarni, 2016, 3859821} and
two observed significant effects (Appendix D). A medium confidence study {Preston, 2018,
4241056} in pregnant women showed a significant decrease in the FTI with increasing first
trimester serum PFOA concentrations. Associations with other thyroid hormones were not
observed among the whole study sample. However, analyses stratified by TPOAb status showed
a borderline significant (p = 0.08) inverse effect of PFOA on TSH among TPOAb-positive
women; no effects were seen in TPOAb-negative women. Another medium confidence study
{Aimuzi, 2020, 6512125} observed a positive association between serum PFOA and early
pregnancy free T4, but this effect was not seen when stratified by TPOAb-status.
Thyroid hormone antibodies were examined in one study {Itoh, 2019, 5915990} which found a
significant effect. A negative association was observed for TPOAb levels in first trimester serum
among mothers in the Hokkaido Study. One cross-sectional study {Dufour, 2018, 4354164} on
mother-child dyads showed evidence of a large increased risk of hypothyroidism in mothers (OR
Q4 vs. Q1 (95% CI): 5.62 (1.64-26.11)), however, there was a great deal of uncertainty in regard
to 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 Odense
Child Cohort (OCC) and showed a significant decrease in serum Cortisol with two-fold increases
in serum PFOA concentrations {Dreyer, 2020, 6833676}. However, diurnal urinary (dU) -
Cortisol, dU-cortisone, and dU-cortisol/cortisone showed non-significant decreases.
C.2.1.5 Findings from the General Adult Population
One study examined thyroid disease among male anglers (age > 50 years) and observed a non-
significant increase in odds of self-reported thyroid disease with increasing serum PFOA
concentrations {Christensen, 2016, 3350721}.
Thyroid function was examined in 12 studies {Blake, 2018, 5080657; Byrne, 2018, 5079678;
Convertino, 2018, 5080342; Crawford, 2017, 3859813; Jain, 2013, 2168068; Jain, 2019,
6315816; Lebeaux, 2020, 6356361; Lewis, 2015, 3749030; Li, 2017, 3856460; Liu, 2018,
4238396; Seo, 2018, 4238334; Zhang, 2018, 5079665} and seven observed significant effects
(Appendix D). A low confidence case-control study {Zhang, 2018, 5079665} examined women
with and without POI found a positive association among controls (i.e., women without POI) for
TSH concentrations with increasing plasma PFOA concentrations. Similarly, TSH levels were
elevated in women with POI which was accompanied by a concomitant negative association with
C-24
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
free T4 concentrations. The thyroid hormone concentrations were within normal ranges in both
cases and controls. Another low confidence case-control study {Heffernan, 2018, 5079713} on
women with and without PCOS found a similar increase in TSH among cases. However, findings
need to be interpreted with caution, since both studies were considered low confidence due to a
lack of information on the control recruitment and selection process.
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
including an inverse association with total T4 in men aged 40 to 60 years, increased total T4 and
decreased TSH in women aged 12 to 20 years, increased free T3 in women aged 20 to 40 years,
and concurrent increases in free and total T3 among women aged 60 years or older. However,
there is no evidence NHANES complex sampling design was accounted for in the analysis which
contributed to a low confidence rating. Jain (2013, 2168068), another low confidence study,
found a significant increase in TSH levels among those NHANES (2007-2008) participants in
the highest tertile (> 5.1 ng/mL) of PFOA exposure compared to the lowest (< 3.3 ng/mL). A
medium confidence follow-up study {Jain, 2019, 6315816} on NHANES (2007-2012)
participants investigated associations with serum PFOA and thyroid hormone concentrations,
stratified by glomerular function (GF) status (GF1, GF-2, GF-3A, and GF-3B/4). Few significant
and borderline significant results were observed; however, the direction of association was
inconsistent across increasing glomerular filtration groups and did not suggest an interaction with
glomerular filtration status. Associations between PFOA and thyroid hormones were inconsistent
across NHANES studies. Lewis et al. (2015, 3749030) and Jain (2013, 2168068) found
significant effects in opposite directions for TSH, however, these effects were in observed in
different NHANES cycles and among different subpopulations. In the 2011-2012 NHANES
participants, Lewis et al. (2015, 3749030) found consistent effects for T3 in women of different
ages, but other results were inconsistent between age and sex groupings.
Inverse associations with TSH and T4 were also observed in 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). One additional low confidence,
cross-sectional study {Byrne, 2018, 5079678} on Alaska natives found a significant positive
association for TSH among all participants and an inverse association with total T3 in men;
however, this population was relatively small (total n = 85; male n = 38) with low exposure
levels (median: 1.01 ng/mL (25th-75th percentile: 0.753-1.44 ng/mL)).
In a controlled trial {Convertino, 2018, 5080342} in which subjects were administered
ammonium perfluorooctanoate (APFO) doses ranging 50-1200 mg for six weeks, {Convertino,
2018, 5080342} report an increase in the average rate of change in free T4. A dose-dependent
increase was also demonstrated by grouping subjects into three treatment bins and showing
increasing mean and median free T4 concentrations. This study, however, was rated as low
confidence because potential confounders were not considered during participant allocation to
treatment groups or in the statistical analysis.
C-25
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.2.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 3 studies from the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} and 6 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOA HESD that investigated the association between PFOA and endocrine effects. Study
quality evaluations for these 9 studies are shown Figure C-12.
Blake etal., 2020, 6305864-
+
+
++
Butenhoffetal., 2004, 1291063-
++
NR
NR
Butenhoff et al., 2012, 2919192-
+
++
NR
De Guise et al., 2021, 9959746 -
+
+
NR
Hu et al., 2010, 1332421 -
++
NR
NR
Loveless et al., 2008, 988599 -
+
+
NR
NTP, 2019, 5400977-
++
++
NR
NTP, 2020, 7330145-
++
++
NR
Sun etal., 2018, 5079802-
+
+
NR
5
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-12. Summary of Study Evaluation for Toxicology Studies of PFOA and
Endocrine Effects
Interactive figure and additional study details available on HAWC.
Available animal toxicity data suggest that PFOA exposure can interfere with male and female
endocrine systems. Overall, studies have reported endocrine organ weight changes, hormone
fluctuations, and organ histopathology across studies of varying durations of oral exposure to
PFOA. Effects typically exhibit a sex-bias depending on the species, endpoint, and exposure
C-26
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
paradigm, likely due to known toxicokinetic differences (see Main PFOA Document). The
thyroid gland and thyroid hormones appear to be affected by PFOA exposure. Effects of PFOA
on gonads and placenta and on reproductive hormones are described in detail in (see Main PFOA
Document).
C.2.2.1 Organ Weight Changes
Significant changes in absolute and relative endocrine organ weights have been observed in
monkeys {Goldenthal, 1978, 1291068} and rats {Butenhoff, 2012, 2919192; Butenhoff, 2004,
1291063; NTP, 2019, 5400977} following oral exposure to PFOA, often with a male-bias in
response (Figure C-13).
Absolute and relative thyroid gland weight was quantified as part of a short-term exposure study
conducted by NTP (2019, 5400977). In that study, male Sprague-Dawley rats received 0, 0.625,
1.25, 2.5, 5, or 10 mg/kg/day PFOA and females received 0, 6.25, 12.5, 25, 50, or 100 mg/kg/day
via gavage for 28 days. Absolute thyroid weight was only significantly increased in males of the
2.5 mg/kg/day exposure group. Thyroid gland weight relative to body weight was elevated in
males administered >_1.25 mg/kg/day PFOA by the end of the study, which may be related to
reductions in mean body weights that were observed in males but not females (Section C.3.2.2),
though body weight in males of the 1.25 mg/kg/day dose group was only modestly reduced by
4.6% compared to controls. No statistically significant effects were observed in females at any
dose and no effects were observed on absolute or relative adrenal gland weight in either sex
{NTP, 2019, 5400977}.
C-27
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Endpoint
Thyroid Gland Weight, Absolute
Study Name
NTP, 2019, 5400977
Study Design Observation Time Animal Descriptioi
-term (28d) 29d Rat. Sprague-Dawley N=10)
Rat, Sprague-Dawley ('{, N=9-10)
Dose (mg/kg/day)
PFOA Endocrine Effects - Organ Weights
| Statistically significant 0 Not statistically significant | I 95% CI |
Thyroid Gland Weight. Relative NTP, 2019, 5400977
n (28d) 29d
Adrenal Gland Weight. Absolute Butenhoff et al.. 2012,2919192 chronic (2y)
drannl Gland Weight, Ratal
avalapmanlal (GD6-17)
Adrenal Gland Weight. Relative to Brain Butenhoff et al.. 2012,2919192 chronic (2y)
Pituitary Gland Weight, Abso
Butenhoff et al... 2004.1291063 reproductive <84d) 1022
reproductive (GD1-PND106) LD22
Rat. Sprague-Dawley ( N=10>
Rat, Sprague-Dawley ('j,
Rat, Sprague-Dawley Cri:Cd(Sd)(Br) N=15)
Rat. Sprague-Dawley Crl:Cd{Sd)(Br) N=14-15)
F1 Mouse, C57Bl_'6n (i', N=B)
Rat. Sprague-Dawley Cri:Cd(Sd)(Br) (o. N=15)
Rat. Sprague-Dawley Cri:Cd(Sd)(Br) f-\ N=14-15)
PC Rat, Crl:CD(SD)IGS BR (•?. N=26-29)
F1 Rat. Crl:CD(SD)IGS BR (- , N =28-29)
Figure C-13. Percent Change in Endocrine Organ Weights Relative to Controls in Rodents
Following Exposure to PFOAa
Interactive figure and additional study details available on HAWC.
GD = gestation day; PND = postnatal day; LD= Lactational Day; Po = parental generation; Fi = first generation; d = day;
y = year.
a CIs for some studies may be too narrow to view at this scale.
Relative pituitary gland weight was elevated in male rhesus monkeys exposed to 3 mg/kg/day via
gavage for 90 days. Changes in body weight were similar to controls for these animals
{Goldenthal, 1978, 1291068}. In male Sprague-Dawley (Crl:COBS@CD(SD)BR) rats, pituitary
weights (absolute and relative to brain or body weight) were reduced following a year-long
dietary exposure to 300 ppm PFOA, which is equivalent to 14.2 mg/kg/day {Butenhoff, 2012,
2919192}. The decrease was consistent across all measures despite slight (i.e., < 10%) non-
significant decreases in both body weight and absolute brain weight. Decrements in pituitary
gland weight were not observed in female rats given the same 300 ppm exposure for one year
(16.1 mg/kg/day equivalent) {Butenhoff, 2012, 2919192}. Another study by Butenhoff et al.
(2004, 1291063) in Sprague-Dawley rats described female-specific reductions in pituitary gland
weight following a multi-lifestage PFOA exposure paradigm. In this study, absolute pituitary
C-28
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
gland weights were reduced in adult Fi females (on lactational day 22 of the F2 generation)
following oral exposure to 3, 10, or 30 mg/kg/day PFOA from GD 0-PND 127 {Butenhoff,
2004, 1291063}. Although relative pituitary weights were not provided, there were not
significant changes in body weights at sacrifice nor absolute brain weights (Section C.4.2),
which implies the reduction in absolute pituitary weight may reflect a specific effect on the
pituitary gland. Fi pup weight was only reduced in the 30 mg/kg/day group during development,
indicating that slower pup growth is not an explanation for the reduced absolute pituitary
weights.
Male-specific reductions in absolute adrenal gland weight and relative to brain weight were
observed by Butenhoff et al. (2012, 919192) after one year of exposure to 300 ppm PFOA
(equivalent to 14.2 mg/kg/day), but was not observed after two years {Butenhoff, 2012,
2919192}. A developmental exposure study by Hu et al. (2010, 1332421) examined relative
body weight of adrenal glands in PND 48 Fi female C57BL/6N mice following maternal
exposure to 0, 0.5, or 1 mg/kg PFOA from GD 6-17. There was an apparent dose-related trend,
but none of the adrenal weights of exposed groups were significantly different from the control
and the study authors did not conduct a trend test.
C.2.2.2 Hormone fluctuations
Several studies have described fluctuations in the levels of hormones secreted from the adrenal,
pituitary, and thyroid glands following short term exposure to rats and mice {NTP, 2019,
5400977; Sun, 2018, 5079802}, exposure during pregnancy to mice {Blake, 2020, 6305864},
and chronic exposure to non-human primates {Butenhoff, 2002, 1276161}.
In the aforementioned 28-day rat study conducted by NTP (2019, 5400977), male-specific
reductions in T4, FT4, and T3 were observed in almost all exposure groups (Table C-3; Figure
C-14); T3 was not significantly affected in the 10 mg/kg/day group, though statistically
significant reductions were observed in all lower dose groups. Notably, these effects in males
occurred at doses lower than those that resulted in decreased body weight, which may be
confounding with hormone responses, as shown in dietary restriction studies in rats {Laws, 2007,
1411456}. T4 and FT4 were significantly reduced in females from the 100 mg/kg/day exposure
group {NTP, 2019, 5400977}. Opposing effects of TSH were observed between the sexes.
Although female TSH concentrations were increased in all exposure groups (6.25-
100 mg/kg/day), male TSH was reduced in the 5 and 10 mg/kg/day exposure groups when
compared to controls {NTP, 2019, 5400977}.
C-29
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOA Endocrine Effects - Thyroid Hormones from NTP, 2019, 5400977
Endpoint Animal Description Dose (mg/kg/day)
O Statistically significant 0 Not statistically significant |—j 95% CI
Thyroid stimulating hormone (TSH) Rat, Sprague-Dawley (¦$, N=10) 0
1
0.625
1.25
\—
i—«—i
•
1
2.5
i i—•—
1
5
i i 9
1
10
1 i—•—i
1
Rat, Sprague-Dawley ( , , N=8-10) 0
i | .
6.25
i
i
! • !
12.5
i h
© i ,
25
50
i
i
! # 1 |
1 & 1 1
100
i
i • M
Thyroxine (T4), Free Rat, Sprague-Dawley (c?, N=10) 0
0.625
o
. i
i
i
1.25
i °
I
2.5
l o
i
5
I o
i
10
1 o
i
Rat, Sprague-Dawley ($, N=9-10) 0
6.25
1 1—
1
1—
> 1 1
•
12.5
1 1 *
1 i
25
1 1 •
1 i
50
1 1—•-
i
100
Thyroxine (T4), Total Rat, Sprague-Dawley 0, N=10) 0
—<
i
>—< !
0.625
,©
i
1
1.25
1©
1
2.5
o
1
5
OI
1
10
'l-#H
1
Rat, Sprague-Dawley (?, N=9-10) 0
1—<
. 1
6.25
—#—1
12.5
25
i nn
I h-9
1
—1 1
50
1 H#H
1
100
Triiodothyronine (T3) Rat, Sprague-Dawley (o , N=10) 0
0.625
1.25
1 1-©-!
! ^
h©H
| " • '
1
>-h ;
i
i
2.5
1 l-^-l
i
5
1 |-#H
i
10
" 1 t
—i 1
Rat, Sprague-Dawley (+', N=9-10) 0
6.25
' i
I
>—i
4)—I
12.5
25
50
100
1 (-•-
1 \—
f 1 1 1 1
1
•—I i
#—) 1
—I 1
1 1 1 1 1 1
-120 -100 -80 -60 -40 -20 0 20 40 60 80 100 120 140
Percent control response (%)
Figure C-14. Percent Change in Thyroid and Thyroid-Related Hormone Levels of Male
and Female Rats Exposed to PFOA for 28 Days as Reported by NTP (2019, 5400977)a'b
Interactive figure and additional study details available on HAWC.
TSH = thyroid stimulating hormone; T3 = triiodothyronine; T4 = thyroxine; CI = confidence interval.
a Some hormone measurements in male rats were below or approaching the limit of quantifications for FT4 (0.3 ng/dL), T4
(0.5 (ig/dL), and T3 (50 ng/dL).
b The red dashed lines indicate a 100% increase or 100% decrease from the control response
C-30
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Blake et al. (2020, 6305864) administered 0, 1, or 5 mg/kg/day to pregnant CD-I mice from GD
1.5 through sacrifice on GD 17.5. On GD 17.5, levels of thyroid hormones in male and female
placentas were determined, including T4, T3, 3,3',5'-triiodothyronine (reverse T3, rT3), ratio of
T3 to T4 (T3:T4), and ratio of rT3 to T4 (rT3:T4). There were no significant effects of PFOA
exposure on rT3, T3, T4, T3:T4, or rT3:T4 ratio.
In a chronic exposure study by Butenhoff et al. (2002, 1276161), male cynomolgus monkeys
were given 0, 3, 10, or 30/20 mg/kg PFOA per day for 26 weeks. The "30/20" notation reflects a
reduction from 30 to 20 mg/kg/day at day 22 of the study due to toxicity in this exposure group.
Only 2 animals from the 30/20 mg/kg/day group survived until sacrifice, which introduces
uncertainty to the results of this dose group, though they are discussed here. Although no change
in TSH was noted in the highest-dose group, it was significantly elevated in both the 3 and
10 mg/kg/day exposure groups by the end of the study, at day 182 (increases of 63% and 118%
changes, respectively). In the lowest exposure group (3 mg/kg/day), T4 was reduced across
multiple timepoints and decreases reached significance in all three dose groups (33%, 29%, and
32% decreases, respectively) at the conclusion of the study. A dose-dependent decrease in FT4
was also observed across multiple time points, with decreases at day 182 of 33%, 38%, and 42%
in the 3, 10, and 30/20 mg/kg/day dose groups, respectively, compared to control levels. Similar
trends were seen in T3 and free triiodothyronine (FT3) levels throughout the study. By day 182,
total and free T3 levels were decreased by 15%, 14%, and 34% and 13%, 17%, and 40%,
respectively, with increasing dose levels.
Prior to this updated assessment, the available literature measuring thyroid hormones was limited
and acute studies were discussed in the 2016 HESD {U.S. EPA, 2016, 3603279}. One such
study in adult male Sprague-Dawley rats given a single oral exposure of PFOA (20 mg/kg)
reported an 80% reduction in T4 and FT4, and a 25% reduction in serum T3 {Martin, 2007,
758419}. This single-dose study supports the thyroid hormone level perturbations, specifically,
the sensitivity of T4 and FT4, that are observed in the current literature update.
Table C-3. Associations Between PFOA Exposure and Thyroid and Thyroid-Related
Hormone Effects in Rodents and Non-human Primates
Endpoint
Study Name
Species
Exposure
Length
Dose (mg/kg/day)
Sex
Change
TSH
NTP (2019,
5400977)
Sprague-
Dawley rat
28 day
0, 0.625, 1.25, 2.5,
5.0, 10 mg/kg/day
M
J. 5—10 mg/mg/day
0, 0.625, 12.5, 25,
50, 100 mg/kg/day
F
t 6.25-
100 mg/kg/day
Butenhoff et al.
(2002 1276161)
Cynomolgus
monkeys
26 weeks
0, 3, 10, or
30/20 mg/kg
M
t 3-10 mg/kg/day
T3 (Total)
Martin et al. (2007
758419)
Sprague-
Dawley rat
single dose
20 mg/kg/day
M
I 20 mg/kg/day
NTP (2019,
5400977)
Sprague-
Dawley rat
28 day
0, 0.625, 1.25, 2.5,
5.0, 10 mg/kg/day
M
4 0.625-
5.0 mg/kg/day
0, 0.625, 12.5, 25,
50, 100 mg/kg/day
F
n.s.
Butenhoff et al.
(2002, 1276161)
Cynomolgus
monkeys
26 weeks
0, 3, 10,
30/20 mg/kg/day
M
1 30/20 mg/kg/day
C-31
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Endpoint
Study Name
Species
ELcn^hC D°se (mg/kg/day)
Sex
Change
Blake et al. (2020,
6305864)
CD-I mice
Developme 0, 1, 5 mg/kg/day
ntal
(GDI.5-
17.5)
M
n.s.
0, 1, 5 mg/kg/day
F
n.s.
FT3
Butenhoff et al.
(2002, 1276161)
Cynomolgus
monkeys
26 weeks 0, 3, 10,
30/20 mg/kg/day
M
4 30/20 mg/kg/day
rT3
Blake et al. (2020,
6305864)
CD-I mice
Developme 0, 1, 5 mg/kg/day
ntal
(GDI.5-
17.5)
M
n.s.
0, 1, 5 mg/kg/day
F
n.s.
T4 (Total)
Martin et al. (2007,
758419)
Sprague-
Dawley rat
single dose 20 mg/kg/day
M
| 20 mg/kg/day
NTP (2019,
5400977)
Sprague-
Dawley rat
28 day 0, 0.625, 1.25, 2.5,
5.0, 10 mg/kg/day
M
4 0.625-
10 mg/kg/day
0, 0.625, 12.5, 25,
50, 100 mg/kg/day
F
| 100 mg/kg/day
Butenhoff et al.
(2002, 1276161)
Cynomolgus
monkeys
26 weeks 0, 3, 10, or
30/20 mg/kg
M
4 3-
30/20 mg/kg/day
Blake et al. (2020,
6305864)
CD-I mice
Developme 0, 1, 5 mg/kg/day
ntal
(GDI. 5-
17.5)
M
n.s.
0, 1, 5 mg/kg/day
F
n.s.
FT4
Martin et al. (2007,
758419)
Sprague-
Dawley rat
single dose 20 mg/kg/day
M
4 20 mg/kg/day
NTP (2019,
5400977)
Sprague-
Dawley rat
28 day 0, 0.625, 1.25, 2.5,
5.0, 10 mg/kg/day
M
4 0.625-
10 mg/kg/day
0, 0.625, 12.5, 25,
50, 100 mg/kg/day
F
4 100 mg/kg/day
Butenhoff et al.
(2002, 1276161)
Cynomolgus
monkeys
26 weeks 0, 3, 10, or
30/20 mg/kg
M
4 10-
30/20 mg/kg/day
Notes: F = female; M = male; n.s. = nonsignificant; T3 = triiodothyronine; T4 = thyroxine; TSH = thyroid stimulating hormone.
Perturbations in adrenal and pituitary hormone levels have been described primarily in rodent
studies (Table C-4). Loveless et al. (2008, 988599) reported elevations in serum corticosterone in
male Crl:CD(SD)IGS BR rats and male Crl:CD-l(ICR)BR mice exposed to 10 or 30 mg/kg/day
PFOA for 29 days, although statistically significant effects were only noted at the 10 mg/kg/day
dose in mice. Increases in rats of the 10 and 30 mg/kg/day groups were 35% and 96% changes,
respectively and in mice were 129% and 131% changes, respectively {Loveless et al., 2008,
988599}. Two studies in mice support that PFOA exposure is associated with elevations in
corticosterone in both males and females. De Guise et al. (2021, 9959746) found that serum
corticosterone was significantly higher in female B6C3F1 mice exposed to 1.88 or
7.5 mg/kg/day PFOA for 28 days (72% and 158%) fold increases, respectively). Sun et al. (2018,
5079802) found that serum corticosterone was elevated in male BALB/c mice exposed to 5 or
20 mg/kg/day PFOA for 28 days (146 and 175%) changes, respectively). This study also
C-32
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
quantified adrenocorticotropic hormone (ACTH). A dose-dependent reduction in ACTH was
observed, however significant effects were only observed at the 20 mg/mg/day dose (-26 and
-58% changes in the 5 and 20 mg/kg/day groups, respectively) {Sun, 2018, 5079802}.
Table C-4. Associations Between PFOA Exposure and Adrenocortical Hormone Effects in
Rodents
Endpoint
Study Name
Species
Exposure
Length
Dose
(mg/kg/day)
Sex
Change
CORT De Guise et al. (2021,
9959746)
B6C3F1 28 days
0, 1.88,
7.5 mg/kg/day
t 1.88 and
7.5 mg/kg/day
Sunetal. (2018,
5079802)
BALB/c 28 days
0, 1.25, 5,
20 mg/kg/day
M
t 5 and
20 mg/kg/day
Loveless et al. (2008,
988599)
Sprague- 29 days
Dawley rat
0,0.3, 1, 10,
30, mg/kg/day
M
n.s.
CD- 29 days
1(ICR)BR
mice
0,0.3, 1, 10,
30, mg/kg/day
M
t 10 mg/kg/day
ACTH Sunetal. (2018,
5079802)
BALB/c 28 days
0, 1.25, 5,
20 mg/kg/day
M
| 20 mg/kg/day
Yofc.v; ACTH = Adrenocorticotropic Hormone; CORT = serum corticosterone; F = female; M = male;
n.s. = nonsignificant.
C.2.2.3 Histopathology
In addition to the neoplastic lesions described in (see PFOA Main Document), several non-
neoplastic lesions have been observed in the thyroid gland and adrenal glands (Figure C-15).
C.2.2.3.1 Thyroid
In the 28-day exposure study, NTP (2019, 5400977) found higher incidences (8/10, minimal
severity) of thyroid follicular cell hypertrophy in female rats following exposure to
100 mg/kg/day PFOA. Three of 10 high-dose males (10 mg/kg/day) also exhibited these
abnormalities. No such lesions were observed in any of the other groups. Although statistical
significance was not achieved, the presence of thyroid follicular cell hypertrophy in both males
and females supports that it is likely an exposure-related effect {NTP, 2019, 5400977}.
In two chronic exposure studies {Butenhoff, 2012, 2919192; NTP, 2020, 7330145}, male and
female Sprague-Dawley rats were fed diets containing PFOA for approximately two years. NTP
(2020, 7330145) used a matrix-type exposure paradigm whereby pregnant rats were administered
PFOA on GD 6 and exposure was continued in offspring postweaning for a total of 107 weeks.
Tissue sections from endocrine organs, including the thyroid gland, were analyzed for histology
in both male and female offspring. Dose groups for this report are referred to as "[perinatal
C-33
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
exposure level (ppm)]/[postweaning exposure level (ppm)]" (e.g., 300/1,000; see Main PFOA
Document for further study design details).
In the thyroid gland, NTP (2020, 7330145) reported higher incidences of follicular cell
hypertrophy in males from the 0/300 ppm group at the 16-week interim evaluation as well as the
terminal evaluation. In females, higher incidences were noted in the 300/1,000 ppm group at the
16-week interim. No differences were observed between groups with combined perinatal and
postweaning exposure compared to groups with postweaning exposure only {NTP, 2020,
7330145}. NTP (2020, 7330145) suggested the elevated incidence of follicular cell hypertrophy
in males could be related to lower concentrations of circulating total T4 and T3, a result that was
observed in the aforementioned NTP 28-day toxicity study {NTP, 2019, 5400977} but were not
assessed in the chronic study. Similarly, Butenhoff et al. (2012, 2919192) observed increased
incidences (13%, n = 49; compared to 2% in controls, n = 50) of thyroid c-cell hypertrophy in
male rats exposed to 30 ppm PFOA for two years (equivalent to 1.3 mg/kg/day), although the
effects did not reach statistical significance nor was there an increase in the 300 ppm males.
Females had an apparent dose-dependent increase in follicular cell hypertrophy with an
incidence of 0/50, 1/49, and 3/49 in the control, 30 ppm, and 300 ppm, respectively; however,
the results were not statistically significant. Although there were sporadic occurrences of
follicular cell hyperplasia in the males, there were no apparent treatment-related effects
{Butenhoff, 2012, 2919192}.
C.2.2.3.2 Adrenal
In a chronic dietary study in rats, the incidence of adrenal gland hyperplasia was 18% (n = 50) in
males exposed to 300 ppm PFOA compared to 4% in controls (n = 49), but the effect did not
reach statistical significance {Butenhoff, 2012, 2919192}. A rat reproductive study by Butenhoff
et al. (2004, 1291063) observed treatment-related microscopic changes in the adrenal glands of
high-dose Fi animals including cytoplasmic hypertrophy and vacuolation of the cells of the
adrenal cortex following exposure to 3, 10, or 30 mg/kg/day {Butenhoff, 2004, 1291063}. In
males, the cells of the adrenal glands were thicker, the zona glomerulosa was more prominent,
and adrenal cortex cells were more vacuolized in 2/10 males from the 10 mg/kg/day exposure
group and 7/10 males from the 30 mg/kg/day group. No effects were observed in females
{Butenhoff, 2004, 1291063}. The adrenal glands appeared normal, and no histopathology was
observed in a study of male cynomolgus monkeys administered up to 30 mg/kg/day PFOA for
6 months by oral tablet {Butenhoff, 2002, 1276161}, or the 28-day and chronic rat studies
conducted by NTP (2019, 5400977; 2020, 7330145).
Non-neoplastic lesions in the pancreas are described in the Main PFOA Document.
C-34
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Endpolnt Study Name Study Design Observation Time Animal Description
Thyroid Gland, Hyperplasia, C-Cell Butenhoffetal., 2012.2919192 chronic <2y) 2y Rat, Sprague-Dawley Cri:Cd(Sd)(Br) (N=43-47)
Rat, Sprague-Dawley Cri:Cd(Sd)(Br) (',, N=49-50)
Thyroid Gland, Hyparplasia, Follicular Cell Butenhoff et al., 2012, 2919192 chronic (2y) 2y Rat, Sprague-Dawley Cri:Cd(Sd)(Br) N=43-47)
Rat, Sprague-Dawley Cri:Cd(Sd)(Br) (',. N=4S-50)
Thyroid Gland, Hypertrophy, Follicular Cell NTP, 2020, 7330145 chronic (GD6-PNW21) 16wk F1 Rat, Sprague-Dawley (c , N=10)
chronic (GD6-PNW107) 16wk F1 Rat. Sprague-Dawley (", N=10)
chronic (PND21-PNW21) 1<3wk F1 Rat. Sprague-Dawley (r\ N=10)
chronic (PND21-PNW107) 16wk F1 Rat. Sprague-Dawley (i. N=10)
chronic (GD6-PNW107) 2y F1 Rat. Sprague-Dawley (i, N=50)
chronic (PND21-PNW107) 2y F1 Rat. Sprague-Dawley (12. N=50)
Thyroid Hypertrophy.'Hyperplasia NTP. 2019,5400977 short-term <28d) 29d Rat, Sprague-Dawley (-J. N=10)
Rat, Sprague-Dawley ($. N=10)
Adrenal Gland. Non-Neoplastic Lesions Butenhoff et al., 2004,1291063 reproductive (84d) LD22 P0 Rat, Crl:CD(SD)IGS BR (v, N=10)
reproductive <64d) 106d P0 Rat, Crl:CO(SD)IGS BR N=10)
reproductive (GD1-PND106) LD22 F1 Rat, Cri:CD(SD)IGS BR (y, N=10)
reproductive (GD0-PND120) PND120 F1 Rat, Crt:CD(SD)IGS BR (J, N=10)
Pituitary Gland, Non-Neoplastic Lesions Butenhoff et al., 2004,1291063 reproductive <84d) LD22 P0 Rat, Crl:CD{SD)IGS BR (V, N=10)
reproductive (64d) 106d P0 Rat, Crl:CD{SD)IGS BR (0', N=10)
reproductive
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mechanistic Pathway
Animal
In Vitro
Grand Total
Cell Growth, Differentiation, Proliferation, Or Viability
1
9
10
Cell Signaling Or Signal Transduction
1
6
6
Extracellular Matrix Or Molecules
0
1
1
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
0
1
1
Hormone Function
2
11
12
Xen obi otic Metabolism
1
1
2
Other
0
2
2
Not ApplicableJNot Specified/Review Article
1
0
1
Grand Total
3
15
17
Figure C-16. Summary of Mechanistic Studies of PFOA 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 PFOA exposure and endocrine effects in
humans based on studies reporting elevated levels of T4 in children and elevated levels of TSH
in adults. The 2016 HESD for PFOA {U.S. EPA, 2016, 3603279} included two studies reporting
positive associations with thyroid disease and one study reporting negative associations. This
updated review supports positive associations with thyroid disease (hypothyroidism). The most
consistent thyroid hormone effects were observed in children, with four studies (2 high and 2
medium confidence) reporting positive associations for T4; however, some inconsistencies across
sexes were also observed, and a large number of studies observed null effects. One study
reporting significant effects on TSH in children {Aimuzi, 2019, 5387078} conducted multi-
pollutant models including other measured PFAS (i.e., PFOS, PFNA, PFDA, PFUA, PFHxS,
PFDoA, and perfluorobutane sulfonate (PFBS)). PFOA was moderately correlated with other
PFAS (r = 0.23-0.56) in cord blood, and estimates were found to be largely unchanged in
multipollutant models. Most results in general population studies indicated positive associations
for TSH. Many high and medium confidence studies generally did not observe significant
associations with endocrine outcomes. Several low confidence studies observed associations, but
the interpretation of these results is limited by several factors related to study quality. Additional
uncertainty exists due to the potential for confounding by other PFAS.
The animal evidence for an association between PFOA exposure and effects in the endocrine
system is considered moderate based evidence from 8 high or medium confidence animal studies.
The strongest evidence of endocrine effects is from perturbations in hormones related to the
thyroid gland. Thyroid hormones appear to be sensitive to PFOA exposure but exhibit highly
C-36
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
complex responses depending on sex, species, and exposure duration. Perturbations were
observed in both sexes, sometimes with opposite effects between the sexes (in the case of TSH).
Reductions in free and total T4 as well as total T3 were noted in both rodents and chronically
exposed non-human primates that in some cases (female rats, male non-human primates)
coincided with compensatory increases in TSH, indicative of classical hypothyroidism.
Reductions in free and total T4, as well as declines in TSH in male rats may suggest
hypothyroxinemia. Elevations in thyroid gland weight were also noted {Butenhoff, 2012,
2919192} in males, as well as increases in thyroid gland follicular cell hypertrophy in male and
female rats {NTP, 2019, 5400977; NTP, 2020, 7330145}, however, the hormones released from
the respective organs (i.e., T4 and FT4) may be more sensitive and direct indicators of toxicity.
Thyroid hormones influence numerous other body systems, notably the nervous system via the
hypothalamic-pituitary-thyroid (HPT) axis, thus effects on other systems may stem from thyroid-
specific targets and vice versa. The available animal evidence supports evidence from human
epidemiological studies indicating that PFOA exposure may affect T4 in children.
Elevations in corticosterone were noted across two animal studies {Sun, 2018, 5079802;
Loveless, 2008, 988599} using male rodents, which coincided with a reduction in ACTH in one
study {Sun, 2018, 5079802}. Such effects may indicate adrenocortical toxicity, which can
involve increased secretion of endogenous glucocorticoids and long-loop feedback on the
hypothalamic-pituitary-adrenal (HPA) axis to reduce ACTH levels {Harvey, 2016, 1201708}.
However, 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
PFOA on adrenocortical hormone levels. Given the perturbations of adrenocortical hormones
and thyroid hormones, it is crucial to interrogate the interaction of multiple systems in order to
evaluate potential dysregulation of the HPA axis and/or HPT axis.
C.2.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOA 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, increased thyroid gland weight, and increased follicular cell hypertrophy in the
thyroid following exposure to doses as low as 0.625 mg/kg/day PFOA. Although a few
associations between PFOA exposure and T4 in children were observed in high and medium
confidence epidemiological studies, there is considerable uncertainty in the results due to
inconsistencies across sexes, age groups, and limited number of studies.
C-37
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-5. Evidence Profile Table for PFOA 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 thyroid-
related hormones and
thyroid disease
4 High confidence
studies
17 Medium confidence
studies
8 Low confidence studies
Studies in adults reported •
positive associations for
the thyroid-related •
hormone TSH (3/8). Sex
differences were observed
in two studies, indicating
increased TSH among
males and decreased TSH
among females. Results
for thyroid hormones (i.e.,
T3 and T4) were generally
mixed among adults;
however, significant
increases in total T3 were
observed (3/5). One study
(1/1) reported increased
risk of thyroid disease in
adult males, but there was
minor concern for
temporality due to the
cross-sectional study
design. Studies in children
observed significant
positive associations
(4/19) and inverse
associations (1/19) for T4,
and one study observed
significant positive
associations for TSH.
Other studies reported
inconsistent or imprecise
results. No clear effect for
hypothyroidism in a single
High and medium
confidence studies
Coherence of
findings across
multiple geographic
locations
• Low confidence studies
• Lnconsistency direction
of effect in adults
which may be
influenced by timing of
outcome sampling (i.e.,
diurnal variations)
• Imprecision of most
findings in children
©oo
Slight
Evidence for endocrine
effects is based on
increased TSH and T3 in
adults, and increased T4 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, increased thyroid
weight, and increased
follicular cell hypertrophy
in the thyroid. Although a
few associations between
PFOA exposure and T4 in
children were observed in
high and medium
confidence epidemiological
studies, there is
considerable uncertainty in
the results due to
inconsistencies across
sexes, age groups, and
limited number of studies.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
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
informative study in
children. In pregnant
women, positive
associations were
observed for TSH (4/8)
and T4 (5/8).
Thyroid hormone
antibodies
1 Medium confidence
study
1 Low confidence study
Studies in children
observed decreased TgAb
among boys born to TA-
negative mothers and
increased TgAb among
girls born to TA-positive
mothers. Among pregnant
women, TPOAb levels
were significantly
decreased.
• Medium confidence
study
• Low confidence study
• Limited number of
studies examining
outcome
Steroid and adrenal
hormones
1 High confidence study
One study in pregnant
women observed a
significant decrease in
serum Cortisol.
• High confidence
study
• Limited number of
studies examining
outcome
Evidence from In Vivo Animal Studies (Section C.2.2)
Evidence Integration
Summary Judgment
Thyroid and thyroid-
related hormones
1 High confidence study
1 Medium confidence
study
Decreased thyroid
hormones were observed
in male (total T4, free T4,
T3) and female (total T4,
free T4) rats following a
28-day exposure (1/1).
Sex-specific PFOA effects
on TSH were observed,
with increased levels in
females and decreased
levels in males. In a
developmental study in
mice (1/1), no significant
effects were observed on
High and medium
confidence studies
Limited number of
studies examining
outcome
(©©O)
Moderate
Evidence was based on
high and medium
confidence studies that
demonstrated decreased
thyroid hormone levels
(free T4, total T4, total
T3), especially in males.
Alterations in
adrenocortical hormone
levels, such as elevated
corticosterone and reduced
C-39
-------
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
the placental thyroid-
related hormones T3, total
T4, rT3, rT3:T4, or T3:T4.
Adrenocortical
hormones
3 Medium confidence
studies
Corticosterone levels were
increased in males (2/2)
and females (1/1)
following short-term
exposure in rodents. One
study observed a dose-
dependent decrease in
ACTH levels in male mice
(1/1).
Medium confidence
studies
Consistent direction
of effect for
corticosterone
levels
Limited number of
studies examining
outcome
ACTH, suggests
perturbation of the HPA
and/or HPT axis.
Increased incidence of
follicular cell hypertrophy
in the thyroid gland
correlated with increased
thyroid gland weight.
Organ weights
2 High confidence
studies
2 Medium confidence
studies
In a 28-day rat study,
increases in absolute and
relative thyroid gland
weights were reported in
males and no significant
effects were observed in
females (1/1). No
significant changes or
transient effects were
observed were observed in
absolute and/or relative
adrenal gland weights
(4/4). Decreased absolute
pituitary gland weights
were observed in only
female rats (1/2).
High and medium
confidence studies
Limited number of
studies examining
outcome
Histopathology
3 High confidence
studies
1 Medium study
Increased follicular cell
hypertrophy was observed
in the thyroid following
short-term and chronic
exposure in rats (2/3). No
changes in pituitary
histopathology were
High and medium
confidence studies
Limited number of
studies examining
outcome
C-40
-------
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
reported in male or female
rats (2/2). No changes in
adrenal histopathology
were reported in female
rats (2/2) but increased
incidence of non-
neoplastic lesions (1/1)
along with a non-
significant increase of
benign
pheochromocytoma and
hyperplasia (1/1) was
observed in male rats.
Notes: TSH = thyroid stimulating hormone; T3 = triiodothyronine; T4 = thyroxine; TgAb = thyroglobulin antibody; TA = thyroid antibodies; TPOAb = thyroid peroxidase antibody;
rT3 = reverse T3; ACTH = adrenocorticotropic hormone; HPA = hypothalamus-pituitary-adrenal; HPT = hypothalamus-pituitary-thyroid.
C-41
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.3 Metabolic/Systemic
EPA identified 71 epidemiological and 24 animal studies that investigated the association
between PFOA and systemic and metabolic effects. Of the epidemiological studies, 9 were
classified as high confidence, 39 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, 5 were classified as high confidence, 17 as medium
confidence, 1 as low confidence, and 1 was considered uninformative (Section C.3.2). 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 PFOA Document).
C.3.1 Human Evidence Study Quality Evaluation and
Synthesis
C.3.1.1 Introduction
Diabetes is a category of diseases caused by either insulin resistance or beta-cell disfunction, or
both. Type 1 diabetes is characterized by insulin deficiency and beta-cell destruction, while type
2 diabetes is characterized by beta-cell disfunction and insulin resistance. Type 2 diabetes is
more common than type 1 diabetes. Gestational diabetes commonly occurs during pregnancy and
is a risk factor for developing diabetes later in life. Diabetes can lead to long-term complications
in several organ systems, including micro- and macro-vascular complications.
Diagnostic criteria for diabetes include hemoglobin Ale (HbAlc) > 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 PFOA concluded that there is no evidence of an
association between PFOA and diabetes, metabolic syndrome, or related outcomes. No
associations were observed between mean serum PFOA up to 91.3-113.0 ng/mL and type 2
diabetes incidence in high-exposure (C8 Health Project) {MacNeil, 2009, 2919319} or
occupational populations {Steenland, 2015, 2851015}. Additionally, the C8 Science Panel
(2012), based on combined data from high-exposure and worker cohorts, concluded that there
was no probable link between PFOA and type II diabetes. One general population study
observed an increased risk of gestational diabetes in women with a mean pre pregnancy serum
PFOA level of 39.4 ng/mL {Zhang, 2015, 2857764}. Serum PFOA was significantly positively
associated with beta-cell function, but not associated with metabolic syndrome, metabolic
syndrome waist circumference, glucose concentration, homeostasis model of insulin resistance,
or insulin levels in adults or adolescents from NHANES {Lin, 2009, 1290820}. No association
C-42
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
was observed between serum PFOA concentrations {Nelson, 2010, 1291110} and insulin
resistance. Another study reported no association between PFOA and metabolic syndrome in
adolescents or adults {Lin, 2009, 1290820}. Overall, these studies show a lack of association of
PFOA with diabetes, metabolic syndrome, and related outcomes.
For this updated review, 71 new epidemiologic studies (72 publications)7 examined the
association between PFOA and metabolic outcomes. Of these, 35 were cohort studies, 6 were
case-control studies, 26 were cross-sectional studies, 2 were nested case-control studies, and 3
were controlled trials. Most studies measured exposure to PFOA using biomarkers in blood. One
study measured exposure to PFOA using biomarkers in blood and in semen {Di Nisio, 2019,
5080655}. Biomarkers in maternal blood were used in 16 studies and cord blood was used in two
studies. Shapiro et al. (2016, 3201206) measured exposure to PFOA in urine and Mancini et al.
(2018, 5079710) estimated dietary exposure to PFOA. 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-four studies examined diabetes (1 in children, 9 in pregnant women), and four studies
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, 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, Homeostatic Model Assessment of Beta-Cell Function (HOMA-B), or Homeostatic
Model Assessment for Insulin Resistance (HOMA-IR) without consideration of diabetes status,
since the treatment of diabetes, particularly in those being treated with hypoglycemic
medications, influences insulin production and secretion.
There are 71 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and metabolic effects. Study quality evaluations for these 71 studies
are shown in Figure C-17, Figure C-18, and Figure C-19.
Based on the considerations mentioned, nine studies were classified as high confidence for all
metabolic outcomes, 39 as medium confidence for all metabolic outcomes, two as medium
confidence for one outcome (anthropometric measurements or diabetes) and low confidence for
multiple other outcomes, two as medium confidence for one outcome (metabolic syndrome or
metabolic function) and low confidence for one other (adiposity or insulin resistance), one as
7 Fassler et al. (2019, 6315820) reports a cross-sectional analysis of participants from the same population as Pinney et al. (2019,
6315819).
C-43
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
medium confidence for multiple outcomes and uninformative for one other (insulin resistance),
14 as low confidence for all metabolic outcomes, and 4 were considered uninformative for all
outcomes. One study (Liu et al., 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 PFOA 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 provided 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 status {Lind, 2014, 2215376}. Low confidence studies presented
concerns with the outcome measures including potential for outcome misclassification
{Christensen, 2016, 3858533; He, 2018, 4238388; Steenland, 2015, 2851015; 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 PFOA exposure and diabetes {Lind, 2014, 2215376}. Other concerns included selection
bias {Fassler, 2017, 6315820}, which resulted from self-selection {Christensen, 2016, 3858533},
failure to provide information on control group selection {Heffernan, 2018, 5079713},
differential recruitment for cases and controls {Lin, 2013, 2850967}, or survival bias
{Steenland,2015, 2851015}. Small sample size was also a concern in some studies {Christensen,
2016, 3858533; Heffernan, 2018, 5079713; Khalil, 2018, 4238547}. 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-44
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Alderete et al.
Ashley-Martin et al.
Ashley-Martin et al.
Blake et al.
Braun et al.
Buck et al.
Cardenas et al.
Cardenas et al.
Chen et al.
Chen et al.
Christensen et al.
Christensen et al.
Christensen et al.
Convertino et al.
Conway et al.
Di Nisio et al.
2019, 5080614-
2016, 3859831 -
2017, 3981371 -|
2018, 5080657
2016, 3859836 -I
2018, 5080288
2017, 4167229-
2019, 5381549
2019, 5080578-
2019, 5387400
2016, 3350721 -
2016, 3858533-
2019, 5080398-
2018, 5080342-
2016, 3859824
2019, 5080655
Domazet et al., 2016, 3981435
Domazet et al., 2020, 6833700 -
Donat-Vargas et al., 2019, 5083542 -
Duan et al., 2020, 5918597-
Fassler et al., 2019, 6315820 -
Fleisch etal.,2017, 3858513
Gyllenhammar et al., 2018, 4238300
Hartman et al., 2017, 3859812
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-17. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Metabolic Effects
Interactive figure and additional study details available on HAWC.
C-45
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
He et al., 2018, 4238388
Heffernan et al., 2018, 5079713-
Huang et al., 2018, 5024212-
Jaacks etal., 2016, 3981711
Jainet al., 2019, 5080621
Jair, 2020, 6833623
Jeddyetal., 2018, 5079850-
Jensen et al., 2018, 4354143
Jensen et al., 2020, 6833719-
Jiang et al., 2014, 2850910
Kang et al., 2018, 4937567-
Karlsen et al., 2017, 3858520 -
Khalil et al., 2018, 4238547 -
Kobayashi et al., 2017, 3981430-
Koshy et al., 2017, 4238478
Lauritzen et al., 2018, 4217244-
Linet al., 2013, 2850967
Lind et al., 2014, 2215376
Liu etal., 2018, 4238396
Liuet al., 2018, 4238514-
Liu etal., 2019, 5881135-
Lopez-Espinosa et al., 2016, 3859832 -
Mancini et al., 2018, 5079710
Manzano-Salgado et al., 2017, 4238509-
Legend
| Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
01 Critically deficient (metric) or Uninformative (overall)
* Multiple judgments exist
Figure C-18. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Metabolic Effects (Continued)
Interactive figure and additional study details available on HAWC.
C-46
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Marks etal., 2019, 5381534
Martinsson et al.,
Matilla-Santander et al.,
Minatoya et al.,
Mitro et al.,
Mora et al.,
Pinney et al.,
Predieri et al., 2015, 3889874 •
Preston et al., 2020, 6833657
Rahman et al., 2019, 5024206
Ren et al.
Shapiro et al., 2016, 3201206-
Starling et al., 2017, 3858473-
Steenland et al., 2013, 1937218
Su etal., 2016, 3860116
Sun etal., 2018, 4241053
Tian et al.,2019, 5080586
Valvi et al., 2017, 3983872
Wang et al., 2018, 5079666
Wang et al., 2018, 5080352
Xu et al.
Yang et al.
Zong et al.
2020, 6833677-
2018,4238462-
2016, 3350666-
1
+
i
+
i
+
i .
+
i
+
i
+
i
+
+
+
+
+
-
+
+
+
- +
+
-
+
+
"
++
++
++
+
+
+
++
•
++
++
+
+
+
++
+
-
++
-
++
+
+
~
B
+
+
+
-
+
~
-
+
-
%
+
+
-
++
+
+
++
+
+
++
+
+
+
+
+
+
+
+
+
p~
+
+
++
+
+
+
+
+
+
+
+
+
nzi
+
++
+
+
++
+
+
+
-
+
+
+
+
+
-
+
+
+
+
~
++
+
++
+
+
+*
+
+
+
-
+
+
-
++
+
+
+
+
+
+
+
+
+
-
+
+
+
+
++
-
+
+
+
+
+
-
+
+
H
+*
+
+
+
-
¦
+
+
+
+
-
Legend
I Good (metric) or High confidence (overall)
+ I 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 PFOA and
Metabolic Effects (Continued)
Interactive figure and additional study details available on HAWC.
C-47
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.3.1.3 Findings from Children and Adolescents
Two medium and two low confidence studies examined blood glucose in children, and only one
reported a positive association with 2-hour glucose. No associations were observed for fasting
glucose. Alderete et al. (2019, 5080614) examined a cohort of obese Hispanic children aged
8-14, from the SOLAR Project and observed a significant association with 2-hour glucose, but
no association with fasting glucose. Two cross-sectional studies reported positive non-significant
associations with fasting glucose, one medium confidence study in 3-18-year old Koreans
{Kang, 2018, 4937567}, and one low confidence study in American obese 8-12 years {Khalil,
2018, 4238547}. Another cross-sectional study in girls ages 6-8 years from the Breast Cancer
and Environment Research Program reported a negative, non-significant association with
glucose levels {Fassler, 2019, 6315820}.
One medium confidence study observed positive, non-significant associations with blood glucose
levels at age 15, using PFOA measured at ages 9 and at age 15 {Domazet, 2016, 3983465}. A
non-significant negative association was observed between PFOA measured at age 15 and blood
glucose measured at age 21 {Domazet, 2016, 3983465}.
Three studies examined the association between PFOA and insulin levels and reported no
associations. One medium confidence study reported a positive, non-significant association with
fasting insulin in obese Hispanic children aged 8-14 {Alderete, 2019, 5080614}. In contrast, two
low confidence studies reported negative non-significant associations between PFOA and fasting
insulin {Fassler, 2019, 6315820; Khalil, 2018, 4238547}
Insulin resistance, as described by the HOMA-IR, was examined in five studies with mixed
results. Alderete et al. (2019, 5080614) observed a positive, non-significant association, while
four low confidence studies reported non-significant negative associations (i.e., decreasing
insulin resistance with increasing serum PFOA) {Khalil, 2018, 4238547; Fassler, 2019,
6315820; Koshy, 2017, 4238478; Fleisch, 2017, 3858513}.
A positive, but non-significant association was observed between PFOA and insulin sensitivity,
measured through both the insulin sensitivity index and the CHECK
Index/Quantitative Insulin Sensitivity Check Index {Fassler, 2019, 6315820}.
One medium confidence study reported 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}.
There was a significant negative association with IGF-1 in girls, and a significant negative
association with IGF-1 in the second quartile of PFOA exposure among boys {Lopez-Espinosa,
2016, 3859832}.
Adiponectin and leptin were both examined in a medium confidence study from the European
Youth Study, and non-significant associations were observed with adiponectin (positive), and
leptin (negative) {Domazet, 2020, 6833700}. Similarly, Fleisch et al. (2017, 3858513) reported a
non-significant negative association with leptin in both early- and mid-childhood. Positive, non-
significant association was observed between maternal blood PFOA and cord blood adiponectin
{Ashley-Martin, 2017, 3981371; Minatoya, 2017, 3981691}.
Three studies examined adiposity, and one reported a significant negative association with fat
mass. One low confidence study observed a significant negative association with log fat mass
C-48
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
and fat mass percentage in girls ages 6-8 years {Fassler, 2019, 6315820}. However, concerns
about selection bias and residual confounding by SES limit confidence in these results.
Chen et al. (2019, 5080578} observed a positive, non-significant association with children's
body fat percentage; non-significance persisted after stratification by child sex. Non-significant
negative associations were observed in the third tertile of PFOA exposure for girls and in the
second tertile of PFOA exposure for boys {Chen, 2019, 5080578}. Similarly, a positive, non-
significant association was observed with children's body fat mass, and non-significance
persisted after stratifying by child sex {Chen, 2019, 5080578}. In a tertile analysis, positive, non-
significant associations were observed in the third tertile of PFOA exposure for all children and
in the third tertile of PFOA exposure for boys; negative, non-significant associations between
PFOA and body fat mass were observed in the second and third tertiles of PFOA exposure
among girls {Chen, 2019, 5080578}. Kmedium confidence cross-sectional study of 9-year old
children in the European Youth Heart Study reported a negative non-significant association with
fat mass {Domazet, 2020, 6833700}.
Seven studies examined BMI measures, with mixed results. Four studies observed no
associations with BMI, and two observed associations with BMI-z-score.
One high confidence study examined the association between cord blood PFOA and age 5 BMI
in the Shanghai Prenatal Cohort {Chen, 2019, 5080578}. There was a negative but non-
significant association between PFOA and BMI (i.e., decreased BMI with higher PFOA
exposure levels). The effect was larger in females (beta = 0.07, 95% CI: -0.4, 0.53) than for
males (beta = 0.2, 95% CI: -0.3, 0.69). Results from a tertile analysis were also non-significant,
even after stratification by sex. For females, BMI increased with increasing tertiles of PFOA,
while BMI decreased with increasing tertiles of PFOA in males. {Chen, 2019, 5080578}. Two
medium confidence studies observed positive, non-significant associations with BMI {Manzano-
Salgado, 2017, 4238509; Braun, 2016, 3859836}. In a sex-stratified analysis, the association
between maternal blood PFOA and BMI at age 7 remained positive among boys but became
negative among girls {Manzano-Salgado, 2017, 42385509}.
Of the three low confidence studies examining BMI, two reported positive, non-significant
associations {Di Nisio, 2019, 5080655; Koshy, 2017, 4238478}, and one reported a negative
non-significant association with BMI {Khalil, 2018, 4238547}.
Six studies examined BMI z-score, two of which reported significant negative associations. Two
studies from the Breast Cancer and the Environment Research Program (one medium, one low
confidence) observed significant negative associations with BMI z-score in girls ages 6-8
{Pinney, 2019, 6315819; Fassler, 2019, 6315820}. Pinney et al. (2019, 6315819) observed a
significant negative association with BMI z-score in girls living in the Greater Cincinnati and the
San Francisco Areas. Karlsen et al. (2017, 3858520) observed a non-significant negative
association with BMI z-score at 18 months and age 5. In children from the POPUP study,
Gyllenhammar et al. (2018, 4238300) observed a positive, significant association with BMI z-
score and 3 and 4-years old children; the association with BMI z-score among 5-year old
children was positive, but not significant.
C-49
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Additionally, a non-significant association was observed with BMI z-score in early- and mid-
childhood {Mora, 2017, 3859823}. Another low confidence study reported a negative, non-
significant association with BMI z-score {Koshy, 2017, 4238478}.
A medium confidence study reported a weak non-significant negative association between serum
PFOA levels and ponderal index at birth in infants from the Hokkaido Study on Environment
and Children's Health {Kobayashi, 2017, 3981430}.
No associations were observed in two low confidence studies examining body weight {Fassler,
2019, 6315820} or being overweight {Koshy, 2017, 4238478}.
Four studies examined waist measurements, and two reported associations. Two studies (one
medium, one low confidence) observed ay significant negative associations with waist-to-height
ratio (i.e., increased wait-to-height ratio as a continuous measure with higher serum PFOA
exposure levels) in girls ages 6-8 {Pinney, 2019, 6315819; Fassler, 2019, 6315820}. However,
one high confidence study observed a positive non-significant association between cord blood
PFOA and waist-to-height ratio in girls ages 5 years {Chen, 2019, 5080578}. Negative non-
significant associations were observed for all children combined, and in boys, and a non-
significant decreasing trend was observed {Chen, 2019, 5080578}.
Two studies (one medium and one low confidence) examined waist-to-hip ratio. The medium
confidence study observed a non-significant negative association {Pinney, 2019, 6315819}
between PFOA and waist-to-hip ratio, while the low confidence study reported a non-significant
positive association between PFOA and waist-to-hip ratio {Fassler, 2019, 6315820}. One
medium confidence study reported a positive, non-significant association with waist-to-hip
circumference. After stratification by sex in the early childhood analysis, a non-significant
negative association was observed among girls. In the mid-childhood analysis, the increase in
waist-to-hip circumference ratio was greater for girls than for boys {Mora, 2017, 3859823}
One high, two medium, and one low confidence study examined waist circumference and
reported one association {Hartman, 2017, 3859812}. The medium confidence study, from the
ALSPAC, assessed data from mother-daughter pairs and observed a significant decrease in
female children's waist circumference {Hartman, 2017, 3859812}. Two medium confidence
studies {Chen et al. 2019 5080578; Mora et al., 2017 3859823} reported a positive, non-
significant association between PFOA and waist circumference. After stratification by sex, non-
statistical significance persisted; associations remained negative for males but were positive for
females {Chen, 2019, 5080578}. A cohort study of maternal-child pairs from the European
Youth Heart Study reported a non-significant percent decrease in waist circumference at 21 years
old with PFOA exposure at age 9 and age 15, and a significant percent decrease in waist
circumference at 21 years old with concurrent PFOA exposure, and a non-significant percent
increase in waist circumference at age 15 with age 9 PFOA exposure {Domazet, 2016,
3981435}.
In the low confidence study, Di Nisio et al. (2019, 5080655) reported a significant difference
between mean waist circumference of Italian male high school students exposed to PFOA
pollution compared to those who were not exposed {Di Nisio, 2019, 5080655}.
C-50
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
There were three studies, each of medium confidence, measuring the association between PFOA
and skinfold thickness. A medium confidence study from the SGA Study reported a non-
significant positive association with tricep skinfold z-score, and a non-significant negative
association with subscapular skinfold thickness z-score among 412 children {Lauritzen, 2018,
4217244}.
Another cohort study, which used a subset of data on children from the European Youth Heart
Study, observed a non-significant percent increase in skinfold thickness at age 15 for increases in
PFOA exposure at 9 years old, as well as a non-significant percent increase in skinfold thickness
at age 21 for increases in PFOA exposure at 9 years old. However, there was a non-significant
percent decrease in skinfold thickness at 21 years old with increase in PFOA exposure from
15 years old {Domazet, 2016, 3981435}.
A cohort study of mother-child pairs was used to assess the association between maternal PFOA
and skinfold thickness {Mora, 2017, 3859823} There was a positive, non-significant association
between PFOA and subscapular-to-triceps skinfold thickness ratio measured in both early
childhood and mid-childhood. After stratification by sex, the effect increased for females, but
decreased non-significantly for males during both early- and mid-childhood. Similarly, the
association between PFOA and the sum of subscapular and tricep skinfold thickness during mid-
childhood decreased for males but increased for females when stratified by sex, but the sum of
subscapular and tricep skinfold thickness during early childhood decreased for females and
increased for males when stratified by sex {Mora, 2017, 3859823}.
C.3.1.4 Findings from Pregnant Women
Eleven studies examined gestational diabetes, and one reported a negative association between
PFOA and gestational diabetes.
A medium confidence study of adults aged 20-60 living in Taiwan reported a significant
negative association with gestational diabetes {Su, 2016, 3860116}.
In a high confidence cohort study from Project Viva of pregnant women, Preston et al. (2020,
6833657) reported a non-significant, null association with gestational diabetes (OR = 1.0; 95%
CI: 0.6, 1.6), but non-significant increased odds of gestational diabetes with increasing quartiles
of PFOA {Preston, 2020, 6833657}.
Two medium confidence case-control studies reported increased, non-significant odds of
gestational diabetes {Wang, 2018, 5079666; Xu, 2020, 6833677}. In pregnant women with no
family history of diabetes, Liu et al. (2019, 588135) reported a non-significant, positive
association between m-PFOA or L-PFOA and odds of gestational diabetes {Liu, 2019,
5881135}. Increased, non-significant odds of gestational diabetes were observed in the second
and third tertiles of L-PFOA exposure, and in the third tertile of m-PFOA exposure; decreased,
non-significant odds of gestational diabetes were observed in the second tertile of m-PFOA
exposure {Liu, 2019, 5881135}. Similarly, nested case-control study conducted by Xu et al.
(2020, 6833677) recruited pregnant women with no history of diabetes and reported increased,
non-significant odds of gestational diabetes across quartiles of PFOA exposure and log-
transformed PFOA exposure.
C-51
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A study from the U.S. National Institute of Child Health and Human Development (NICHD)
Fetal Growth Study reported a non-significant increased risk of gestational diabetes among all
women, women with a family history of type 2 diabetes, and women with an overweight pre-
pregnancy BMI {Rahman, 2019, 5024206}. A non-significant decreased risk of gestational
diabetes was observed among pregnant women without a family history of type 2 diabetes and
among women who did not have an overweight pre-pregnancy BMI {Rahman, 2019, 5024206}.
Three medium and one low confidence studies reported negative, non-significant associations
with gestational diabetes {Shapiro, 2016, 3201206; Wang, 2018, 5080352; Valvi, 2017,
3983872; Zong, 2016, 3350666}.
Seven studies evaluated blood glucose and related measures, with mixed results. Two studies
reported an association with oral glucose tolerance test results; no associations were reported for
fasting glucose, impaired glucose tolerance, or hyperglycemia.
A medium confidence study of pregnant women with and without gestational diabetes reported
increased, but non-significant odds of increased fasting blood glucose with increasing tertiles of
n-PFOA {Wang et al., 2018 5079666}. Liu et al. (2019, 5881135) observed a positive, non-
significant associations between both sum m-PFOA and L-PFOA and fasting glucose. Three
medium confidence cohort studies observed negative, non-significant associations with fasting
blood glucose {Wang, 2018, 5080352; Jensen, 2018, 4354143; Starling, 2017, 3858473}.
Overall oral glucose tolerance test results were evaluated in one study {Wang, 2018, 5080352}.
When modeled continuously, there was a positive, non-significant association between PFOA
and OGTT glucose. No significant difference was observed in mean oral glucose tolerance test
results between tertiles of PFOA {Wang, 2018, 5080352}.
Two medium confidence studies examined 1-hour blood glucose, and both reported positive
significant associations. Ren et al. (2020, 6833646) observed a significant increase in 1-hour
plasma glucose levels and Liu et al. (2019, 5881135) reported a significant positive association
between serum L-PFOA and glucose homeostasis at 1-hour, and a negative, non-significant
association between sum m-PFOA and 1-hour glucose.
Two medium confidence studies examined 2-hour blood glucose. A significant positive
association was observed between L-PFOA and 2-hour glucose, but the positive association
between sum m-PFOA and 2-hour glucose was not significant {Liu, 2019, 5881135}. A medium
confidence study from the Odense Child Cohort reported a negative non-significant association
between serum PFOA and 2-hour glucose among 158 women at high risk for gestational diabetes
{Jensen, 2018, 4354143}.
Three studies examined impaired glucose tolerance. In a subset of women from Project Viva
Preston et al. (2020, 6833657) observed decreased, non-significant odds of impaired glucose
tolerance. This was also observed in a tertile analysis, but the odds of impaired glucose tolerance
were greater with increasing tertiles of PFOA {Preston, 2020, 6833657}. A medium confidence
study also reported decreased odds of impaired glucose tolerance {Shapiro, 2016, 3201206}.
The single low confidence study observed non-significant increased odds of impaired glucose
tolerance with PFOA increasing continuously, but non-significant decreased odds of impaired
glucose tolerance with increasing quartiles of PFOA {Matilla-Sandtander, 2017, 4238432}.
C-52
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
One high confidence study examined isolated hyperglycemia in pregnant women from the
Project Viva cohort {Preston, 2020, 6833657}. When analyzed continuously, increasing PFOA
did not affect the odds of hyperglycemia. A quartile analysis showed non-significant decreased
odds of hyperglycemia with increasing quartiles of PFOA {Preston, 2020, 6833657}.
Two studies (one high confidence and one medium confidence) evaluated blood glucose levels
{Preston, 2020, 6833657; Ren, 2020, 6833646}. Both studies reported a non-significant positive
association with blood glucose levels. After stratifying by age, Preston et al. (2020, 6833657)
reported a non-significant negative association with blood glucose among women aged 35 and
older. In the medium confidence study, results from an age-stratified analysis showed non-
significant decreased odds of high plasma glucose for women at 20-23 gestational weeks {Ren,
2020, 6833646}.
Two studies evaluated insulin resistance measures; neither reported any associations.
There were two studies of medium confidence evaluating insulin levels {Jensen, 2018, 4354143;
Wang, 2018, 5080352}. One of these studies reported a non-significant negative association with
fasting insulin levels {Jensen, 2018, 4354143}, while the other observed a non-significant
positive association with fasting insulin levels {Wang, 2018, 5080352}.
Two medium confidence studies assessed insulin resistance. One reported a non-significant
negative association {Jensen, 2018, 4354143}, while the other observed a non-significant
positive association {Wang, 2018, 5080352} with insulin resistance. Wang et al. (2018,
5080352) reported no significant difference in mean insulin resistance between tertiles of PFOA.
One medium confidence study evaluated insulin sensitivity (measured using the Matsuda index)
and observed a positive, non-significant association {Jensen, 2018, 4354143}.
A non-significant percent decrease in beta-cell function was observed {Jensen, 2018, 4354143}.
Adiponectin and leptin were both examined in a high confidence study from Project Viva, and no
significant associations were observed A non-significant negative association with adiponectin
and a non-significant positive association with leptin were reported {Mitro, 2020, 6833625}.
After stratification by age during pregnancy, non-significant positive associations with leptin
persisted; a positive, non-significant association with adiponectin was observed among women
under age 35 during pregnancy {Mitro, 2020, 6833625}.
Three medium confidence cohort studies examined gestational weight gain, with one reporting an
association. Ashley-Martin et al. (2016, 3859831) used data from mother-infant pairs from the
Maternal-Infant Research on Environmental Chemicals (MIREC) to estimate the odds of having
high cord blood PFOA ( > 0.39 ng/mL) per increase in gestational weight gain. ORs were
significant for both 1 kg increase in gestational weight gain and interquartile range (IQR)
increase in gestational weight gain {Ashley-Martin et al. (2016, 3859831).
Jaacks et al. (2016, 3981711) observed a positive, non-significant association with gestational
weight gain among 218 mothers, mothers with a BMI < 25; a negative association was reported
among mothers with a BMI > 25. Increased, non-significant odds of excessive gestational weight
gain were observed with increasing PFOA and decreased, non-significant odds of inadequate
weight gain were reported {Jaacks, 2016, 3981711}.
C-53
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Another study reported a positive, non-significant association with gestational weight gain
among all women who were underweight or of normal weight and among under- or normal-
weight mothers of daughters. Negative, non-significant associations with gestational weight gain
were observed among overweight or obese mothers of all children, of boys, and of girls, and
among normal or underweight mothers of sons {Marks, 2019, 5381534}.
One study evaluated anthropometric measurements and PFOA from the Project Viva cohort
study and followed 801 pregnant women to 3 years postpartum {Mitro, 2020, 6833625}.
Positive, non-significant associations were reported with 3-year postpartum arm circumference,
subscapular skinfold thickness, tricep skinfold thickness, and 3-year postpartum waist
circumference. After stratification by age during pregnancy, there was a significant increase in
waist circumference measured at 3 years postpartum among women who were 35 or older during
pregnancy {Mitro, 2020, 6833625}.
One high confidence cohort study evaluated BMI. A significant positive association with BMI
among 786 pregnant women was reported {Mitro, 2020, 6833625}. Statistical significance did
not persist after stratification by age (under 35/age 35 and older) {Mitro, 2020, 6833625}.
C.3.1.5 Findings from the General Adult Population
Eight studies investigated the relationship between PFOA and diabetes in the general population,
and three reported a positive association.
A medium confidence study from the E3N cohort reported a non-significant increased risk of
type 2 diabetes in the 7th and 8th deciles of PFOA exposure, and increased risk of type 2
diabetes was observed in the 4th-6th deciles of PFOA exposure. {Mancini, 2018, 5079710}.
Another medium confidence study, from the Nurses' Health Study II, reported a significant
association with type 2 diabetes among n female nurses {Sun, 2018, 4241053}.
One high confidence cohort study from the Diabetes Prevention Program followed adults at
increased risk of type 2 diabetes and observed an increased, but non-significant risk of diabetes
per doubling of PFOA {Cardenas, 2017, 4167229; Cardenas, 2019, 5381549}. After
stratification by sex, a non-significant negative association was observed among men {Cardenas,
2017, 4167229}. Non-significant negative associations were also observed in analyses by tertiles
{Cardenas, 2019, 5381549}.
Another medium confidence study reported non-significant increased odds of type 2 diabetes
were observed in the 2nd tertile of PFOA exposure, while non-significant decreased odds were
observed in the 3rd tertile of PFOA exposure {Donat-Vargas, 2019, 598342}.
Significant decreased odds of type 1, type 2, and uncategorized diabetes were observed in
participants in the C8 Health Project {Conway, 2016, 3859824}. After stratifying by age,
significant decreased odds of type 1, type 2, and uncategorized diabetes were observed among
adults, significant decreased odds of type 1 diabetes were observed for children with type 1
diabetes, but non-significant increased odds of type 2 and uncategorized diabetes were observed
among children {Conway, 2016, 3859824}.
Among the three low confidence studies, one reported a non-significant negative association with
diabetes {Lind, 2014, 2215376}, while two overlapping NHANES studies reported non-
C-54
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
significant positive associations with diabetes {He, 2018, 4238388} and prediabetes
{Christensen, 2016, 3858533}. Significantly increased odds of diabetes were observed for males,
non-significant increased odds were observed for females {Christensen, 2016, 3858533}. Low
confidence ratings resulted from concerns with potential for outcome misclassification
{Christensen, 2016, 3858533; He, 2018, 4238388}, self-selection into the study, residual
confounding by SES {Christensen, 2016, 3858533}, and failure to establish temporality between
exposure and outcome {He, 2018, 4238388}.
Four studies (three medium confidence and one low confidence) evaluated metabolic syndrome;
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 MetS as defined
by the Adult Treatment Panel III criteria (OR = 1.89, 95% CI: 0.93, 3.86). Two medium
confidence studies used overlapping data from NHANES and reported non-significant negative
associations with metabolic syndrome. Liu et al. (2018, 4238514) 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.
A low confidence study observed significant increased odds of metabolic syndrome for
participants with serum n-PFOA > 1.90 ng/mL compared to those with serum PFOA
<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 five studies examining the association between PFOA and glucose, and three
reported associations with fasting blood glucose, and one reported an association with 2-hour
glucose.
A medium confidence study of adults aged 19-87 years from China reported a significant
positive association with fasting blood glucose {Duan, 2020, 5918597}. Similarly, a study using
NHANES data on adults from 1999-2014 observed a significant positive correlation between
fasting glucose and serum PFOA {Huang, 2018, 5024212}. Su et al. (2016, 3860116) reported a
statistically significant decrease in fasting blood glucose for both increasing quartiles of PFOA
and per doubling of PFOA among Taiwanese adults aged 20-60.
Another cohort study, which followed adults at high risk of type 2 diabetes, observed a positive,
non-significant increase in 30-minute glucose per doubling of PFOA, while a negative, non-
significant association was observed between with 2-hour glucose {Cardenas, 2017, 4167229}.
A non-significant negative association with 2-hour glucose was reported per doubling in PFOA
among Taiwanese adults aged 20-60, but a significant decrease in 2-hour glucose was observed
for increasing quartiles of PFOA {Su, 2016, 3860116}.
One study reported non-significant decreased odds of elevated glucose with increasing tertiles of
PFOA {Christensen, 2019, 5080398}. Odds were adjusted for PFDA, PFOS, PFHxS, 2-(N-
methyl-PFOSA) acetate (MPAH), PFNA, perfluoroundecanoic acid (PFUnDA) simultaneously.
The association between PFOA and resting metabolic rate was assessed in the POUNDS LOST
trial, a clinical trial of overweight and obese adults aged 30-70. A non-significant positive
correlation between PFOA and resting metabolic rate was observed {Liu, 2018, 4238396}. In the
first 6 months of the trial, resting metabolic rate decreased non-significantly across all tertiles of
C-55
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOA exposure for both men and women. Neither the trend across tertiles nor the interaction
between PFOA and sex were significant {Liu, 2018, 4238396}. In months 6-24 of the trial,
resting metabolic rate decreased significantly for males, and non-significantly for females. No
statistical significance was observed for the interaction between PFOA and sex {Liu , 2018,
4238396}.
Twelve studies examined insulin resistance measures; of these studies, one found reported
significant associations with fasting insulin, insulin resistance, insulinogenic index 1, fasting
plasma insulin, 30-minute insulin, fasting proinsulin, and insulin (corrected response), and one
reporting associations with the ratio of proinsulin to insulin.
The single high confidence study used a subset of data on adults at high risk of type 2 diabetes
from the Diabetes Prevention Program {Cardenas, 2017, 4167229}. A positive, significant
association was observed between PFOA and fasting insulin {Cardenas, 2017, 4167229}. Two
low confidence studies examined fasting insulin, and both reported non-significant negative
associations with fasting insulin {Chen, 2019, 5387400; He, 2018, 4238388}.
Two medium confidence studies reported negative, non-significant associations with insulin
levels {Sun, 2018, 4241053; Domazet, 2016, 3981435}. In contrast, another medium confidence
observed a positive, non-significant association with insulin levels {Liu, 2018, 4238514}.
Nine studies examined insulin resistance, 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 statistically significant, positive association with insulin resistance
{Cardenas, 2017, 4167229}. A medium confidence study of adults in NHANES observed a non-
significant increase in insulin resistance with increase in PFOA {Liu, 2018, 4238514}. However,
Donat-Vargas et al. (2019, 5083542) reported a non-significant negative association with insulin
resistance in both continuous and tertile analyses. In a sensitivity analysis, a non-significant
negative association was observed between insulin resistance and baseline PFOA second tertile,
and between insulin resistance and PFOA measured at the end of follow-up for both the second
and third tertile of PFOA exposure. A non-significant positive association with insulin resistance
was reported in the third tertile of baseline PFOA exposure {Donat-Vargas, 2019, 5083542}.
In a medium confidence study, a non-significant decrease in insulin resistance (measured as
HOMA-IR) was observed at age 15 and 21 years old per increase in PFOA exposure from
9 years old {Domazet, 2016, 3981435}. At age 21, there was a non-significant increase in
HOMA-IR per increase in PFOA measured at age 15 {Domazet, 2016, 3981435}.
Three low confidence studies that examined the association between PFOA and insulin
resistance. Non-significant negative associations between PFOA and insulin resistance were
observed in continuous analyses {Lind, 2014, 2215376; Chen, 2019, 5387400}. In a sex-
stratified tertile analysis, a non-significant negative association was observed with log-HOMA-
IR among males, with non-significant increasing HOMA-IR observed with increasing quartiles
of PFOA {He, 2018, 4238388}. HOMA-IR decreased non-significantly with increasing quartiles
of PFOA among females {He, 2018, 4238388}. These studies were given low confidence ratings
due to failure to account for diabetes status {Lind, 2014, 2215376}, or use of medications that
impact insulin levels in HOMA-IR analyses {Chen, 2019, 5387400}, and failure to account for
the complex sampling design of NHANES in statistical analyses {He, 2018, 4238388}.
C-56
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The association between plasma PFOA and insulinogenic index 1 was investigated in a high
confidence study from the Diabetes Prevention Program. A significant positive association was
observed with insulinogenic index among adults at high risk for type 2 diabetes {Cardenas, 2017,
4167229}.
In a high confidence study, Cardenas et al. (2017, 4167229) reported significant associations
were observed between PFOA and fasting plasma insulin, 30-minute insulin, fasting proinsulin,
and insulin (corrected response).
In a low confidence study, a significant positive association was reported for the ratio of
proinsulin to insulin and PFOA {Lind, 2014, 2215376}.
Five studies examined beta cell function and two reported a significant association. A high
confidence study from the Diabetes Prevention Program reported a significant positive
association with beta cell function (measured as HOMA-B) among adults at high risk for type 2
diabetes {Cardenas, 2017, 4167229}. A significant positive association with beta-cell function
was reported in a medium confidence study of adults from NHANES {Liu, 2018, 4238514}. Two
medium confidence studies reported negative, non-significant associations with HOMA-B
{Donat-Vargas, 2019, 5083542; Domazet, 2016, 3981435}.
One low confidence study reported a positive, non-significant association with HOMA-B {Chen,
2019, 5387400}. This study was given a low confidence rating due to failure to exclude
participants using medications that could impact beta-cell function.
Five studies examined adiponectin, and one observed an association. A high confidence study
from the Health Outcome Measures of the Environment (HOME) study reported non-significant
positive association between maternal blood PFOA and adiponectin in children {Ashley-Martin,
2017, 3981371}. In contrast, a significant negative association with adiponectin was observed
among adults in the Diabetes Prevention Program {Cardenas, 2017, 4167229}. A medium
confidence study reported a negative non-significant correlation between PFOA and plasma
adiponectin {Sun, 2018, 4241053}.
Two high confidence studies reported non-significant positive associations with adiponectin; no
statistically significant effects were observed after stratifying by infant sex in either study {Buck,
2018, 5080288; Minatoya, 2017, 3981691}.
Five studies examined associations with leptin. One study reported a significant association.
Three high quality studies examined leptin {Buck, 2018, 5080288; Minatoya, 2017, 3981691;
Ashley-Martin, 2017, 3981371}, all of which sampled mother-child pairs and observed positive,
non-significant associations with children's leptin concentrations {Buck, 2018, 5080288;
Minatoya, 2017, 3981691; Ashley-Martin, 2017, 3981371}.
Two medium confidence studies examined leptin. One study, from the POUNDS LOST clinical
trial, followed overweight and obese adults. A positive, significant correlation was observed
between plasma PFOA and leptin concentrations {Liu, 2018, 4238396}.
A non-significant, slightly positive association was observed between PFOA and soluble leptin
receptors {Liu, 2018, 4238396}
C-57
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Eight studies examined hemoglobin and five reported an association. A high confidence study on
participants in the Diabetes Prevention Program reported a significant positive association with
HbAlc {Cardenas, 2017, 4167229}. Two medium confidence studies reported positive, non-
significant associations with HbAlc {Duan, 2020, 5918597; Sun, 2018, 4241053}. One medium
confidence study of PFOA and HbAlc among 10,859 NHANES participants reported a negative,
significant spearman correlation between serum PFOA and plasma hemoglobin {Huang, 2018,
5024212}.
Another medium confidence cross-sectional study assessed the association between plasma
PFOA and HbAlc among adults aged 20-60 {Su, 2016, 3860116}. A negative, non-significant
association between HbAlc and continuous PFOA was reported, but a significant decrease in
average HbAlc was observed with increasing quartiles of PFOA {Su, 2016, 3860116}. In the
POUNDS LOST trial, a clinical trial of overweight and obese adults, negative, significant
correlation was observed between PFOA and HbAlc {Liu, 2018, 4238396}. Additionally, a
medium confidence cross-sectional analysis of adults from NHANES reported a significant
negative association with HbAlc {Liu, 2018, 4238514}.
One low confidence study reported a statistically significant negative association with HbAlc
among women with PCOS, and a non-significant positive association with HbAlc among
women without PCOS {Heffernan, 2018, 5079713}. Another low confidence study reported no
significant association between PFOA and glycated hemoglobin {Chen, 2019, 5387400}. Low
confidence ratings were given to these studies due to failure to exclude participants using
medications that could impact HbAlc {Chen, 2019, 5387400} and concerns with participant
selection and residual confounding {Heffernan, 2018, 5079713}.
Eight studies evaluated body weight measures, and six reported an association.
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, non-significant association with weight loss in months 6-24 of the trial {Liu et al., 2018
4238396}. A significant increase in average weight gain during months 6-24 of the trial was
observed with increasing tertiles of PFOA {Liu, 2018, 4238396}.
Seven studies evaluated being overweight and one reported a significant association. A cohort
study of mothers and children from the Faroe Islands followed mother-child pairs reported an
increased, significant risk of being overweight at age 5 with increase in maternal PFOA and a
non-significant increased risk of being overweight at 18 {Karlsen, 2017, 3858520}. In a tertiles
analysis, a non-significant negative association was observed with being overweight at 18
months, and a non-significant positive association was observed with being overweight at age 5
{Karlsen, 2017, 3858520}. A significant increased risk of being obese at age 5 was observed in
the highest tertile of maternal PFOA exposure {Karlsen, 2017, 3858520}.
A medium confidence study reported significantly greater serum PFOA among obese adults
compared to non-obese adults {Jain, 2019, 5080621}. Five medium confidence studies evaluated
maternal PFOA and risk of being overweight or obese in their children; these studies reported
increased, non-significant risk or odds of being overweight {Braun, 2016, 3859836; Lauritzen,
2018, 4217244; Martinsson, 2020, 6311645; Manzano-Salgado, 2017, 4238509; Mora, 2017,
3859823}. In a sex-stratified analysis, Mora et al. (2017, 3859823) observed an increased, non-
C-58
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
significant relative risk of being overweight or obese among boys, but a decreased, non-
significant risk among girls.
In the low confidence studies, significant associations were seen between PFOA and being
overweight {Tian, 2019, 5080586} and being obese {Yang, 2018, 4238462}. One study was
given a low confidence rating due to concerns with BMI being related to PFOA; although this
was acknowledged by the authors, this was not accounted for in the analysis {Tian, 2019,
5080586}. Low confidence ratings were also given due to concerns with outcome
misclassification and residual confounding by SES {Yang, 2018, 4238462}.
One study observed a significant negative association with weight-for-age z-score among
children {Braun, 2016, 3859836}. A significant interaction between maternal PFOA and age was
observed in the second tertile of maternal PFOA exposure, but not in the third tertile of maternal
PFOA exposure {Braun, 2016, 3859836}.
Five studies evaluated body fat measures, and one reported an association. Four studies of
medium confidence evaluated body fat {Hartman et al., 2017, 3859812; Mora, 2017, 3859823;
Braun, 2016, 3859836; Liu, 2018, 5881135}. A negative, non-significant association was
observed between maternal plasma PFOA and body fat percentage in young girls in the
ALSPAC, and this association persisted after stratification by age at menarche {Hartman, 2017,
3859812}. However, the negative association between maternal plasma PFOA and trunk fat
percentage in young girls was significant {Hartman, 2017, 3859812}. Three medium confidence
studies reported positive, non-significant associations with body fat measures {Mora, 2017,
3859823; Braun, 2016, 3859836; Liu, 2018, 5881135}.
Two medium confidence studies evaluated fat mass, and no associations were reported. Non-
significant, positive associations with fat mass were reported among children {Jeddy, 2018,
5079850} and overweight and obese adults {Liu, 2018, 5881135}.
Fifteen studies assessed BMI, and one reported a significant association.
In the HOME study, a cohort study of mother-child pairs, PFOA exposure was measured during
pregnancy and BMI was recorded at age 8 {Braun, 2016, 3859836}. Significant positive
associations with BMI z-score were observed in the second tertile of maternal PFOA exposure,
and a negative, non-significant association was observed in the third tertile of maternal PFOA
exposure {Braun, 2016, 3859836}. Additionally, significant increases in BMI z-score between
ages 2 and 8 were observed in both the second and third tertile of maternal PFOA exposure
{Braun, 2016, 3859836}. Two medium confidence studies of mother-child pairs observed
positive, but non-significant association between maternal serum PFOA child's BMI z-score
{Lauritzen, 2018, 4217244; Jensen, 2020, 6833719}.
Two high confidence studies and three medium confidence studies observed positive, non-
significant associations with BMI {Cardenas, 2017, 4167229; Chen, 2019, 5387400; Mora,
2017, 3859823; Domazet, 2016, 3981435; Liu, 2018, 4238396}. After sex-stratification, a
negative, non-significant association with BMI was observed among male children in mid-
childhood {Mora, 2017, 3859823}. Domazet et al. (2016, 3981435) reported a non-significant
positive association between PFOA measured at age 15 and BMI at age 21.
C-59
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
In a medium confidence cohort study from the ALSPAC, a significant negative association with
BMI was observed among mother-child pairs {Hartman, 2017, 3859812}. In a medium
confidence study from the Fernald Community Cohort, a repeated-measures analysis reported a
non-significant percent decrease in BMI was observed per IQR increase in PFOA, while a latent-
analysis reported a non-significant percent increase in BMI per IQR increase in PFOA {Blake,
2018, 5080657}. In a sex-stratified analysis, non-significant percent decreases were observed for
both males and females {Blake, 2018, 5080657}.
In the single low confidence study, Tian et al. (2019, 5080586) observed a statistically significant
increase in BMI with increase in PFOA. In a sex-stratified analysis, a statistically significant
positive association was reported between PFOA and BMI among men; the association between
PFOA and BMI among women was positive, but not significant {Tian, 2019, 5080586}. This
study was given a low confidence rating due to concerns with BMI being related to PFOA;
although this was acknowledged by the authors, this was not accounted for in the analysis.
Four studies examined anthropometric measurements, and one reported significant association
with waist circumference. One medium confidence study reported a negative, non-significant
association with hip-circumference {Chen, 2019, 5387400}. Three medium confidence studies
evaluated waist measurements and observed positive, non-significant associations with waist
circumference {Chen, 2019, 5387400; Braun, 2016, 3859823; Liu, 2018, 4238396}
A low confidence study from the Isomers of C8 Health project evaluated waist circumference
among adults. A significant, positive association with waist circumference was observed. After
stratification by sex, the association with waist circumference among men remained significant,
but was not significant among women {Tian, 2019, 5080586}. Significant increased odds of
increased waist circumference were observed in the overall study population and among men;
odds of increased waist circumference were increased but non-significant among women {Tian,
2019, 5080586}. This study was given a low confidence rating due to concerns with waist
circumference being related to PFOA; although this was acknowledged by the authors, this was
not accounted for in the analysis.
C.3.1.6 Findings from Occupational Studies
There was one occupational study, which came from the C8 Health Project {Steenland, 2013,
1937218}. A decreased, non-significant risk of type 1 diabetes was observed in the second and
fourth quartiles of PFOA exposure in both lagged and untagged analyses were observed. A non-
significant increased risk of type 1 diabetes was observed in the third quartile in both lagged and
untagged analyses {Steenland, 2013, 1937218}.
C3.2 Animal Evidence Study Quality Evaluation and
Synthesis
C.3.2.1 Metabolic Homeostasis
There is 1 study from the 2016 PFOAHESD {U.S. EPA, 2016, 3603279} and 5 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOA HESD that investigated the association between PFOA and metabolic effects. Study
quality evaluations for these 6 studies are shown in Figure C-20.
C-60
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
s^AS 0^c0V^
Blake et al., 2020, 6305864-
Butenhoff et al., 2012, 2919192-
Cope et al., 2021, 10176465-
—1—
+
I
+
++
B
B
B
++
B
+
++
¦
++ ++
++
+
+
++
++
B
B
++ ++
B
NTP, 2019, 5400977-
++
++
++
++
++
B
++
++
NTP, 2020, 7330145-
++
++
++ ++
++
++ ++
++
++
van Esterik et al., 2015, 2850288 -
++
++
B
B
++ ++
B
B
S
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-20. Summary of Study Evaluation for Toxicology Studies of PFOA and Metabolic
Effects
Interactive figure and additional study details available on HAWC.
PFOA has been observed to cause perturbations in metabolic homeostasis in rodents. However,
there appears to be differences in responses depending on species, length of exposure, and sex.
Overall, the effects on metabolic parameters following PFOA exposure are inconclusive.
In a 28-day study conducted by NTP (2019, 5400977), glucose was significantly decreased in
male Sprague-Dawley rats following exposure to > 2.5 mg/kg/day PFOA. No significant
response was observed in the female rats treated with up to 100 mg/kg/day PFOA. In a single-
dose study in male Sprague-Dawley rats, Elcombe et al. (2010, 2850034) similarly observed a
significant decrease in serum glucose after administration of 300 ppm PFOA in feed (equivalent
to approximately 19 mg/kg/day) for 28 days. However, a chronic study by Butenhoff et al.
(2012, 2919192) observed increases in glucose when measured beginning at 3 months. The
authors exposed Sprague-Dawley rats to 30 or 300 ppm PFOA in feed for a 24-month period.
Serum samples for clinical chemistry measurements were taken at 3, 6, 12, 18, and 24 months.
For males in the 30 ppm group (-1.3 mg/kg/day PFOA), glucose levels were significantly higher
than controls at 3, 6, and 12 months, then returned to baseline control levels at 18 and 24 months.
Male rats in the 300 ppm group (-14.2 mg/kg/day PFOA) had significantly higher serum glucose
levels than the control groups at the 3- and 24-month time points. In female rats, effects on
serum glucose were only observed at the 6-month timepoint; in both the 30 and 300 ppm groups
C-61
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(-1.6 and -16.1 mg/kg/day PFOA, respectively), serum glucose levels were significantly lower
than controls. In a two-year study conducted by NTP (2020, 7330145), no effects on glucose
levels were reported in male and female Sprague-Dawley rats (See Main PFOA Document for
further study design details).
In CD-I mice, four independent studies investigated the effects of gestational PFOA exposure on
adult offspring {Hines, 2009, 194816; Quist, 2015, 6570066; Cope, 2021, 10176465} or
pregnant dams {Blake, 2020, 6305864} and found no effect on glucose levels or glucose
tolerance. Interestingly, Hines et al. (2009, 194816) observed weight gain in female offspring
exposed to lower doses of PFOA (0.01, 0.1, and 0.3 mg/kg/day but not 1 mg/kg/day or controls)
from GD 1-17. This weight gain was correlated with mid-life (21-33 weeks of age) increased
serum insulin and leptin levels in the 0.01 and 0.1 mg/kg/day groups, but not glucose tolerance in
early (15-16 weeks of age) or late (70-74 weeks of age) adulthood. These results indicate
potential susceptibility to metabolic dysfunction later in life after low-dose gestational PFOA
exposure. However, in a similar study, Quist et al. (2015, 6570066) exposed pregnant mice to 0,
0.01, 0.1, 0.3, or 1 mg/kg/day from GD 1-17 and observed no statistical differences in serum
glucose or insulin levels in female offspring at postnatal week 13 (PNW 13). Blake et al. (2020,
6305864) also saw no effect on dam serum glucose with gestational exposure to 1 or
5 mg/kg/day PFOA from GD 1.5-11.5 or GD 1.5-17.5. Cope et al. (2021,10176465) exposed
dams to 0.2, 1.0, or 2.0 PFOA mg/kg/day from GD 1.5-17.5 and observed a slightly elevated but
non-significant fasting glucose level in male pups fed low fat diets (LFD) and female pups fed
either LFD or high fat diets (HFD) at PND 54-58. The study also reported a dose-dependent
increase in insulin levels, which caused a 43.1% decrease in QUICKI score in males pups
exposed to 1 mg/kg PFOA. No significant effect on glucose tolerance was observed.
Body mass composition in male pups fed LFD was altered with significant increases in fluid
mass at 0.1 mg/kg/day, and fat mass, fluid mass, and percent fluid at 1.0 mg/kg/day {Cope,
2021, 10176465}. No significant changes were observed in male pups fed with HFD at any
PFOA dose group. Female pup fed LFD or HFD had no significant changes to body mass
composition at any PFOA dose group.
C.3.2.2 Survival, Clinical Observations, Body Weight, and
Food/Water Consumption
There are 8 studies from the 2016 PFOAHESD {U.S. EPA, 2016, 3603279} and 14 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOA HESD that investigated the association between PFOA and systemic effects. Study
quality evaluations for these 22 studies are shown in Figure C-21.
C-62
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Biegel et al., 2001, 673581
Butenhoffet al„ 2004, 1291063
Butenhoff etal., 2012, 2919192
Cope etal., 2021, 10176465
Crebelli etal., 2019, 5381564
De Guise etal., 2021, 9959746
Dewittetal., 2008,1290826
Guoetal., 2019, 5080372
Guoetal., 2021, 7542749
Guoetal., 2021, 9960713
Guoetal., 2021, 9963377
Lau etal., 2006, 1276159
Li etal., 2017, 4238518
Loveless et al., 2008, 988599
NTP, 2019, 5400977
NTP, 2020, 7330145
Perkins et al., 2004,1291118 -
Shi etal., 2020, 7161650-
Wolf etal., 2007, 1332672-
Yan etal., 2014, 2850901 J
Yu etal., 2016, 3981487-
Zhang etal., 2020, 6505878- +
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
C-63
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure C-21. Summary of Study Evaluation for Toxicology Studies of PFOA and Systemic
Effects
Interactive figure and additional study details available on HAWC.
Available animal toxicity data suggest that PFOA exposure can elicit whole-body toxicity, which
is reflected by changes in survival, body weights, food consumption, and other clinical
observations. Reductions in survival precipitated only at higher doses of PFOA in a single non-
human primate study. Reductions in terminal body weight and reductions in weight gain are
consistently observed across studies of varying durations of oral exposure to PFOA. Prior to this
updated assessment, the available literature measuring clinical outcomes, food and water
consumption, body weight, and survival primarily consisted of acute studies {U.S. EPA 2016,
3603279}. Many of the findings were consistent with those in more recent literature and are
included herein.
C3.2.2.1 Survival
Although one subchronic toxicity study in non-human primates exposed to >_30 mg/kg/day for
90 days PFOA showed reductions in survival {Goldenthal, 1979, 7692862}, survival rates were
not affected in rodent studies across study durations and doses {NTP, 2019, 5400977; NTP,
2020, 7330145; Perkins, 2004, 1291118; Crebelli, 2019, 5381564; Thomford, 2001, 5432382}.
Interestingly, survival was increased in two studies: Butenhoff et al. (2012, 2919192) and Biegel
et al. (2001, 673581). Butenhoff et al. (2012, 2919192) fed male Sprague Dawley rats 0, 30, or
300 ppm PFOA via the diet (equivalent to 0, 1.3, or 14.2 mg/kg/day) for two years and observed
that survival was increased in males at the highest dose (Figure C-22). No significant effect was
observed in female rats (exposure equivalents of 0, 1.6, or 16.1 mg/kg/day) in this study (Figure
C-22). Similarly, Biegel et al. (2001, 673581) observed increased survival in male Crl:CD BR
(CD) rats fed 300 ppm (equivalent to 13.6 mg/kg/day) PFOA each day at the end of another two-
year study. In other studies of rats, mice, and non-human primates included in this updated
assessment, all animals survived to the end of study (Figure C-22).
Endpoint Study Name
Survival Crebelli et al„ 2019. 5381564
NTP, 2019, 5400977
Perkins et al., 2004. 1291118
NTP, 2020, 7330145
Study Design
subchronic (5wk)
short-term (28d)
subchronic (13wk)
chronic (GD6-PNW107)
Observation Time
5wk
2y
chronic (PND21-PNW107) 2y
Butenhoff et al., 2012. 2919192 chronic (2y) 2y
Biegel et al,, 2001,673581 chronic (2y) 2y
Animal Description
Mouse. C57BI/6 N=6-8)
Rat, Sprague-Dawley {•¦. N=10)
Rat, Sprague-Dawley (;¦, N=10)
Rat, Sprague-Dawley Crt:Cd Br (, N=55)
F1 Rat Sprague-Dawley N=50)
F1 Rat. Sprague-Dawley (¦". N=S0)
Fl Rat. Sprague-Dawley N=5Q)
F1 Rat Sprague-Dawley N=50>
Rat, Sprague-Dawley Cri:Cd(Sd)(Br) (. N=50)
Rat, Sprague-Dawley Cri:Cd('Sd)(Br) (v, N=50)
Rat, Crl:Cd Br {,-% N=156>
PFOA Whole Body Effects - Survival
nge'STi
mficant increase ~ Significant decrease
Concentration (mg/kg/day)
Figure C-22. Effects on Survival in Rodents Following Exposure to PFOA (logarithmic
scale)
PFOA 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; Fi = first generation; d = day; wk = week; y = year.
C-64
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C3.2.2.2 Clinical Observations
Clinical observations have been reported in some animal studies of oral exposure to PFOA. Two
28-day studies described clinical assessments following 28 days of oral PFOA exposure via
gavage in Sprague Dawley rats. Whereas NTP (2019, 5400977) did not observe any treatment-
related clinical observations in 7- to 9-week old Sprague Dawley rats exposed to PFOA (0-
10 mg/kg/day, males; 0-50 mg/kg/day, females) for 28 days, Cui et al. (2009, 757868) described
adverse clinical signs in male Sprague Dawley rats exposed to 5 mg/kg/day, including cachexia
and lethargy in the third week of study. In a 5-week study in male C57BL/6 mice exposed to 0.1-
5 mg/kg/day PFOA, no signs of overt toxicity were observed {Crebelli, 2019, 5381564}.
The aforementioned study by Butenhoff et al. (2004, 1291063) reported that there were low
incidences of dehydration, urine-stained abdominal fur, and ungroomed fur in at least three of the
30 Po male, but not female rats exposed for 70 days in the 30 mg/kg/day exposure group. No
effects were noted in lower exposure groups (1-10 mg/kg/day), nor in the Fi offspring at the end
of study.
The chronic exposure study by Butenhoff et al. (2012, 2919192) checked for palpable masses
daily during the two-year exposure, but the incidence was indistinguishable from controls in all
exposure groups.
C.3.2.2.3 Body Weight in Adults
Reductions in body weight and/or reductions in weight gain have been observed in non-human
primates as well as across rodent studies of varying exposure lengths (short-term, subchronic,
chronic), species (rats or mice), and strains of mice.
In a short-term exposure study, Dewitt et al. (2008, 1290826) found that mean body weight was
reduced in female C57BL/6N mice exposed to 15 or 30 mg/kg/day PFOA in drinking water for
15 days; no effects were observed at or below 7.5 mg/kg/day (Figure C-23). Six independent
studies reported body weights from BALB/c mice exposed to various doses (ranging from 0.4-
20 mg/kg/day) of PFOA via gavage for 28 days; all exposures began around 6-8 weeks of age
{Li, 2017, 4238518; Guo, 2019, 5080372; Yan, 2014, 2850901; Yu, 2016, 3981487; Guo, 2021,
9963377; Guo, 2021, 7542749}. Of these, Yu et al. (2016, 3981487) was the only study that did
not observe any changes in body weight; mice were exposed to 0.5 or 2.5 mg/kg/day PFOA
(Figure C-23). Significant reductions in body weight that differed by more than 10% of control
were observed only at the highest doses tested in the other studies: 2.5 mg/kg/day in Li et al.
(2017, 4238518), 10 mg/kg/day in Guo et al. (2019, 5080372; 2021, 9963377; 2021, 7542749),
and 5 or 20 mg/kg/day in Yan et al. (2014, 2850901) (Figure C-23). Two studies reported weight
reductions in ICR mice exposed for approximately one month. Zhang et al. (2020, 6505878)
observed that 5 mg/kg/day, but not 0.5 or 2 mg/kg/day, PFOA was sufficient to reduce body
weight in female ICR mice after 28 days of exposure (Figure C-23). Males were not evaluated.
Son et al. (2008, 1276157) observed similar results in male ICR mice exposed to 17.63 or
47.21 mg/kg/day for 21 days. Females were not evaluated.
Another short-term exposure study by Loveless et al. (2008, 988599) in CD-I mice administered
0, 0.3, 1, 10, 30 mg/kg/day for 28 days via gavage noted that mean terminal body weights at the
end of study were 86 and 78% of control at 10 or 30 mg/kg/day, respectively. In another study,
6- to 8-week old C57BL/6 mice were exposed to 0, 0.1, 1, or 5 mg/kg/day PFOA in drinking
C-65
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
water for 5 weeks. Whereas untreated control mice gained an average of 5.1 ± 0.2 g over the
course of the 5-week study, mice treated with 5 mg/kg/day PFOA gained significantly less
weight (3.0 ± 0.1 g) {Crebelli, 2019, 5381564}. Shi et al. (2020, 7161650) had similar findings
for the 8-week old C57BL/6J male mice that were dosed with 0, 0.5, 1, and 3 mg/kg/day in
drinking water for 5 weeks. Mice at all dose levels (0.5, 1, or 3 mg/kg/day) were reported to
show a marked decline in body weight gains starting around day 21. The highest dose group
(3 mg/kg/day) was reported to have a lower body weight compared to the control group, which
was demonstrated by both a significant lower body weight on day 35 and a significant difference
in body weight gain over the study period. De Guise et al. (2021, 9959746) exposed B6C3F1
female mice to 0, 1.88, and 7.5 mg/kg/day PFOA via drinking water for 4 weeks. Mice in the
high-dose group had significantly lower body weight compared to the mice in the control group
from exposure day 14 to 28.
Five short-term studies have determined the effect of PFOA on body weight in rats. Loveless et
al. (2008, 988599) applied the aforementioned exposure paradigm for CD-I mice in male
Crl:CD(SD)IGS BR rats. Mean terminal body weights at the end of the 28-day study were 10
and 25% lower than control at 10 and 30 mg/kg/day, respectively. Another study exposed male
and female Sprague Dawley rats to PFOA for 28 days (0, 0.625, 1.25, 2.5, 5, or 10 mg/kg/day for
males, 0, 6.25, 12.5, 25, 50, or 100 mg/kg/day for females) {NTP, 2019, 5400977}. The mean
body weights of 0.625, 1.25, and 2.5 mg/kg/day males and all treated females were within 10%
of the respective vehicle control groups throughout the study. At the end of study, mean body
weights of the 5 and 10 mg/kg/day males were 12% to 19% lower, respectively, than those of the
vehicle control group. No effects on terminal body weight were observed in females.
The remaining three short-term PFOA exposure studies {Pastoor, 1987, 3748971; Cui, 2009,
757868; Rigden, 2015, 7907801} in rats also suggest a decrease in body weight following PFOA
exposure, and are discussed in greater detail in the 2016 PFOA HESD {U.S. EPA 2016,
3603279}. Briefly, Pastoor et al. (1987, 3748971) reported a 17% decrease in body weight from
controls in male Crl:CD (SD) BR rats that had been exposed to 50 mg/kg PFOA for 7 days.
Females were not evaluated. Cui et al. (2009, 757868) found that terminal body weight was
significantly reduced in male Sprague Dawley rats exposed to 20 mg/kg/day PFOA for 28 days,
but the magnitude of this change (in comparison to controls) was less than 10%. No effects were
observed at the 5 mg/kg/day group and females were not evaluated. Rigden et al. (2015,
7907801) exposed male Sprague Dawley rats to 0, 10, 33, or 100 mg/kg/day PFOA via gavage
for three days and recorded body weights each day throughout exposure as well as for four days
after the end of exposure. Although body weight decreased on the last day of exposure in the 33
and 100 mg/kg/day exposure groups, growth resumed and the trajectory mirrored that of all other
groups including controls during the 4 days after exposure.
In a subchronic exposure study, Perkins et al. (2004, 1291118) weighed male Sprague Dawley
rats weekly over the course of a 13-week exposure to 0, 0.06, 0.64, 1.94, or 6.5 mg/kg/day. Body
weight change and absolute body weight at study termination were both reduced in the highest
exposure group (Figure C-23). Another subchronic study in rhesus monkeys (two per sex per
group) reported reductions in body weight following exposure to 30 or 100 mg/kg/day PFOA for
13 weeks {Goldenthal, 1979, 7692862}. The reduction in weight loss preceded death in one
monkey of each sex. Changes in body weight were similar to controls in the other dose groups (3
or 10 mg/kg/day).
C-66
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Absolute body weights of parental (P)-generation male and female Sprague Dawley rats were
measured in a reproductive toxicity study by Butenhoff et al. (2004, 1291063); six-week old rats
were exposed to 0, 1, 3, 10, or 30 mg/kg/day PFOA via gavage for at least 70 days prior to
mating and until sacrificed. During the peripubertal period (through test day 15), body weight
relative to the control group was reduced in males exposed to 10 or 30 mg/kg/day. Terminal
body weight was reduced in Po males following 106 days of exposure at dosages of 3 mg/kg/day
and above, and the changes were greater than 10% in groups exposed to 10 or 30 mg/kg/day
(Figure C-23). Body weights for the Po females were not significantly different (and generally
within 10% from control) during the precohabitation period, body weights in the Po females at
other time points are discussed in the Main PFOA Document.
Two chronic exposure studies reported opposing effects on body weights in male rats that were
fed chow laden with 300 ppm (equivalent to 13.6 mg/kg/day) PFOA for two years {Butenhoff,
2012, 2919192; Biegel, 2001, 673581}. Whereas the Butenhoff et al. (2012, 2919192) study was
performed in Sprague Dawley rats and evaluated the effects of PFOA on body weight in each
sex, Biegel et al. (2001, 673581) used Crl:CD BR rats and only looked at males. Butenhoff et al.
(2012, 2919192) reported reduced body weights in males and females whereas Biegel et al.
(2001, 673581) reported a 34% increase in body weight.
Of note, a few studies observed that the reductions in body weight and/or body weight change
began around day 14-15 of exposure in BALB/c mice {Li, 2017, 4238518} and in Sprague
Dawley rats {NTP, 2019, 5400977}. Although this observation was specific to males in one 28-
day rat study {NTP, 2019, 5400977}, it was common to both sexes in BALB/c mice {Li, 2017,
4238518}. Zhang et al. (2020, 6505878) observed a trending reduction in body weight in female
ICR mice at day 15 of exposure to 5 mg/kg/day PFOA, however the effect did not reach
significance until day 25 and males were not tested. More data are required to understand
whether the reductions in body weight are more common in a particular sex.
C-67
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Endpoint
Study Name
Study Design
Observation Time
Animal Description
Body Weight, Absolute
Dewitt et al.. 2008, 1290826
short-term (15d)
16d
Mouse, C57BL/6n (V. N=8)
Li et al., 2017,4238518
short-term (28d)
28d
Mouse. BALB/c (c , N=6)
Mouse, BALB/C (J, N=6)
Yan et al., 2014, 2850901
short-lerm (28d)
28d
Mouse, BALB/c (.{ , N=16)
Yuetal.,2016. 3981487
short-term (28d)
28d
Mouse. BALB/c (•<. N=5)
Zhang et al.. 2020, 6505878
short-term (28d)
28d
Mouse, ICR (2, N=8)
Guo et al., 2021, 7542749
short-term (28d)
28d
Mouse, BALB/c N=12)
Loveless et al.. 2008, 988599
short-term (29d)
0d
Mouse. Crl:CD-1(ICR)BR 0, N=20)
7d
Mouse, Cr1:CD-1(ICR)BR N=20)
14d
Mouse, Crl:CD-1(ICR)BR (o, N=20)
21d
Mouse. Crl:CD-1(ICR)BR N=20)
28d
Mouse, Crl:CD-1(ICR)BR (.y, N=20)
Guo et al., 2019, 5080372
short-term (4wk)
4wk
Mouse. BALB/c (<*, N=12)
De Guise et al.. 2021. 9959746
short-term (4wk)
4wk
Mouse. B6C3F1 ($, N=12-16)
Butenhoff et al., 2004, 1291063
reproductive (64d)
106d
P0 Rat. Crl:CD(SD)IGS BR (. N=15)
Butenhoff el al., 2012, 2919192
chronic (2y)
2y
Rat, Sprague-Dawley Cri:Cd(Sd)(Br) , N=35-44)
Rat, Sprague-Dawley Cri:Cd(Sd)(Br) ($, N=15)
Body Weight Change
Shi etal., 2020. 7161650
subchronic (5wk)
0-5wk
Mouse, C57BL/6J {.?, N=8)
Crebelli et al., 2019, 5381564
subchronic (5wk)
0-5wk
Mouse, C57BI/6 ( ;, N=6-8)
Perkins etal., 2004, 1291118
subchronic (13wk)
0-13wk
Rat, Sprague-Dawley Cr1:Cd Br (
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
5 mg/kg/day groups, which were significantly heavier than controls at an earlier timepoint, PND
21 (See Main PFOA Document). In Cope et al. (2021, 10176465), male pups of CD-I mice
exposed to 0.1 or 1 mg/kg/day PFOA did not display significant difference in body weight
except for an increase in the 1 mg/kg/day PFOA group at PNW 17 in the low-fat diet group. In
female pups, no significant differences were observed among any exposed group.
Absolute body weights in adult Fi-generation rats were also measured in the aforementioned
study by Butenhoff et al. (2004, 1291063). Pomale and female Sprague Dawley rats were
exposed to 0, 1, 3, 10, or 30 mg/kg/day PFOA for at least 70 days prior to mating and until
sacrificed and their offspring (Fi generation) were dosed similarly beginning at weaning.
Relative body weights were reduced in Fi male and female juvenile (PND 35) rats, as well as
peripubertal Fi (PND 56) male rats from the 30 mg/kg/day group. Additionally, male rats from
the 10 mg/kg/day group had significantly reduced body weight (post-weaning) beginning at PND
77 and lasting through the end of the study. A dose-dependent reduction in body weight at the
end of the study (PND 120) was observed in Fi males (Figure C-24). Effects on maternal body
weight and on offspring prior to weaning are described in (see Main PFOA Document).
Two rodent studies evaluated the relative sensitivities of body weight to perinatal and/or
postnatal exposure of PFOA. NTP (2020, 7330145) evaluated the effects on body weight
following perinatal and/or postweaning exposure to PFOA in Sprague Dawley rats. In that study,
pregnant rats were exposed to 0, 150, or 300 ppm PFOA to constitute a perinatal exposure in
offspring, and postnatal exposures (0, 150, or 300 ppm for males, 0, 300, or 1,000 ppm for
females) were continued during the postweaning period for two years (See Main PFOA
Document for further study design details). Body weights at the 16-week interim period tended
to be lower in all Fi gestational (GD 6-PNW 21; GD 6-PNW 107) and post-weaning (PND 21-
PNW 21; PND 21-PNW 107) exposure groups and reached significance in all male exposure
groups. At the end of the two-year study, there were no consistent effects of PFOA exposure on
Fi males. However, absolute body weight was reduced in Fi females exposed during gestation
plus after weaning (GD 6-PNW 107) as well as after weaning alone (PND 21-PNW 107).
Similar findings come from Wolf et al. (2007, 1332672), who investigated the relative
contributions of gestational and lactational exposures to PFOA in CD-I mice. Pregnant mice
were given 0 or 5 mg/kg/day PFOA at staggered intervals of gestational development (GD 7-17,
10-17, 13-17, or 15-17) and/or 0, 3, or 5 mg/kg/day during the lactational period (PND 1-22).
Body weights were determined in male and female pups on PND 22 and PND 92. While no
reductions in absolute body weight in any group at PND 92 were observed, an elevation in body
weight was noted in PND 92 mice exposed to 3 mg/kg/day from GD 1-17, which had been
significantly decreased from control when measured on PND 22 (see Main PFOA Document).
C-69
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOA Whole Body Effects - Body Weight
Endpoint
Study Name
Study Design
Observation Time
Animal Description
£ No significant change A Significant increase Significant decrease
Body Weight, Absolut©
Lau et al„ 2006. 1276159
developmental (GD1-18)
PNW6.5
F1 Mouse, CD-1 { N=4)
¦ , , ^7
F1 Mouse. CD-1 ( N=4)
PNW60 F1 Mouse, CD-1 ( '', N=4)
F1 Mouse, CD-1 ( N=4)
Song et al., 2018, 5079725
developmental (GD1-17)
F1 Mouse, Kunming {•<', N=7-9)
V ¦ •
Wolf et al., 2007, 1332672
developmental (GD1-17)
F1 Mouse, CD-1 (•?, N=11-14)
F1 Mouse, CD-1 (i, N=11-14)
developmental (GD1-PND22)
PND92
F1 Mouse, CD-1 N=12-14)
F1 Mouse. CD-1 ( -, N=12-14)
developmental (PND1-22)
PND92
F1 Mouse, CD-1 (.?, N=11-14)
F1 Mouse. CD-1 < -, N=11-14)
Cope etal., 2021,10176465
developmental (GD1,5-17.5)
F1 Mouse, CD-1 N=7)
F1 Mouse, CD-1 { *. N=8)
Butenhoff et al.. 2004,1291063
reproductive (GD0-PND120)
F1 Rat, Crl:CD(SD)lGS BR (J. N=29-30)
rNUI
NTP. 2020, 7330145
chranic (GD6-PNW21)
16wk
F1 Rat, Sprague-Dawley ( N=10)
chronic (GD6-PNW107)
16wk
F1 Rat, Sprague-Dawley ( ¦', N=10)
V7 V7 y
F1 Rat, Sprague-Dawley f ::, N=10)
chronic (PND21-PNW21)
16wk
F1 Rat, Sprague-Dawley N=10)
V7 TJ7
chronic (PND21-PNW107)
16wk
F1 Rat, Sprague-Dawley (N=10)
F1 Rat, Sprague-Dawley < :, N=10)
chronic (GD6-PNW107)
F1 Rat, Sprague-Dawley { ?, N=22-24)
F1 Rat, Sprague-Dawley N=19-22)
chronic (PND21-PNW107)
2y
F1 Rat, Sprague-Dawley ( •', N=21-24)
F1 Rat. Sprague-Dawley { ¦•, N=18-19)
Body Weight Change
Cope etal.. 2021, 10176465
developmental (QD1.5-17.5)
F1 Mouse. CD-1 (-•')
F1 Mouse, CD-1 (•'¦)
Body Weight. Percent of Control
Butenhoff et al.. 2004,1291063
reproductive (GD0-PND120)
pkir-voE
F1 Rat, Crl:CD(SD)IGS BR (c', N=29-30)
reproductive (GD1-PND106)
PND35
F1 Rat, Cr1:CD(SD)IGS BR (2, N=28-29)
TT , , y
reproductive (GD0-PND120)
F1 Rat, Crl:CD(SD)IGS BR (
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Studies of varying exposure durations in rats have also reported food and/or water consumption
that in some cases support a relationship between reduced intake and weight loss. The study by
Rigden et al. (2015, 7907801) noted a slight decrease in food consumption (data were not
provided) and suggested dehydration related to decreased water consumption as an explanation
for weight loss due to increased urine volume during the final two days of exposure. In another
study of male Sprague Dawley rats exposed to 0, 5, or 20 mg/kg/day PFOA for 28 days via
gavage exhibited decreased food consumption at the 5 mg/kg/day dose {Cui, 2009, 757868}
However, this level of exposure did not coincide with an effect on weight loss. Elcombe et al.
(2010, 2850034) also recorded food consumption (per gram basis) in male Sprague Dawley rats
fed 300 ppm PFOA for 28 days. Rats exposed to PFOA consumed less food by day 28. No
differences in food consumption were observed in another study of male Sprague Dawley rats
fed 0, 1, 10, 30, or 100 ppm (equivalent to 0, 0.06, 0.64, 1.94, 6.5 mg/kg/day) for 13 weeks,
despite reductions in body weight at the highest exposure level {Perkins, 2004, 1291118}.
Females were not used in this study. Biegel et al. (2001 673581) and Butenhoff et al. (2012,
2919192) reported elevated food consumption in male rats exposed to 300 ppm PFOA for two
years. Butenhoff et al. (2012, 2919192) also evaluated female rats and reported inconsistent
trends of reduced food consumption that appeared to be related to variations in body weight
(Figure C-25).
The reproductive toxicity study in Sprague Dawley rats by Butenhoff et al. (2004, 1291063)
recorded food consumption of Po males as well as their male and female Fi offspring at PND 35
following exposure to 0, 1, 3, 10, or 30 mg/kg/day PFOA via gavage. Mean absolute feed
consumption (as a percent of control) of male Po rats was reduced in the highest exposure group
for a majority of the time across 106 days of study. However, given the aforementioned
reductions in body weight for these animals, feed consumption relative to body weight was
actually elevated at the 3, 10, and 30 mg/kg/day doses. For Fi males and females, absolute feed
consumption was reduced at the 30 mg/kg/day dose (Figure C-25).
Two non-human primate studies covered in the 2016 PFOA HESD {U.S. EPA 2016, 3603279}
reported reductions in food consumption. Male cynomolgus monkeys displayed overt toxicity,
including reduced food consumption, after just 12 days of oral exposure to 30 mg/kg/day PFOA.
As a result, the exposure was reduced to 20 mg/kg/day on day 22 for the remainder of the 26-
week study {Butenhoff, 2002, 1276161}. Male cynomolgus monkeys were used in another study
that evaluated health effects including food consumption post exposure to 0, 2, or 20 mg/kg/day
PFOA for 4 weeks. Low/no food consumption was observed in one male cynomolgus monkey
from the 20 mg/kg/day exposure group {Thomford, 2001, 5432382}.
PFOA Whole Body Effects - Food Consumption
Endpoint Study Name Study Design Observation Time Animal Description | # No significant chango A Significant incroaso ~ Significant docroaso
Feed Consumption. Absolute Shi et al., 2020.7161650 subchronic (5wk) 5wk Mouse, C57BL/6J (3\ N=8)
Butenhoff et al., 2004, 1291063 reproductive <64d) 1-106d P0 Rat, Crl:CD{SD)IGS BR (->, N=30)
reproductive (GD0-PND120) PND22-120 F1 Rat, Crl:CD(SD)IGS BR (
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.3.3
Mechanistic Evidence
Mechanistic evidence linking PFOA exposure to adverse metabolic and systemic outcomes are
discussed in Sections 3.3.2, 3.3.3, and 3.4.5 of the 2016 PFOAHESD {U.S. EPA, 2016,
3603279}. There are 35 and 33 studies from recent systematic literature search and review
efforts conducted after publication of the 2016 PFOA HESD that investigated the mechanisms of
action of PFOA that lead to metabolic and systemic effects, respectively. A summary of these
metabolic and systemic studies is shown in Figure C-26 and Figure C-27, respectively.
Additional mechanistic synthesis will not be conducted since evidence suggests but is not
sufficient to infer that PFOA leads to metabolic and systemic effects.
Mechanistic Pathway
In Vitro Grand Total
Big Data. Non-Targeted Analysis
2
1
2
5
Cell Growth, Differentiation, Proliferation, Or Viability
4
0
13
16
Cell Signaling Or Signal Transduction
1 1
4
6
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
8
2
8
18
Hormone Function
3
5
3
11
Inflammation And Immune Response
2
0
1
3
Oxidative Stress
2
1
3
6
Xenobiotic Metabolism
0
0
4
4
Other
1
0
0
1
Not Applicable/Not Specified/Review Article
1
0
0
1
Grand Total
12
7
17
35
Figure C-26. Summary of Mechanistic Studies of PFOA and Metabolic Effects
Interactive figure and additional study details available on Tableau.
C-72
-------
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
0
0
•
4
Cell Growth, Differentiation, Proliferation, Or Viability
1
0
mm
g
Cell Signaling Or Signal Transduction
1
1
10
Extracellular Matrix Or Molecules
0
0
1
1
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation 1 1
11
Hormone Function
0
0
1
1
Inflammation And Immune Response
0
0
3
3
Oxidative Stress
1
1
9
Xenobiotic Metabolism
0
1
2
3
Other
1
0
3
4
Not Applicable/Not Specified/Review Article
1
0
0
1
Grand Total
4
2
27
33
Figure C-27. Summary of Mechanistic Studies of PFOA 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 PFOA exposure and metabolic effects in
humans based on observed effects for diabetes, gestational weight gain, leptin, and adiposity
measures in high and medium confidence studies. However, there are generally imprecise and
inconsistent findings across 72 epidemiological studies. Stronger evidence exists for diabetes and
some adiposity measures relative to other metabolic outcomes.
The available human epidemiological evidence supports an association between PFOA and
diabetes, including gestational diabetes. Five studies reported positive associations with
gestational diabetes, and five studies reported positive associations with type-2 diabetes in the
general population. There is evidence of a positive association with leptin in adults (fours
studies) and in pregnant women (one study), while in children the findings are mixed (two
studies). This suggests that age may be a factor in the association between PFOA and leptin.
Three epidemiological studies observed positive associations with gestational weight gain among
pregnant women, with one association being significant. Four general population studies
reported a positive association with waist circumference, and four studies of children reported
non-significant positive associations with waist circumference, and one study reported inverse
associations in children. There is evidence of an association between PFOA exposure and body
fat and being overweight, particularly in adults, but findings are imprecise and inconsistent.
Findings for an association between PFOA 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.
C-73
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The animal evidence for an association between PFOA and systemic or metabolic effects is
indeterminate. Although some alterations related to glucose homeostasis were reported in the 5
high or medium confidence studies available animal toxicity literature for metabolic effect, the
results were often inconsistent when comparing between species, sexes, length of exposure, and
life stages. In addition, the effects on body weight, clinical observations, and mortality from 21
high or medium confidence studies indicate that the systemic effects occur only at the high doses
tested. In male rats, changes in serum glucose levels appear to be influenced by exposure
duration, with short-term exposure resulting in decreased serum glucose levels and chronic
exposure resulting in increased serum glucose levels. In mice, there was no evidence of altered
glucose levels due to PFOA exposure and conflicting reports of changes in serum insulin levels
in studies with similar exposure paradigms.
Evidence from animal studies suggests that exposure to PFOA of varying durations can elicit
adverse whole-body effects, which primarily manifest as reductions in body weight that are not
always explained by decreased food and/or water consumption or other clinical signs of toxicity.
The effects are consistent across studies of varying exposures to PFOA, across species, and
across sex. Reductions in body weight may serve as an early indicator of later PFOA toxicity
because it can reflect poor health in the whole organism.
C.3.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOA 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, leptin, and adiposity effects observed in
high and medium confidence studies in humans exposed to median PFOA levels between 1.4 and
68 ng/mL. Although there is some evidence of negative effects of PFOA exposure on metabolic
syndrome, there is considerable uncertainty in the results due to inconsistency across studies and
limited number of studies.
C-74
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-6. Evidence Profile Table for PFOA 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
3 High confidence
studies
15 Medium confidence
studies
7 Low confidence studies
Significant increases in
FBG (3/13), including one
high confidence study.
However, other studies
reported contrasting
findings, including
significant decreases
(1/13) in FBG levels after
the OGTT, or imprecise
results. Findings for FBG
levels in children and
pregnant women were
inconsistent between
studies across confidence
levels.
1 High and medium
confidence studies
• Low confidence studies
• Lmprecision of most
findings
• Lnconsistent direction of
effects
• Potential for selection
bias and residual
confounding by SES
Diabetes (and
gestational diabetes)
3 High confidence
studies
18 Medium confidence
studies
6 Low confidence studies
Findings in adults were
mixed. Positive
associations indicating
increased risk were
observed in several studies
(4/11); however,
significant negative
associations indicating
decreased risk were also
observed (3/11). Findings
in adults (5/9) also showed
reduced HbAlc, with a
few reaching significance
(3/9). In pregnant women,
risk of diabetes or
gestational diabetes was
typically increased (6/10),
1 High and medium
confidence studies
• Low confidence studies
• Lnconsistent direction of
effect
• Lmprecision of findings
• Potential for outcome
misclassification, self-
selection, residual
confounding by SES, and
concerns about
temporality
©OO
Slight
Evidence for metabolic
effects is based on
increases in fasting blood
glucose, increased risk of
diabetes, and increases in
adiposity in adults and
pregnant women. 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,
leptin, and adiposity and
there was limited animal
evidence. Although there is
some evidence of negative
effects of PFOA 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-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
reaching significance in
one study. The only study
examining diabetes in
children was considered
uninformative.
Insulin levels
1 High confidence study
8 Medium confidence
studies
7 Low confidence studies
In adults, studies reported
HOMA-IR was
significantly increased
(2/10), but findings from
other studies (6/10)
indicated non-significant
decreases. HOMA-B was
also reported to be
significantly increased
(2/3). Findings for fasting
insulin were mixed, but
significant increases were
observed (2/9). In
pregnant women, findings
indicated decreased
HOMA-IR (2/3),
including one significant
study. In children, findings
for HOMA-IR were
primarily inverse (3/5),
but findings were
generally imprecise for
fasting insulin levels.
1 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
Adiponectin and leptin
5 High confidence
studies
3 Medium confidence
studies
In adults, one study
observed significant
increased leptin (1/1)
while two studies (2/2)
reported decreased
1 High and medium
confidence studies
• Low confidence studies
• Lnconsistent direction of
effects
• Lmprecision of findings
C-76
-------
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
3 Low confidence studies adiponectin, with one
reaching significance.
Findings in children were
mixed for leptin, including
one study reporting
significant decreased
leptin (1/3), but non-
significant positive
associations were
observed for adiponectin.
Findings in pregnant
women were mixed or
imprecise.
Adiposity
8 High confidence
studies
23 Medium confidence
studies
6 Low confidence studies
In adults, findings for BMI
were largely positive
(4/9), with two studies
reporting significant
increases in BMI or risk of
being overweight/obese.
WC was reported to be
significantly increased in
one study, but findings
from other studies in
adults were imprecise. In
children, results were
mixed with two studies
(2/17) reporting
significant positive
associations with
measures of BMI and two
studies (2/17) reporting
significant inverse
associations with
measures of BMI. Other
anthropometric measures
1 High and medium
confidence studies
• Low confidence studies
• Lnconsistent direction of
effects
• Lmprecision of findings
• Potential for residual
confounding by SES,
study sensitivity issues
due to some small sample
sizes
C-77
-------
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
were also mixed, however,
two studies reported
significant decreases in
WC and waist-to-height
ratio.
Metabolic syndrome
Findings for metabolic • Medium confidence
• Low confidence study
4 Medium confidence
syndrome in adults were studies
• Imprecision of findings
studies
mixed, however, two •
• Potential for selection
1 Low confidence study
studies (2/5) observed
bias, outcome
positive associations of
misclassification, and
increased risk of MetS
residual confounding by
with large effect sizes.
SES
Evidence from In Vivo Animal Studies (Section C.3.2.1 and Section C.3.2)
Glucose homeostasis
2 High confidence
studies
3 Medium confidence
studies
Most studies reported no
significant effects on
glucose levels (4/5) or
glucose tolerance (1/1) in
rodents, however, one 28-
day study in rats reported
a dose-dependent decrease
in glucose levels in males.
One developmental mouse
study observed no
significant changes in
insulin levels for either
sex (1/1).
• High and medium
confidence studies
• Dose-response
relationship
• Inconsistent direction of
effects across exposure
durations, sex, and
species
• Limited number of studies
examining outcomes
Body weight
5 High confidence
studies
16 Medium confidence
studies
Reduction in absolute
body weights (19/21),
body weight change (6/7),
and body weight as a
percentage of control (4/4)
were reported following
short-term, subchronic,
and chronic exposure in
• High and medium
confidence studies
• Consistent direction of
effects
Confounding variables
such as decreases in food
consumption
OOO
Indeterminate
Alterations related to
glucose homeostasis were
reported in 5 high or
medium confidence
studies were inconclusive
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
water consumption) from
20 high or medium
C-78
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Evidence Stream Summary and Interpretation
Studies and
Interpretation
Body mass composition
1 Medium confidence
study
Survival and mortality
3 High confidence
studies
3 Medium confidence
studies
Clinical observations
2 High confidence
studies
2 Medium confidence
studies
Food and water
consumption
2 High confidence
studies
4 Medium confidence
studies
Summary and Key
Findings
rats and mice. In rats,
body weight in males
appeared to be more
sensitive to the effects of
PFOA.
One developmental mouse
study reported
significantly increased
fluid mass, fat mass, and
percent fluid mass in
males only (1/1).
Increased survival was
observed in male rats only
following PFOA exposure
(2/6). No significant
effects on mortality were
observed for females
(3/3).
Clinical observations were
observed in rodent studies
(2/4). Observations
included dehydration,
urine-stained abdominal
fur, and/or ungroomed fur
in male rats. Ataxia in
females was reported in a
mouse study.
No significant exposure
related effect on food
consumption (4/5) nor
water consumption (2/2)
was observed in rodents
for either sex.
Factors that Increase
Certainty
• Medium confidence
study
• High and medium
confidence studies
• High and medium
confidence studies
Factors that Decrease
Certainty
• Limited number of studies
examining specific
outcome
• Limited number of studies
examining outcome
• Lnconsistent direction of
effect across studies and
sex
• Limited number of studies
examining outcomes
• Qualitative and subjective
data reporting
Evidence Stream
Judgment
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.
Evidence Integration
Summary Judgment
• High and medium • Limited number of studies
confidence studies examining outcomes
• Consistent direction of
effects across studies
Note: FBG = fasting blood glucose; OGTT = oral glucose tolerance testing; SES = social economic status; HbAlc = hemoglobin Ale; 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-79
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.4 Nervous
EPA identified 38 epidemiological and 11 animal studies that investigated the association
between PFOA and nervous effects. Of the epidemiological studies, 3 were classified as high
confidence, 30 as medium confidence, and 5 were considered low confidence (Section C.4.1). Of
the animal studies, 3 were classified as high confidence, 6 as medium confidence, and 2 were
considered low confidence (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
PFOA Document).
C.4.1 Human Evidence Study Quality Evaluation and
Synthesis
C.4.1.1 Introduction
The 2016 Health Assessment {U.S. EPA, 2016, 3603279} reported mixed results from the
literature reviewed and emphasized 2012 C8 Science Panel (2012, 1430770) conclusions, which
reported no probable link between PFOA exposure and neurodevelopmental disorders in
children, including attention deficit hyperactivity disorder (ADHD) and learning disabilities.
Among the studies reviewed for the 2016 Health Assessment, evidence of a significant positive
association for child PFOA levels and parent reported ADHD was observed in children aged 12-
15 in the general population {Hoffman, 2010, 1291112}, and a positive association with ADHD-
like behaviors and decreased executive function in children in a highly exposed community
{Stein, 2014, 2721873}. The relationship between PFOA exposure and ADHD-related behavior
was also observed in a single country from the INUENDO cohort, showing a significant increase
in hyperactivity among children ages 7 to 9 with elevated PFOA exposure {Hoyer, 2015,
2851038}. A significant increase in risk of development of cerebral palsy in males associated
with maternal PFOA was observed in a case-control study of maternal PFOA levels of
participants within the DNBC {Liew, 2014, 2852208}. Studies on outcomes such as Apgar
score, fine motor skills, gross motor skills, cognitive skills, behavioral problems, and
coordination problems did not find significant evidence for an effect of PFOA exposure {Fei,
2008, 1290822; Fei, 2011, 758428}. Data interpretations within these studies were limited in
some cases by use of a cross-sectional analysis {Fei, 2008, 1290822; Hoffman, 2010, 1291112;
Stein, 2014, 2721873}, potential random misclassification error resulting from using current
PFOA levels as proxy measures of etiologically relevant exposures {Hoffman, 2010, 1291112;
Stein, 2014, 2721873}, outcomes defined by parental report {Fei, 2008, 1290822; Fei, 2011,
758428; Hoyer, 2015, 2851038; Hoffman, 2010, 1291112} or parent and teacher report {Stein,
2014, 2721873}, and limited sample sizes in some sub-analyses {Hoyer, 2015, 2851038}.
For this updated review, 36 studies (38 publications)8 investigated the association between
PFOA and neurological outcomes. Two were conducted in high-exposure communities
{Spratlen, 2020, 6364693; Stein, 2013, 2850964}. One publication {Vuong, 2020, 6356876}
was conducted in pregnant women. The remainder were conducted on the general population.
Study designs included 3 case-control {Ode, 2014, 2851245; Long, 2019, 5080602; Shin, 2020,
8 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-80
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
6507470}, 2 nested case-control {Liew, 2015, 2851010; Lyall, 2018, 4239287}, 26 cohort
(Appendix D). The studies measured PFOA in different matrices, including blood, cord blood,
breast milk {Forns, 2015, 3228833; Lenters, 2019, 5080366}, maternal serum, amniotic fluid
{Long, 2019, 5080602}, and maternal plasma. Eight 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 examined in these studies included clinical conditions such as ADHD, autism spectrum
disorder (ASD), multiple sclerosis (MS), and hearing loss. Neurological function was also
assessed by performance on numerous neuropsychological tests evaluating neurological
domains, including development, general intelligence (i.e., intelligence quotient (IQ)), social-
emotional, executive function, ADHD and attention, ASD and intellectual disability (ID),
memory, and visuospatial performance.
C.4.1.2 Study Quality
There are 38 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and nervous effects. Study quality evaluations for these 38 studies are
shown in Figure C-28 and Figure C-29.
Of the 38 studies identified since the 2016 assessment, three studies {Niu, 2019, 5381527;
Oulhote, 2016, 3789517; Harris, 2018, 4442261} were classified as having high confidence, 30
studies as medium confidence, and five as 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 PFOA. Small sample size,
temporality and reporting concerns were cited as limitations in Weng et al. (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.
C-81
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
\ec^0
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
,0©
Oulhote etal.,2016, 3789517
Oulhote et a!., 2019, 6316905 -
Quaak et al., 2016, 3981464 -
Shin et al., 2020, 6507470-
Shrestha et al., 2017, 3981382 -
Skogheim et al., 2019, 5918847 -
Spratlen et al.
Steenland et al.
Stein et al.
Strom et al.
Vuong et al.
Vuong et al.
Vuong et al.
Vuong et al.
Vuong et al.
Vuong et al.
Wang et al.
Weng et al.
Zhang et al.
2020,6364693 -
2013, 1937218
2013, 2850964
2014, 2922190
2016, 3352166
2018, 5079675
2018, 5079693-
2019, 5080218-
2020, 6356876 -
2020, 6833684
2015, 3860120-
2020, 6718530-
2018, 4238294-
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-29. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Neurological Effects (Continued)
Interactive figure and additional study details available on HAWC.
C-83
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.4.1.3 Findings in 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 two high-exposure
community cohort studies {Stein, 2013, 2850964; Spratlen, 2020, 6364693} examined
developmental outcomes in children. In a high confidence study {Niu, 2019, 5381527} from the
Shanghai Minhang Birth Cohort Study (SMBCS), maternal PFOA concentrations
(median = 19.9 ng/mL) during pregnancy were consistently associated with increased risk of
problems with personal-social skills in 4-year old girls (but not in boys), as assessed by the ASQ.
In boys, significant decreases in risk for problems with gross motor development were observed,
and the risk of problems with problem solving skills were non-significantly elevated. Results
from a medium confidence study {Goudarzi, 2016, 3981536} reported prenatal PFOA
(median =1.2 ng/mL) concentrations were associated with statistically significantly lower
Mental Development Index (MDI) scores for female (but not male) infants at 6 months of age. In
contrast, no apparent trends with neurodevelopmental indices from the Bayley Scales of Infant
Development (BSID-II) at one year of age were reported in a high-exposure community study of
children prenatally exposed to the WTC Disaster {Spratlen, 2020, 6364693}. Adverse
associations at 2 and 3 years were not observed, however, a significant positive association was
reported for the MDI at 3 years {Spratlen, 2020, 6364693}. A medium confidence study {Jeddy,
2017, 3859807} using data from the ALSPAC observed inconsistent patterns of association
between prenatal PFOA concentrations (median = 3.7 ng/mL) and neurodevelopmental indices in
15-month olds as assessed by an adapted version of the MacArthur Communicative
Development Inventories for Infants (MCDI). An inverse association was reported for
intelligibility scores among 38-month olds, but there were no associations with maternal PFOA
for language or communicative scores in 38-month olds. Results varied by maternal age at
delivery, as a statistically significantly inverse association was observed for vocabulary
comprehension and production scores in 15-month infants with mothers younger than 25 years of
age, and a significant inverse association for intelligibility scores in children 38 months of age
with mothers older than 30 years of age {Jeddy, 2017, 3859807}. Results did not suggest an
adverse association between estimated or measured PFOA exposures and performance on
neuropsychological tests (NEPSY-II) in a high-exposure community study of children
participating in the C8 Health Project {Stein, 2013, 2850964}. In one low confidence study,
which assessed perinatal PFOA breast milk exposures (median = 40 ng/mL) and child
neuropsychological development at 6, 12 and 24 months of mother-child pairs in the HUMIS
{Forns, 2015, 3228833}, no association was reported between perinatal PFOA exposures and
early neuropsychological development.
Eleven studies evaluated cognitive function and IQ measures among children, with most
conducted within the general population {Vuong, 2020, 6833684; Zhang, 2018, 4238294; Stram,
2014, 2922190; Harris, 2018, 4442261; Oulhote, 2019, 6316905; Skogheim, 2019, 5918847;
Vuong, 2019, 5080218; Liew, 2018, 5079744; Wang, 2015, 3860120; Lyall, 2018, 4239287}
and two within high-exposure communities {Stein, 2013, 2850964; Spratlen, 2020, 6364693}. In
a medium confidence study {Stein, 2013, 2850964} of children from the C8 Health Project, girls
aged 6 to 12 years with measured childhood PFOA (median = 35.0 ng/mL) exposure above the
median had a 4.1 score decrease in the Wechsler Individual Achievement Test-II (WIAT-II)
Numerical Operations scaled score as compared with girls below the median. A significant 4.9
score increase was observed among boys for the same measure. Overall, children in the highest
C-84
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
vs. the lowest quartile of estimated in utero PFOA (110.8-649.2 ng/mL vs. 4.5- < 11.7 ng/mL)
had significant increases in full scale IQ. Across all administered tests, no consistent adverse
associations between measured childhood PFOA (median = 35.0 ng/mL) and cognitive function
{Stein, 2013, 2850964} were observed. Positive associations between prenatal PFOA
(median = 5.2 ng/mL) and reading skills were reported in a medium confidence study in children
aged eight years utilizing data from the HOME study {Vuong, 2020, 6833684}. Childhood
serum PFOA concentrations at ages three and eight were statistically significantly associated
with higher children's reading scores at ages five and eight years, respectively in a medium
confidence prospective study of data within the HOME study {Zhang, 2018, 4238294}. No
significant associations between prenatal PFOA and offspring scholastic achievement were
reported in a medium confidence prebirth cohort study of children (up to age 20) participants
within the Danish Fetal Origins Cohort {Stram, 2014, 2922190}. Maternal prenatal PFOA
(median = 3.3 ng/mL) concentrations were statistically significantly associated with lower
cognitive function as assessed by the Boston Naming Test with cues in a medium confidence
study of children aged seven years {Oulhote, 2019, 6316905}.
Skogheim et al. (2019, 5918847) examined cognitive dysfunction in preschool children from the
Norwegian Mother, Father, and Child Cohort Study (MoBa) and evidence was inconsistent.
Significant decreases in non-verbal working memory were observed only in the highest quintile
and significant increases in verbal working memory only in the third quintile od PFOA prenatal
exposure (median = 2.5 ng/mL) {Skogheim, 2019, 5918847}. No adverse associations between
prenatal (geometric mean = 5.2 ng/mL) and childhood (geometric mean = 2.4 ng/mL) PFOA and
cognitive function at eight years were reported, and a statistically significant increase of 4.1
points in working memory associated with an increase in prenatal PFOA was reported in a
medium confidence study utilizing data from the HOME study {Vuong, 2019, 5080218}. Child
sex modified the positive association {Vuong, 2019, 5080218}, with higher full-scale IQ in
female children, and no association in male children. In another medium confidence study in a
highly exposed community study, statistically significant sex-specific trends between exposures
and some cognitive outcomes (verbal and full-scale IQ) at four and six years were observed,
suggesting stronger positive associations for females compared to males {Spratlen, 2020,
6364693}. No consistent associations between prenatal PFOA and child IQ at five years of age
were reported in a medium confidence study of children from the DNBC {Liew, 2018,
5079744}. Data from a medium confidence study {Wang, 2015, 3860120} on the Taiwan
Maternal and Infant Cohort Study showed no consistent associations between maternal serum
PFOA (median = 2.5 ng/mL) and IQ measurements in children five or eight years of age.
Six studies examined the potential relationship between PFOA and social-emotional and
behavioral regulation problems {Quaak, 2016, 3981464; Oulhote, 2019, 6316905; Ghassabian,
2018, 5080189; Vuong, 2018, 5079693; Oulhote, 2016, 3789517; Weng, 2020, 6718530}. The
relationship between prenatal PFOA (median = 870.0 ng/L) exposures and behavioral
development at age 18 months using the Child Behavior Checklist 1.5-5 (CBCL 1.5-5) was
explored in a high confidence study utilizing data from the Dutch cohort Linking Maternal
Nutrition to Child Health (LINC) {Quaak, 2016, 3981464}. Results indicated prenatal exposure
to PFOA was statistically significantly negatively associated with externalizing behavior
problems in boys, indicating less problems. Statistically significant associations between serum
PFOA (median = 4.1 |ig/L) in children aged five years and total Strengths and Difficulties
Questionnaire (SDQ) behavioral survey scores assessed at age seven were reported in a high
C-85
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
confidence study {Oulhote, 2016, 3789517}. Maternal prenatal PFOA concentrations
(median = 3.3 ng/mL) were positively associated with total SDQ scores, indicating more
behavioral problems, in a medium confidence study of children seven years of age {Oulhote,
2019, 6316905}. Higher newborn PFOA levels (median =1.1 ng/mL) in dried blood spots were
associated with difficulties in prosocial behavior, but not total behavioral difficulties, as assessed
by the maternal completed SDQ at age 7 in another medium confidence study {Ghassabian,
2018, 5080189}. Evidence was mixed and insufficient to support an overall association between
prenatal PFOA (median = 5.2 ng/mL) and inattention, impulsivity as assessed by the Connors'
Continuous Performance Test-II (CCPT-II) in a medium confidence study {Vuong, 2018,
5079693}. A low confidence study on adolescents reported no significant correlations between
prenatal PFOA levels (mean = 2.9 ng/mL) and brain activity in regions associated with impulsive
behavior as assessed by MRI imaging in teenage offspring {Weng, 2020, 6718530}.
One medium confidence study {Stram, 2014, 2922190} from the Danish Fetal Origins Cohort
examined the association between prenatal PFOA exposure and depression among offspring with
20 years of follow-up. No significant association was observed between clinical depression and
maternal PFOA (3.8 ng/mL) levels.
Two medium confidence studies {Vuong, 2016, 3352166; Vuong, 2018, 5079675} examined the
relationship between PFOA 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 both 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. No associations were observed between prenatal PFOA levels and
executive function {Vuong, 2016, 3352166}. In analyses using childhood (8 years old) serum
PFOA levels {Vuong, 2018, 5079675}, results indicated higher PFOA levels were significantly
associated with increased odds of being at risk of having clinical impairments—specifically for
the metacognition index at age eight.
Six medium confidence studies among the general population {Stram, 2014, 2922190; Liew,
2015, 2851010; Quaak, 2016, 3981464; Skogheim, 2019, 5918847; Lenters, 2019, 5080366},
and one in a high-exposure community {Stein, 2013, 2850964}, examined ADHD and measures
of attention in children. A medium confidence study of participants in the C8 Health Study
observed consistently lower Clinical Confidence Index scores, indicating less probability of
ADHD, on the CCPT-II in children (mean age = 9.9 years) associated with increased estimated
in utero PFOA levels (median = 43.7 ng/mL) and increased measured childhood PFOA
(median = 35.0 ng/mL) {Stein, 2013, 2850964}. Stram et al. (2014, 2922190) investigated the
association between prenatal PFOA exposure and ADHD among offspring (follow-up to age 20)
of participants within the Danish Fetal Origins Cohort. No association between prenatal PFOA
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 PFOA exposures (case median = 4.1 ng/mL; control median = 4.0 ng/mL) and ADHD in
children. No consistent evidence was observed to suggest that prenatal PFOA exposures increase
the risk of ADHD. Quaak et al. (2016, 3981464) explored the relationship between prenatal
PFOA 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 observed
between prenatal PFOA exposures and ADHD scores in the whole population as well as within
C-86
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
the sex-stratified analyses. One medium confidence study {Lenters, 2019, 5080366} examined
early life high PFOA exposures in breast milk in relation to ADHD among children (range: 7.2-
14.1 years old) from the HUMIS and observed positive non-significant associations with odds of
ADHD (OR: 1.35, 95% CI: 0.87, 2.11), but not consistently in various models.
Two low confidence studies {Ode, 2014, 2851245; Lien, 2016, 3860112} examined ADHD and
ADHD-related measures, but no significant associations were observed. Lien et al. (2016,
3860112) evaluated the association between cord blood PFOA (mean =1.6 ng/mL) exposures
and neurobehavioral symptoms related to ADHD among 7-year old participants from the Taiwan
Birth Panel Study and the Taiwan Early-Life Cohort. No significant associations or trends were
observed; however, the direction of association was primarily negative. Ode et al. (2014,
2851245) investigated the association in a case-control study between cord blood PFOA
(median =1.8 ng/mL for cases; 1.83 ng/mL for controls) exposures and ADHD diagnosis in
childhood (age range 5-17 years), but no consistent pattern was observed. Deficiencies identified
in these studies included the reliability of exposure measures, limited study sensitivity, and
potential for residual confounding.
Six medium confidence studies evaluated PFOA exposures in relation to autism, autistic
behaviors, and ID {Braun, 2014, 2345999; Liew, 2015, 2851010; Oulhote, 2016, 3789517;
Long, 2019, 5080602; Lyall, 2018, 4239287; Shin, 2020, 6507470}. A two-fold increase in
serum PFOA (median = 4.06 |ig/L) at age five was associated with significantly higher SDQ
autism screening scores at age seven in a high confidence study {Oulhote, 2016, 3789517}. In a
medium confidence study from the HOME study, increasing maternal serum PFOA
concentrations (median = 5.5 ug/L) were non-significantly associated with fewer autistic
behaviors in children 4 to 5 years of age as assessed by maternal completed Social
Responsiveness Scale (SRS) scores {Braun, 2014, 2345999}. No consistent evidence of an
association between maternal plasma PFOA (median = 3.9 ng/mL for cases; 4.0 ng/mL for
controls) and diagnosed childhood autism was reported in a medium confidence study of mother-
child pairs with an average of ten years of follow-up within the DNBC {Liew, 2015, 2851010}.
No association was observed in a medium confidence case-control study of amniotic fluid PFOA
(median = 0.3 ng/mL for cases; 0.3 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 PFOA (geometric mean = 3.6 ng/mL for ASD cases;
3.3 ng/mL for ID cases; 3.7 ng/mL for controls) was inversely associated with diagnosed ASD
and ID in a medium confidence study of children aged 4.5-9 years {Lyall, 2018, 4239287}. No
significant association was observed in a medium confidence study of modeled prenatal maternal
PFOA (median =1.1 ng/mL for ASD cases; 1.2 ng/mL for controls) and clinically confirmed
ASD among children (age 2-5 years) in the Childhood Autism Risk from Genetics and
Environment (CHARGE) study {Shin, 2020, 6507470}.
The effects on visuospatial performance were evaluated in one high confidence study {Harris,
2018, 4442261} which observed associations, and one medium confidence study {Vuong, 2018,
5079693} which observed no associations. In participants from Project Viva {Harris, 2018,
4442261} observed that children scored consistently lower on visual-motor tests (Wide Range
Assessment of Visual Motor Abilities) with increasing prenatal PFOA exposure. No clear
patterns were observed using early childhood (median age = 3.2 years) test performance, but
significant inverse associations for mid-childhood (median age = 7.7 years) test performance
C-87
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
were observed for the second (4.1-5.6 ng/mL) and fourth ( > 7.7 ng/mL) quartiles of prenatal
PFOA exposure. 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 the General Adult Population
The effects of PFOA 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 a significant association between serum PFOA and performance on tests for
memory and learning corresponding to a 6% higher (better memory and learning) mean score.
Findings of a medium confidence study {Shrestha, 2017, 3981382}, described above, indicated
higher serum PFOA in adults was associated with significantly better performance executive
function measured by the Wisconsin Card Sorting Test (WCST).
Two studies {Berk, 2014, 2713574; Shrestha, 2017, 3981382} examined the effects of PFOA
exposure on depression among adults. No associations were reported in a medium confidence
study of depression, assessed by the Beck Depression Inventory (BDI), and serum PFOA
(median = 8.1 ng/mL) in a cross-sectional study of adults aged 55 to74 years {Shrestha, 2017,
3981382}. One low confidence NHANES study {Berk, 2014, 2713574} observed a lower
prevalence of depression with increasing PFOA exposure as assessed by the nine-item
depression module of the Patient Health Questionnaire (PHQ-9).
Only one medium confidence study {Vuong, 2020, 6356876} examined social-emotional effects
in pregnant women. No evidence was reported to support an adverse relationship between serum
PFOA during pregnancy and maternal depressive symptoms assessed by the Beck Depression
Inventory-II (BDI-II) from pregnancy to 8 years postpartum.
Two medium confidence studies explored the relationship between PFOA and memory
impairment {Gallo, 2013, 2272847; Shrestha, 2017, 3981382} and observed mixed effects. Gallo
et al. (2013, 2272847) observed statistically significant inverse associations with memory
impairment in adults from the C8 Health Project. However, no adverse effects of PFOA on
memory impairment were observed in adults (ages 55-74 years) in New York State {Shrestha,
2017, 3981382}.
Two medium confidence cross-sectional studies investigated PFOA and hearing impairment in
analyses of adult NHANES participants and observed mixed effects. Li (2020, 6833686)
reported significant positive associations between PFOA and hearing impairment, while Ding
and Park (2020, 6711603) reported no significant associations.
C.4.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 3 studies from the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} and 8 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOA HESD that investigated the association between PFOA and nervous effects. Study quality
evaluations for these 11 studies are shown in Figure C-30.
C-88
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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
Butenhoffet al., 2004, 1291063-
++
NR
NR
++
++
+
Butenhoff et al., 2012, 2919192-
+
++
NR
-
+
++
Goulding et al., 2017, 3981400-
+
+
NR
+
+
+
Guo et al., 2019, 5080372-
+
+
NR
++
+
+
Li et al., 2017, 4238518-
+
NR
NR
+
+
+
Macon et al., 2011, 1276151 -
++
++
+
+
+
+
NTP, 2019, 5400977-
++
++
NR
++
++
++
NTP, 2020, 7330145-
++
++
NR
++
++
++
Shi et al., 2020, 7161650-
+
+
-
+
+
+
Yuet al., 2016, 3981487-
+
+
NR
++
+
NR
van Esterik et al., 2015, 2850288 -
++
NR
NR
++
+
-
++ ++ ++
Figure C-30 Summary of Study Evaluation for Toxicology Studies of PFOA and Nervous
Effects
Interactive figure and additional study details available on HAWC.
There are few studies available that evaluate neurotoxicity, including neurodevelopmental
toxicity, with short-term, chronic, or gestational exposure to PFOA in experimental models.
From the available literature, there is little evidence of morphological changes or damage that
can be attributed to PFOA exposure. However, there is some evidence suggesting that PFOA
exposure may be associated with behavioral and physiological effects, areas of research that may
warrant further analysis. Additionally, several single-dose studies indicate that
neurodevelopmental endpoints may be sensitive indicators of PFOA toxicity.
C-89
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Absolute and/or relative brain weights, as well as brain histopathology, were reported in studies
using mice, rats, and monkeys; these studies generally reported null or inconsistent results across
dose-groups, generations, sexes, or studies {Perkins, 2004, 1291118; Yu, 2016, 3981487;
Goldenthal, 1978, 1291068; Yahia, 2010, 1332451; Macon, 2011, 1276151; Butenhoff, 2004,
1291063; Butenhoff, 2012, 2919192}. Statistically significant changes in brain weight were
often not consistent across sexes or generations, were transient, were not dose-dependent, or
occurred at relatively high doses compared to other health outcomes. For example, in a 2-year rat
feeding study, Butenhoff et al. (2012, 2919192) observed significantly increased absolute brain
weights in males from the low dose group (1.3 mg/kg/day) but not the high dose group
(14.2 mg/kg/day) or either female treatment groups. In a rat reproductive study, Butenhoff et al.
(2004, 1291063) observed no change in absolute brain weight in P0 males or females and no
change in females from the Fi generation, but reported a significant decrease in absolute brain
weight in the high-dose Fi males (30 mg/kg/day) at PND 120. Similarly, Macon et al. (2011,
1276151) reported a transient significant decrease in absolute brain weight in Fi male mice
exposed to 1 and 3 mg/kg/day during gestation at PND 63 (time points measured ranged from
PND 7-84). There were no differences in absolute brain weight in females or in relative brain
weight in either sex. However, sample sizes in control females were too low on PNDs 63 and 84
to conduct statistical analysis. Dam mice in the highest dose group reported by Yahia et al.
(2010, 1332451) in a gestational study (10 mg/kg/day) had significantly decreased absolute brain
weight (approximately 7% decrease) and no statistical difference in relative brain weight. A 28-
day study in male mice with doses up to 2.5 mg/kg/day {Yu, 2016, 3981487} and a 13-week
study with interim sacrifices at 4 and 7 weeks in male mice with doses up to 6.5 mg/kg/day
{Perkins, 2004, 1291118} also found no evidence of altered absolute or relative brain weights
after PFOA exposure. One monkey study with a limited sample size (n=2/sex/group) reported
decreased absolute brain weight in females dosed with 10 mg/kg/day PFOA for 90 days (highest
dose tested that did not induce mortality) {Goldenthal, 1978, 1291068}. There were no
significant effects on brain weight in males from the same study. Despite several noted changes
in brain weight, there were no reports of altered brain histopathology due to PFOA exposure in
the available literature {Butenhoff, 2004, 1291063; Yahia, 2010, 1332451; Butenhoff, 2012,
2919192; Li, 2017, 4238518; NTP, 2019, 5400977; NTP, 2020, 7330145}. In a subchronic study
in male C57BL/6J mice, Shi et al. (2020, 7161650) observed increased neuronal apoptosis and
cell shrinkage, though no quantitative data were provided.
Goulding et al. (2017, 3981400) assessed behavioral effects in Fi male offspring gestationally
exposed to 0, 0.1, 0.3, or 1 mg/kg/day PFOA from GD 1-17. The authors conducted different
behavioral assays across multiple periods of development through adulthood (~3 weeks-
6 months of age). Significant effects were only observed in the highest dose group
(1 mg/kg/day). Ambulatory activity in an open-field chamber, reported as the number of
photocell breaks, was measured on PND 18-20. There was a significant increase in the number
of photocell breaks in the 1 mg/kg/day dose group on PND 18, however, this response was not
observed on PND 19 or PND 20. On PND 60, Goulding et al. (2017, 3981400) reported no
significant effects due to PFOA exposures in the auditory startle response, habituation, prepulse
startle inhibition, and running wheel tests. The running wheel assay was repeated at PND 72 with
similar results. On PND 168, mice were monitored for ambulatory activity following an acute
injection of methamphetamine; the authors reported a significantly decreased number of
photocell breaks in the 1 mg/kg/day group compared to controls. A few studies report clinical
C-90
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
signs of toxicity that exhibit neurotoxicity including ataxia in potentially moribund animals
{Goldenthal, 1978, 1291068; Butenhoff, 2012, 2919192}.
Yu et al. (2016, 3981487) analyzed tissue concentrations of four neurotransmitters in the brains
of male mice exposed to 0, 0.5, or 2.5 mg/kg/day PFOA for 28 days. Concentrations of
dopamine, serotonin, and norepinephrine were significantly altered in the 0.5 mg/kg/day dose
group compared to controls but not the high dose group; dopamine and serotonin were both
increased while norepinephrine was decreased. Glutamate concentrations in the 2.5 mg/kg/day
dose group were significantly decreased compared to controls. Guo et al. (2019, 5080372) also
reported a significant reduction in glutamate concentrations in male mice exposed to 10
mg/kg/day PFOA, but not to 0.4 or 2 mg/kg/day, for 28 days.
Several studies reported on additional behavioral and neurochemical effects. Onishchenko et al.
(2011, 758427) and Sobolewski et al. (2014, 2851072) observed behavioral effects including
altered locomotor activity, exploratory behavior, circadian activity, and motor coordination in
mouse offspring following gestational or perinatal exposure to single dose levels of PFOA (0.3
and 0.1 mg/kg/day in the respective studies). Cheng et al. (2013, 2304777) administered 10 ppm
PFOA to pregnant rats from GD 1-PN D21 and similarly observed altered motor coordination
and locomotor activity in male and female offspring. This study did not report drinking water
consumption or body weights of the dams. Johansson et al. (2008, 1276156; 2009, 757874) also
observed behavioral (spontaneous behavior and locomotion) and neurochemical effects (altered
cholinergic system responses and brain enzyme and protein levels) in adult mouse offspring after
a single PFOA dose of either 0.58 or 8.7 mg/kg on PND 10.
C.4.3 Mechanistic Evidence
Mechanistic evidence linking PFOA exposure to adverse nervous outcomes is discussed in
Sections 3.2.4 and 3.4.1 of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279}. There are 21
studies from recent systematic literature search and review efforts conducted after publication of
the 2016 PFOA HESD that investigated the mechanisms of action of PFOA that lead to nervous
effects. A summary of these studies is shown in Figure C-31. Additional mechanistic synthesis
will not be conducted since evidence suggests but is not sufficient to infer that PFOA leads to
nervous effects.
C-91
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mechanistic Pathway
Animal
Human
In Vitro
Grand Total
Big Data, Non-Targeted Analysis
D
0
0
3
Cell Growth, Differentiation, Proliferation, Or Viability
2
0
6
Cell Signaling Or Signal Transduction
0
g
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
2
0
1
3
Hormone Function
1
0
2
3
Inflammation And Immune Response
0
1
0
1
Oxidative Stress
0
0
4
Xenobiotic Metabolism
0
0
1
1
Other
1
0
0
1
Not Applicable/Not Specified/Review Article
0
1
4
Grand Total
12
1
10
21
Figure C-31. Summary of Mechanistic Studies of PFOA and Nervous Effects
Interactive figure and additional study details available on Tableau.
C.4.4 Evidence Integration
There is slight evidence on an association between PFOA exposure and nervous effects in
humans. The epidemiological studies reviewed since the 2016 Health Assessment provide mostly
mixed results on the associations between PFOA and neurological outcomes. There were no new
neurological studies identified that evaluated cerebral palsy. Outcomes investigated include those
of depression, memory impairment, hearing impairment, ASD, and ID.
The recent epidemiological studies provide limited indication of adverse effects of PFOA on
neurodevelopment, neuropsychological development {Goudarzi, 2016, 3981536; Niu, 2019,
5381527}, cognitive development {Harris, 2018, 4442261; Oulhote, 2019, 6316905}, and
executive function {Vuong, 2018, 5079675}. Results for IQ were largely non-significant and
inconsistent. There was no evidence of an association with depression; only two studies observed
effects of PFOA on hearing {Li, 2020, 6833686} and memory impairment {Gallo 2013,
2272847}. Overall, results for neurodevelopmental, neuropsychological, cognitive, and executive
function outcomes were somewhat mixed and limited in number of studies.
The recent epidemiological studies also provide limited indication of adverse effects of PFOA on
behavioral problems, ADHD, ASD, and ID. There was suggestive evidence of an association
between PFOA exposure and behavioral problems associated {Oulhote, 2016, 3789517; Outhote,
2019, 6316905; Ghassabian, 2018, 5080189}; however, overall results were mixed. Of the
multiple studies examining associations between PFOA and ADHD, only one {Lenters, 2019,
5080366} observed associations with PFOA in a high-exposed population. No adverse
associations of ID with PFOA were observed. Oulhote et al. (2016, 3789517) observed a two-
fold increase in serum PFOA at age five was associated with significantly higher SDQ autism
screening scores at age seven, but no associations between PFOA and autism screening scores
were observed in other studies. However, many studies have methodological concerns, as PFOA
exposures in cases and controls within the ADHD and ASD studies were often either similar to
C-92
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
or had mean control exposures greater than cases in many 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 animal evidence for an association between PFOA exposure and neurological effects in
animals is slight. In animal models, some changes in absolute brain weight were noted after
PFOA exposure however, the changes in brain weight were not associated with histopathological
effects. There is limited, but compelling evidence from several single-dose studies indicating
neurodevelopmental consequences of PFOA exposure during perinatal periods, though these
studies cannot be modeled for this assessment due to the exposure paradigm. In a multi-dose
study, Goulding et al. (2017, 3981400) assessed neurodevelopmental consequences of PFOA
exposure, but the observed effects in neonates were transient and therefore, are difficult to
interpret. This study also reported a suppression of ambulatory activity in mice from the high
dose group following an acute injection of methamphetamine on PND 168. The biological
significance of the alterations in neurotransmitter levels observed in a separate study is unclear
{Yu, 2016, 3981487}; however, these effects indicate a potential alteration of neural signaling
and could be an additional outcome related to PFOA neurotoxicity or a potential toxicological
mechanism underlying the observed behavioral changes.
C.4.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOA exposure has the potential to cause nervous system effects
in humans under relevant exposure circumstances (Table C-7). This conclusion is based
primarily on effects on neurodevelopment, neuropsychological and cognitive development,
executive function, and behavioral problem in studies in humans exposed to median PFOA
ranging from 12 to 5.2 ng/mL, and on evidence from animal models showing alterations in
neurodevelopment, neurobehavior, and neurotransmitter levels following exposure to doses as
low as > 0.3 mg/kg/day PFOA. There is considerable uncertainty in the results due to
inconsistency across studies and limited number of studies.
C-93
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-7. Evidence Profile Table for PFOA Nervous System 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.4.1)
Neurodevelopment
1 High confidence study
4 Medium confidence
studies
1 Low confidence study
Findings were mixed «
both across and within
studies, often by sex. A
high confidence study
reported significant
associations with
development problems
for both sexes, but with
different skills. Two
medium confidence
studies reported
significant associations
with developmental
effects, but results were
inconsistent. Significant
inverse associations were
found only in 6-month
neonates in one study and
only in girls in another
study. Remaining studies
did not report consistent
associations.
High and medium
confidence studies
• Low confidence study
• Lnconsistent direction
of effects within and
across studies
• Small magnitude of
effects in significant
associations
Cognitive Function
11 Medium confidence
studies
Cognitive function • Medium confidence
findings were mixed both studies
across and within studies,
often by sex and timing
of exposure measure. Of
11 studies examining
children, studies observed
significant positive
associations with
cognitive function
measures such as reading,
• Lnconsistent direction
of effects within and
across studies
©OO
Slight
Evidence for nervous system
effects is based on high
confidence studies reporting
significant adverse findings,
including for
neurodevelopmental,
behavioral, attention, autism,
and visuospatial outcomes,
which sometimes varied by
sex and direction and
magnitude of effect.
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:
Human evidence indicted
effects on
neurodevelopment,
neuropsychological and
cognitive development,
executive function, and
behavioral problems.
Animal evidence indicated
alterations in
neurodevelopment,
neurobehavior, and
neurotransmitter levels.
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-94
-------
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
full scale IQ, and verbal
ability (4/11), while
others reported
significant inverse
associations (2/11). Other
non-significant results in
these studies were mixed.
The remaining studies
observed inconsistent or
no effects.
Social-emotional and
behavioral regulation
1 High confidence study
4 Medium confidence
studies
1 Low confidence study
Six studies examined •
social-emotional and
behavioral effects in
children, with mixed
results. One high
confidence study
observed significant
associations with
behavioral and peer
relationship problems at
age seven alongside non-
significant mixed
associations for other
behavioral measures. One
medium study reported
significant inverse
associations with
externalizing behaviors in
boys at 18 months.
Another medium
confidence study found
significant positive
associations with total
SDQ scores, indicating
increased behavioral
problems with increased
High and medium
confidence studies
• Low confidence study
• Lnconsistent direction
of effects across and
within studies
C-95
-------
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
exposure. The remaining
studies reported non-
significant, mixed
associations.
Evidence Integration
Summary Judgment
Depression
3 Medium confidence
studies
1 Low confidence study
Two medium confidence • Medium confidence
studies reported results
for depression in general
population adults. An
additional study of
medium confidence
reported results for
depression among
pregnant women
exclusively. All three
studies reported positive
associations, though none
reached significance. A
low confidence study
found an inverse
relationship.
studies
• Low confidence study
• Lnconsistent direction
of effects across
studies
Executive function
3 Medium confidence
studies
Two studies examined
executive function
impacts among children
from the HOME Study.
One study observed
significant associations
with increased odds of
metacognition
impairments, while the
other observed no
associations. In one
medium confidence study
of adults, exposure was
associated with increased
executive function.
• Medium confidence
studies
• Lnconsistent direction
of effects across age
groups and studies in
same cohort
• Limited number of
studies examining
outcome
C-96
-------
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
1 High confidence study
7 Medium confidence
studies
2 Low confidence studies
• High and medium
confidence studies
Studies examining
attention-related effects,
such as ADHD,
inattention, and • Large magnitude of
hyperactivity, occurred in effects
children only. One
medium confidence and
one low confidence study
reported significant
associations, though the
observed effects were in
opposite directions. The
remaining studies
reported no or non-
significant associations.
• Low confidence studies
• Lnconsistent direction
of effects across
studies
Autism, autistic
behaviors, and
intellectual disability
1 High confidence study
5 Medium confidence
studies
Six studies examined
autism-related outcomes
among children. One
high confidence study
observed significant
positive associations
between age 5 exposures
and autism screening
scores at age 7. One
medium confidence study
observed significant
inverse associations with
autism and with
intellectual disability in
the overall study
population. The
remaining medium
confidence studies
reported findings that
were inconsistent.
• High and medium
confidence studies
• Lnconsistent direction
of effects across
studies, ages, and
exposure windows in
study with most
significant association
• Small magnitude of
effect in significant
associations
C-97
-------
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
Visuospatial
performance
1 High confidence study
1 Medium confidence
study
Two studies reported on •
visuospatial performance
in children. One high
confidence study •
observed significant
inverse associations with
visual-motor performance
in mid-childhood but
significant positive
associations with visual-
spatial and visual-motor
performance in early
childhood. The medium
confidence study reported
no significant
associations in
childhood.
High and medium
confidence studies
Large magnitude of
effect
• Inconsistent direction
of effects across studies
and age groups
• Limited number of
studies examining
outcome
Memory impairment
4 Medium confidence
studies
Two studies examined •
memory effects in
children, with one
medium confidence study •
reporting significant
inverse associations with
non-verbal working
memory for the highest
exposure category. Two
studies examined
memory impacts among
adult populations. In one
medium confidence
study, a significant
inverse association with
memory impairment was
reported. The other
medium confidence study
Medium confidence
studies
Large magnitude of
effect
• Inconsistent direction
of effects across
studies
C-98
-------
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 no significant
associations.
Hearing impairment
Two medium confidence • Medium confidence
• Inconsistent direction
2 Medium confidence
studies examined hearing studies
of effects across
studies
impairment among
studies
adults. One study • Large magnitude of
observed significant effect
• Limited number of
positive associations with
studies examining
hearing impairment for
outcome
the highest exposure
group, while the other
reported inconsistent
non-significant
associations.
Evidence from In Vivo Animal Studies (Section C.4.2)
Organ weights
1 High confidence study
3 Medium confidence
studies
Significant effects for
absolute brain weight
were found only in
developmental studies
and only in males. One
developmental study in
mice reported transient
reductions in absolute
brain weight, while a
developmental study in
rats reported decreased
absolute brain weight as
well as decreased body
weight. One chronic
exposure study in rats
found that absolute brain
weight was increased in
only the low-dose group,
and one short-term study
• High and medium •
confidence studies
• Consistent direction of
some findings across •
studies
Incoherence of findings
in other neurological
endpoints
Confounding variables
such as decreases in
body weights
Evidence Integration
Summary Judgment
©oo
Slight
Changes in absolute brain
weight, were noted after
PFOA exposure; however,
the changes in brain weight
were not associated with
histopathological effects.
One study found transient
neurobehavioral effects in
neonates following
developmental exposure and
such findings are difficult to
interpret. The same study
also found neurobehavioral
changes in adulthood. The
biological significance of the
C-99
-------
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
in mice found no
effects.
Histopathology
3 High confidence
studies
1 Medium confidence
studies
No changes in brain
histopathology were
reported in rats (4/4).
• High and medium
confidence studies
• Consistent direction of
effects across studies
• Limited number of
studies examining
outcomes
alterations in
neurotransmitters levels in a
separate study is unclear.
However, these effects
indicate a potential alteration
of neural signaling and could
be an additional outcome
related to PFOA
neurotoxicity or a potential
toxicological mechanism
underlying the observed
behavioral changes.
Neurobehavior
1 Medium confidence
study
A developmental study in# Medium confidence
male mice observed a
transient increase in
locomotor activity level
during the pre-weaning
period and no changes in
startle reactivity or
prepulse inhibition
sm
study
• Limited number of
studies examining
outcome
Neurotransmitters
2 Medium confidence
studies
Two studies observed
alterations of
neurotransmitter
concentrations in male
mice following short-
term PFOA exposure and
observed decrease
glutamate (2/2) and
norepinephrine (1/1) and
an increase in dopamine
(1/1) and serotonin
sm
• Medium confidence
studies
• Consistent direction of
effects across studies
• Limited number of
studies examining
specific outcomes
• Biological significance
of effects is unclear
Notes: ADHD = attention deficit hyperactivity disorder; HOME = Health Outcomes and Measures of the Environment; IQ = intelligence quotient; SDQ = Strengths and
Difficulties Questionnaire.
C-100
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.5 Renal
EPA identified 23 epidemiological and 7 animal studies that investigated the association between
PFOA and renal effects. Of the epidemiological studies, 1 was classified as high confidence, 2 as
medium confidence, 19 as low confidence, and 1 was considered uninformative (Section C.5.1).
Of the animal studies, 3 were classified as high confidence, and 4 were considered medium
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 PFOA Document).
C.5.1 Human Evidence Study Quality Evaluation and
Synthesis
C.5.1.1 In traduction
PFOA has the potential to affect the kidney's function of tubular resorption because of it uses
tubular transporters for excretion and resorption {U.S. EPA, 2016, 3603279}. Biomarkers of
renal function include blood urea nitrogen (BUN), serum creatinine, estimated glomerular
filtration rate (eGFR), and uric acid levels. eGFR is a marker of non-malignant renal disease.
The 2016 HESD for PFOA {U.S. EPA, 2016, 3603279} concluded there was evidence of a
suggestive association between PFOA and two renal outcomes (i.e., uric acid levels and eGFR)
based on one occupational {Costa, 2009, 1429922}, two studies in high-exposed communities
{Steenland, 2010, 1290810; Watkins, 2013, 2850974}, and one general population study
(Shankar, 2011, 2919232). Kidney function was measured by eGFR, hyperuricemia, and uric
acid levels. However, given the cross-sectional study designs, reverse causality as an explanation
could not be ruled out. The report also concluded there was no probable link between PFOA
exposure and kidney disease based on three occupational studies {Steenland, 2015, 2851015;
Steenland, 2012, 2919168; Raleigh, 2014, 2850270}.
For this updated review, 23 studies examined the association between PFOA and renal health
outcomes. Five studies were in children and adolescents {Geiger, 2013, 2919148; Kataria, 2015,
3859835; Qin, 2016, 3981721; Khalil, 2018, 4238547}, two in pregnant women {Nielsen, 2020,
6833687; Gyllenhammer, 2018, 4238300}, one study was in occupational workers {Rotander,
2015, 3859842} and the remainder of the studies were in general population. Seventeen of the
studies utilized a cross-sectional study design; the remaining studies included five cohort study
designs {Blake, 2018, 5080657; Conway, 2018, 5080465; Dhingra, 2016, 3981521;
Gyllenhammer, 2018, 4238300; Nielsen, 2020, 6833687}, and one controlled trial {Convertino,
2018, 5080342} (Appendix D). All studies measured PFOA 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 those studying
populations in the United States, five studies utilized data from the NHANES {Geiger, 2013,
2919148; Jain, 2019, 5080378; Jain, 2019, 5381566; Kataria, 2015, 3859835; Lee, 2020,
6833761; Scinicariello, 2020, 6833670}. Outcomes evaluated in these studies included clinical
conditions, such as chronic kidney disease (CKD) and gout, and biomarkers of renal function,
including uric acid, eGFR, albumin, and creatinine.
C-101
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.S.I.2 Study Quality
Several considerations were specific to evaluating the quality of studies examining kidney
function and kidney disease. Since PFOA is removed from the blood by the kidney, cross-
sectional analyses using serum PFOA as the exposure measure are problematic if individuals
with compromised kidney function are included: PFOA 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 23 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and renal effects. Study quality evaluations for these 23 studies are
shown in Figure C-32.
Of the 23 studies identified since the 2016 assessment, one was classified as high confidence
{Dhingra, 2016, 3981521}, two as medium confidence {Dhingra, 2017, 3981432;
Gyllenhammar, 2018, 4238300}), 19 as low confidence, and one as uninformative {Seo, 2018,
4238334}. The main concerns with the low confidence studies included potential for residual
confounding, selection bias, and reverse causality. Other concerns included small sample sizes
{Khalil, 2018, 4238547; Nielsen, 2020, 6833687}, selective reporting of significant results {Lee,
2020, 6833761}, and potential for selection bias {Lin, 2013, 2850967; Rotander, 2015,
3859842}. Additionally, low confidence studies utilizing cross-sectional analyses of kidney
function with serum PFOA were impacted by the potential for reverse causation.
Seo et al. (2018, 4238334) was considered uninformative due to use of bivariate statistical
analyses, limiting the ability to interpret the results. Additionally, other potential sources of bias
were identified, including a lack of information on participant recruitment and selection,
unexplained discrepancies in sample sizes, and missing details on outcome assessment methods.
C-102
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
vM® J**
^ <3>&Cy>6<
Arrebola et al., 2019, 5080503
Blake et al., 2018, 5080657 -
Chen etal., 2019. 5387400-
Convertino et al., 2018, 5080342 -
Conway et al., 2018, 5080465 -
Dhingra et al., 2016, 3981521 -
Dhingra et al., 2017, 3981432 -
Geiger et al.,2013, 2919148
Gyllenhammar et al., 2018, 4238300
Jain and Ducatman, 2019, 5080378
Jain and Ducatman, 2019, 5381566-
Kataria et al., 2015, 3859835-
Khalil etal., 2018, 4238547-
Lee etal., 2020, 6833761 -
Lin etal., 2013, 2850967-
Liu etal., 2018, 4238514
Nielsen et al., 2020, 6833687 -
Qin etal., 2016, 3981721-
Rotander et al., 2015, 3859842 -
Scinicariello et al., 2020, 6833670 -
Seo etal., 2018, 4238334-
Wang et al.,2019, 5080583-
Zeng et al.,2019, 5918630-
B
B
++
B
++
++
++
EE
19
++
B
¦
++
9
++
B
++
B
B
B
++
++
11
B
B
++
a
B
B
B
++
B
B
B
B
B
Bi
I
n
++
++
B 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-32. Summary of Study Evaluation for Epidemiology Studies of PFOA and Renal
Effects
Interactive figure and additional study details available on HAWC.
C-103
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.S.I.3 Findings in Children and Adolescents
Three low confidence studies examined 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). Geiger et al. (2013, 2919148) used NHANES data from 1999-2000 and
2003-2008 to assess the association between serum PFOA levels and hyperuricemia in children
aged 12 to 18 years. A statistically significant positive association was observed between
increasing quartiles of PFOA and hyperuricemia (p-trend = 0.0071), and serum uric acid (p-
trend = 0.0001). An overlapping NHANES (2003-2010) study {Kataria, 2015, 3859835} also
observed a significant positive association for uric acid for the highest quartile of PFOA
exposure (> 4.7 ng/mL) compared to the lowest (< 2.5 ng/mL). Qin et al. (2016, 3981721)
reported significant positive associations with uric acid and hyperuricemia in children aged 12 to
15 years from the GBCA in Taiwan. Positive associations were observed when the highest
compared to the lowest PFOA quartiles. When stratified by sex, the associations were only
evident among boys, including an increasing trend (p-trend = 0.033) {Qin, 2016, 3981721}.
One low confidence study {Kataria, 2015, 3859835} reported on GFRs among children (12-
19 years old) from NHANES (2003-2010). A negative association was reported between PFOA
and eGFR, where the fourth quartile was associated with a statistically significant decrease in
eGFR compared to the lowest exposure quartile, and the second and third quartiles showed a
non-significant decrease.
Two low confidence studies investigated associations between PFOA and serum creatinine
among children and adolescents {Khalil, 2018, 4238547; Kataria, 2015, 3859835}. Kataria et al.
(2015, 3859835) reported a significant positive association with serum creatinine in the highest
PFOA quartile when compared with the lowest quartile. Khalil et al. (2018, 4238547) observed
weak, non-significant negative association with serum creatinine in obese children (8-12 years).
C.S.I.4 Findings from the General Adult Population
Three studies examined CKD and no significant associations were observed {Conway, 2018,
5080465; Dhingra, 2016, 3981521; Wang, 2019, 5080583}. CKD was defined as an eGFR
of < 60 mL/min/1.73 m2. A high confidence C8 Health Project community study {Dhingra,
2016, 3981521} observed positive non-significant increases in the risk of CKD in both
retrospective and prospective analyses, and among non-diabetic participants. In retrospective
analyses, the magnitude of effect was diminished and inconsistent when modeling exposure
using increasing lag periods (5-, 10-, and 20-year lag). In contrast, negative associations were
observed in two low confidence studies {Wang, 2019, 5080583; Conway, 2018, 5080465}.
Analyses of participants in the Isomers of C8 Health Project in China {Wang, 2019, 5080583}
observed a significant negative association with odds of CKD. Analyses of diabetic individuals
in the U.S.-based C8 Health Project {Conway, 2018, 5080465} also showed significantly
reduced odds, but this effect was not observed in non-diabetic participants. However, a concern
for reverse causality makes interpretation of the results difficult in both low confidence studies.
Gout was examined in one low confidence study {Scinicariello, 2020, 6833670} on adults from
NHANES (2009-2014) and a significant increased trend in risk of self-reported gout across
PFOA quartiles was observed (p-value = 0.01). The observed effects were similar when
stratifying by CKD status.
C-104
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Seven low confidence general population studies {Arrebola, 2019, 5080503; Chen, 2019,
5387400; Lin, 2013, 2850967; Scinicariello, 2020, 6833670; Seo, 2018, 4238334; Zeng, 2019,
5918630; Jain, 2019, 5080378} and one low confidence occupational study {Rotander, 2015,
3859842} examined uric acid levels, and three of these studies reported specifically on
hyperuricemia {Arrebola, 2019, 5080503; Scinicariello, 2020, 6833670; Zeng, 2019, 5918630}.
Significant findings were found in three studies, indicating a positive association with uric acid
or increased odds of hyperuricemia, while non-significant positive associations were observed
for uric acid in three general population confidence studies and one occupational study.
A low confidence NHANES (2009-2014) study {Scinicariello, 2020, 6833670} on adults
reported a significant positive association between serum PFOA and serum uric acid in quartile
analyses, and the trend was significant (p-trend = 0.0001). The association remained when
restricted to participants without CKD, but the association was not consistent among those with
CKD. Analyses of hyperuricemia were similar. A significant increasing trend in the odds of
hyperuricemia was observed among the whole sample and those without CKD. Similarly, a
positive association with serum uric acid was observed in a low confidence study on participants
from the Isomers of C8 Health Project {Zeng, 2019, 5918630}. In addition, a significant positive
association was observed for hyperuricemia and total-PFOA exposure {Zeng, 2019, 5918630}.
Results were similar among men and women in sex-stratified analyses. Utilizing NHANES data
from 2007-2014, a low confidence study {Jain, 2019, 5080378} assessed the associations
between serum PFOA and uric acid across gender and stages of GF. For males, serum PFOA and
uric acid were positively associated (p < 0.01) at stage GF-1 and GF-2 and negatively associated
(p < 0.01) at stage GF-3B/4. For females, all associations were positive but only reached
significance for GF-1 and GF-3A. Two low confidence study {Chen, 2019, 5387400; Lin, 2013,
2850967} did not observe associations with plasma uric acid in Croatian adults aged 44-
56 years, or in adolescents and young adults aged 12 to 30 years in the Young Taiwanese Cohort
Study. A low confidence study {Arrebola, 2019, 5080503} from the BIO AMBIENT.ES study
observed a non-significant increase in risk of hyperuricemia.
One low confidence occupational study examined serum uric acid levels among firefighters with
past exposure to AFFF {Rotander, 2015, 3859842}. Uric acid levels were elevated with
increasing PFOA exposure in firefighters, but the result did not reach significance.
One medium and two low confidence studies in high exposed populations examined eGFR, and
two studies reported negative associations {Blake, 2018, 5080657; Dhingra, 2017, 3981432},
while one reported a positive association {Wang, 2019, 5080583}. Dhingra et al. (2017,
3981432) reported a significant negative association with measured but not modelled PFOA and
a negative trend in eGFR across measured serum PFOA quintiles in women from the Women
from C8 Science Panel Project. The study used modelled PFOA as an approach to demonstrate
that cross-sectional analyses using measured PFOA are affected by reverse causation {Dhingra,
2017, 3981432}. Blake et al. (2018, 5080657) observed negative non-significant associations in
participants of the Fernald Community Cohort (FCC) with high exposure to PFAS from their
household water supplies. Wang et al. (2019, 5080583) observed positive associations in a high-
exposed population from the Isomers of C8 Health Project.
The evidence on PFOA and renal effects among pregnant women was limited. Only two studies
on pregnant women examined effects on eGFR {Nielsen, 2020, 6833687; Gyllenhammer, 2018,
4238300}. One medium confidence study {Gyllenhammer, 2018, 4238300} assessed the
C-105
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
relationship between maternal PFOA during pregnancy and maternal eGFR three weeks after
delivery, calculated using both creatinine- and cystatin C-based estimates of GFR. A significant
positive relationship between cystatin C-GFR and maternal PFOA was reported
(P = 0.004 ± 0.002, p = 0.022). Changes in kidney function during pregnancy were evaluated in a
small group of pregnant women (n = 73) using creatinine-GFR and cystatin C-GFR in a low
confidence study (Nielsen, 2020, 6833687), but no significant effects were observed using partial
Spearman rank correlations. While the medium confidence study in pregnant women reported a
positive association between PFOA and eGFR {Gyllenhammer, 2018, 4238300}, given the
limited number of studies, there is not enough evidence to determine conclusive associations
between PFOA renal function among pregnant women and an occupational group of firefighters.
Four low confidence studies examined albumin and creatinine as biomarkers for renal function
{Convertino, 2018, 5080342; Chen, 2019, 5387400; Jain, 2019, 5381566; Lee, 2020, 6833761}.
The four studies provided differing conclusions. Jain and Ducatman (2019, 5381566) reported
statistically significant positive with serum and urine creatinine, and serum albumin in NHANES
(2005-2014) participants. Statistically significant negative associations were observed 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 PFOA, stratification by stages of GF had
inconsistent effects. However, Lee et al. (2020, 6833761) observed a decreased risk of
albuminuria (defined as urine albumin-to-creatinine ratio > 30 mg/g) Chen et al. (2019, 5387400)
did not observe significant associations with plasma creatinine. Convertino et al. (2018,
5080342) did not observe any associations with serum creatinine during a phase 1 controlled trial
assessing the chemotherapeutic potential of APFO.
One low confidence study {Liu, 2018 4238514} examined serum proteins among NHANES
(2013-2014) participants and reported a significant positive association using linear PFOA
exposure levels. The result was similar for total PFOA but did not reach significance.
C.5.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 2 studies from the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} and 5 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOA HESD that investigated the association between PFOA and renal effects. Study quality
evaluations for these 7 studies are shown in Figure C-33.
C-106
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
_l I
Blake et al., 2020, 6305864 -
+
+
++
+
+
++
B
++
B
Butenhoff etal., 2004, 1291063-
++
NR
NR
++
++
B
++
++ ++
++
Butenhoff et al., 2012, 2919192-
+
++
NR
-
+
++
++
++
B
Guo et al., 2019, 5080372-
+
+
NR
++
B
B
++
B
B
NTP, 2019, 5400977-
++
++
NR
++
++
++
B
++ ++
++
NTP, 2020, 7330145-
++
++
NR
++
++
++
++
++ ++
++
Shi et al., 2020, 7161650-
+
+
+
B
B
B
++
++*
B
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-33. Summary of Study Evaluation for Toxicology Studies of PFOA and Renal
Effects
Interactive figure and additional study details available on HAWC.
The available data suggest the renal system may be adversely affected by PFOA exposure, but
the evidence primarily comes from studies conducted in rats. Two studies in mice {Blake, 2020,
6305864; Shi, 2020, 7161650} and one study in monkeys {Butenhoff, 2002, 1276161} reported
no effects on the renal system. In contrast, several short-term and chronic studies reported
significant increases in absolute and/or relative kidney weights in rats {Cui, 2009, 757868; NTP,
2019, 5400977; Butenhoff, 2004, 1291063; Butenhoff, 2012, 2919192; NTP, 2020, 7330145}
and/or alterations in serum biomarkers of renal function {Cui, 2009, 757868; NTP, 2019,
5400977; NTP, 2020, 7330145; Guo, 2019, 5080372}. However, only two studies reported
concurring histological changes in the kidney {Cui, 2009, 757868; NTP, 2020, 7330145}.
Effects on kidney weight were predominately observed in male rats rather than female rats,
regardless of study design and exposure duration (Figure C-34 (absolute kidney weight), Figure
C-35 (relative kidney weight in males), Figure C-36 (relative kidney weight in females)). This is
true of both absolute and relative kidney weight metrics. However, across both sexes, several
studies observed statistically significant decreases in absolute kidney weight at the highest doses
tested (Figure C-34), which often corresponded to doses resulting in reduced body weight (See
C-107
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOA Main Document and Section C.3.2). These changes in body weight may influence the
interpretation of absolute and relative kidney weight changes.
Endpolnt Study Name Study Design Observation Time Animal Description
Kidney Weight. Absolute Butenhoffetal.. 2012.2919192 chronic <2y) 2y Rat. Sprague-Dawley Cri:Cd(Sd)(Br) (•?. N=15) <
Kidney Weight. Left, Absolute Butenhoffetal.. 2004.1291063 reproductive (64d) 106d P0 Rat, Crl:CD(SD)IGS BR N=29-30)
reproductive (GD0-PND120) PND120 F1 Rat, Crl:CD(SD)IGS BR N=29-30)
Kidney Weight. Right. Absolute Butenhoffetal.. 2004.1291063 reproductive (64d) 106d P0 Rat, Crl:CD(SD)IGS BR N=29-30)
reproductive (GD0-PND120) PND120 F1 Rat, Crl:CD(SD)IGS BR (o. N=29-30)
NTP. 2019, 5400977 short-term (28d) 29d Rat, Sprague-Dawley N=10)
NTP. 2020,7330145 chronic (GD6-PNW21) 16wk F1 Rat, Sprague-Dawley N=10) <
chronic (GD6-PNW107) 16wk F1 Rat, Sprague-Dawley (•', N=10) <
chronic (PND21-PNW21) 16wk F1 Rat, Sprague-Dawley {-¦ , N=10) <
chronic (PND21-PNW107) 16wk F1 Rat, Sprague-Dawley (-f, N=10)
Kidney Weight. Absolute Blake et al.. 2020, 6305864 developmental (GD1.5-11.5) GD11.5 P0 Mouse. CD-1 (i, N=6)
developmental (GD1.5-17.5) GD17.5 P0 Mouse, CD-1 d. N=6)
Butenhoffetal.. 2012.2919192 chronic (2y) 2y Rat. Sprague-Dawley Cri:Cd(Sd)(Br) (-, N=15) <
Kidney Weight. Left, Absolute Butenhoffetal.. 2004.1291063 reproductive (84d) LD22 P0 Rat, Crl:CD(SD)IGS BR U-, N =26-29)
reproductive (GD1-PND106) LD22 F1 Rat, Crl:CD(SD)IGS BR (^. N=27-29)
Kidney Weight. Right. Absolute Butenhoff et al.. 2004.1291063 reproductive (84d) LD22 P0 Rat, Crl:CD(SD)lGS BR (-, N=26-29)
reproductive (GD1-PND106) LD22 F1 Rat, Crl:CD(SD)IGS BR (^. N=28-29)
NTP. 2019,5400977 short-term <28d) 29d Rat. Sprague-Dawley (2. N=9-10) <
NTP. 2020, 7330145 chronic (GD6-PNW107) 16wk F1 Rat, Sprague-Dawley N=10) <
chronic (PND21-PNW107) 16wk F1 Rat, Sprague-Dawley IN=10) -
001 o!l 1 10 100
Concentration (mg/kg/day)
PFOA Renal Effects - Absolute Kidney Weight
I 0 No significant change A Significant increase Significant decrease
AAA
Figure C-34. Absolute Kidney Weights in Rodents Following Exposure to PFOA
(logarithmic scale)
PFOA 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; Po = parental generation; Fi = first generation; PND = postnatal day; PNW = postnatal week; d = day;
wk = week; y = year.
NTP (2019, 5400977) observed dose-dependent increases in the absolute and relative kidney
weights of male Sprague Dawley rats treated with PFOA for 28 days. Absolute and relative
kidney weights were increased in all treated groups (doses of 0.625-10 mg/kg/day), though the
increase in absolute weight was only significant for the three middle dose groups (1.25, 2.5, and
5 mg/kg/day). The highest dose group (10 mg/kg/day) resulted in the largest increase in relative
kidney weight of approximately 36% control weight. The lack of a clear dose-response trend in
absolute kidney weights was likely related to decreased body weights observed at doses
>2.5 mg/kg/day. Despite the increases observed in kidney weights, there were no significant
histological changes observed in the kidneys of PFOA-treated rats {NTP, 2019, 5400977}. Cui
et al. (2009, 757868) similarly observed increased relative kidney weights in male rats
administered 5 or 20 mg/kg/day for 28 days, though the increases were not dose-dependent
(absolute weights were not reported); however, histological changes were observed in the
kidneys of the high-dose group, including turbidness and tumefaction in the epithelia of the
proximal convoluted tubule (reported qualitatively without incidence data).
A similar trend in kidney weight was observed for male rats in a two-generation reproduction
study {Butenhoff, 2004, 1291063}. Adult Po and Fi males had significantly increased absolute
kidney weights at 1, 3, and 10 mg/kg/day, but decreased kidney weights at the highest dose level
of 30 mg/kg/day. Relative kidney weights were significantly increased in all treated males
(increases of 16%-27% and 11%—19% change in Po and Fi males, respectively). Kidney weights
relative to brain weights were increased at 1, 3, and 10 mg/kg/day, but not 30 mg/kg/day. In the
high-dose male group, absolute and relative kidney weight changes occurred in a pattern
C-108
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
typically associated with decrements in body weight. However, in the lower-dose groups of
males, significant increases in absolute kidney weight and relative-to-body and brain weights
appear to be treatment-related and are consistent with the results reported for male rats in the 28-
day study by NTP (2019, 5400977). Increased kidney weights observed following exposure to
PFOA may be a response to the challenge of providing transporters for renal removal of the
foreign molecule {U.S. EPA, 2016, 3603279}. Increased kidney weight can be regarded as an
adaptive response to the transport challenge. It is beneficial for the individual but adverse in the
sense that it signifies the need to upregulate tubular transporters in the kidney to excrete the
foreign material and a reflection of PFOA bioaccumulation in serum and tissues. Butenhoff et al.
(2004, 1291063) did not report conducting kidney histopathology in this reproductive study.
Two chronic dietary studies in Sprague Dawley rats evaluated effects on the renal system, but the
results were not consistent across studies. Butenhoff et al. (2012, 2919192) observed increased
relative kidney weight in male rats administered 300 ppm in the diet (equivalent to 14.2
mg/kg/day) after one year of exposure, but no changes in absolute or relative kidney weight or
histopathology were observed after two years of exposure. In contrast, a two-year study by NTP
(2020, 7330145) observed altered kidney weights and increased incidences of nonneoplastic
lesions in the kidneys of male rats exposed to postweaning dietary concentrations of 20, 40, 80,
150, or 300 ppm with or without perinatal exposure to 150 or 300 ppm (See Main PFOA
Document for study design details). At the 16-week interim evaluation, absolute kidney weights
were increased in males of the 0/20 and 300/20 ppm groups (perinatal/postweaning
concentrations, equivalent to postweaning doses of 1.1, and 1.0 mg/kg/day, respectively) and
decreased in males of the 0/300 and 300/300 ppm groups (31.7 and 32.1 mg/kg/day,
respectively), but not significantly altered compared to controls in any of the intermediate dose
groups. However, relative kidney weights were significantly increased in all treated groups
(range of 21%—35% increases across all groups); body weights were also significantly reduced in
all treatment groups (dose-dependent range of 9%-45% decreases across all groups).
Substantially reduced body weights in treated males makes interpretation of kidney weight
effects difficult.
C-109
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
End point
Kidney Weight, Relative
Study Name Study Design Observation Time
Shietal.. 2020,7161650 subclironic (5wk) Swk I
Animal Description
5. C57BU6J N=8>
PFOA Renal Effects - Relative Kidney Weight in Males
Dose (mg/kg/day) | @ Statistically significant # Not statistically significant I—195% CI |
Butanhoff at al., 2012. 2919192 chronic (2y)
Idney Weight, Left, Relative Butanhoff at
i ©productive (64d)
Rat, Sprague-Dawlay Cri:Cd(SdXBr)( ;.', N=15) l
P0 Rat, Crl:CD(SD)IGS BR ( , N=29-30)
reproductive (GD0-PND120) PND120 F1 Rat, Cri:CD(SD)IGS BR N=29-30)
Kidney Weight, Right, Relative
Butanhoff at al„ 2004, 12810(33
reproductive (64d)
reproductive (GDO-PND120)
106d
PND120
PO Rat, Crl:CD
0
reproductive (GD0-PND120)
PND120
F1 Rat. Crl;CD(SD)IGS BR (T. N=29-30)
30
0
Kidney Weight, Right. Relative to Brain
Butenhoff et al.. 2004, 1291063
reproductive (64IGS BR (¦ , N=29-30>
10
30
0
1
10
30
reproductive (GD0-PND120)
F1 Rat. Crl:CD(SD)IGS BR fcf. N=29-30)
0
3
30
-•-I
—•—I
10 20 30 40 50
Figure C-35. Percent Change In Relative Kidney Weights of Male Rats Following Exposure
to PFOA
Interactive figure and additional study details available on HAWC.
GD = gestation day; Po = parental generation; Fi = first generation; PND = postnatal day; PNW = postnatal week; d = day;
wk = week; y = year; CI = confidence interval.
C-110
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Female rats were generally less sensitive to changes in kidney weights compared to male rats,
with most differences occurring in the highest dose groups only (Figure C-35, Figure C-36). NTP
(2019, 5400977) observed dose-dependent increases in absolute and relative kidney weights of
female rats treated with PFOA for 28 days. Absolute kidney weight was only increased at the
highest dose of 100 mg/kg/day (11% increase) while relative kidney weight was increased at 50
and 100 mg/kg/day (7% and 17% increases, respectively). Similar to males from this study, there
were no significant histological changes observed in the kidneys of PFOA-treated rats {NTP,
2019, 5400977}. In contrast, in a two-generation reproduction study {Butenhoff, 2004,
1291063}, absolute and relative kidney weights of Po females were significantly decreased at
30 mg/kg/day (decreases of approximately 5%-8% change), and no effects were observed on
kidney weight in Fi females. There were no significant effects on the body weight of these
animals at terminal sacrifice.
Butenhoff et al. (2012, 2919192) observed an increase in absolute kidney weight (11% change)
in female rats administered 30 but not 300 ppm PFOA in the diet for two years (equivalent to 1.6
and 16.1 mg/kg/day, respectively). In contrast, the authors reported a significant increase in
relative kidney weights of female rats administered 300 but not 30 ppm (15% change). That dose
group also experienced an approximately 12% decrease in body weight by the time of terminal
sacrifice, but the change was not statistically significant. The authors reported no change in renal
histopathology in female rats {Butenhoff, 2012, 2919192}. A second two-year feeding study by
NTP (2020, 7330145) found alterations in absolute kidney weight and increased incidences of
nonneoplastic lesions in the kidneys of female rats exposed to postweaning dietary
concentrations of 300 or 1,000 ppm with or without perinatal exposure to 300 ppm (See Main
PFOA Document for study design details). At the 16-week interim evaluation, absolute kidney
weights were decreased in females of the 0/1,000 ppm and 300/1,000 ppm groups (equivalent to
63.4 and 63.5 mg/kg/day postweaning doses); however, relative kidney weights were unaltered
in females. Body weights were significantly reduced in females exposed to 1,000 ppm
postweaning (by 12%). Decreased absolute kidney weights observed in females exposed to
1,000 ppm were likely related to reduced body weights as there was no change in relative kidney
weight.
C-lll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOA Renal Effects - Relative Kidney Weight in Females
Endpoint
Kidney Weight, Relative
Study M ante Study Design Observation l ime Animal Description
Blake et al„ 2020,6305864 developmental <001.511.5) GDI 1.5 TO Mouse. CD-1 (9. N=6)
etopmental (01)1.5-17.51 GIN 7.
Dose (mg/kg/day) ^^^^^tatiMieaii^ignifican^^^^o^misticaii^ignificam^^^S^^r
buffet »l„ 2012,2910102 chronic
;y Weight. Left, Relative Butcnhoft'ct al„ 2004, 1291063 reproductive (84dj LD22
reproductive (GIJO-PXDI06) l.l>22
NTP. 2019, 5400977
(28d) 29d
Kidney Weiglit, Relative to Hrain Kntenhoffct nl., 2012, 29191
Kidney Weigh I, I .eft. Relutiv
Rat. Sprague-Dawley Cri:Cd(Sd)fBr) I?. N=IS)
a. Crl:CD(SD)IGS BR (9, N=26 29)
Rat,Cri:('JXSI))IG.S KR (9, N=27-29)
Kidney Weight.Right. Relative Batenhoflet «L. 2004,1291063 reproductive lS4d) LD22 TORat.Crl:CD(SD)IGS BR (9. N=26-2V
reproductive(GD0-PKDI06) I.D22 Fl Rat, OtCDtSDlTGS BR l9: N=28-29)
Rat. Spraguc Dawley (9. N=9 10)
NTP. 2020,7330145 chronic (GD6-PNW107) 16wk Fl Rat. Spraeue-Dawley (5,K=10l 0
chronic (PND21 PNW107) 16wk Fl Rat.Sprague Dawley (9. N=10l
Rat, Spragnc-Dawley fri:Cd(SdXBr> (9, N=15)
ilcnhorfelal.,20(M. 1291063 reproductive f84d) I.D22 P0 Rat, Crl:CD(SDlKiS BR (9, N=26-29) 0
reproductive tGDO PND106) LD22 F! Rat.Crl:CD(SD)IGSBR l9, N=27 2'
Kidney Weight, Right, Relative to Brain Butenhoff el al., 2004. 1291063 reproductive f84d> I.D22 TO Rat, Cri:CD(SDlIGS BR 19- N=26-29>
it. Cri;CD(SD)IGS BR (9, !">
-15 -10 -5 0 5 10 15 20 25 30
Figure C-36. Percent Change in Relative Kidney Weights of Female Rodents Following
Exposure to PFOA
Interactive figure and additional study details available on HAWC.
GD = gestation day; Po = parental generation; Fi = first generation; PND = postnatal day; PNW = postnatal week; d = day;
wk = week; y = year; CI = confidence interval.
Histopathological examination of male rats at the 16-week interim of a 2-year dietary study
showed increased incidences of renal tubule mineralization in the 0/150, 0/300, and 300/300 ppm
C-112
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
groups compared to the 0/0 ppm control group (incidences of 40%, 50%, and 60%, respectively,
compared to 0% incidence in the control group) {NTP, 2020, 7330145}. No other significant
histological changes were observed in males, and the male groups were removed from that study
shortly after the interim. However, examination of female rats revealed treatment-related
increased incidences of renal tubule mineralization, hyperplasia of the urothelium that lines the
renal papilla, and necrosis of the renal papilla (that was observed only after 2 years). As shown
in Table C-8, these lesions were mainly found in the female groups with the highest postweaning
exposure (1,000 ppm, equivalent to approximately 63 mg/kg/day).
Table C-8. Incidences of Nonneoplastic Lesions in the Kidneys of Female Sprague-Dawley
Rats as Reported by NTP (2020, 7330145)
Postweaning Dose
Perinatal Dose
0 ppm 300 ppm 1,000 ppm
16 Weeks
Renal Tubule, Mineralization
0 ppm 2/10 (20%) (1.0)a
150 ppm -
300 ppm -
Renal Papilla Urothelium, Hyperplasia
0 ppm 0/10 (0%)
150 ppm -
300 ppm -
Renal Papilla, Necrosis
0 ppm 0/10 (0%)
150 ppm -
300 ppm -
107 Weeks
Renal Tubule, Mineralization
0 ppm 5/50 (10%) (1.2)
150 ppm -
300 ppm -
Renal Papilla Urothelium, Hyperplasia
0 ppm 4/50 (8%) (1.0)
150 ppm -
300 ppm -
Renal Papilla, Necrosis
0 ppm 0/50 (0%)
150 ppm -
300 ppm -
Notes:
* Statistically significant at p < 0.05; ** p < 0.01.
a Average severity grade of lesion in affected animals: 1 = minimal; 2 = mild; 3 = moderate; 4 = marked
In a second similar study conducted by NTP in male rats only due to high mortality in the initial
study, relative kidney weights of all groups exposed to postweaning dietary concentrations of 20,
C-113
1/10 (10%) (1.0)
2/10 (20%) (1.0)
0/10 (0%)
0/10 (0%)
0/10 (0%)
0/10 (0%)
7/10* (70%) (1.0)
5/10 (50%) (1.2)
4/10* (40%) (1.3)
4/10* (40%) (1.0)
0/10 (0%)
0/10 (0%)
6/50 (12%) (1.3) 16/50** (32%) (1.0)
8/50 (16%) (1.0)
8/50 (16%) (1.5)
21/50** (42%) (1.0) 40/50** (80%) (1.9)
8/50 (16%) (1.0)
45/50** (90%) (1.8)
0/50 (0%) 12/50** (24%) (2.3)
0/50 (0%)
22/50** (44%) (2.1)
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
40, or 80 ppm (equivalent to approximately 1, 2, or 4.6 mg/kg/day) for 16 weeks were
significantly greater than the 0/0 ppm control group, but absolute kidney weights were
significantly increased only in the groups exposed to 20 ppm postweaning {NTP, 2020,
7330145}. Body weights were significantly decreased in all treated groups (by 9%-21%), and
that could explain why absolute kidney weights did not achieve statistical significance in the
higher dose groups in these growing rats. These patterns in kidney weights are similar to those
observed for male rats in the studies by NTP (2019, 5400977) and Butenhoff et al. (2004,
1291063). There were no significant histological changes in the kidneys for male rats found at
the interim or two-year terminal evaluations.
In contrast to results found in studies with rats, no treatment-related effects were reported for
relative kidney weight in male mice administered PFOA for 5 weeks {Shi, 2020, 7161650},
kidney weight and histopathology in female mice administered PFOA during gestation {Blake,
2020, 6305864}, or kidney weight and histopathology in male monkeys administered PFOA for
6 months by oral capsule {Butenhoff, 2002, 1276161}. One short-term study in rats {NTP, 2019,
5400977} and three chronic studies in rats or monkeys also examined the urinary bladder for
histopathology after exposure to PFOA, and no treatment-related effects were reported
{Butenhoff, 2012, 2919192; NTP, 2020, 7330145; Butenhoff, 2002, 1276161}.
Several studies analyzed clinical chemistry and urinalysis endpoints related to renal toxicity,
though there is uncertainty regarding adversity of the observed effects. In two separate studies,
NTP observed increased concentrations of BUN in male and female rats following 28 days or
16 weeks of exposure {NTP, 2019, 5400977; NTP, 2020, 7330145}. However, without
concomitant increases in blood creatinine concentrations, NTP concluded that the slight
increases in urea nitrogen were likely due to a decrease in water intake {NTP, 2019, 5400977;
NTP, 2020, 7330145}. In fact, creatinine concentrations were significantly decreased in male
rats administered > 0.625 mg/kg/day in the 28-day study, though NTP considered this change to
be related to decreased food intake and body weight rather than a direct treatment effect {NTP,
2019, 5400977}.
Butenhoff et al. (2012, 2919192) also observed slight increases in BUN in male and female rats,
but only at the 3- and 6-month evaluations of the 2-year study; creatinine was not measured. No
significant differences were observed in serum BUN, serum creatinine, or urinary creatinine in
female mice administered PFOA during gestation {Blake, 2020, 6305864} or in male monkeys
administered PFOA for 6 months {Butenhoff, 2002, 1276161}. However, a 28-day study in male
mice found significant, dose-dependent decreases in BUN and increases in serum ammonia
levels in all treated groups (0.4-10 mg/kg/day) compared to controls; the authors of this study
suggest these changes are signs of urea cycle dysfunction caused by PFOA {Guo, 2019,
5080372}.
Two studies found that the activity of creatine kinase was decreased in male rats administered
PFOA for 28 days or up to 2 years {NTP, 2019, 5400977; Butenhoff, 2012, 2919192}. NTP
considered this effect to be treatment-related but not toxicologically relevant {NTP, 2019,
5400977}. No effects on creatine kinase were observed in male or female rats at the 16-week
interim evaluation of the NTP chronic dietary study {NTP, 2020, 7330145}.
No apparent treatment-related effects were observed on urinalysis endpoints (e.g., volume, pH,
specific gravity, protein, blood) measured in male or female rats over the course of two years of
C-114
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
treatment {Butenhoff, 2012, 2919192} or in male monkeys over the course of 6 months of
treatment {Butenhoff, 2002, 1276161}.
C.5.3 Mechanistic Evidence
Mechanistic evidence linking PFOA exposure to adverse renal outcomes is discussed in Sections
3.1.1.4, 3.2.5, 3.3.4, and 3.4.3 of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279}. There are
4 studies from recent systematic literature search and review efforts conducted after publication
of the 2016 PFOA HESD that investigated the mechanisms of action of PFOA that lead to renal
effects. A summary of these studies is shown in Figure C-37. Additional mechanistic synthesis
will not be conducted since evidence suggests but is not sufficient to infer that PFOA leads to
renal effects.
Mechanistic Pathway Animal In Vitro Grand Total
Big Data, Non-Targeted Analysis
1
0
1
Cell Growth, Differentiation, Proliferation, Or Viability
1
0
1
Cell Signaling Or Signal Transduction
2
1
3
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
1
1
2
Grand Total
3
1
4
Figure C-37. Summary of Mechanistic Studies of PFOA 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 PFOA exposure and renal effects in humans
based on mixed evidence of decreased renal function. The 2016 HESD for PFOA {U.S. EPA,
2016, 3603279} concluded there was evidence of an association between PFOA and two renal
outcomes (i.e., uric acid levels and eGFR) based on one occupational study {Costa, 2009,
1429922}, two studies in higher exposed communities {Steenland, 2010, 1290810; Watkins,
2013, 2850974}, and one general population study {Shankar, 2011, 2919232}. In this updated
review, there was some evidence of associations with decreased kidney function, although
reverse causality (i.e., increases in serum perfluoroalkyl levels could be due to a decrease in
glomerular filtration and shared renal transporters for perfluoroalkyls and uric acid) cannot be
ruled out. There were mixed results across the measures of renal function. A positive association
was observed for CKD in a high confidence study in a C8 Health Project population including
non-diabetics {Dhingra, 2016, 3981521}; while two low confidence studies reported negative
associations {Wang, 2019, 5080583; Conway, 2018, 5080465}. The results were also
inconsistent when assessing eGFR, in three highly exposed population studies, with two
reporting negative associations {Blake, 2018, 5080657; Dhingra, 2017, 3981432} and one
positive association {Wang, 2019, 5080583}. Regarding hyperuricemia and uric acid levels,
results varied across gender and stages of GF. In children, there were mixed results for
associations between PFOA and creatinine and uric acid. One low confidence study reported a
C-115
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
statistically significant decrease in eGFR in adolescents across PFOA quartiles {Kataria, 2015,
3859835}.
The animal evidence for an association between PFOA exposure and renal toxicity is slight
based on 7 high or medium confidence animal studies that suggests the kidney can be a target of
PFOA toxicity, although changes in kidney weight or histopathology have only been observed in
rats. Clinical chemistry and urinalysis endpoints do not provide strong evidence of damage to
kidney structure or function; however, kidney weights, particularly in male rats, were
significantly increased following short-term and chronic exposure. The observed increases in
kidney weights may indicate an adaptive response that is adverse in the sense that it signifies the
need to upregulate tubular transporters in the kidney to excrete the foreign material and is a
reflection of PFOA bioaccumulation in serum and tissues. However, kidney weights appear to be
heavily influenced by changes in body weight which impacts the ability to interpret and model
these responses.
Studies in animals generally found no histological changes correlating with increased kidney
weight. The NTP chronic study {NTP, 2020, 7330145} in rats provides the most convincing
evidence that the kidney can be damaged by exposure to PFOA, although the doses with effects
observed were relatively high (approximately 18 and 63 mg/kg/day in females and 16 and
32 mg/kg/day in males). Renal lesions were mainly observed in treated females, except for
increased tubule mineralization which was observed in both sexes. Cui et al. (2009, 757868) also
observed kidney damage in male rats treated with 20 mg/kg/day PFOA for 28 days, but the
incidences of specific lesions were not reported. The mechanisms of this kidney damage are
unknown, but it may be related to direct cytotoxicity from the high concentration of PFOA in the
urine {NTP, 2020, 7330145}.
C.5.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOA exposure has the potential to cause renal effects in
humans under relevant exposure circumstances (Table C-9). This conclusion is based primarily
on effects on measures of kidney function observed in studies in humans exposed to median
PFOA ranging from 3.5 to 11.9 ng/mL, and on evidence in rats showing increased kidney
weights and renal lesions following exposure to doses as low as 1 mg/kg/day and 16 mg/kg/day
PFOA, respectively. Although there is some evidence of negative effects of PFOA 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-116
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-9. Evidence Profile Table for PFOA 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.5.1)
Uric acid
11 Low confidence
studies
Studies in children
observed significant
increases in uric acid
(3/3) and hyperuricemia
(2/2) with increasing
exposure to PFOA. In
studies of adults,
significant increases were
observed in studies of the
general population (3/7),
while non-significant
increases were reported
in other general
population studies (2/7)
and an occupational
study (1/1). Significant
increases in the odds of
hyperuricemia were also
observed (2/7) in adults.
• Consistent direction of • Low confidence studies
effect among children and
adults
Serum and urinary
biomarkers
7 Low confidence
studies
Increases in serum • No factors noted
albumin were observed in
adults (2/2), but urinary
albumin was observed to
be decreased (1/1).
Significant increases in
serum creatinine (1/1)
were observed in adults,
along with increased
urinary creatinine (1/1),
leading to a decreased
albumin-creatinine ratio.
Results for urinary total
protein and urea were not
• Low confidence studies
• Limited number of studies
examining outcome
• Lncoherence of findings
related to serum and urine
albumin levels
©OO
Slight
Several studies of
medium and low
confidence found
evidence of decreased
kidney function among
children and adults,
including increased uric
acid and hyperuricemia
and decreased eGFR.
Overall, findings were
inconsistent, with
opposing directions of
effect observed for some
outcomes. Uncertainties
remain due to the mixed
_results, limited studies
evaluating albumin, gout,
and proteins, and
concerns about reverse
causality in lower
confidence studies.
©OO
Evidence Suggests
Primary basis:
Human evidence indicted
effects on kidney function
and animal evidence
indicated increased
kidney weight and renal
lesions in rats. Although
there is some evidence of
negative effects of PFOA
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-117
-------
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
consistent (2/2). A
limited number of studies
evaluated effects in
children, and one (1/2)
observed increases in
serum creatinine at the
highest levels of
Glomerular filtration
Results for GFR were • Medium confidence
• Low confidence studies
rate
mixed. One study in studies
2 Medium confidence
children (1/1) reported a
• Lnconsistent direction of
studies
significant decrease in
effect in studies of adults
4 Low confidence
eGFR at the highest
studies
exposure level. In adults
decreases in eGFR were
observed in two studies
(2/3), and a significant
increase in eGFR was
observed in one study
(1/3). In studies of
pregnant women, a
positive association with
GFR was observed (1/2).
Chronic kidney Three studies examined •
disease CKD in adults who were
1 High confidence study both diabetic and non-
2 Low confidence diabetic. The high
studies confidence study reported
non-significant increased
odds of CKD. The two
low confidence studies
found significant
decreases in CKD (2/2),
with one of those results
reported for diabetic
adults (1/3).
High confidence study • Low confidence studies
• Lnconsistent direction of
effect across studies, which
may be due to reverse
causality in low confidence
studies
• Lmprecision of findings
C-118
-------
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
Gout Significantly increased • No factors noted
1 Low confidence study odds of self-reported gout
were observed in
NHANES adults (1/1) at
higher levels of exposure.
The association remained
in analyses stratified by
CKD status.
• Low confidence studies
• Limited number of studies
examining outcome
• Potential outcome
misclassification due to
self-reported outcome
Evidence from In Vivo Animal Studies (C.5.2)
Kidney weight
3 High confidence
studies
3 Medium confidence
studies
Kidney weights were <
significantly changed
following short-term and
chronic exposure in
several studies,
particularly in male rats;
however, concurrent
decreases in body weight
may have influenced
results. No effects on
absolute or relative
kidney weight were
reported in studies in
mice (2/2). Absolute
kidney weight in male
rats was increased at
lower doses and
decreased at higher doses
following PFOA
exposure (3/4). Absolute
kidney weight in female
rats was either increased
(2/4) or decreased (2/4).
Changes in relative
kidney weight were also
observed in rats. For
High and medium
confidence studies
• Lnconsistent direction of
results
©OO
Slight
Evidence Integration
Summary Judgment
1 Changes in body weight Evidence was based on 7
may limit ability to interpret/?/g/? and medium
these responses confidence studies.
Kidney weights,
» particularly in male rats,
were changed following
short-term and chronic
exposure. Most studies
found no histological
changes correlating with
increased kidney weight,
but one chronic study
provides convincing
evidence that the kidney
can be damaged by
exposure to PFOA. Renal
lesions were mainly
observed in exposed
females, except for
increased tubule
mineralization which was
observed in both sexes.
C-119
-------
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 rats, only increases
in relative kidney weight
were observed (3/4). For
female rats, increases
(2/4) and decreases (1/4)
were observed.
Histopathology
2 High confidence
studies
2 Medium confidence
studies
Most studies found no
histopathological changes
in the kidneys of treated
animals (3/4), including
one developmental study
in mice, one short-term
study in rats, and one
chronic study in rats.
However, one high
confidence chronic study
found evidence of kidney
damage in male and
female rats following
PFOA exposure.
Increased hyperplasia and
necrosis of the renal
papilla were observed in
female rats. Increased
renal tubule
mineralization was noted
in both sexes.
• High and
medium confidence
studies
noted
Clinical chemistry and
urinalysis endpoints do
not provide strong
evidence of damage to
kidney structure or
function. Changes in
No factors clinical chemistry
parameters such as
increased serum 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.
Serum biomarkers
2 High confidence
studies
3 Medium confidence
studies
Changes in serum BUN
were observed in several
studies (3/5); however,
increases in BUN may be
contributed to decreased
water consumption. A
decrease in serum
creatinine was observed
(1/3) but may be
• High and
medium confidence
studies
• Incoherence
of findings in serum
biomarkers of renal
function
• Changes in
water consumption,
food intake, and
body weight may
limit ability to
C-120
-------
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
attributed to decreased
food intake and body
weight. Decreased serum
creatine kinase (2/3) and
increased serum
ammonia (1/1) were also
noted.
interpret these
responses
Urinalysis
2 Medium confidence
studies
One study in rats
measured several urinary
endpoints at different
timepoints over two years
of exposure to PFOA and
found no exposure-
related changes. No
changes in urinary
creatinine were observed
in mice exposed to PFOA
during gestation.
• Medium
confidence studies
• Limited
number of studies
examining outcome
Notes: BUN = blood urea nitrogen; CKD = chronic kidney disease; eGFR = estimated glomerular filtration rate; GFR = glomerular filtration rate; NHANES = National Health and
Nutrition Examination Survey.
C-121
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.6 Hematological
EPA identified 8 epidemiological and 3 animal studies that investigated the association between
PFOA 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, and 2 were considered medium 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 PFOA Document).
C. 6.1 Human E vide nee Study Quality E valuation and
Synthesis
C.6.1.1 In traduction
The mechanisms for PFOA 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}. PFOA 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 PFOA 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
PFOA {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 PFOA 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 PFOA and
hematological health outcomes (Figure C-38). 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 PFOA using biomarkers in blood. Samples were taken from
pregnant women, children, adolescents, or adults. Most included studies were cross-sectional,
meaning exposures and outcomes were evaluated during the same period. 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, the NHANES {Etzel, 2019, 5043582; Jain, 2020,
6333438; Jain, 2020, 6833623}. Etzel et al. (2019, 5043582) (N = 7,040) used 2003-2010
NHANES data for adolescents and adults 12 and over {Etzel, 2019, 5043582}, and Jain (2020,
6333438) (N= 11,251) and Jain (2020, 6833623) (N = 10,644), used 2003-2016 NHANES data
for adults 20 years and older {Jain, 2020, 6333438; Jain, 2020, 6833623}. Also in the United
States, Khalil et al. (2018, 4238547) used data on 48 obese children at 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
C-122
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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. Convertino et al. (2018, 5080342) conducted a six-week trial
with experimental exposure to APFO among late-stage cancer patients at two medical centers in
Glasgow and Dundee, Scotland.
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 socioeconomic. In particular, the
duration of breastfeeding is expected to be associated with both PFOA exposure and nutrition
intake {Abraham, 2020, 6506041}. The blood matrix (whole blood vs. plasma or serum) could
also affect the interpretation of results. Measuring PFOA and serum lipids concurrently was
considered adequate in terms of exposure assessment timing. Given the long half-life of PFOA
(median half-life = 2.7 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 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and hematological effects. Study quality evaluations for these 8
studies are shown in Figure C-38.
Based on the considerations mentioned, three studies were classified as medium confidence, two
as low confidence, and three as uninformative. The low confidence had deficiencies in
confounding and limited sample sizes. Convertino et al. (2018, 5080342) did not control for
confounding, although this concern is somewhat attenuated by the prospective trial study design
wherein investigators manipulated the exposure levels. Khalil et al. (2018, 4238547) was
affected by a small sample size, and potential residual confounding attributable to differences in
participants' socioeconomic status (SES). Three studies were rated as uninformative for
hematological outcomes. For Jain (2020, 6833623), the use of PFOA 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-123
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
e\ec
,Ge
o^° Co^ ^ c,e\eu <=,\»OT 0s1®
Abraham et al., 2020, 6506041 -
I
+
I
+
++
--
I
I
+
i
+
~
Chen et al., 2019, 5387400-
+
+
++
+
+
+
+
+
Convertino et al., 2018, 5080342 -
+
++
+
-
+
+
+
-
Etzel et al., 2019, 5043582-
++
+
+
+
++
+
+
+
Jain, 2020, 6333438-
+
+
++
+
-
+
+
+
Jain, 2020, 6833623-
+
+
+
+
+
+
+
~
Jiang et al., 2014, 2850910-
-
++
+
¦
-
+
-
~
Khalil et al., 2018, 4238547-
-
+
+
+
+
-
~
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-38. Summary of Study Evaluation for Epidemiology Studies of PFOA 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 or vitamin D deficiency and observed no
associations. In adolescents and adults from NHANES (2003-2010), Etzel et al. (2019, 5043582)
observed non-significant positive prevalence ORs for vitamin D deficiency and decreases in
levels 25-hydroxy vitamin D pre doubling of PFOA. A low confidence study, Khalil et al. (2018,
4238547) observed a non-significant positive association between PFOA exposure and 25-
hydroxy vitamin D levels in 8-12-year old United States children.
In adults from NHANES (2003-2016), Jain (2020, 6333438) observed small statistically
significant increases in whole blood hemoglobin (WBHGB) levels (Appendix D). This was true
C-124
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
for participants with or without anemia, and the magnitude of the association was larger among
those anemics. For example, associations (slopes) between PFOA and WBHGB for anemic
females were more than five times higher as compared to nonanemic females (beta = 0.03413 vs.
0.00605). 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 females, association between WBHGB and PFOA
concentrations were uniformly positive across worsening stages of renal failure. For anemic
males, association between WBHGB and PFOA concentrations were positive except at GF-3A
(45 < eGFR < 60 mL/min/1.73 m2). Overall, the association between WBHGB and PFOA
followed U-shaped distributions. Hemoglobin levels were also examined in pregnant women in
Jiang et al. (2014, 2850910). Significant positive correlations were observed between total PFOA
and hemoglobin levels (r = 0.192, p < 0.05) and albumin (r = 0.251 ,p<0 .01), although these
results did not consider the influence of confounding factors and should be interpreted with
caution.
Chen et al. (2019, 5387400) observed non-significant decreases in serum calcium levels among
Croatian adults.
Several markers of liver function blood clotting tests were examined in a phase 1 dose-
calculation trial using APFO. In this low confidence study, Convertino et al. (2018, 5080342),
observed no clear differences in plotted probabilistic fibrinogen, prothrombin time (PPT), or
activated partial thromboplastin time (aPPT) at various PFOA concentrations.
C.6.2 Animal Evidence Study Quality Evaluation and
Synthesis
There is 1 study from the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} and 2 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOA HESD that investigated the association between PFOA and hematological effects. Study
quality evaluations for these 3 studies are shown in Figure C-39.
C-125
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
,Ge
Butenhoff et al., 2012, 2919192-
+
++
l
NR
I
I
+
Guo et al., 2021, 7542749-
+
+
NR
+
+
NTP, 2019, 5400977-
++
++
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
++
B
Figure C-39. Summary of Study Evaluation for Toxicology Studies of PFOA 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. Five oral studies in rodents or monkeys with short-term to chronic exposure durations
evaluated the effects of PFOA on the hematological system. Significant changes in some
erythron parameters following PFOA exposure to rats at dose levels as low as 0.625 mg/kg/day
{NTP, 2019, 5400977} and increases in aPPT and PPT in monkeys exposed to 30 mg/kg/day
{Butenhoff, 2002, 1276161} suggest the potential for the hematological system to be a target of
PFOA toxicity.
In a 28-day study, significant decreases in erythrocyte count, hematocrit, and hemoglobin
(> 1.25 mg/kg/day), reticulocytes (> 0.625 mg/kg/day), and mean cell volume (10 mg/kg/day)
were observed in male Sprague Dawley rats (Figure C-40) {NTP, 2019, 5400977}; however, the
majority of these effects, except reticulocyte counts, were within 10% of control levels.
Significant decreases in erythrocyte count (100 mg/kg/day), hematocrit (> 6.25 mg/kg/day), and
hemoglobin (> 12.5 mg/kg/day) were observed in female rats from the same 28-day study, but
the effects were also within 10% of control levels (Figure C-40) {NTP, 2019, 5400977}.
Loveless et al. (2008, 988599) administered PFOA to male Sprague-Dawley rats or male CD-I
mice at dose levels 0, 0.3, 1, 10, or 30 mg/kg/day for 29 days. In rats, hemoglobin and hematocrit
were significantly decreased (91%-92% of control) at 10 and 30 mg/kg/day and a significant
increase in reticulocytes (197% of control) was observed with 30 mg/kg/day. No other altered
hematological effects were reported in rats or mice, though there was a slight increase in
granulocytic bone marrow hyperplasia in mice dosed with 10 or 30 mg/kg/day.
C-126
-------
DRAFT FOR PUBLIC COMMENT
NTP, 2019, 5400977 Hematological Effects (Percent control response)
Endpoint Animal Description Dose (mg/kgfday) | © Statistically signilicanl © Not slalisUmlly significant I—¦ 95% CI |
Erythrocytes Rat. Sprague-Dawley (<¦'. N=10) 0
H
K©
H
0.625
1.25
h©-
2.5
h©
5
h©-
10
h©--
Rat, Sprague-Dawley (2. N=9-10) 0
•
6.25
©
12.5
i
25
•
50
•
100
o
Hematocrit (HCT) Rat. Sprague-Dawley (<"<'. N=10) 0
~
0.625
•
1.25
I©
2.5
h©
5
b©
10
-©(
Rat. Sprague-Dawley (2. N=9-10) O
t
6.25
o
12.5
o
25
o
50
o
100
o
Hemoglobin (HGB) Rat. Sprague-Dawley (r . N=10) 0
«
~
0.625
•
1.25
©
2.5
0
5
1©
10
Rat. Sprague-Dawley (2. N=9-10) 0
©
<
~
6.25
•
12.5
o
25
o
50
o
100
o
Manual Hematocrit (HCT) Rat, Sprague-Dawley (..', N=10) 0
HH
0.625
©
1.25
0
2.5
o
5
t-©H
10
o
Rat. Sprague-Dawley (2. N=9-10) 0
©
6.25
©
12.5
•
25
©
50
~
100
o
Mean Cell Volume (MCV) Rat, Sprague-Dawley (,.:', N=10) 0
©
0.625
»
1.25
1
I
2.5
i
5
•
10
0
Rat. Sprague-Dawley (2. N=9-10) 0
«
»
6.25
1
1
12.5
•
25
50
1
100
I
!
Reticulocytes Rat, Sprague-Dawley ( i , N=10) 0
y—i
0.625
© 1
1.25
»—•—1
2.5
1—©—1
5
1 ©
10
1—•—1
6.25
1—
1 Q
25
50
100
—•—
0 -40 -30 -20 -10
10 20 30 40 50 60
Persian! control response {%)
Figure C-40. Hematological Effects in Male and Female Sprague Dawley Rats Dosed with
PFOA for 28 Days as Reported by NTP (2019, 5400977)
Interactive figure and additional study details available on HAWC.
HCT = hematocrit; HGB = hemoglobin; MCV = mean cell volume; CI = confidence interval.
MARCH 2023
C-127
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Dietary administration of 30 or 300 ppm PFOA (equivalent to 1.3 or 14.2 mg/kg/day in males
and 1.6 or 16.1 mg/kg/day in females) to male and female Sprague-Dawley rats for 2 years
produced mild or inconsistent effects on hematology {Butenhoff, 2012, 2919192}. The authors
provided data on red blood cell counts, hemoglobin, and hematocrit at 3, 6, 12, 18, and
24 months, though only time points prior to 52 weeks are considered as clinical pathology testing
in aging rodents may be affected by naturally occurring disease {Weingand, 1992, 670731}. In
males, Butenhoff et al. (2012, 2919192) reported significant decreases in red blood cell counts in
both dose groups at 6 months and in the 14.2 mg/kg/day group at 12 months. These decreases did
not exceed 10% change from controls. Similarly, the authors reported significant decreases in
hematocrit in both dose groups at 3 months and with 14.2 mg/kg/day at 12 months, but these
changes also did not exceed 10% difference from controls. There was no observed effect on
hemoglobin levels at any time point. In females, significant changes were often noted in the
1.6 mg/kg/day dose group but not the 16.1 mg/kg/day group. For example, minimal decreases in
hemoglobin, hematocrit, and red blood cell counts were observed at 6 months in the
1.6 mg/kg/day group but not the high dose group. Dose-dependent decreases in these three
parameters were observed at the 12-month time point, though the magnitude of change did not
exceed 10% difference from controls. Discussions on other parameters related to immune system
function from this study are provided in (see PFOA Main Document).
In a 28-day study, significant decreases in serum levels of hemoglobin, bilirubin, platelets, and
iron were observed in 6 to 8-week old mice exposed to PFOA (0.4 - 10 mg/kg/day) via oral
gavage {Guo, 2021, 7542749}. Dose dependent reductions in platelets were significantly
reduced in animals by day 7 (> 2mg/kg/day) and all treatment groups by day 28. Reductions in
hemoglobin were measured as early as day 7 in the highest dose tested (10 mg/kg/day), but by
day 21 all exposure groups (> 0.4 mg/kg/day) experienced depletion. This decrease in
hemoglobin also correlated to a dose dependent reduction in serum iron content and significantly
elevated bilirubin (10 mg/kg/day). Guo et al. (2021, 7542749) considered that reductions in
hemoglobin, iron, and platelets and elevation of bilirubin are consistent with the pathophysiology
of anemia.
In a 90-day study with rhesus monkeys, significant increases in aPPT and PPT were observed at
30 mg/kg/day at 1-month analyses {Goldenthal, 1978, 1291068}; at 3 months, the same effects
were seen in the lone surviving monkey at 30 mg/kg/day (early mortality at the high dose level
of 100 mg/kg/day precluded hematological analyses). A 182-day oral (capsule) study in male
cynomolgus monkeys reported no hematological findings at dose levels up to 20 mg/kg/day
{Butenhoff, 2002, 1276161}.
C 6.3 Mechanistic Evidence
There was no mechanistic evidence linking PFOA exposure to adverse hematological outcomes
in the 2016 PFOA HESD {U.S. EPA, 2016, 3603279}. There are 4 studies from recent
systematic literature search and review efforts conducted after publication of the 2016 PFOA
HESD that investigated the mechanisms of action of PFOA that lead to hematological effects. A
summary of these studies is shown in Figure C-41. Additional mechanistic synthesis will not be
conducted since evidence is inadequate to infer that PFOA leads to hematological effects.
C-128
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mechanistic Pathway
Human
In Vitro
Grand Total
Atherogenesis And Clot Formation
Oxidative Stress
Cell Growth, Differentiation, Proliferation, Or Viability
Big Data, Non-Targeted Analysis
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
2
1
1
1
1
Grand Total
3
4
Figure C-41. Summary of Mechanistic Studies of PFOA and Hematological Effects
Interactive figure and additional study details available on Tableau.
The evidence evaluating an association between PFOA exposure and hematological effects in
humans is considered indeterminate based on limited number of studies and inconsistent and
non-significant findings. Many of the outcomes were not studied in more than one study, making
coherence hard to establish. Two studies that examined 25-hydroxy vitamin D levels reported
mixed non-significant effects. There is evidence of an association between increased PFOA and
slightly increased WBHGB, particularly among anemic adults from a large NHANES study
{Jain, 2020, 6333438}. Increases in hemoglobin and albumin may also affect pregnant women
{Jiang 2014, 2850910}. However, it is unclear whether the observed changes are clinically
adverse.
The animal evidence for potential hematological effects is indeterminate. There is limited data
on the hematological system being a target for PFOA in animal models, inconsistent results
between sexes and species, and generally minimal effects observed (within 10% of the control).
In the four studies that reported effects on red blood cells in rats {NTP, 2019, 5400977;
Loveless, 2008, 988599; Butenhoff, 2012, 2919192; Guo, 2021, 7542749}, results were all
within 10% of the controls except for the decrease in reticulocytes observed in male rats in NTP
(2019, 5400977).
C.6.4.1 Evidence Integration Judgment
Overall, there is inadequate evidence to assess whether PFOA exposure can cause hematological
effects in humans under relevant exposure circumstances (Table C-10).
C.6.4 Evidence Integration
C-129
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-10. Evidence Profile Table for PFOA 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 examined
changes in serum 25-
hydroxy vitamin D.
Results in both children
and adults were
inconsistent across
exposure groups and
largely imprecise.
• Medium confidence
study
• Inconsistent direction of
effects across studies
• Low confidence study
• Imprecision of findings
• Limited number of studies
examining outcome
Anemia and whole
blood hemoglobin
(WBHGB)
1 Medium confidence
study
• Medium confidence
study
One study observed
significant associations
with increased WBGHB
particularly among anemic • Consistent direction of
adults. findings across
subpopulations
• Limited number of studies
examining outcome
OOO
Indeterminate
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
1 Medium confidence
study
One study observed a non- • Medium confidence
significant inverse study
association with serum
calcium concentrations.
• Limited number of studies
examining outcome
Liver function blood
clotting
1 Low confidence study
Associations with
concentrations of
probabilistic fibrinogen,
PPT, and aPTT were
imprecise.
• No factors noted
• Low confidence study
• Limited number of studies
examining outcome
Evidence from In Vivo Animal Studies (Section C.6.2)
Complete blood count In a chronic and short- <
1 High confidence study term exposure study,
2 Medium confidence decreased hematocrit
studies levels (2/2) were observed ¦
in male and female rats
but this includes transient
High and medium
confidence studies
Dose-dependent
response
• Inconsistent direction of
effects across studies
OOO
Indeterminate
• Limited number of studiesEvidence was limited and
examining outcome inconsistent with direction
of effect for hematological
C-130
-------
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
effects at only the 3 month
timepoint in the chronic
study (1/1). One short-
term study in rats reported
a dose-response decrease
in hematocrit in males but
not females (1/1). Most
studies found exposure
associated decreases in
hemoglobin (2/3) after 28
days in males (2/3) and
females (1/2). RBC was
decreased (2/2) in rats of
both sexes at the highest
dose in an short-term
study (1/1), and at the 6-
month timepoints in a
chronic study (1/1). One
short-term study in rats
found decreased mean cell
volume in males only at
the highest dose tested.
One study in male mice
found a dose-dependent
decrease in platelets
following a 28-day
exposure to PFOA.
endpoints in animal
models.
Serum iron
1 Medium confidence
study
One 28-day study in male • Medium confidence
mice observed a dose- study
dependent decrease in
serum iron levels (1/1). • Dose-dependent
response
• Limited number of studies
examining outcome
Notes: aPTT = activated partial thromboplastin time; PPT = prothrombin time; RBC = red blood cell; WBHGB = whole blood hemoglobin.
C-131
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.7 Respiratory
EPA identified 5 epidemiological and 4 animal studies that investigated the association between
PFOA and respiratory effects. Of the epidemiological studies, all 5 were classified as medium
confidence (Section C.7.1). Of the animal studies, 2 were classified as high confidence, and 2
were considered medium 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 PFOA 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 or sensitivity.
In the 2016 Health Assessment for PFOA {U.S. EPA, 2016, 3603279}, no epidemiological
evidence on pulmonary function was available; the C8 Science Panel concluded there was no
probable link between PFOA exposure and respiratory health effects (e.g., chronic obstructive
pulmonary disease (COPD)) {C8 Science Panel, 2012, 1430770}.
For this updated review, six new epidemiologic studies investigated the association between
PFOA and respiratory effects: five studies targeting the general population reported on several
lung function outcomes, and one occupational study examined COPD {Steenland, 2015,
2851015} (Appendix D). All studies measured PFOA 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
conducted in Taiwan {Qin, 2017, 3869265}; one was a cross-sectional study of adolescents and
young adults residing near the WTC {Gaylord, 2019, 5080201}, and one was an occupational
cohort study of workers and former workers at a chemical plant in West Virginia {Steenland,
2015, 2851015}. Five studies examined lung function measures in children and young adults,
including forced expiratory volume in one second (FEV1), forced vital capacity (FVC) and
FEV1/FVC ratio, forced expiratory flow at 25%-75% (FEF 25%-75%), peak expiratory flow
rate (PEF) measured, lung volume and resistance at oscillation frequencies of 5 Hz or 20Hz, lung
function at birth and severity of obstructive airways disease.
Studies that examined respiratory illnesses or symptoms reflecting immune system responses
(e.g., asthma and allergies) and respiratory tract infections (e.g., cough) are discussed in the Main
PFOA Document.
C-132
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C. 7.1.2 Study Quality
There are 5 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and respiratory effects. Study quality evaluations for these 5 studies
are shown in Figure C-42.
All five studies identified since the last assessment were classified as medium confidence. The
medium confidence 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}.
vo#1®.AS*6'
Agier et al., 2019, 5043613-
I
+
I
+
++
I
+
I
+
l
I
+
+
Gaylord et al., 2019, 5080201 -
+
+
+
-
+
+
+
+
Impinen et al., 2018, 4238440 -
+
++
+
+
+
-
+
-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-42. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Respiratory Effects
Interactive figure and additional study details available on HAWC.
C. 7.1.3 Findings from 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, all reported negative associations (i.e., decrease in FEV1
with higher PFOA levels). In children ages 6-12 years, Agier et al. (2019, 5043613) reported
C-133
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
statistically significant associations with prenatal exposure (beta per log2 increase PFOA = -1.4,
95% CI: -2.7, -0.1), but not for postnatal exposure. Qin et al. (2017, 3869265) observed
statistically significant associations for children ages 10-15 years with asthma (beta per In
increase PFOA = -0.10, 95% CI: -0.19, -0.02), and in boys with asthma, but not in girls with
asthma. There was also a significantly decreasing trend by quartiles of PFOA in children with
asthma (p-trend = 0.002), but not observed in children without asthma. Negative non-significant
associations were observed in two of the four studies {Manzano-Salgado, 2019, 5412076;
Gay lord, 2019, 5080201}.
For other lung function measures examined, there was limited evidence of associations.
Manzano-Salgado et al. (2019, 5412076) reported a statistically significant association between
maternal PFOA concentrations and FVC at age 4 (beta = -0.17, 95% CI: -0.34, -0.01), but not
for FVC at age 7 or for other measures of lung function, at either age 4 or age 7. Qin et al. (2017,
3869265) observed statistically significant associations for FEF25-75% (beta = -0.223, 95% CI:
-0.4, -0.045) and a significant decreasing trend with quartiles of PFOA (p-value = 0.014) in
children with asthma, but not for FVC or PEF or for any lung function measures in children
without asthma. Impinen et al. (2018, 4238440) reported a statistically significant association
between prenatal PFOA exposure and 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 PFOA = 1.43, 95% CI: 1.03, 1.98). The study also reported a non-significant increase in
odds of reduced lung function at birth, as measured by tidal flow volume. Other lung function
measures (i.e., FVC, FVC/FEV1, lung resistance, total lung capacity, functional residual
capacity, and residual volume) in adolescents and young adults residing near the WTC were all
inversely associated with PFOA exposure, but none were significant {Gaylord, 2019, 5080201}.
C.7.1.4 Findings from the General Adult Population
One occupational cohort study {Steenland, 2015, 2851015} assessed incidence of COPD and
cumulative PFOA exposure in adult workers and former workers at a chemical plant in West
Virginia. The study observed a non-significant increased risk of COPD in no-lag models, but no
clear pattern of association in 10-year lag models.
C 7.2 Animal Evidence Study Quality Evaluation and
Synthesis
There are 2 studies from the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} and 2 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOA HESD that investigated the association between PFOA and respiratory effects. Study
quality evaluations for these 4 studies are shown in Figure C-43.
C-134
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Butenhoffetal., 2012, 2919192-
NTP, 2019, 5400977-
NTP, 2020, 7330145-
Perkins etal.,2004, 1291118-
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-43. Summary of Study Evaluation for Toxicology Studies of PFOA and
Respiratory Effects
Interactive figure and additional study details available on HAWC.
There is evidence suggesting oral PFOA exposure may adversely affect the nasal, olfactory, and
pulmonary systems, though the database examining respiratory toxicity is generally limited.
Adverse histopathological effects in the lung and nose were observed in short-term and chronic
studies in adult rats {Cui, 2009, 757868; Butenhoff, 2012, 2919192; NTP, 2019, 5400977}.
However, several other studies, including two chronic toxicity studies in rats and one
developmental toxicity study in mice, did not report treatment-related alterations in the
respiratory system of adults or neonates after treatment with PFOA {Perkins, 2004, 1291118;
Yahia, 2010, 1332451; NTP, 2020, 7330145}.
In a 2-year rat feeding study, Butenhoff et al. (2012, 2919192) observed significantly increased
incidences of alveolar macrophages and pulmonary hemorrhage in males in the high-dose group
(300 ppm, equivalent to 14.2 mg/kg/day) (Table C-l 1). However, the incidences of perivascular
mononuclear infiltrate and interstitial pneumonia were decreased in both exposure groups.
Incidence of perivascular mononuclear infiltrate was also reduced in females, though only in the
low-dose group (1.6 mg/kg/day, 4% incidence compared to 26% in controls). There was also a
significant increase in the incidence of lung vascular mineralization in females, though this was
again observed only in the low-dose group (44%, 76%, and 52% incidence in the 0, 1.6, and
16.1 mg/kg/day groups, respectively). Altered lung histopathology in males was considered a co-
critical effect for this study in derivation of candidate RfDs for PFOA {U.S. EPA, 2016,
3603279}, though Butenhoff et al. (2012, 2919192) questioned whether these effects were
directly related to PFOA treatment. Two additional chronic dietary studies in rats found no
treatment-related effects on lung weight or histopathology {Perkins, 2004, 1291118; NTP, 2020,
C-135
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
7330145}. NTP (2020, 7330145) reported significant effects on lung weight in males and
females that were considered secondary to decreased body weight and not direct toxicological
effects of PFOA.
Table C-ll. Incidences of Non-Neoplastic Pulmonary Lesions in Male Rats as Reported by
Butenhoff et al. (2012, 2919192)
Dose
Pulmonary Lesion
0 ppm
(0 mg/kg/day)
30 ppm
(1.3 mg/kg/day)
300 ppm
(14.2 mg/kg/day)
Alveolar Macrophages
10/49 (20%)
16/50 (32%)
31/49 (63%)*
Hemorrhage
10/49 (20%)
14/49 (29%)
22/50 (44%)*
Vascular Mineralization
43/49 (88%)
43/49 (88%)
47/50 (94%)
Perivascular Mononuclear
21/49 (43%)
3/49 (6%)*
7/50 (14%)*
Infiltrate
Interstitial Pneumonia
16/49 (33%)
5/49 (10%)*
7/50 (14%)
Notes:
* Statistically significant atp < 0.05.
Cui et al. (2009, 757868) observed pulmonary congestion and focal or diffuse thickening of
epithelial walls in the lungs of male rats gavaged with 5 or 20 mg/kg/day PFOA for 28 days
(incidence data not provided). While NTP (2019, 5400977) did not report alterations in lung
weight or histopathology after dosing for 28 days, there were several effects on the nasal cavity
and olfactory system that were not suggestive of gavage-related reflux (Figure C-44). Chronic
active inflammation of the nasal respiratory epithelium was observed in both males and females,
though these effects did not exhibit a linear dose-response relationship. Similarly, olfactory
epithelial inflammation and degeneration were observed in females. Increases in nasal and
olfactory hyperplasia were thought to be a result of the observed epithelial degradation and/or
inflammation {NTP, 2019, 5400977}. Interestingly, these nasal and olfactory effects were
observed across multiple PFAS (PFOA, perfluorohexanoic acid (PFHxA), PFNA, PFBS,
PFHxS) in toxicity studies conducted by NTP (2019, 5400977; 2019, 5400978), though not in
the chronic PFOA feeding study {NTP, 2020, 7330145}. No other studies identified during this
assessment reported examinations of nasal or olfactory systems in animal models.
C-136
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOA Respiratory Effects - Non-neoplastic Nasal Lesions
Endpoint Animal Description Dose
¦¦ Significant trend No significant trend
Nose, Respiratory Epithelium, Inflammation Chronic Active Rat, Sprague-Dawley (o, N=10) 0
0.625
1.25
2.5
5
10
Respiratory Epithelium, Hyperplasia Rat, Sprague-Dawley (o, N=10) 0
0.625
1.25
2.5
5
10
Nose, Respiratory Epithelium, Inflammation Chronic Active Rat, Sprague-Dawley ( , N=10) 0
6.25
12.5
25
50
100
Olfactory Epithelium, Degeneration Rat, Sprague-Dawley (9, N=10) 0
6.25
12.5
25
50
100
Olfactory Epithelium, Hyperplasia Rat, Sprague-Dawley (2, N=10) 0
6.25
12.5
25
50
100
Olfactory Epithelium, Inflammation, Suppurative Rat, Sprague-Dawley ( , N=10) 0
6.25
12.5
25
50
100
Respiratory Epithelium, Hyperplasia Rat, Sprague-Dawley N=10) 0
6.25
12.5
25
50
100
1 1 1 1 1 1 1—
0 10 20 30 40 50 60 70 80 90 100
Incidence (%)
Figure C-44. Incidence of Nonneoplastic Nasal Lesions in Male and Female Sprague-
Dawley Rats Following 28-day Oral Exposure to PFOA, as Reported by NTP (2019,
5400977)
Interactive figure and additional study details available on HAWC.
Statistical significance reached at p < 0.05.
C-137
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
There is one available study in mice that assessed potential pulmonary effects of PFOA
exposure. In this developmental toxicity study, Yahia et al. (2010, 1332451) saw no effect on the
lungs of maternal or neonatal mice after up to 10 mg/kg/day PFOA treatment from GD 0-18.
C. 7.3 Mechanistic Evidence
Mechanistic evidence linking PFOA exposure to adverse respiratory outcomes is discussed in
Section 3.3.4 of the 2016 PFOA HE SD {U.S. EPA, 2016, 3603279}. There are 3 studies from
recent systematic literature search and review efforts conducted after publication of the 2016
PFOA HESD that investigated the mechanisms of action of PFOA that lead to respiratory
effects. A summary of these studies is shown in Figure C-45. Additional mechanistic synthesis
will not be conducted since evidence suggests but is not sufficient to infer that PFOA leads to
respiratory effects.
Mechanistic Pathway Animal In Vitro Grand Total
Cell Growth, Differentiation, Proliferation, Or Viability
1
2
3
Cell Signaling Or Signal Transduction
0
1
1
Inflammation And Immune Response
0
1
1
Oxidative Stress
0
1
1
Grand Total
1
2
3
Figure C-45. Summary of Mechanistic Studies of PFOA and Respiratory Effects
Interactive figure and additional study details available on Tableau.
C.7.4 Evideri ce In tegra tion
The evidence of an association between PFOA exposure and respiratory effects in humans is
slight, with an indication of decreased lung function among infants, children, and adolescents.
However, the results are inconsistent and there are a small number of studies examining
respiratory effects, particularly in adults. While there is some evidence of detrimental respiratory
health effects, particularly in children with asthma, the available epidemiological evidence
examining PFOA exposure and respiratory health is limited.
The animal evidence for an association between PFOA exposure and respiratory effects is
indeterminate, based on inconsistencies in the available evidence in the high and medium
confidence studies. While the increases in alveolar macrophages and hemorrhaging reported by
Butenhoff et al. (2012, 2919192) are suggestive of pulmonary damage, these results were not
observed in two other chronic feeding studies in rats {Perkins, 2004, 1291118; NTP, 2020,
7330145}. The authors of the study also call into question whether those effects were related to
PFOA treatment {Butenhoff, 2012, 2919192}. NTP (2019, 5400977) provides data suggestive of
nasal toxicity due to PFOA exposure, though the positive results in males do not follow a linear
dose-response and are difficult to interpret. The significant effects in females (i.e., olfactory
epithelium degeneration and inflammation) occur at relatively high doses (50 mg/kg/day)
compared to effects seen for other health outcomes. Therefore, it does not appear that respiratory
effects are sensitive or replicable outcomes of PFOA toxicity.
C-138
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C. 7.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOA exposure has the potential to cause respiratory effects in
humans under relevant exposure circumstances (Table C-12). This conclusion is based on
evidence of an association between PFOA and detrimental respiratory health effects, particularly
in children with asthma, in a small number of epidemiologic studies with median exposure levels
from 0.50 - 2.4 ng/mL; however, limited number of studies and issues with inconsistency across
studies raise considerable uncertainty. Moreover, evidence in animals is sparse and largely
uninterpretable regarding its relevance to humans.
C-139
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-12. Evidence Profile Table for PFOA 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
5 Medium confidence
studies
Studies in infants,
children, and adolescents
report significant
decreases in FVC (1/5)
and in FEV1 and FEF25-
75% among those with
asthma (1/5). Studies in
children observed
significantly decreased
FEV1 associated with
prenatal and cross-
sectional exposures (2/5).
Other studies observed
non-significant decreases
inFEVl (2/5).
• Medium confidence
studies
• Consistent direction of
effect among infants,
children, and
adolescents
• No factors noted
Obstructive disease
1 Medium confidence
study
1 Low confidence study
One study in infants
reported significantly
increased odds of low
severity obstructive airway#
disease. An occupational
study of adult workers in a
chemical plant observed
no association with
COPD.
• Medium confidence
study
¦ Low confidence study
• Lmprecision of observed
effect across exposure
groups in the
occupational study
©OO
Slight
Several studies of medium
confidence found evidence
for decreases in lung
function measures among
infants, children, and
adolescents, though other
medium confidence studies
did not observe significant
effects. Few studies
examined obstructive
disease effects. Those
_studies that did were of
lower confidence and
showed imprecision of
findings across exposure
groups. Uncertainty
remains about respiratory
outcomes among adults in
occupational settings and
Limited number of studies in the general population,
examining outcome
©OO
Evidence Suggests
Primary basis:
No evidence in animals and
human evidence indicted
detrimental respiratory
health effects, particularly
in children with asthma.
However, limited number
of studies and issues with
imprecision across studies
raise considerable
uncertainty.
Human relevance, cross-
stream coherence, and
other inferences:
No specific factors are
noted.
Evidence from In Vivo Animal Studies (Section C.7.2)
Histopathology
2 High confidence
studies
2 Medium confidence
studies
Two studies evaluating • High and Medium
chronic and short-term confidence studies
exposure to PFOA in male
and female rats found •
increases in non-neoplastic
lesions and inflammation
in the lungs and nose
• Lnconsistent direction of
results
OOO
Indeterminate
Evidence was based on 4
high and medium
confidence studies and
C-140
-------
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
(2/4). Two additional
chronic exposure studies
in rats reported no changes
in histopathological
endpoints in the lungs
jm
Organ weight
2 High confidence
studies
1 Medium confidence
study
Studies evaluating rat lung <
weight found that short-
term exposure to PFOA
had no effect (2/3). One
study found that lung
weight increased in male
and female rats after
chronic PFOA exposure,
however, this was
attributed to decreased
body weight and not
considered a toxicological
effect (1/3).
High and Medium
confidence studies
»Inconsistent direction of
results
» Changes in body weight
may limit ability to
interpret these responses
Evidence Integration
Summary Judgment
provided inconsistent
results. One study suggests
alveolar macrophages and
hemorrhaging increased,
while two other chronic
_studies reported no change.
Nasal toxicity reported in
one study did not occur in
a dose dependent manner,
while another required
relatively high doses to
occur. Lung weight was
increased in one chronic
exposure but occurred with
decreased body weight.
Notes: COPD = chronic obstructive pulmonary disease; FEF25-75% = forced expiratory flow at 25-75%; FEV1 = forced expiratory volume; FVC = forced vital capacity.
C-141
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.8 Musculoskeletal
EPA identified 8 epidemiological and 1 animal studies that investigated the association between
PFOA and musculoskeletal effects. Of the epidemiological studies, 6 were classified as medium
confidence and 2 were considered low confidence (Section C.8.1). The animal study was
considered 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
PFOA Document).
C 8.1 Human E vide nee Study Quality E valuation and
Synthesis
C.8.1.1 In traduction
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 PFOA {U.S. EPA, 2016, 3603279} did not previously evaluate
musculoskeletal health outcomes in humans. The C8 Science Panel {C8 Science Panel, 2012,
1430770} concluded there is no probable link between PFOA and osteoarthritis.
For this updated review, nine studies (nine publications) examined the association between
PFOA exposure and musculoskeletal health outcomes. Different study designs were used; one
was a cohort study {Jeddy, 2018, 5079850}, one used cross-sectional and prospective analyses
{Hu, 2019, 6315798}, and the remainder were cross-sectional. All studies measured PFOA in
blood components (i.e., blood, plasma, or serum), and one study {Di Nisio, 2019, 5080655}
measured PFOA in semen. Three studies {Khalil, 2016, 3229485; Lin, 2014, 5079772; Uhl,
2013, 1937226} used data from participants in 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 mineral
density; bone area, thickness (e.g., endosteal and periosteal thickness), or circumference; bone
mineral content (BMC); bone stiffness; ultrasound attenuation and speed of sound (indicators of
bone quality); lean body mass; height; arm span; bone fracture; and plasma concentrations of P-
C-telopeptides of type I collagen, a marker for bone turnover.
C.8.1.2 Study Quality
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}.
C-142
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The 2016 HESD for PFOA {U.S. EPA, 2016, 3603279} did not previously evaluate
musculoskeletal health outcomes in humans. The C8 Science Panel {C8 Science Panel, 2012,
1430770} concluded there is no probable link between PFOA and osteoarthritis.
There are 8 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and musculoskeletal effects. Study quality evaluations for these 8
studies are shown in Figure C-46.
Different study designs were used; one was a cohort study {Jeddy, 2018, 5079850}, one used
cross-sectional and prospective analyses {Hu, 2019, 6315798}, and the remainder were cross-
sectional. All studies measured PFOA in blood components (i.e., blood, plasma, or serum), and
one study {Di Nisio, 2019, 5080655} measured PFOA in semen. Three studies {Khalil, 2016,
3229485; Lin, 2014, 5079772; Uhl, 2013, 1937226} used data from participants in 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 mineral density; bone area, thickness (e.g., endosteal and
periosteal thickness), or circumference; BMC; bone stiffness; ultrasound attenuation and speed
of sound (indicators of bone quality); lean body mass; height; arm span; bone fracture; and
plasma concentrations of P-C-telopeptides of type I collagen, a marker for bone turnover
Three cross-sectional or retrospective studies {Di Nisio, 2019, 5080655; Khalil, 2018, 4238547;
Steenland, 2015, 2851015} classified as low confidence had deficiencies in participant selection,
confounding, outcome measurement, and study sensitivity. Participant selection was considered
a deficiency mainly due to underreporting about participation rates and participant characteristics
relative to non-participants (e.g., those who died before the retrospective study was conducted).
Other deficiencies included potential for outcome misclassification when the musculoskeletal
outcome (taking medication for osteoarthritis) was not validated using medical records
{Steenland, 2015, 2851015}; potential for residual confounding by SES; small sample sizes and
limited ranges of participant exposure to PFOA {Di Nisio, 2019, 5080655; Khalil et al., 2018,
4238547}.
C-143
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
xec^
,0®
Cluett et al., 2019, 5412438-
++
i
+
++
i
+
++
i
+
i
+
+
Di Nisio et al., 2019, 5080655 -
-
+
+
-
+
+
-
-
Hu et al., 2019, 6315798-
+
++
++
-
++
+
+
+
Jeddy et al., 2018, 5079850 -
+
B
++
-
+
+
+
+
Khalil et al., 2016, 3229485-
++
+
+
+
+
+
Khalil et al., 2018, 4238547-
-
++
-
+
+
-
-
Lin et al., 2014, 5079772-
+
+
+
+
+
+
Uhl et al., 2013, 1937226-
+
+
+
+
++
+
+
+
Figure C-46. Summary of Study Evaluation for Epidemiology Studies of PFOA 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 (Appendix D).
While the medium confidence studies observed few statistically significant associations between
PFOA and the musculoskeletal health outcomes examined, the associations supported a harmful,
rather than beneficial, direction of effect. Cluett et al. (2019, 5412438) observed a statistically
significant inverse association with the areal bone mineral density (aBMD) z-score (a
standardized measure of bone mineral amount relative to bone area) in children aged 6-10 years,
with a greater magnitude of effect for females and was not significant for males. Inverse
C-144
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
significant associations were also observed for BMC z-score. Jeddy et al. (2018, 5079850)
observed a statistically significant inverse association between prenatal PFOA exposure and
height in 17-year old girls. A statistically significant inverse association was also observed with
whole-body bone area, but this was no longer 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. Khalil et al. (2016, 322948) observed higher odds of osteoporosis
in women aged 12-80 years from NHANES (2009-2010). Uhl et al. (2013, 1937226) observed
statistically significantly increased odds of osteoarthritis in women aged 20-84 years in
NHANES cycles (2003-2008). This was most apparent among younger premenopausal women
aged 20-49, who may have differing susceptibility to endocrine disruption. An overlapping
NHANES study {Lin, 2014, 5079772} observed no statistically significant associations with
history of bone fractures in women aged 20 and older. 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
five of six sites examined: the spine, total hip, femoral neck, hip trochanter, 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 PFOA exposure and
arm span.
C.8.1.5 Findings from Occupational Studies
One low confidence study of occupational exposure {Steenland, 2015, 2851015} reported
limited, conflicting evidence related to osteoarthritis in predominantly male workers: participants
with elevated PFOA exposure had lower odds of self-reported osteoarthritis after a 10-year time
lag, but this finding was not supported across exposure quartiles.
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 PFOA HESD {U.S., EPA, 2016, 3603279} that investigated the
C-145
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
association between PFOA and musculoskeletal effects. Study quality evaluations for this 1
study is shown in Figure C-47.
&
^.e9'
G°
1 c,e\® e^° o^° ^ o^e<
van Esterik et al., 2015, 2850288-
++
NR
+*
++
+
-
s
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
Multiple judgments exist
Figure C-47. Summary of Study Evaluation for Toxicology Studies of PFOA and
Musculoskeletal Effects
Interactive figure and additional study details available on HAWC.
Limited data are available on the effect of PFOA on the musculoskeletal system other than
developmental skeletal defects resulting from gestational exposure that are discussed in Section
3.4.4.2 of the Main PFOA Document. EPA did not identify any publications that reported
musculoskeletal effects outside of those associated with developmental toxicity from the 2016
PFOA HESD {U.S., EPA, 2016, 3603279} 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 PFOA exposure to adverse musculoskeletal
outcomes in the 2016 PFOA HESD {U.S. EPA, 2016, 3603279}. There are 8 studies from recent
systematic literature search and review efforts conducted after publication of the 2016 PFOA
HESD that investigated the mechanisms of action of PFOA that lead to musculoskeletal effects.
A summary of these studies is shown in Figure C-48. Additional mechanistic synthesis will not
be conducted since evidence suggests but is not sufficient to infer that PFOA leads to
musculoskeletal effects.
C-146
-------
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
0
7
Cell Signaling Or Signal Transduction
1
3
4
Extracellular Matrix Or Molecules
1
1
2
Oxidative Stress
0
2
2
Grand Total
2
7
8
Figure C-48. Summary of Mechanistic Studies of PFOA 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 PFOA exposure and musculoskeletal effects
in humans based on observed effects on bone mineral density and bone health in several medium
confidence studies. Additionally, there is limited evidence of negative effects of PFOA on
skeletal size (height and arm span) and connective tissue disorders (osteoarthritis). No
epidemiological studies examined the relationship between PFOA and muscular disorders. No
musculoskeletal health outcome epidemiology studies were previously reviewed in the 2016
HESD for PFOA {U.S. EPA, 2016, 3603279}.
Although relatively few studies have investigated musculoskeletal health outcomes related to
PFOA exposure, some shared conclusions can be drawn. The observed associations in
epidemiological studies were primarily between increased PFOA exposure and decreased bone
mineral density (consistently among various skeletal sites), bone mineral density relative to bone
area, height in adolescence, osteoporosis, and osteoarthritis. These issues with bone density may
correspond with the reports of reduced ossification and skeletal deformities in developmental
animal models with gestational PFOA exposure (See Main PFOA Document). Rarer outcomes,
such as fracture, were not observed to be associated with PFOA 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 led to reduced confidence in most studies; common issues included cross-sectional
design or potential for residual confounding.
The animal evidence for an association between PFOA exposure and effects in the
musculoskeletal system is considered indeterminate based on lack of information in animal
models. There is one low confidence study where there was some change in bone length.
C.8.4.1 Evidence Integration Judgment
Overall, evidence suggests that PFOA 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 PFOA ranging from 0.99 to 5.4 ng/mL. Although there is some evidence of
C-147
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
negative effects of PFOA 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-148
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-13. Evidence Profile Table for PFOA Musculoskeletal Effects
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
Evidence from Studies of Exposed Humans (Section C.8.1)
Bone parameters
5 Medium confidence
studies
1 Low confidence study
Decreases in bone mineral • Medium confidence
content (BMC) were
observed in two studies
(2/6) on children and
adults. Reductions in bone
mineral density (BMD)
were also observed in
children and adults (4/6),
including site specific
BMD measures. However,
there was some
inconsistency in direction
of effect when 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.
studies
• Imprecision of findings
across studies, including
for bone area association,
due to wide confidence
intervals and measures of
BMD
• Inconsistent direction of
effect based on sex
• Low 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 osteoporosis.
Uncertainties remain due
to inconsistent or imprecise
results, and limited
evidence for fractures, size
measures, and odds of
osteoarthritis or
osteoporosis.
Fractures
1 Medium confidence
study
Study authors reported no • Medium confidence
significant association study
with incidence of bone
fractures.
• Imprecision of findings
• Limited number of studies
examining outcome
Size measures
1 Medium confidence
study
1 Low confidence study
Studies among children
found significantly
decreased height (1/2), but
results for arm span were
not precise in a study of
high school students in a
high-exposure community
(1/2).
• Medium confidence
study
• Imprecision of findings
• Limited number of studies
examining outcome
• Low confidence study
®oo
Evidence Suggests
Primary basis:
No evidence in animals and
human evidence indicated
effects on bone mineral
density and bone health.
Although there is some
evidence of negative effects
of PFOA 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-149
-------
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
Lean body mass
1 Medium confidence
study
Study authors reported no • Medium confidence
significant association study
among adolescent
females.
• Limited number of studies
examining outcome
Osteoarthritis
1 Medium confidence
study
1 Low confidence study
Findings for osteoarthritis • Medium confidence • Lmprecision of findings
were mixed. Significantly
increased odds of
osteoarthritis were
observed among females
ages 20-84 in both
continuous and categorical
analyses, among the
highest exposure group of
females ages 20-49, and
among all adults ages 20-
49 (1/2). The risk of
osteoarthritis was
decreased in an
occupational study, but
findings were not precise.
study
• Lnconsistent direction of
effect based on study
population
• Limited number of studies
examining outcome
• Low confidence study
Osteoporosis
1 Medium confidence
study
Significant increases for
the odds of osteoporosis
were observed in a study
of females 12-80 years of
age.
• No factors noted
• Lmprecision of findings
from categorical analyses
• Limited number of studies
examining outcome
Notes: BMC = bone mineral content; BMD = bone mineral density.
C-150
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.9 Gastrointestinal
EPA identified 4 epidemiological and 3 animal studies that investigated the association between
PFOA and gastrointestinal effects. Of the epidemiological studies, 1 was classified as medium
confidence and 3 were considered low confidence (Section C.9.1). Of the animal studies, 1 was
classified as high confidence, and 2 were considered medium 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 PFOA Document).
C. 9.1 Human E vide nee Study Quality E valuation and
Synthesis
C.9.1.1 In traduction
PFOA exposure may affect gastrointestinal 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}. Gastrointestinal outcomes only assessed in the context of
immune system health, including ulcerative colitis and Crohn's disease, are discussed in the Main
PFOA Document. However, some research suggests an overall immunosuppressive effect of
PFOA 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
gastrointestinal 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 PFOA and
gastrointestinal health outcomes {Dalsager, 2016, 3858505; Hammer, 2019, 8776815; Xu, 2020,
6315709; Timmermann, 2020, 6833710}. PFOA was measured in serum or blood, and the
outcomes measured included diarrhea and vomiting, and IBD biomarkers zonulin and
calprotectin. Dalsager et al. (2016, 3858505) measured PFOA in pregnant women in Denmark
and collected self-reported health outcomes for their children (< 4 years). Hammer et al. (2019,
8776815) examined a subset of the general population in the Faroe Islands enrolled in the
Children's Health and the Environment in the Faroes (CHEF) study. Xu et al. (2020, 6315709)
examined child and adult residents of Ronneby, Sweden, exposed to PFAS in drinking water, as
well as unexposed individuals from a nearby town. Timmermann et al. (2020, 6833710)
examined a subset of 4-18-month old children from a 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 gastrointestinal
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 gastrointestinal disorders has developed substantially over recent decades and diagnostic
C-151
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
indicators continue to rapidly evolve. Causal factors in developing gastrointestinal 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
gastrointestinal conditions lead to increased intake of PFOA from food packaging or preparation
methods, increased PFOA absorption through the gastrointestinal tract, or reduced fecal
excretion. Measuring PFOA and gastrointestinal outcomes concurrently was considered adequate
in terms of exposure assessment timing. Given the long half-life of PFOA (median half-
life = 2.7 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 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and gastrointestinal effects. Study quality evaluations for these 4
studies are shown in Figure C-49.
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 gastrointestinal 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 PFOA 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-152
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
. c,& rt\l
,
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOA and gastrointestinal effects. Study quality evaluations for these 3 studies are shown in
Figure C-50.
.a
Butenhoff et al., 2012, 2919192
Chang etal., 2020, 6320656
NTP, 2020, 7330145
B
++
B
B
++
++
Legend
0
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
NR
Not reported
Figure C-50. Summary of Study Evaluation for Toxicology Studies of PFOA and
Gastrointestinal Effects
Interactive figure and additional study details available on HAWC.
The only information available to assess the gastrointestinal tract is histopathological evaluations
(Figure C-51). In many cases, this was evaluated in the control and high-dose groups only.
Chronic studies in rats suggest that oral exposure to PFOA may increase the incidence of non-
neoplastic lesions in the gastrointestinal tract {NTP, 2020, 7330145; Chang, 2020, 6320656}.
However, shorter durations may not elicit the response as noted in a study where no
histopathological findings were observed in the duodenum, jejunum, or ileum of the small
intestine or the cecum, colon, or rectum of the large intestine of rats after 28 days. Likewise, no
adverse effects were seen in the forestomach and glandular stomach or salivary gland {NTP,
2019, 5400977}.
NTP (2020, 7330145) used a matrix-type exposure paradigm whereby pregnant rats were
administered PFOA on GD 6 and exposure was continued in offspring postweaning for a total of
107 weeks. Dose groups for this report are referred to as "[perinatal exposure level
(ppm)]/[postweaning exposure level (ppm)]" and ranged from 0/0-300/300 ppm in males and
0/0-300/1000 ppm in females (see Main PFOA Document for further study design details). At
the 16-week interim evaluation, incidences of chronic active inflammation of the glandular
stomach submucosa were increased in all male treated groups compared to the control; however,
statistical significance was only achieved in the 0/300 ppm group. No significant differences
were noted in groups with and without perinatal exposure and no effects were seen in females at
interim sacrifice. At the 2-year evaluation, females of the 0/1000 and 300/1000 ppm groups
C-154
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
exhibited increased incidences of ulcer, epithelial hyperplasia, and chronic active inflammation
of the submucosa of the forestomach when compared to controls. In addition, a single case of
squamous cell papilloma was noted in both exposure groups (NTP, 2020, 7330145).
In a dietary study, male and female rats fed 30 or 300 ppm PFOA for two years exhibited no
stomach abnormalities during histopathological examination. In the salivary glands of male rats,
significant increases in chronic sialadentitis were noted at 30 ppm (27%) and 300 ppm (30%).
However, study authors reported this as being associated with antemortem viral infection. This
effect was not observed in females (Butenhoff et al., 2012, 2919192).
Eiidpolnl
sa, Inflammation. C
Large Intestine. Cam Nun Neoplastic Lesions
Large Intestine. Colon. Non-Neoplastic Lesions
Large Intestine. Rectum. Non-Neoplastic Lesions
Salivary Gland. Sialadenitis. Chronic
Salivary (Hand*. Siakulcnlilu, Chronic
Stomach. Horestrvmach, Kpithelium. Hyperplasia
NTP, 2020. 7330145
Chang el al.. 2020.6320636
Cluing ci al.. 2020.6320656
Clung el al.. 2020.6320656
BulcnholTd al.. 2012. 2919192
Bulcnholl et al„ 2012. 2919192
NTP, 2020, 7330145
NTP. 2020,7330145
Study Design
ronic (PND2I-PNWI07)
ronic (GD6-PNW107)
ronic (PND2I -PNW2I)
ronic (GD6 PNW21)
ronic (26 wk)
ronic <26 «k)
ronic (26 uk>
runic (2y)
ronic (2y)
ronic (PND2I-PNWI07)
ronic (GD6 -PNWI07)
ronic (PND2I-PNWI07)
ronic (GD6-PNWI07)
Animal DeH-rlption
Fl Rat. Spraguc-Dawlcv (9, N=50)
Fl Rat. Spraguc-DauIcy (9, N=50)
Fl Rat. Spraguc-Dawley (tf, N-10)
Fl Ra(. Spraguc Dawlcv (
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
lesions consistent with Oesophagostomnm spp. were observed but were considered common in
the intestinal tract of non-human primates at the time the study was conducted {Chang, 2020,
6320656}.
NTP conducted a 28-day study in which 10 or 100 mg/kg/day of PFOA were orally administered
to male or female rats, respectively. No histopathological findings were noted in the duodenum,
jejunum, or ileum of the small intestine or the cecum, colon, or rectum of the large intestine.
Likewise, no adverse effects were seen in the forestomach and glandular stomach or salivary
gland (NTP, 2019, 5400977).
C.9.3 Mechanistic Evidence
There was no mechanistic evidence linking PFOA exposure to adverse gastrointestinal outcomes
in the 2016 PFOA HESD {U.S. EPA, 2016, 3603279}. There are 5 studies from recent
systematic literature search and review efforts conducted after publication of the 2016 PFOA
HESD that investigated the mechanisms of action of PFOA that lead to gastrointestinal effects. A
summary of these studies is shown in Figure C-52. Additional mechanistic synthesis will not be
conducted since evidence is inadequate to infer that PFOA leads to gastrointestinal effects.
Mechanistic Pathway Animal Human In Vitro Grand Total
Cell Growth, Differentiation, Proliferation, Or Viability
1
0
1
2
Cell Signaling Or Signal Transduction
1
0
0
1
Fatty Acid Synthesis, Metabolism, Storage, Transport, Binding, B-Oxidation
0
0
1
1
Inflammation And Immune Response
0
0
1
1
Other
0
1
1
2
Grand Total
1
1
3
5
Figure C-52. Summary of Mechanistic Studies of PFOA and Gastrointestinal Effects
Interactive figure and additional study details available on Tableau.
C.9.4 Evidence Integration
The evidence evaluating an association between PFOA and gastrointestinal health effects in
humans is indeterminate based on a paucity of research and the quality of the available studies.
In the 2016 HESD for PFOA {U.S. EPA, 2016, 3603279}, gastrointestinal outcomes from
epidemiological studies were only assessed in the context of immune system health, with limited
evidence of associations with gastroenteritis. The available research has not demonstrated
conclusive effects of PFOA exposure and gastrointestinal health effects, including vomiting, or
diarrhea.
The animal evidence for an association between PFOA exposure and gastrointestinal tract effects
is indeterminate based on limited data in animal models. The only significant non-neoplastic
lesions observed were noted in the stomachs of male rats treated at 0/300 ppm and female rats
C-156
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
treated at high doses (0/1000 ppm and 300/1000 ppm) in a 2-year feeding study {NTP, 2020,
7330145}. Additionally, lack of significant effects in rat colon and cynomolgus monkey cecum,
colon, and rectum indicated no signs of ulcerative colitis {Chang, 2020, 6320656}.
C.9.4.1 Evidence Integration Judgment
Overall, there is inadequate evidence to assess whether PFOA exposure can cause
gastrointestinal effects in humans under relevant exposure circumstances (Table C-14).
C-157
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-14. Evidence Profile Table for PFOA Gastrointestinal 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.9.1)
Diarrhea and vomiting
1 Medium confidence
study
1 Low confidence study
Two studies examining
diarrhea observed non-
significant increased
associations with PFOA in
young children. One study
also observed decreased
incidence of vomiting, but
odds of vomiting across
exposure tertiles in
children ages 1-4 years
were non-significant and
inconsistent. No studies
were conducted in adults.
• Medium confidence
study
• Low confidence study
• Lnconsistent directions 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
Both studies examining
IBD observed no
associations with PFOA.
Non-significant decreases
in IBD incidence or IBD
biomarkers were observed
in association with PFOA.
• 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 two studies
reporting increases in
diarrhea and vomiting and
two other studies reporting
decreases in IBD.
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.
Evidence from In Vivo Animal Studies (Section C.9.2)
Histopathology
1 High confidence study
2 Medium confidence
studies
One chronic exposure
study found evidence of
increased incidence of
nonneoplastic lesions
including ulcer, epithelial
hyperplasia, and/or
inflammation in male and
female rats. Two chronic
exposure studies found no
1 High and medium
confidence studies
• Limited number of studies
examining outcome
• Lnconsistent direction of
effects across animal
models
OOO
Lndeterminate
Evidence was limited to
three studies that
demonstrated unexplained
inconsistency across
animal models regarding
gastrointestinal toxicity.
C-158
-------
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
evidence of nonneoplastic
lesions within the
gastrointestinal tract in
both sexes in rats or in
male monkeys.
Notes: IBD = inflammatory bowel disease.
C-159
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.10 Dental
EPA identified 2 epidemiological studies that investigated the association between PFOA 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 PFOA Document).
C.10.1 Human Evidence Study Quality Evaluation and
Synthesis
C.10.1.1 In traduction
PFOA 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 PFOA 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 PFOA exposure and
dental caries in children and adolescents {Puttige Ramesh, 2019, 5080517; Wiener, 2019,
5386081}. Dental caries 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 the 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 PFOA was assessed via biomarkers in blood.
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 PFOA and dental outcomes
concurrently was considered adequate in terms of exposure assessment timing. Given the long
half-life of PFOA (median half-life = 2.7 years) {Li, 2018, 4238434}, current blood
concentrations are expected to correlate well with past exposures.
C-160
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
There are 2 studies from recent systematic literature search and review efforts conducted after
publication of the 2016 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and dental effects. Study quality evaluations for these 2 studies are
shown in Figure C-53.
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.
I I I I I I 1__
Puttige Ramesh et al., 2019, 5080517
Wiener etal., 2019, 5386081
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-53. Summary of Study Evaluation for Epidemiology Studies of PFOA and Dental
Effects
Interactive figure and additional study details available on HAWC.
C. 10.1.3 Findings
Results for the studies that examined dental outcomes are presented in Appendix D. Both studies
observed non-significantly increased odds of dental caries with increased PFOA exposure
children and adolescents {Puttige Ramesh, 2019, 5080517; Wiener, 2019, 5386081}. Puttige
Ramesh et al. (2019, 5080517) also observed increased odds of dental caries in the third highest
quartile of exposures, but decreased odds in the second and highest quartiles compared to the
lowest. Analyses did not account for age, but considered gender, race, education level of
parent/guardian, family-poverty-to-income ratio, blood lead level, and serum cotinine level (an
C-161
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
indicator of exposure to smoking). Wiener and Waters (2019, 5386081) adjusted the analysis 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. No studies of dental health
outcomes were available for pregnant women, adults, or occupational workers.
C.10.2 Animal Evidence Study Quality Evaluation and
Synthesis
In the available literature, there is no reported biological consequence of PFOA exposure on
dental outcomes in animals.
C.10.3 Mechanistic Evidence
There was no mechanistic evidence linking PFOA exposure to adverse dental outcomes in the
2016 PFOA HESD {U.S. EPA, 2016, 3603279}. There are no studies from recent systematic
literature search and review efforts conducted after publication of the 2016 PFOA HESD that
investigated the mechanisms of action of PFOA that lead to dental effects. Additional
mechanistic synthesis will not be conducted since evidence is inadequate to infer that PFOA may
cause dental effects.
C. 10.4 Eviden ce In tegra tion
The evidence evaluating an association between PFOA exposure and dental effects in humans is
indeterminate based on the limited number of available studies and imprecision of observed
results. Dental outcomes were not previously reviewed in the 2016 HESD for PFOA {U.S. EPA,
2016, 3603279}. The present epidemiological review identified only two dental studies in
humans in which prevalence of dental caries was evaluated. Both studies observed non-
significantly increased odds of dental caries {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 PFOA exposure and dental effects is
indeterminate because there are no available studies in animal models that examine dental effects
due to PFOA exposure.
C. 10.4.1 Evidence Integration Judgment
Overall, there is inadequate evidence to assess whether PFOA exposure can cause dental effects
in humans under relevant exposure circumstances (Table C-15).
C-162
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-15. Evidence profile table for PFOA Dental Effects
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
Evidence from Studies of Exposed Humans (Section C.10.1)
Dental caries
2 Medium confidence
studies
Two studies observed • Medium confidence • Limited number of studies
non-significant increase in studies examining outcome
odds of dental caries. No
significant associations • Imprecision of findings
observed in studies of
children from NHANES.
ooo
Indeterminate
Evidence was limited to
two studies that reported
non-significant positive
associations with dental
caries in children and
adolescents, but results are
imprecise. Uncertainties
remain regarding effects in
adults in 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.
C-163
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.ll Ocular
EPA identified 1 epidemiological and 2 animal studies that investigated the association between
PFOA and ocular effects. The 1 epidemiological study was classified as medium confidence
(Section C. 11.1). Of the animal studies, 2 were classified as high confidence (Section C.ll .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 PFOA Document).
C.ll.l Human E vide nee Study Quality E valuation and
Synthesis
C.ll. 1.1 In traduction
For this updated review, there is one epidemiological study that investigated the association
between PFOA and ocular effects {Zeeshan, 2020, 6315698}.
This cross-sectional study was conducted in Shenyang, China 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
PFOA serum concentrations among the 1202 study participants were 6.06 ng/mL
(Q1 = 3.97 ng/mL, Q3 = 9.12 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 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and ocular effects. Study quality evaluation for this 1 study is shown
in Figure C-54.
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-164
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure C-54. Summary of Study Evaluation for Epidemiology Studies of PFOA 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 PFOA in adults aged 22-
96 years, who had lived for at least 5 years in in Shenyang, China (Appendix D). Ocular
outcomes examined included VI, vitreous disorder, synechia, macular disorder, corneal pannus,
and ACD, and combined eye disease (aggregating all 9 ocular conditions examined). A positive
statistically significant association between VI and total serum PFOA was observed (OR: 1.80;
95% CI: 1.37, 2.37). When stratified by age, the association between combined eye disease and
total serum PFOA was statistically significant for participants aged < 65 years (OR: 1.25; 95%
CI: 1.01, 1.56) but not for the older participants (OR: 1.19; 95% CI: 0.71, 1.98). 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 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and ocular effects. Study quality evaluations for these 2 studies are
shown in Figure C-55.
C-165
-------
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)
B Critically deficient (metric) or Uninformative (overall)
Not reported
o^°°V
NTP, 2019, 5400977-
++
++
I
NR
++
NTP, 2020, 7330145-
++
++
NR
++
Figure C-55. Summary of Study Evaluation for Toxicology Studies of PFOA and Ocular
Effects
Interactive figure and additional study details available on HAWC.
Eye irritation studies in rabbits suggest that PFOA acts as an ocular irritant {Gabriel, 1976,
4442370}; however, no adverse lesions were noted in eye tissues during histopathological
examination in repeated-dose oral toxicity studies in rats. In a 28-day oral toxicity study where
only control and high-dose groups were evaluated, no histopathological findings were noted in
eyes of male rats treated with 10 mg/kg/day or female rats treated with 100 mg/kg/day {NTP,
2019, 5400977}. In a chronic exposure study, male and female Sprague-Dawley rats were fed
diets containing PFOA for approximately two years (See Main PFOA Document further study
design details). Observation of gross abnormalities and histopathological examination of eye
tissues were conducted in pups at 16 weeks and 2 years with no treatment related abnormalities
noted {NTP, 2020, 7330145}.
C.11.3 Mechanistic Evidence
There was no mechanistic evidence linking PFOA exposure to adverse ocular outcomes in the
2016 PFOA HESD {U.S. EPA, 2016, 3603279}. There is 1 study from recent systematic
literature search and review efforts conducted after publication of the 2016 PFOA HESD that
investigated the mechanisms of action of PFOA that lead to ocular effects. A summary of these
studies is shown in Figure C-56. Additional mechanistic synthesis will not be conducted since
evidence is inadequate to infer that PFOA leads to ocular effects.
C-166
-------
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-56. Summary of Mechanistic Studies of PFOA and Ocular Effects
Interactive figure and additional study details available on Tableau.
The evidence evaluating an association between PFOA exposure and ocular effects in humans is
considered indeterminate based on a limited number of studies. In the 2016 Health Assessment
for PFOA {U.S. EPA, 2016, 3603279}, no epidemiological evidence of an association between
PFOA exposure and ocular health effects was observed. One recent epidemiological study
reported an association between PFOA exposure and visual impairment 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 PFOA.
The animal evidence for an association between PFOA and ocular effects is indeterminate due to
the limited evidence available in animal models. In two available studies in animal models that
assess ocular toxicity, there were no observed ocular effects with short-term or chronic PFOA
exposure in male or female rats.
C. 11.4.1 Evidence Integration Judgment
Overall, there is inadequate evidence to assess whether PFOA exposure can cause ocular effects
in humans under relevant exposure circumstances (Table C-16)
C.11.4 Evidence Integration
C-167
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-16. Evidence Profile Table for PFOA Ocular Effects
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
Evidence from Studies of Exposed Humans (Section C.ll.l)
Eye disease
1 Medium confidence
study
Histopathology
2 High confidence
studies
The only study examining • Medium confidence
eye disease was a cross- study
sectional study that
observed significant
positive associations
between visual impairment
and serum PFOA. The
association was also
significant for combined
eye disease, but only in
participants aged <65
years.
• Limited number of studies
examining outcome
OOO
Indeterminate
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)
OOO
Inadequate Evidence
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.
No changes in ocular <
histopathology were
reported in one 28-day and
one chronic study in male
and female rats.
High confidence
studies
• Limited number of studies
examining outcome
OOO
Lndeterminate
Evidence was limited to
two studies reporting no
findings of ocular toxicity.
C-168
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.12 Dermal
EPA identified 1 epidemiological and 2 animal studies that investigated the association between
PFOA and dermal effects. The 1 epidemiological study was classified as medium confidence
(Section C.12.1). Of the animal studies, 2 were classified as high 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 PFOA 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 PFOA exposure. In the Puberty Cohort, a large sub-cohort of the DNBC in Denmark,
Ernst et al. (2019, 5080529) examined the association between prenatal PFOA exposure and
pubertal development in. 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 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and dermal effects. Study quality evaluation for this 1 study is shown
in Figure C-57.
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-169
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure C-57. Summary of Study Evaluation for Epidemiology Studies of PFOA and
Dermal Effects
Interactive figure and additional study details available on HAWC.
C. 12.1.3 Findings
Results for the studies that examined dermal outcomes are presented in Appendix D. Ernst et al.
(2019, 5080529) observed negative associations between prenatal PFOA exposure and age at the
occurrence of acne. Significant negative associations were observed for girls per doubling of
PFOA (P: -5.16; 95% CI: -8.50, -1.82), and in the highest tertile of PFOA exposure compared
to the lowest (P: -6.09; 95% CI: -12.10, -1.70) {Ernst, 2019, 5080529}. Associations in boys
were negative and non-significant.
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 PFOA HESD {U.S. EPA, 2016, 3603279} that investigated the
association between PFOA and dermal effects. Study quality evaluations for these 2 studies are
shown in Figure C-58.
C-170
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
NTP, 2019, 5400977-
++
++
I
NR
NTP, 2020, 7330145-
++
++
NR
Legend
Good (metric) or High confidence (overall)
Adequate (metric) or Medium confidence (overall)
Deficient (metric) or Low confidence (overall)
H Critically deficient (metric) or Uninformative (overall)
Not reported
vOO
,ce
Figure C-58. Summary of Study Evaluation for Toxicology Studies of PFOA and Dermal
Effects
Interactive figure and additional study details available on HAWC.
There is no evidence in the literature that oral PFOA exposure results in dermal toxicity in
animal models. An NTP (2019, 5400977) study explored histopathology of the skin following 28
days of oral gavage of up to 10 mg/kg/day PFOA in male and up to 100 mg/kg/day PFOA in
female Sprague Dawley rats. They observed no lesions of dermal tissue. Similarly, in a
subsequent report, NTP (2020, 7330145) reported no lesions in dermal tissue in male or female
Sprague Dawley rats that received PFOA via feed for 2 years (See Main PFOA Document for
study design details).
C.12.3 Mechanistic Evidence
There was no mechanistic evidence linking PFOA exposure to adverse dermal outcomes in the
2016 PFOA HESD {U.S. EPA, 2016, 3603279}. There are 2 studies from recent systematic
literature search and review efforts conducted after publication of the 2016 PFOA HESD that
investigated the mechanisms of action of PFOA that lead to dermal effects. A summary of these
studies is shown in Figure C-59. Additional mechanistic synthesis will not be conducted since
evidence is inadequate to infer that PFOA may cause dermal effects.
C-171
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Mechanistic Pathway
In Vitro
Grand Total
Cell Growth, Differentiation, Proliferation, Or Viability
2
Extracellular Matrix Or Molecules
Inflammation And Immune Response
Oxidative Stress
2
Grand Total
2
2
Figure C-59. Summary of Mechanistic Studies of PFOA and Dermal Effects
Interactive figure and additional study details available on Tableau.
The evidence evaluating an association between PFOA exposure and dermal effects in humans is
indeterminate based on the limited number of studies available. The 2016 HESD for PFOA {U.S.
EPA, 2016, 3603279} did not report on the association between oral PFOA exposure and dermal
effects. In this updated review, one epidemiological study examined the association between
PFOA exposure and dermal effects during puberty and observed an inverse association with age at
the occurrence of acne, which was significant only in girls, suggesting earlier occurrences of acne
with increasing PFOA exposure.
The animal evidence for an association between PFOA exposure and dermal effects is
indeterminate. There are two high confidence studies that evaluated the skin as part of the
histopathological evaluation that observed no dermal lesions. There are no reported biological
consequences of oral PFOA exposure on dermal tissue in animal models.
C. 12.4.1 Evidence Integration Judgment
Overall, there is inadequate evidence to assess whether PFOA exposure can cause dermal effects
in humans under relevant exposure circumstances (Table C-17).
C.12.4 Evidence Integration
C-172
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-17. Evidence Profile Table for PFOA 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 a
significant inverse
association with age of
acne onset in adolescents,
but this was significant
only in girls.
• Medium confidence
study
• Limited number of OOO
studies examining Indeterminate
outcome
• Inconsistent directions of Evidence was limited to
effects across sexes one study reporting
associations that vary in
significance by sex.
Evidence from In Vivo Animal Studies (Section C.12.2)
Histopathology
2 High confidence
studies
No changes in dermal
histopathology were
reported in one 28-day
and one chronic study in
male and female rats.
• High confidence
studies
• Limited number of
studies examining
outcome
OOO
Indeterminate
Evidence was limited to
two studies 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-173
-------
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
01234567
OR for Q2
(0.07-0.16
Later Souzaetal.. Maternal Case Median=0.20 ng/mL — , .. . 1.26
' na/mL) vs 01 f<
pregnancy 2020 Blood Control (25th-75th _ . ,
^ 0 07ng/mL)
percentile: 0.16-0.38
ng/mL)
1
1
1
1
1
1
1
1 •
1
1
1
1
1
1
OR for Q3
(0.16-0.22
ng/mL) vs Q1 {<
0.07 ng/mL)
1
1
1
1
1
1
1
1 •
1
1
1
1
1
1
ORfor Q4( >
0.22 ng/mL) vs
01 (< 0.07 281
ng/mL)
1
1
1
1
1
1
1
1 •
1
1
1
1
1
1
Median (25th-75th ^
Xuetal., „ J„1 Cross percentiles)= 1.05 ;^er
Cord Blood ng/mL (0.779-L33 " h9 ppoi,
ng/mL) changS ,n PFOA)
1
1
1
1
1
1
1
~ 1
1
1
1
1
1
1
01234567
Figure D-l. Odds of Small-for-gestational-age in Children from Low Confidence Epidemiology Studies Following Exposure to
PFOA
D-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Interactive figure and additional study details available on Tableau.
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 PFOA Exposure and Developmental Effects in Recent Epidemiological Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Ashley-Martin
et al. (2017,
3981371)
High
Canada, Cohort
2008-2011
Pregnant women Maternal blood
(enrolled if <14 Early pregnancy
weeks gestation,
>18 years of
age) and their
infants at
recruitment and
from MIREC
N = 1,509
1.7 (1.2-2.4)
BW (z-score): Regression
adequate, coefficient per
inadequate, and loglO-unit
excess weight gain increase in
PFOA
BW: -0.1 (-0.34,0.13)
Females: -89.51 (-263.4,
84.38)
Males:-35.51 (-198.99,
127.97)
BW z-core:
Adequate weight gain: -0.36
(-0.85,0.11)
Excess weight gain: -0.08 (-
0.44, 0.27)
Inadequate weight gain: -
0.08 (-0.78, 0.63)
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, eachPFAS0
Bach et al.
(2016, 3981534)
High
Denmark, Cohort
2008-2013
Pregnant women Maternal serum
and their infants Early pregnancy
from the Aarhus 2.0 (1.5-2.6)
Birth Cohort
N = 1,507
BL (cm), BW
(g, z-score),
gestational
length
(weeks), HC
(cm), PTB
Regression
coefficient per
IQR increase in
PFOA and by
quartiles
OR per 0.1-unit
increase in
PFOA, per IQR
increase, and
by quartiles
BL: 0.1 (-0.1,0.2)
Q2: 0 (-0.4, 0.4)
Q3:0(-0.4, 0.4)
Q4: 0.1 (-0.3,0.4)
BW (g): 7 (-10, 23)
Q2: 3 (-54, 59)
Q3: 15 (-42, 72)
Q4: 9 (-47, 64)
BW (z-score): 0.02 (-0.02,
0.06)
D-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Q2: 0.009 (-0.13, 0.14)
Q3: 0.04 (-0.09, 0.17)
Q4: 0.02 (-0.1, 0.16)
Gestational length: 0.1 (0,
0.2)
Q2: 0 (-0.3, 0.2)
Q3: 0.1 (-0.2,0.3)
Q4: 0.1 (-0.2,0.4)
HC: 0.1(0,0.2)
Q2: 0 (-0.2, 0.3)
Q3: 0.1 (-0.2,0.4)
Q4: 0.1 (-0.1,0.4)
Results: Lowest quartile used as reference.
Confounding: Maternal age, pre-pregnancy BMI and educational level, GA
Bell et al. (2018, United Cross-sectional Singleton and Blood
5041287) States, twin infants born Later pregnancy
High 2008-2010 in from Upstate Singletons: 1.10
KIDS (0.69-1.63)
N = 2,071 Twins: 1.01
singletons; 1,040 (0.69-1.53)
twins
BL (cm), BW (g),
GA (weeks), HC
(cm), ponderal
index
Regression
coefficient per
log(PFOA+l)
unit increase
BL
S: 0.02 (-0.13, 0.17)
T: 0.21 (-0.11,0.52)
BW
S:-11.55 (-35.72, 12.62)
T: 18.48 (-17.18, 54.13)
GA
S: 0.01 (-0.07, 0.08)
T: -0.01 (-0.12,0.11)
HC
S: 0.04 (-0.17,0.26)
T: 0.12 (-0.22, 0.46)
Ponderal index
S: -0.01 (-0.03, 0.01)
T: -0.01 (-0.04, 0.02)
Results: S = Singletons; T = Twins
D-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location Population, Exposure Matrix,
' Design Ages, Sample Timing, Outcome Comparison
Years N Levels3
Resultsb
Comparison: Logarithm base not specified.
Confounding: Maternal age, maternal BMI, maternal education, infertility treatment, parity
Bjerregaard-
Olesen et al.
(2019, 5083648)
High
Denmark, Cohort Pregnant women Maternal serum
2011-2013 and their Early pregnancy
children from IQR = 0.92
FETOTOX
N = 671
BL (cm), BW (g),
HC (cm)
Regression BL
coefficient per 0.1 (-0.1, 0.2)
IQR increase in Females: -0.2 (-0.5, 0)
PFOA Males: 0.2 (0, 0.3),
Interaction p-value = 0.008
BW
18 (-9, 45)
Females: -23 (-78, 31)
Males: 31 (6, 56)
HC
0.1 (0,0.2)
Females: -0.1 (-0.3, 0.1)
Males: 0.2(0.1,0.3),
Interaction p-value = 0.004
Confounding: Age at delivery, pre-pregnancy BMI, educational level, smoking, alcohol intake, GA at birth
Buck Louis et al.
(2018, 5016992)
High
United
States,
2009-2013
Cohort
Pregnant women Maternal blood
(age range 18- Early pregnancy
40 years) with
singleton
pregnancies
from the NICHD
Fetal Growth
Studies
N = 2,106
1.985 (1.297-
3.001)
BL (cm), BW (g),
GA at delivery
(weeks), HC (cm),
umbilical
circumference
(cm), upper arm
length (cm), upper
thigh length (cm)
Regression
coefficient per
SD increase in
log-PFOA
BL: -0.23 (-0.35,-0.1)
BW: -5.9 (-28.75, 16.94)
GA: 0.01 (-0.08,0.1)
HC:-0.04 (-0.12, 0.03)
Umbilical circumference: -
0.06 (-0.19, 0.07)
Upper arm length: -0.02 (-
0.07, 0.03)
Upper thigh length: -0.19 (-
0.26,-0.12)
NICHD = National Institute of Child Health and Human Development
Comparison: Logarithm base not specified.
Confounding: Maternal age, education, pre-pregnant body mass index, serum cotinine, infant sex, chemical-maternal race/ethnic
interaction, mode of delivery
D-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Chu et al. (2020,
6315711)
High
China,
2013
Cohort
Pregnant women Maternal serum
(aged 18-45 Later pregnancy
years) and
infants from
Guangzhou
Birth Cohort
Study
N = 372
1.538(0.957-
2.635)
Girls: 1.497
(0.920-2.642)
Boys: 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
PFOA or by
quartiles
BW
-73.64 (-126.39, -20.88)
Girls: -56.04 (-129.32,
17.24)
Boys: -71.8 (-148.61, 5.00)
p-value for interaction by sex
= 0.958
GA
-0.21 (-0.44, 0.02)
Girls: -0.53 (-0.83, -0.23)
Boys: 0.17 (-0.16, 0.51)
p-value for interaction by sex
= 0.002
LBW
1.16 (0.52, 2.58)
Q2: 0.61 (0.14,2.69)
Q3: 0.27 (0.05, 1.42)
Q4: 1.00 (0.23,4.35)
p-trend = 0.007
PTB
1.49 (0.94, 2.36)
Q2: 0.71 (0.23,2.14)
Q3: 1.60 (0.60,4.23)
Q4: 1.84 (0.72,4.71)
p-trend = 0.273
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
and their estimated fetal per twofold
children from 2.35 (1.6-3.30) weight at 12 increase in
INMA study PFOA
AC
12 wk: 0.8 (-2.4, 4.0)
Girls: 2.9 (-1.7, 7.2)
Boys: -1.5 (-6.0, 2.8)
D-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
N = 1,230 (Girls
= 597, Boys =
633)
weeks, 20 weeks,
and 34 weeks
20 wk:-0.5 (-3.7, 2.8)
Girls: 2.7 (-1.9, 6.9)
Boys:-3.1 (-7.5, 1.2)
34 wk: (1.1 (-2.1,4.3)
Girls: 1.2 (-3.2, 5.4)
Boys: 1.1 (-3.3, 5.4)
FL
12 wk: 1.9 (-1.4, 5.2)
Girls: 4.2 (-0.5, 8.3)
Boys: -0.6 (-5.0, 3.8)
20 wk:-1.4 (-4.6, 1.9)
Girls: 0.2 (-4.3, 4.6)
Boys: -3.0 (-7.5, 1.3)
34 wk:-0.2 (-3.5, 3.1)
Girls:-1.8 (-6.3, 2.7)
Boys: 1.2 (-3.4, 5.5)
BPD
12 wk: -0.5 (-5.6, 4.5)
Girls: 3.9 (-0.7, 8.2)
Boys:-4.7 (-11.1, 1.8)
20 wk: 0.0 (-3.2, 3.3)
Girls: 2.9 (-1.5, 7.3)
Boys:-2.6 (-7.1, 1.8)
34 wk: 1.9 (-1.3, 5.1)
Girls: 1.6 (-2.9, 6.0)
Boys: 2.2 (-2.4, 6.6)
Estimated Fetal Weight
12 wk: 1.2 (-2.1,4.4)
Girls: 3.3 (-1.4, 7.5)
Boys: -1.2 (-5.7, 3.2)
20 wk: -0.8 (-4.0, 2.4)
Girls: 2.0 (-2.5, 6.4)
Boys: -3.5 (-8.0, 0.9)
D-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
34 wk: 1.3 (-1.9,4.5)
Girls: 0.7 (-3.8, 5.0)
Boys: 2.1 (-2.4, 6.4)
INMA = INfancia y Medio Ambiente (Environment and Childhood) Project
Confounding: Cohort, parity, maternal age, country of birth, smoking at week 12, maternal pre-pregnancy BMI, studies,
season of last menstrual period
Darrow et al.
(2013, 2850966)
High
United
States
2005-2011
Cohort
Pregnant women Maternal serum at
from the C8HP enrollment
exposed through 14.3(8.0-29.8)
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
LBW, BW (g),
PTB
OR (LBW,
PTB),
regression
coefficient
(BW) per ln-
unit increase in
LBW
All births
Per ln-unit increase: 0.94
(0.75, 1.17)
Per IQR increase: 0.95 (0.85,
1.06)
PFOA, per IQR Q2: 0.94 (0.45, 1.98)
increase in Q3: 0.99 (0.48, 2.05)
PFOA, or by Q4: 1.25 (0.63, 2.46)
quintiles Q5: 0.92 (0.44, 1.95)
First prospective birth
Per ln-unit increase: 1.07
(0.78, 1.47)
Per IQR increase: 0.99 (0.87,
1.12)
Q2: 0.82 (0.23, 2.85)
Q3: 1.03 (0.35,3.06)
Q4: 1.86 (0.67,5.14)
Q5: 1.06 (0.32, 3.54)
BW
All births
Per ln-unit increase: -8 (-28,
12)
Per IQR increase: -5 (-13, 2)
Q2: 35 (-33, 105)
Q3:-9 (-79,61)
Q4: 4 (-65, 72)
Q5: 0 (-68, 69)
p-trend = 0.701
D-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
First prospective birth
Per ln-unit increase: 5 (-22,
33)
Per IQR increase: 1 (-10, 11)
Q2: 135 (34, 276)
Q3: 26 (-71, 124)
Q4: 56 (-37, 149)
Q5: 74 (-20, 169)
p-trend = 0.622
PTB
All births
Per ln-unit increase: 0.93
(0.78, 1.1)
Per IQR increase: 0.95 (0.88,
1.04)
Q2: 1.56 (0.88,2.76)
Q3: 1.19(0.66,2.14)
Q4: 1.21 (0.67,2.19)
Q5: 1.01 (0.55, 1.86)
p-trend = 0.629
First prospective birth
Per ln-unit increase: 1.09
(0.86, 1.37)
Per IQR increase: 1.01 (0.92,
1.1)
Q2: 1.11 (0.42,2.89)
Q3: 1.30 (0.51,3.27)
Q4: 1.49 (0.62, 3.61)
Q5: 1.32 (0.53,3.32)
p-trend = 0.409
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.
D-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Eick et al.
(2020, 7102797)
High
United
States
2014-2018
Second trimester Maternal serum
pregnant women from the second
from the CIOB trimester
cohort
BW (g)
N = 461
GA, BW (z-
score), PTB
N = 506
0.76(0.46-1.12)
Cohort Second trimester Maternal serum BW (g, z- Regression BW (g)
score), GA coefficient by T2: 62.93 (-42.94, 168.8)
(weeks), PTB tertile T3: 86.07 (-36.31, 208.45)
PTB: OR by BW (z-score)
tertile T2: 0.13 (-0.10, 0.35)
T3: 0.12 (-0.14,0.37)
GA
T2: -0.29 (-0.74,0.17)
T3:-0.10 (-0.63, 0.43)
PTB
T2: 1.79 (0.75, 4.28)
T3: 2.37 (0.88,6.38)
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, and food
insecurity.
Gardener et al.
(2021,7021199)
High
United
States
Recruitment
: 2009
Cohort
Pregnant women Maternal serum
in third trimester from primarily third
(ages 18-49)
and children at
birth from the
Vanguard Pilot
Study of the
NCS
GA at birth
N = 433
BW
N = 403
trimester
1.4 (0.9-2.0)
GA at birth
(weeks), BW
(z-score), GA
<37 weeks
GA at birth and GA at birth
BW:
Mean by
quartile
GA <37 weeks
andBW: OR
by quartile
Mean
Ql: 38.94 (38.60,39.27)
Q2: 38.53 (38.19, 38.88)
Q3: 38.67 (38.35,38.98)
Q4: 38.85 (38.49, 39.20)
p-trend = 0.79
BW
Mean
Ql: -1.35 (-4.69,2.02)
Q2: 0.41 (-3.00, 3.86)
Q3: 0.75 (-2.38,3.91)
Q4: 1.95 (-1.5,5.41)
p-trend = 0.20
OR
D-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Q2: 1.2 (0.56, 2.59)
Q3: 0.84 (0.40, 1.80)
Q4: 0.91 (0.41,2.02)
p-trend = 0.62
GA <37 weeks
OR
Q2: 3.17(0.94, 10.7)
Q3: 3.14(0.95, 10.31)
Q4: 1.38 (0.32, 5.97)
p-trend = 0.53
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,
2008-2009
Cohort
BW (g)
Mother-newborn Cord blood
pairs from
FLEHSII 1.52 nL (1.10-
N = 248 2.10 nL)
FLEHS II = Flemish Environmental and Health Study II
Confounding: GA, child's sex, smoking of the mother during pregnancy, parity, maternal pre-pregnancy
Regression
coefficient per
IQR increase in
PFOA
-34.5 (-129.02, 60.02)
BMI
Huo et al. (2020, China, Cohort
6835452) 2013-2016
High
Mothers (aged > Maternal blood GA (weeks), PTB Regression
20 years) and
their children
from the
Shanghai Birth
Cohort
N = 2,849
Later pregnancy
11.85 (9.20-
15.26)
(indicated, non-
spontaneous,
spontaneous, and
overall)
coefficient
(GA) and OR
(PTB) per ln-
unit increase in
PFOA and per
tertile
GA:0 (-0.14, 0.13)
Tl: 0.11 (-0.31,0.54)
T2: -0.69 (-1.75, 0.37)
T3: 0.03 (-0.29,0.35)
OR T2: 0.11 (-0.03,0.24)
OR T3:-0.01,-0.15, 0.12)
PTB, overall: 0.92 (0.61,
1.33)
Females: 0.82 (0.44, 1.55)
Males: 1.02 (0.59, 1.78)
PTB, indicated: 1.71 (0.8,
3.67)
T2: 0.96 (0.44,2.11)
D-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
T3: 1.02 (0.47,2.22)
PTB, non-spontaneous:
Females: 2.64 (0.83, 8.39)
Males: 1.23 (0.44, 3.39)
PTB, spontaneous: 0.73
(0.45, 1.19)
T2: 0.71 (0.43, 1.17)
T3: 0.76 (0.46, 1.22)
Females: 0.54 (0.26, 1.13)
Males: 0.95 (0.49, 1.81)
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
Lauritzen et al. Norway and Cohort
(2017,3981410) Sweden,
High 1986-1988
Mother-infant
pairs from
NICHD SGA
N = 424 (265
from Norway,
159 from
Sweden (78
girls, 81 boys))
Maternal serum
Later pregnancy
Norway: 1.62
(Range = 0.31—
7.97)
Sweden: 2.33
(Range = 0.60-
6.70)
BL (cm), BW (g),
GA (weeks), HC
(cm), SGA
Regression
coefficient or
OR (SGA) per
ln-unit increase
in PFOA
BL
-0.49 (-0.99, 0.02); p-value
= 0.06
NO: -0.1 (-0.7, 0.4); p-value
= 0.656
SE: -1.3 (-2.3,-0.3); p-
value = 0.01
SE-girls: -0.8 (-2.4, 0.8); p-
value = 0.34
SE-boys: -1.6 (-2.9, -0.4)
BW
-81.7 (-202, 39.2); p-value =
0.185
NO: 37 (-99, 174); p-value =
0.59
SE: -359 (-596, -122), p-
value = 0.003
SE-girls:-156 (-541,228);
p-value = 0.419
D-12
-------
DRAFT FOR PUBLIC COMMENT
T Population,
Reference, Location, . .
^ , Design Ages9
Confidence Years
MARCH 2023
Exposure Matrix,
Sample Timing, Outcome Comparison Resultsb
Levels"
SE-boys: -526 (-828, -222);
p-value = 0.001
GA
-0.20 (-0.34, 0.14); p-value
= 0.255
NO: -0.2 (-0.6, 0.2); p-value
= 0.431
SE: -0.3 (-0.9, 0.3); p-value
= 0.318
SE-girls: -0.1 (-1.1, 0.9); p-
value = 0.802
SE-boys: -0.4 (-1.2, 0.5); p-
value = 0.365
HC
-0.02 (-0.32, 0.27)
NO: 0.2 (-0.2, 0.5); p-value
= 0.354
SE: -0.4 (-1.0, 0.1); p-value
= 0.115
SE-girls: -0.1 (-1.0, 0.7); p-
value = 0.728
SE-boys: -0.6 (-1.3, 0.1); p-
value = 0.103
SGA
1.21 (0.69, 2.11)
NO: 0.66 (0.33, 1.33); p-
value = 0.246
SE: 5.25 (1.68, 16.4); p-value
= 0.004
SE-girls: 4.73 (0.79, 28.3); p-
value = 0.089
SE-boys: 6.55 (1.14,37.45);
p-value = 0.035
D-13
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
NICHD SGA = The US National Institute of Child Health and Human Development (NICHD) Scandinavian Successive SGA Births
Study
Outcome: SGA defined as BW below the 10th percentile for GA, sex, and parity.
Results: NO = Norway; SE = Sweden
Confounding: Maternal age, height, pre-pregnancy BMI, education, parity, smoking status at conception, interpregnancy interval,
offspring sex
Lind et al. Denmark Cohort
(2017,3858512) 2010-2012
High
Infants
prenatally
exposed to
PFAS from the
Odense Child
Cohort
N = 212 girls,
299 boys
Maternal serum
Early pregnancy
1.7(1.1-2.3)
BW (g), HC (cm),
gestational length
(days)
Regression
coefficient per
ln-unit increase
in PFOA or by
quartiles
BW
Males: -5 (-92, 82)
p-trend by quartiles = 0.88
Females: 6 (-90, 102)
p-trend by quartiles = 0.88
HC
Males: (-0.3, 0.3)
p-trend by quartiles = 0.80
Females: 0.1 (-0.3, 4)
p-trend by quartile = 0.72
Gestational length
Males
Continuous: -0.7 (-2.9, 1.5)
Q2: 1.0 (-2.4,4.4)
Q3: 2.7 (-0.9,6.3)
Q4: -0.9 (-4.6, 2.7)
p-trend by quartiles = 0.63
Females
Continuous: -1.5 (-4.3, 1.3)
Q2
Q3
Q4
1.1 (-2.7,4.9)
-2.7 (-6.3, 1.2)
-3.6 (-8.0, 0.8)
p-trend by quartiles = 0.04
BW and HC:Quartile analysis
did not show any statistically
significant associations
D-14
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Results: Lowest quartile used as reference.
Confounding: Age at examination, weight for age z-score, pre-pregnancy BMI, parity, smoking
Luo et al. (2021
9959610)
High for B W
Medium for
birth length,
ponderal index
Manzano-
Salgado et al.
(2017, 4238465)
High
China,
2017-2019
Cohort
Mother-newborn Maternal blood
pairs within three days of
N = 224 delivery
3.51 (2.23-4.80)
Regression
coefficient per
ln-unit
BW (g), BL
(cm), ponderal
index (kg/m3)
increase in
PFOA
Confounding: Maternal age, prepregnancy BMI, education, parity, environmental tobacco smoke exposure, alcohol drinking, GA, and
newborn sex.
BW: -62.37 (-149.08, 24.35)
BL: 0.08 (-0.36, 0.52)
Ponderal index: -0.61 (-1.15,
-0.06), p-value <0.05
Spain,
2003-2008
Cohort
Mother (aged
>16 years)-child
pairs from
INMA
N = 1,202
Maternal plasma BL (cm), BW (g)
Early pregnancy GA (weeks), HC
Mean = 2.35
(SD = 1.25)
Regression
coefficient and
(cm), LBW, LBW OR per
at term, PTB, SGA doubling of
PFOA and per
quartiles
BL:-0.01 (-0.15, 0.14)
Q2: 0.01 (-0.28, 0.29)
Q3:-0.06 (-0.36, 0.24)
Q4: -0.03 (-0.34, 0.28)
Females: 0.04 (-0.16, 0.24)
Males: 0.01 (-0.18, 0.21)
BW: -9.33 (-38.81,20.16)
Q2: -29.6 (-92.82, 33.63)
Q3: -32.99 (-97.08, 31.09)
Q4:-32.77 (-97.65, 32.11)
Females: 13.81 (-26.67,
54.3)
Males: -24.75 (-66.71,
17.22)
GA:-0.05 (-0.16, 0.07)
Q2: -0.05 (-0.29, 0.2)
Q3: 0.03 (-0.23, 0.28)
Q4:-0.12 (-0.37, 0.17)
Females: -0.08 (-0.24, 0.08)
Males:-0.04 (-0.2, 0.13)
HC:-0.07 (-0.17, 0.03)
Q2: -0.01 (-0.22,0.19)
D-15
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Q3: 0.04 (-0.17, 0.25)
Q4:-0.16 (-0.38, 0.06)
Females: 0.03 (-0.1, 0.17)
Males: -0.13 (-0.27, 0)
LBW: 0.9 (0.63, 1.29)
Females: 0.76 (0.48, 1.21)
Males: 1.12(0.64, 1.99)
LBW at term: 0.85 (0.53,
1.34)
Females: 0.62 (0.36, 1.06)
Males: 1.67 (0.72, 3.86),
interaction p-value = 0.05
PTB: 0.92 (0.72, 1.19)
Females: 1.19(0.62,2.31)
Males: 0.74 (0.43, 1.25)
SGA: 0.92 (0.72, 1.19)
Females: 0.72 (0.5, 1.04)
Males: 1.18(0.82, 1.69),
interaction p-value = 0.08
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
Sagiv et al.
(2018, 4238410)
High
United
States,
1999-2002
Cohort
Pregnant women Maternal blood
and infants from Early pregnancy
Project Viva 5.8 (IQR =3.8)
N = 1,644
BW-for-GA (z-
score), gestational
length (weeks),
PTB
Regression
coefficient per
IQR increase Q2
and by quartiles Q3
Q4
PTB:
OR per IQR
increase and by
quartiles
BW-for-GA: -0.02 (-0.08,
0.03)
:-0.04 (-0.17, 0.09)
:-0.12 (-0.25, 0.02)
: -0.07 (-0.21, 0.07)
Gestational length: -0.05 (-
0.16,0.06)
Q2: 0.05 (-0.22, 0.32)
D-16
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Q3:0 (-0.28, 0.28)
Q4: -0.04 (-0.33, 0.24)
PTB: 1 (0.9, 1.3)
Q2: 1.1 (0.6, 2)
Q3: 1.1 (0.6, 1.9)
Q4: 1.2 (0.7,2.2)
BW-for-GA and gestational
length: no statistically
significant associations by
sex
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
5.5 (3.8-7.7)
BW (z-score),
length-for-age (z-
score), rapid
weight gain,
weight-for-age (z-
score), weight-for-
length (z-score)
Regression
coefficient by
tertile (per
doubling in
PFOA)
Rapid weight
gain: Relative
risk by tertile
BW
T2: 0.18 (-0.06,0.42)
T3: -0.15 (-0.4,0.1)
Length-for-age
T2: 0.19 (-0.2,0.5)
T3:-0.32 (-0.72, 0.07)
Weight gain
T2: 1.08 (0.78, 1.5)
T3: 0.8 (0.56, 1.15)
Weight-for-age
T2: -0.02 (-0.34, 0.29)
T3:-0.46 (-0.78,-0.14), p-
trend < 0.01
Weight-for-length
T2: -0.31 (-0.56,-0.06)
D-17
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
T3:-0.34 (-0.59,-0.08), p-
trend = 0.02
BW, length-for-age, and
weight gain: no statistically
significant trends
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.
(2017, 3858473)
High
United
States,
2009-2014
Cohort
Pregnant women Maternal serum
(aged >16 years)
Adiposity (% fat
mass), BW (g)
and infants from
Healthy Start at
birth
N = 628
1.1 (0.7-1.6)
Regression
coefficient per
ln-unit increase
PFOA and by
tertiles
Adiposity: -0.43 (-0.91,
0.04)
T2: -0.34 (-1.06, 0.38)
T3:-0.97 (-1.74,-0.2)
BW:-51.4 (-97.2,-5.7)
T2: -15.9 (-84.9, 53.2)
T3: -92.4 (-166.2, -18.5)
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
(aged >16 years)
and infants from
Healthy Start
assessed up to 5
months
N = 415 (202
girls, 213 boys)
Maternal serum
1.0 (0.7-1.6)
Adiposity (%),
weight-for-age z-
score (WAZ),
weight-for-length
z-score (WLZ),
WAZ and WLZ
growth from birth
to 5 months, rapid
growth in WAZ or
WLZ
Regression
coefficient per
ln-unit increase
in PFOA and
by tertiles
Rapid growth:
OR per ln-unit
increase in
PFOA
Adiposity: 0.76 (-0.03, 1.55)
T2: 1.4(0.18,2.62)
T3: 1.16 (-0.18,2.49)
Females: 0.27 (-0.85, 1.4)
T2: 1.71 (-0.06, 3.48)
T3: 0.03 (-1.77, 1.83)
Males: 1.53 (0.35,2.71)
T2: 1.2 (-0.56, 2.97)
T3: 2.81 (0.79,4.84)
p-value for sex interaction =
0.07
WAZ: 0.01 (-0.14, 0.15)
D-18
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
T2: 0.17 (-0.05,0.39)
T3: 0.08 (-0.16,0.32)
Females: -0.14 (-0.34, 0.06)
T2: 0.01 (-0.3,0.33)
T3:-0.18 (-0.51, 0.14)
Males: 0.17 (-0.05, 0.39)
T2: 0.31 (-0.01,0.63)
T3: 0.38 (0.01,0.75)
No statistically significant
interaction by sex
WLZ: 0.01 (-0.16,0.18)
T2: 0.1 (-0.16,0.35)
T3: 0.07 (-0.21, 0.35)
Females:-0 11 (-0.34,0.12)
T2: -0.01 (-0.38, 0.35)
T3:-0.17 (-0.55, 0.2)
Males: 0.14 (-0.11, 0.39)
T2: 0.17 (-0.21,0.55)
T3: 0.33 (-0.1,0.76)
WAZ, growth from birth:
0.07 (-0.08, 0.21)
WAZ, rapid growth: 1.25
(0.77, 2.04)
WLZ, growth from birth:
0.09 (-0.10, 0.27)
WLZ, rapid growth: 1.43
(0.92, 2.22)
Outcome: Rapid growth defined as change in WAZ or WLZ >0.67 between birth and 5 months
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
Tanner et al.
(2020, 6322293)
High
Sweden,
Recruitment
:2007-
Cohort
Mother-infant Maternal serum Age of infant PGV Regression Age of infant PGV: 0.58
pairs from (months), infant coefficient per (0.17, 0.99), p-value = 0.01
SELMA study growth slope loglO-unit
D-19
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
2010;
followed up
to 5.5 years
N = 1,334
GM = 1.6
(Range = 0.2-
21.1)
(loglO), infant
PGV (loglO),
infant spurt
duration (loglO),
infant weight
plateau (kg)
increase in Growth slope: -0.06 (-0.11,
PFOA -0.01), p-value = 0.02
PGV: -0.02 (-0.05, 0.02)
Spurt duration: 0.06 (0.01,
0.11), p-value = 0.02
Weight plateau: 0.81 (0.21,
1.41), p-value = 0.01
SELMA = Swedish Environmental Longitudinal Mother and Child, Asthma and Allergy
Outcome: PGV = peak growth velocity
Confounding: Sex, PTB, mother's age, weight, parity, and smoking
Valvi et al. Faroe
(2017,3983872) Islands
High 1997-2000
Cross-sectional Pregnant women Maternal serum
and their Later pregnancy
children 3.31 (2.54-3.99)
N = 604 (288
girls, 316 boys)
HC (cm), body Regression
length (cm), coefficient per
BW (g) doubling of
PFOA
HC
0 (-0.22, 0.23)
Girls: 0.10 (-0.23,0.44)
Boys: -0.05, (-0.36, 0.26)
p-value for sex interaction =
0.90
Body length
0.03 (-0.29, 0.35)
Girls: -0.01 (-0.48, 0.46)
Boys: 0.02 (-0.42, 0.47)
p-value for sex interaction =
0.64
BW
-11 (-88, 67)
Girls: 58 (-48, 164)
Boys: -71 (-184, 42)
p-value for sex interaction =
0.04
Confounding: Maternal age at delivery, education, parity, pre-pregnancy BMI, smoking during pregnancy, child sex
Wang et al.
(2016, 3858502)
High
Taiwan
Recruitment
20002001,
Cohort
Children from Maternal serum HC (cm), BL (cm), Regression HC
Taiwan Maternal Later pregnancy BW (kg), SGA, coefficient per Girls: 0.11 (-0.26, 0.47)
and Infant Girls: 2.34 height z-score at ln-unit increase Boys: 0.06 (-0.24, 0.36)
Cohort Study, (1.57-3.43) each age, average
D-20
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing, Outcome
Levels"
Comparison
Resultsb
assessment
assessed at ages
Boys: 2.37 childhood height
in PFOA or by
BL
up to age 11
2, 5, 8, and 11
(1.35-3.47) z-score, weight z-
quartiles
Girls:
-0.32 (-0.92, 0.28)
years
score, average
SGA: OR per
Boys:
0.31 (-0.22,0.84)
N = 106 girls,
childhood weight
ln-unit increase
117 boys
z-score
in PFOA
BW
Girls:
-0.08 (-0.18,0.01)
Boys:
0.04 (-0.05,0.12)
SGA
Girls:
1.48 (0.63, 3.48)
Boys:
0.63 (0.32, 1.13)
Girls' analysis by quartiles:
no statistically significant
associations
Height and weight z-scores
by age: NR, no significant
interactions for either sex (p-
value > 0.10)
Outcome: SGA defined as BW below the 10th percentile for GA by sex using 1998-2002 Taiwan nationwide singleton BW charts.
Results: Lowest quartile used as reference.
Confounding: Family annual income, maternal age at delivery, maternal education, maternal previous live children, maternal pre-
pregnancy BMI
Whitworth et al.
Norway Cohort
Pregnant women Maternal plasma
PTB, BW (z-
Regression
PTB
(2012, 2349577)
2003-2004
and their
Around 17 weeks of
score), LGA,
coefficient and
Q2: 0.3 (0.1, 1.3)
High
children from
gestation
SGA
OR per unit
Q3: 0.7 (0.2,2.4)
MoBa
2.2 (1.7-3.0)
increase in
PFOA, or by
Q4: 0.1 (0.03,0.6)
p-trend = 0.02
PTB, LGA,
quartile
SGA
LGA
N = 901
Q2: 0.9 (0.5, 1.7)
BW
Q3: 1.0 (0.5, 1.9)
N = 849
Q4: 0.6 (0.3, 1.4)
p-trend = 0.33
D-21
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
SGA
Q2: 0.8 (0.3,2.3)
Q3: 1.3 (0.5,3.2)
Q4: 1.0 (0.3,2.8)
p-trend = 0.92
BW
Per unit increase: -0.03
(-0.10,0.04)
Q2
Q3
Q4
-0.06 (-0.28,0.16)
-0.08 (-0.32,0.16)
-0.21 (-0.45,0.04)
p-trend = 0.10
MoBa = Norwegian Mother and Child Cohort Study
Results: Lowest quartile used as reference.
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, prepregnancy BMI, parity. Additional confounding for BW: Weight gain at 17 weeks.
Wikstrom et al. Sweden Cohort
(2020,6311677) 2007-2010
High
Infants exposed Maternal serum BW (g), BW-SDS, Regression BW
prenatally to
PFAS from the
SELMA study
N = 1533 (732
girls, 801 boys)
Early pregnancy
1.61 (1.11-2.30)
SGA
coefficient
(BW, BW-
SDS) or OR
(SGA) per ln-
unit increase in
PFOA or by
quartiles
Per increase:
-68 (-112,-24)
Q2: 27 (-35, 89)
Q3: -41 (-106, 23)
Q4: -90 (-159, -91)
Girls
Per increase:
-86 (-145, -26)
Q2: 30 (-55, 115)
Q3: -36 (-124, 52)
Q4:-136 (-231,-40)
Boys
Per increase: -49 (-113,
15)
Q2
Q3
Q4
26 (-66, 116)
-44 (-139, 50)
-47 (-147, 54)
D-22
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
BW-SDS
Per increase: -0.152 (-0.251,
-0.052)
Q2: 0.065 (0.076, 0.206)
Q3:-0.088 (-0.235,0.058)
Q4: -0.204 (-0.362, -0.047)
Girls
Per increase: -0.191 (-0.325,
-0.057)
Q2: 0.065 (-0.124,0.255)
Q3:-0.088 (-0.285,0.109)
Q4:-0.299 (-0.513,-0.085)
Boys
Per increase: -0.111 (-0.258,
0.036)
Q2
Q3
Q4
0.065 (-0.144,0.274)
-0.086 (-0.302), 0.131)
-0.117 (-0.348,0.114)
SGA
Per increase: 1.43 (1.03,
1.99)
Q2: 0.77 (0.45,
Q3: 0.96 (0.57,
Q4: 1.44 (0.86,
Girls
Per increase: 1.96(1.18,
3.28)
Q2: 1.00 (0.40,
Q3: 1.64 (0.71,
Q4: 2.33 (1.00,
Boys
Per increase: 1.16(0.75,
1.78)
Q2: 0.67 (0.34,
Q3: 0.66 (0.33,
1.32)
1.61)
, 2.40)
,2.51)
3.83)
5.43)
1.31)
1.29)
D-23
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Q4: 1.04 (0.54,2.01)
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.
Sweden Nested case- Pregnant women Serum Miscarriage
OR per
Per doubling: 1.48(1.09,
(2021, 7413606)
2007-2010 control from the First trimester
doubling in
2.01); p-value <0.05
High
SELMA study Case: 2.00 (1.44-
PFOA, or by
Q2: 1.69 (0.8, 3.56)
N= 1,527 2.76)
quartile
Q3: 2.02 (0.95,4.29)
Control: 1.64 (1.13,
Q4: 2.66 (1.26, 5.65)
2.32)
SELMA = Swedish Environmental Longitudinal Mother and Child, Asthma and Allergy
Results: Lowest quartile used as reference.
Confounding: Parity, age, and cotinine (tobacco smoke) exposure
Xiao et al.
Denmark Cohort Pregnant women Maternal blood BL, BW, and
Regression
BL z-score
(2019, 5918609)
1994-1995 and their Later pregnancy cranial
coefficient per
-0.14 (-0.40, 0.13)
High
children GM = 2.37 |ig/g circumference (z-
log2-unit
Girls: -0.02 (-0.37, 0.32)
N=171 (range: 0.8-6.9 scores)
increase in
Boys: -0.27 (-0.65,0.10)
l-ig/g)
PFOA
BW z-score
-0.29 (-0.56, -0.01)
Girls: -0.20 (-0.57, 0.16)
Boys: -0.39 (-0.79, -0.01)
Cranial circumference z-
score
-0.17 (-0.48, 0.15)
Girls:-0.30 (-0.74, 0.13)
Boys: -0.03 (-0.46,0.15)
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, China Cross-sectional Parents and their Maternal and
9960202) 2010-2013 children at birth paternal serum
High from LWBC within three days of
N = 369 birth
BW (g)
Regression
coefficient per
ln-unit increase
in PFOA
BW by maternal exposure
Model A: -25.2 (-75.29,
24.89)
D-24
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Maternal: 42.83
(Range = 1.16-
602.79)
Paternal: 103.38
(Range = 1.24-
2,077.93)
BW by paternal exposure
Model A: -5.67 (-54.05,
42.72)
LWBC = Laizhou Wan Birth Cohort
Confounding: All models adjusted for characteristics of parent with measured exposure: age, education, BMI (before pregnancy for
maternal exposure). Maternal exposure models additionally adjusted for parity. "Adjusted" models additionally adjusted for other
parent's exposure and characteristics.
Yeung et al.
(2019, 5080619)
High
United
States
Recruitment
2008-2010,
assessment
up to age 3
Cohort
Children aged 0- Blood
3 from Upstate
KIDS study 1.1(0.7-1.6)
N = 1,954
singletons (S)
(930 girls, 1,024
boys) and 902
twins (T)
BMI, BMI z-score,
length (cm), length
z-score, obesity,
weight (g), weight
z-score, rapid
weight gain,
weight-for-length
(WFL) z-score
Regression
coefficient or
OR (rapid
weight gain,
obesity) per
log-SD increase
in PFOA or by
quartiles
BMI
S:-0.11 (-0.17,-0.05); p-
value < 0.05
S-girls: -0.18 (-0.27, -0.09);
p-value <0.05
S-boys: -0.05 (-0.12,0.03)
T: 0.04 (-0.06, 0.14)
BMI z-score
S: -0.08 (-0.12,-0.04); p-
value < 0.05
Q2: -0.189 (-0.30, -0.07); p-
value < 0.05
Q3: -0.22 (-0.33,-0.10); p-
value < 0.05
Q4: -0.24 (-0.35,-0.12); p-
value < 0.05
S-girls:-0.13 (-0.19,-0.07);
p-value <0.05
Q2:-0.16 (-0.32, 0.01)
Q3: -0.23 (-0.39,-0.06); p-
value < 0.05
Q4:-0.33 (-0.50,-0.16); p-
value < 0.05
D-25
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
S-boys: -0.04 (-0.09, 0.02)
Q2:-0.21 (-0.37,-0.05); p-
value < 0.05
Q3: -0.20 (-0.37,-0.03); p-
value < 0.05
Q4:-0.16 (-0.32, 0.01)
T: 0.05 (-0.03,0.12)
Q2: 0.23 (0.03, 0.42); p-value
<0.05
Q3: 0.21 (0.01, 0.40); p-value
<0.05
Q4: 0.19 (-0.02, 0.39)
Length
S: 0.13 (0.02, 0.25); p-value
<0.05
S-girls: 0.19(0.01,0.37)
S-boys: 0.09 (-0.06, 0.25)
T: 0.16 (-0.03, 0.34)
Length z-score
S: 0.05 (0.001, 0.11); p-value
<0.05
S-girls: 0.07 (-0.004, 0.15)
S-boys: 0.04 (-0.03,0.11)
T: 0.07 (-0.01,0.15)
Weight
S: -12.57 (-49.47, 24.33)
S-girls: -30.22 (-84.05,
23.60)
S-boys: 6.60 (-44.69, 57.89)
T: 94.04 (33.82, 154.26); p-
value < 0.05
Weight z-score
D-26
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
S: -0.03 (-0.07, 0.01)
S-girls:-0.05 (-0.11,0.01)
S-boys: -0.01 (-0.06, 0.05)
T: 0.09 (0.03, 0.16); p-value
<0.05
WFL z-score
S: -0.08 (-0.12, -0.04); p-
value < 0.05
S-girls:-0.13 (-0.19,-0.06);
p-value <0.05
S-boys: -0.04 (-0.0p, 0.02)
T: 0.04 (-0.04, 0.12)
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. Denmark, Cohort
Pregnant women
Maternal plasma
BW (z-score,
Regression
At birth
(2010, 1429893) 1996-2002
and their
First and second
g); weight at 5
coefficient
BW
Medium
children
trimesters
and 12 months
per unit
z-score: -0.024 (-0.046,
followed up at
(z-score, g);
increase in
-0.002); p-value <0.05
birth, 5 months,
5.21 (0.5-21.9)
height at 5 and
PFOA
g: -12.8 (-24.5,-1.2); p-
and 12 months
12 months (z-
value <0.05
from DNBC
score, cm);
BMI at 5 and
Boys
z-score: -0.018 (-0.051,
N at birth =
12 months
0.015)
1114 (552 boys,
(kg/m2, z-
g: -9.5 (-26.6, 7.6)
562 girls)
score)
Girls
z-score: -0.03 (-0.058,
0.001)
g: -15.2 (-31.1, 0.7)
D-27
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location, Population, Exposure Matrix,
„ .... Design Ages, Sample Timing, Outcome Comparison Resultsb
Commence Years ¦ . ,,
Cd s N Levels3
Weight
5 months follow-up
z-score: -0.009 (-0.031,
0.012)
g: -9.4 (-28.6, 9.9)
Boys
z-score: -0.032 (-0.063, -
0.001); p-value <0.05
g: -30.2 (-59.3,-1.1); p-
value <0.05
Girls
z-score: 0.009 (-0.020,
0.038)
g: 7.9-17.7,33.4)
12 months follow-up
z-score: -0.015 (-0.038,
0.007)
g: -19.0 (-44.9, 6.8)
Boys
z-score: -0.036 (-0.069,
-0.003); p-value <0.05
g: -43.1 (-82.9, -3.3); p-
value <0.05
Girls
z-score: 0.002 (-0.029,
0.034)
g: 2.5 (-30.9, 36.0)
Height
5 months follow-up
z-score: 0.017 (-0.007,
0.040)
cm: 0.044 (-0.017,0.105)
Boys
D-28
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
z-score: 0.0015 (-0.020,
0.050)
cm: 0.039 (-0.050, 0.127)
Girls
z-score: 0.018 (-0.014,
0.049)
cm: 0.047 -0.038,0.132)
12 months follow-up
z-score: 0.016 (-0.009,
0.042)
cm: 0.049 (-0.026,0.124)
Boys
z-score: 0.011 (+-0.027,
0.048)
cm: 0.032 (-0.079, 0.143)
Girls
z-score: 0.021 (-0.013,
0.056)
cm: 0.064 (-0.039,0.166)
BMI
5 months follow-up
z-score: -0.015 (-0.040,
0.010)
kg/m2: -0.025 (-0.067,
0.017)
Boys
z-score: -0.04 (-0.078,
-0.003); p-value <0.05
kg/m2: -0.067 (-0.129,
-0.004); p-value <0.05
Girls
z-score: 0.007 (-0.027,
0.041)
kg/m2: 0.012 (-0.045, 0.069)
12 months follow-up
D-29
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
z-score: -0.025 (-0.052,
0.002)
kg/m2: -0.042 (-0.086,
0.002)
Boys
z-score: -0.046 (-0.086,
-0.006); p-value <0.05
kg/m2: -0.078 (-0.0144,
-0.011); p-value<0.05
Girls
z-score: -0.006 (-0.043,
0.030)
kg/m2: -0.01 (-0.068, 0.048)
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, prepregnancy body mass index, 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 Cross-sectional
Pregnant women Cord blood at birth
BW (g), HC
Regression
BW
(2007, 1290833)
States
and their 1.6(1.2-2.1)
(cm), BL
coefficient
Per ln-unit increase: -104
Medium
2004-2005
newborns from
(cm),
per ln-unit
(-213, 5)
Baltimore
ponderal
increase in
Per IQR increase: -58 (-119,
THREE Study,
index (g/cm3
PFOA,
3)
N = 293
* 100), GA
regression
(days)
coefficient
HC
per IQR
Per ln-unit increase:
increase in
-0.41(-0.76, -0.07), p-value
PFOA
<0.05
Per IQR increase: -0.23
(-0.42, -0.04), p-value <0.05
BL
Per ln-unit increase: -0.10
(-0.64, 0.44)
D-30
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Per IQR increase: -0.06
(-0.36, 0.24)
Ponderal index
Per ln-unit increase: -0.07
(-0.138, -0.001), p-value
<0.05
Per IQR increase: -0.039
(-0.077, -0.001), p-value
<0.05
GA
Per ln-unit increase: 1.1
(-1.2,3.4)
Per IQR increase: 0.9 (-1.1,
2.9)
Fei et al. (2008,
1290822)
Medium
Confounding: GA, maternal age, BMI, race, parity, smoking, baby sex, height, net weight gain, diabetes, hypertension. Additional
confounding for head circumference: delivery mode.
Denmark Cohort Pregnant women Maternal plasma ] Apgarscore ORforQ4vs. 1.14 (0.57,2.25)
Recruitment and their First trimester <10 Q1
1996-2002, children at 6 and
Assessment 18 months from
6-18 the DNBC
months later N = 1,400
DNBC = Danish National Birth Cohort
Confounding: Maternal age, maternal occupation and educational status, pregnancy body mass index (BMI), smoking and alcohol
consumption during pregnancy, gestational weeks at blood drawing, child's sex.
Fei et al. (2008, Denmark Cohort
Pregnant women Maternal plasma
Placental
Regression
Placental weight
2349574) 1996-2002
and their between 4-14 weeks
weight (g),
coefficient
Per unit increase: -2.06
Medium
newborns from gestation
BL (cm), HC
per unit
(-5.39, 1.28)
the DNBC 5.21(3.91-6.97)
(cm),
increase in
Q2:-11.4 (-34, 11.2)
abdominal
PFOA, or by
Q3:-13.6 (-36.8, 9.7)
Placental weight
circumferenc
quartile
Q4:-21.3 (-46.1, 3.4)
N = 1,337
e (cm)
Birth length
Birth length
D-31
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
N = 1,376
Head
circumference
N = 1,347
Abdominal
circumference
N = 1,325
Per unit increase: -0.069
(-0.113,-0.024)
Q2
Q3
Q4
-0.21 (-0.51,-0.09)
-0.04 (-0.35, 0.27)
-0.49 (-0.81,-0.16)
Head circumference
Per unit increase: -0.03
(-0.064, 0.004)
Q2
Q3
Q4
-0.09 (-0.32,0.14)
-0.23 (-0.47,0.01)
-0.14 (-0.39, 0.12)
Abdominal circumference
-0.059 (-0.106,-0.012)
Q2
Q3
Q4
-0.07 (-0.38, 0.25)
-0.16 (-0.49, 0.16)
-0.29 (-0.63, 0.06)
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, prepregnancy body
mass index, gestational week at blood drawing
Stein et al.
(2009, 1290816)
Medium
United
States
2005-2006
Cohort
Pregnant women Maternal serum
and their infants within 5 years after
from the C8HP pregnancy
21.2(10.3-49.8)
Birth defects
N = 1,505
PTB
N = 1,571
LBW
N = 1,589
Birth defects,
PTB, LBW
OR per IQR
increase in
PFOA
Birth defects
1.1 (0.8, 1.6)
PTB
0.8 (Oi
LBW
0.7 (0.5,
1.1)
1.0)
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."
D-32
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Outcome: PTB defined as birth at <37 weeks gestation; LBW defined as <5.5 pounds at birth.
Confounding: Maternal age, parity, education level at interview, smoking status at interview, PFOS levels.
Savitz et al.
United Cohort
Pregnant women Modeled
PTB, term
OR per 100
PTB
(2012, 1276141)
States
from the C8HP 1990-1994
LBW, birth
ng/mL
Per 100 ng/mL: 0.97 (0.93,
Medium
1990-2005
N= 11,737 6.0(4.5-27.6)
defects
increase in
1.02)
estimated
Q3: 1.0 (0.9, 1.2)
1995-1999
PFOA, OR
Q4: 1.0 (0.8, 1.1)
10.7(5.1-50.4)
by quintile,
Q5: 1.0 (0.8, 1.1)
OR per IQR
Per IQR: 0.96 (0.89, 1.05)
2000-2005
increase in
15.9 (5.9-56.2)
estimated
Term LBW
PFOA
Per 100 ng/mL: 0.96 (0.79,
1.16)
Q3: 1.2 (0.8, 1.9)
Q4: 1.2 (0.7, 1.9)
Q5: 0.8 (0.4, 1.4)
Per IQR: 0.89 (0.66,
1.2)
Arbuckle et al.
(2020, 6356900)
Medium
Birth defect
Per 100 ng/mL: 0.97 (0.9,
1.06)
Q3: 1.0 (0.7, 1.3)
Q4: 1.1 (0.8, 1.4)
Q5: 1.0 (0.8, 1.3)
Per IQR: 1.0(0.86, 1.16)
C8HP = C8 Health Project
Outcome: PTB defined as birth 3 or more weeks before the due date; LBW defined as <5.5 pounds at birth.
Results: Lowest two quintiles used as reference. Quintile ranges defined as follows: <40lh percentile = 3.9-<6.8; 60th percentile = 16.6;
80th percentile = 63.1.
Confounding: Exposure year, maternal age, parity, education level at interview, smoking status at interview.
Canada,
2008-2011
Cohort
Pregnant women Maternal blood Anoclitoris
Regression ACD: 0.78 (-0.25, 1.82)
(age range =17-
42 years) and
their infants
from MIREC
N = 205
1.70(1.10-2.50)
distance (ACD,
mm),
anofourchette
distance (AFD,
mm), anopenile
coefficient
per ln-unit
increase in
PFOA and by
quartiles
Q2: 0.88 (-0.79, 2.54)
Q3: 0.48 (-1.22, 2.17)
Q4: 1.06 (-0.65, 2.76)
AFD: 0.06 (-1.2, 1.32)
D-33
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location,
Confidence Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing, Outcome
Levels"
Comparison
Resultsb
distance (APD,
mm), anoscrotal
distance (ASD,
mm)
Q2
Q3
Q4
AP
Q2
Q3
Q4
-0.69 (-2.66, 1.28)
0.73 (-1.27, 2.74)
-0.56 (-2.6, 1.48)
D: 0.1 (-0.94, 1.14)
-0.76 (-2.65, 1.12)
-0.02 (-1.91, 1.88)
-0.51 (-2.5, 1.48)
ASD: 1.36 (0.3,2.41)
Q2: 0.23 (-1.67, 2.13)
Q3: -0.43 (-2.34, 1.47)
Q4: 1.77 (-0.23, 3.77)
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.
(2022, 9959688)
Medium
United
States
2014-2018
Cohort
Mother-infant
pairs from the
Emory
University
African
American
Vaginal, Oral,
and Gut
Microbiome in
Pregnancy Study
N = 370
Maternal serum,
Early pregnancy,
0.71 (0.45-1.07)
BW (g), SGA
BW:
Regression
coefficient
per doubling
in PFOA and
by quartiles
SGA: Odds
ratio per
doubling in
PFOA and by
quartiles
BW
Per doubling: -14 (-49, 21)
Q2: -126 (-241, -10)
p < 0.05
Q3: -44 (-162, 73)
Q4: -107 (-227, 13)
p-trend = 0.23
SGA
Per doubling: 1.20 (0.97,
1.49)
Q2: 2.22 (1.10, 4.50) p< 0.05
Q3: 2.44 (1.21, 4.92) p< 0.05
Q4: 2.23 (1.10, 4.54) p< 0.05
p-trend = 0.06
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. Taiwan,
(2012, 1332466) 2004-2005
Cross-sectional Mother-infant Cord blood at birth B W (g), BL B W, BL, GA,
pairs from TBPS (cm), GA HC, ponderal
BW:-19.2 (-63.5, 23.1)
BL: -0.003 (-0.21, 0.21)
D-34
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Medium
N = 429
GM (SD) =
1.84 (2.23)
(weeks), HC
(cm),
ponderal
index
(g/cm3),
PTB, LBW,
SGA
index:
Regression
coefficient
per ln-unit
increase in
PFOA
PTB, LBW,
GA: 0.06 (-0.14,0.26)
Head circumference: -0.05 (-
0.22,0.17)
Ponderal index: -0.01 (-0.04,
0.02)
PTB: 0.64 (0.4, 1.02)
SGA: OR per LBW: 0.53 (0.18, 1.55)
ln-unit SGA: 1.24 (0.75, 2.05)
increase in
PFOA
TBPS = Taiwan Birth Panel Study
Outcome: PTB defined as GA <37 weeks. LBW defined as a BW <2,500 g. SGA defined as a BW below the 10th percentile for GA.
Confounding: Maternal age, prepregnancy body mass index, education level, log (Ln)-transformed cord blood cotinine levels, type of
delivery, parity and infant sex
Chen et al.
(2017, 3981292)
Medium
Taiwan,
2004-2005
Cohort
Mother-infant Cord blood
pairs from the
Taiwan Birth
Panel Study
(TBPS)
N = 429
BMI (z-score,
kg/m2), height
(z-score, cm),
weight (z-score,
kg)
Regression At Birth
coefficient
per ln
increase in
PFOA
BMI: -0.09 (-0.2, 0.02)
Females: 0.02 (-0.13, 0.17)
Males: -0.2 (-0.36, -0.04)
Height: -0.04 (-0.16, 0.08)
Females: -0.007 (-0.18,
0.17)
Males:-0.05 (-0.22, 0.12)
Weight:-0.07 (-0.18, 0.03)
Females: 0.02 (-0.14, 0.17)
Males:-0.15 (-0.3,-0.006)
Population: Infants were followed up at 4, 6, 13, 24, 60, 84, and 108 months
Results: Regression coefficients reported at birth; BMI, height, and weight (overall and stratified by infant sex) at follow-up points were
not statistically significant
Confounding: Maternal age, pre-pregnancy BMI, education level, ln-cord blood cotinine, infant sex, PTB, postnatal ETS exposure,
breastfeeding
D-35
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Chen et al.
(2021, 7263985)
Medium
China
Recruitment
:2013-2015
Cohort
Mother-child
pairs from the
SBC,
Ages >20,
N = 214
(95 male
children, 119
female children)
Maternal plasma
from the first
trimester
15.2 (11.08-20.88)
BW (g), BL
(cm), HC (cm)
Regression
coefficient per
ln-unit increase
in PFOA
BW
33.7 (-83.9, 151.3)
BL
-0.27 (-0.61, 0.07)
Males
-0.21 (-0.73, 0.32)
Females
-0.21 (-0.74, 0.33)
HC
-45.9 (-113.9, 22.0)
SBC = Shanghai Birth Cohort
Confounding: Maternal age, BMI, educational level, occupation, income, fetal sex, parity, GA, smoking, and alcohol.
Darrow et al.
(2014, 2850274)
Medium
United
States,
Recruitment
2005-2006;
Follow-up:
2008-2011
Cohort
Pregnant women Serum collected Primary
OR per ln-unit Primary Analysis: 1.01 (0.
with known
PFAS exposure
(ages >20 years)
from C8HP
N = 1438 (first
pregnancy =
1129)
before pregnancy
15.6(9.0-31.9)
analysis increase in
miscarriage, PFOA and by
first pregnancy quintiles
miscarriage
1.16)
Q2: 0.84 (0.53, 1.32)
Q3:1.08 (0.69, 1.69)
Q4: 1.08 (0.69, 1.68)
Q5: 1.00 (0.63, 1.58)
First Pregnancy: 1.04 (0.89,
1.21)
Q2: 1.03 (0.62, 1.71)
Q3: 1.27 (0.78,2.08)
Q4: 1.34(0.81,2.20)
Q5: 1.07 (0.64, 1.77)
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)
Results: Lowest quintile used as reference.
Confounding: Maternal age, educational level, smoking status, BMI, self-reported diabetes, time between conception, and serum
measurement
D-36
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Design
Years
Population,
Ages,
N
Exposure Matrix,
Sample Timing, Outcome
Levels"
Comparison Resultsb
de Cock et al.
The Cohort
Mother-child
Cord blood BMI (kg/m2), HC
Regression BMI, HC, height, and weight:
(2014, 2713590) Netherlands
pairs
(cm), height
coefficient for no statistically significant
Medium
Recruitment
N = 89
870 ng/L (Range (cm), weight
quartiles of associations
:2011-2013
= 300-2,700 (kg)
PFOA
Follow-up
ng/L)
at 1, 2, 4, 6,
9, and 11
months after
birth
Confounding: BW, GA, maternal height
de Cock et al.
(2016, 3045435)
Medium
The
Netherlands,
2011-2013
Cross-sectional
Mother-infant
pairs
N = 64
Cord blood
870 ng/L (Range
= 200-2,700
ng/L)
BW (g)
Results: Lowest tertile used as reference.
Confounding: GA, maternal BMI, maternal height,
intake
Regression T2: 24.6 (-270.12, 319.33)
coefficient by T3: 191.3 (-137.17, 519.73)
tertiles Females
T2: 238.1 (-183.42, 659.57)
T3:-10.8 (-487.87, 466.34)
Males
T2: -184.8 (-623.06, 253.41)
T3: 168.4 (-239.18, 575.92)
No statistically significant
associations or trends by
tertiles
maternal age at birth, and parity, paternal BMI, paternal height, education, fish
Govarts et al.
(2018, 4567442)
Medium
Cohort
Mother-child
pairs from
FLEHS I and II,
HUMIS, LINC,
and PCB Cohort
N = 662
Cord blood
SGA
550 ng/L (299-
1,200 ng/L)
OR per IQR
increase of
PFOA
1.637(0.971,2.761)
Belgium,
the
Netherlands,
Norway,
and
Slovakia
2002-2012
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.
D-37
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Confounding: Maternal education, maternal age at delivery, maternal height, maternal pre-pregnancy BMI, smoking during pregnancy,
parity, child's sex
Gyllenhammar
etal. (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 = 381
Maternal serum
Later pregnancy
2.3 (1.6-3.0)
BL (SD scores),
BW (SD scores),
gestational
length (days),
HC (SD scores),
length (SD
scores), weight
(SD scores)
Regression BL: 0.0014 (-0.1435, 0.1478)
coefficient BW:-0.0579 (-0.1852,
per IQR 0.0695)
increase in Gestational length: -0.2201
maternal (-1.5028, 1.055)
PFOA HC:-0.0219 (-0.1648,
0.121)
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. Canada Cohort
Pregnant women Maternal serum B W (g, z-
BW, GA:
BW: -37.4 (-86.0, 11.2)
(2010, 1290814) Recruitment
(>18 years of
collected at 15-16 score), length
Regression
T2: 20.54 (-100.51, 141.57)
Medium : 2005-2006
age) and their
weeks gestation of gestation
coefficient per
T3: 14.80 (-107.29, 136.89)
Follow-up
singleton
(weeks), SGA,
ln-unit or per
BW (g per unit): -12.4 (-
at delivery:
children
GM(SD)= 1.3 (2.9) PTB
unit increase in
32.8, 8.0)
2006-2007
delivered at or
PFOA and by
after 22 weeks
tertiles
BW (z-score): -0.078 (-0.19,
gestation
0.032)
N = 252
SGA, PTB:
T2: 0.055 (-0.22, 0.33)
Relative risk by
T3: 0.031 (-0.25,0.31)
tertiles
GA: -0.06 (-0.28,0.15)
T2: -0.012 (-0.54, 0.52)
T3: -0.086 (-0.62,0.45)
SGA:
T2: 0.55 (0.16, 1.83)
T3: 0.99 (0.25,3.92)
PTB:
T2: 0.88 (0.28, 2.78)
T3: 1.31 (0.38,4.45)
D-38
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Outcome: SGA defined as BW <10lh 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, height, smoking. Additional confounding for BW: Infant gender,
GA at birth. Additional confounding for PTB: Infant gender.
Hjermitslev et
al. (2020,
5880849)
Medium
Greenland,
Recruitment
:2010-
2011,2013-
2015
Cohort
Pregnant women Maternal serum BW (g), GAat Regression BW:-119 (-202,-36.6), p-
(>18 years of
age) and their
children from
ACCEPT
N = 256
Early pregnancy,
later pregnancy
1.06 (Range =
0.10-7.26)
birth (weeks),
HC (cm), PTB
coefficient and
OR per ln-unit
increase in
PFOA
value = 0.005
Females: -161 (-283, -40.1),
p-value = 0.01
Males:-81.2 (-194, 31.2)
GA: 0.45 (0.17,0.74), p-
value = 0.002
Female: 0.48, p-value =
0.019
Male: 0.42, p-value = 0.043
HC:-0.14 (-0.42, 0.14)
Females:-0.51 (-0.88,0.15)
Males: 0.22 (-0.56, 0.12)
PTB
OR: -0.146, p-value = 0.011
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.
(2020, 6833719)
Medium
Denmark,
2010-2012
and follow-
up at 18
months of
age
Cohort
Pregnant women Maternal serum
and infants at 3
and 18 months
of age from
Odense Child
Cohort
N = 593
1.62 (0.67-4.03)
Ponderal index
standard
deviation score
(SDS)
Regression
coefficient
per unit
increase in
PFOA
0.07 (0.01
0.02
0.13), p-value =
Outcome: Ponderal index (kg/m3) was calculated as weight (kg) divided by the length cubed (m3)
Results: PFOA pooled 3 and 18 months
Confounding: Maternal age, parity, pre-pregnancy BMI, pre-pregnancy BMI2, education, smoking, sex, visit, adiposity marker at birth
Kashino et al. Japan,
(2020,6311632) 2003-2009
Cohort
Mother-infant
pairs from the
Plasma
Later pregnancy
Birth HC (cm), BL Regression HC: 0.053 (-0.189, 0.295)
(cm), BW (g) coefficient
D-39
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Medium
Hokkaido Study 2.0 (1.3-3.3)
on Environment
and Children's
Health
N = 1,949
per loglO-unit Females: 0.039 (-0.32,
increase in 0.398)
PFOA Males: 0.099 (-0.228, 0.425)
Length: -0.032 (-0.309,
0.246)
Females: -0.013 (-0.4,
0.373)
Males: -0.041 (-0.442, 0.36)
BW: -18.7 (-69.8, 32.4)
Females:-1.8 (-75.1, 71.5)
Males:-29.5 (-101.3, 42.3)
HC, BL, andBW: 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
Kobayashi et al.
(2017, 3981430)
Medium
Japan, Cross-sectional Pregnant women Maternal serum BL (cm), BW (g) Regression
2002-2005 at 22-35 weeks Later pregnancy coefficient
1.4(0.9-2.1) perln-unit
increase in
PFOA
Length: 0.01 (-0.37, 0.4)
BW:-494. (-130.4, 31.6)
gestation and
infants from
Hokkaido Study
on Environment
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
Length and BW: no
statistically significant
associations
Kobayashi et al.
Japan Cohort
Mother-child
Maternal blood in
BL (cm), BW
Regression BL
(2022,
Recruitment
pairs from the
the third trimester
(g)
coefficient -0.408 (-1.112, 0.307), p-
10176408)
:2002-2005
Sapporo Cohort
1.3 (0.8-1.8)
per loglO-unit value = 0.262
Medium
Females
D-40
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
of the Hokkaido
Birth Cohort
N= 372 (198
female children,
174 male
children)
1.2 (0.8-1.7)
Males
1.4 (0.9-1.8)
increase in Females:-0.608 (-1.538,
PFOA 0.302), p-value = 0.187
Males: -0.077 (-1.253,
1.099), p-value = 0.897
BW
-107.1 (-232.5, 18.4), p-
value = 0.094
Females: -183 (-361.9, -
4.1), p-value = 0.045
Males: -55.8 (-235.4, 123.8),
p-value = 0.540
Confounding: Maternal age (continuous), pre-pregnancy BMI (continuous), maternal smoking in the third trimester (yes/no), maternal
alcohol consumption during pregnancy (yes/no), parity (primiparous/multiparous), educational level, annual household income,
cesarean section (yes/no), maternal blood sampling period, GA (continuous), and infant sex.
Kwon et al.
(2016, 3858531)
Medium
Korea,
2006-2010
Cohort
Pregnant women Cord blood
and infants from
EBGRC 0.91(0.68-1.15)
N = 268
BW (g)
Regression
coefficient per
log-unit
increase in
PFOA
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
-77.93 (-153.56,-2.3), p-
value = 0.04
Lenters et al.
(2016, 5617416)
Medium
Greenland,
Poland, and
Ukraine
2002-2004
Cohort
Pregnant women Maternal serum
and singleton Later pregnancy
infants from GM = 1.421 (2-
INUENDO SD ln-PFOA =
N= 1,250 1.175)
INUENDO = Biopersistent Organochlorines in Diet and Human Fertility
Confounding: Study population, maternal age, pre-pregnancy BMI, parity
BW at term (g)
Regression
coefficient
per 2-SD
increase in ln-
PFOA
-68.94 (-134.25, -3.63), p-
value = 0.039
Liew et al.
(2016, 6387285)
Medium
Denmark, Case-control
1996-2002
Females from the Plasma,
Danish National Cases:
Birth Cohort, 3.96 (3.02, 5.22)
N = 438 Controls:
3.56 (2.76, 4.66)
Miscarriage
OR per
doubling of
PFOA and by
quartiles
1.4 (1, 1.9)
Q2: 1 (0.5, 1.8)
Q3: 1.4 (0.8,2.6)
Q4: 2.2 (1.2,3.9)
p-value for trend <0.01
D-41
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Liu et al. (2020,
6833609)
Medium
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,
2005-2009
Cohort
Females from the
LIFE study,
Ages <24, 24-29,
30-34, >35,
N = 344
Pregnancy loss
HR per log-unit 0.93 (0.75,
increase in
PFOA
1.16)
Serum,
Pregnant
women:
3.3 (2.2, 4.9)
Infertile
females:
3.2 (2.5, 4.3)
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
China,
2009-2013
Nested case-
control
Pregnant women Maternal blood
and infants
N = 519 0.79 (0.51-1.17)
PTB (spontaneous) OR per log 10-
unit increase
in PFOA and
by quartiles
1.08 (0.41, 1.6), p-value =
0.538
Q2: 1.22 (0.68,2.16)
Q3: 0.87 (0.48, 1.6)
Q4: 1.02 (0.55, 1.88)
No statistically significant
association by quartiles
Population: Cases, n = 144; controls, n = 375
Exposure Level: Cases: 0.74 (0.51-1.17); controls: 0.80 (0.51-1.18)
Results: Lowest quartile used as reference.
Confounding: Sampling time, maternal age, pre-pregnancy BMI, occupation, parity, gravidity, spontaneous abortion history, child
gender, folic acid use, passive smoking, fasting status, medication use
Maisonet et al. Great Cohort
Pregnant women Maternal serum
BW (g), BL
Regression
BW:
(2012, 1332465) Britain
and their during pregnancy
(cm), GA
coefficient by
T2: -56.81 (-153.05, 39.43)
Medium Recruitment
singleton girls (median 15 weeks)
(weeks),
tertiles
T3:-133.45 (-237.37,-
assessed at birth
ponderal
29.54)
1991 1992,
and 2, 9, and 20 3.7 (Range = 1.0-
index
p-trend = 0.0120
Followed-up
months from 16.4)
(g/cm3),
until 20
ALSPAC
weight at 20
BL:
months of
months (g)
T2: 0.14 (-0.34,0.61)
age
BW
T3:-0.44 (-0.96, 0.08)
D-42
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
N = 422
BL
N = 356
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)
p-trend = 0.0978
GA:
T2: -0.25 (-0.61,0.12)
T3:-0.34 (-0.73, 0.05)
p-trend = 0.0833
Ponderal Index:
T2:-0.06 (-0.12, 0.01)
T3: 0.02 (-0.05,0.09)
p-trend = 0.5920
Weight at 20 months:
T2: -184.21 (-465.9, 97.48)
T3: 128.4 (-180.94, 437.74)
p-trend = 0.4147
Upper tertile of B W:
T2: 15.13 (-573.62, 603.87)
T3:-27.39 (-785.4, 730.61)
p-trend = 0.9430
Middle tertile of BW:
T2: -121.55 (-708.11,
465.01)
T3: 169.83 (-497.87, 837.54)
p-trend = 0.6149
Lower tertile of B W:
T2:-21.13 (-827.99, 785.72)
T3: 248.27 (-570.54,
1,067.08)
p-trend = 0.5488
ALSPAC = Avon Longitudinal Study of Parents and Children
Results: Lowest tertile used as reference.
Confounding: BW: maternal smoking during pregnancy, maternal prepregnancy BMI, previous live births, and GA; BL additionally
adjusted for maternal education. GA: GA when maternal serum sample was obtained. Ponderal index: maternal prepregnancy BMI,
previous live births, and GA when maternal serum sample was obtained. Weight at 20 months (all tertiles): height at 20 months, BW,
D-43
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
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-
Salgado et al.
(2017, 4238509)
Medium
Spain,
2003-2008
Cohort
Mother (aged
>16 years)-child
pairs from
INMA assessed
at birth and 6
months
N = 1,154 (568
girls, 586 boys)
Maternal blood
GM = 2.32
(1.63-3.31)
Rapid growth,
weight gain (z-
score)
Relative risk
and regression
coefficient per
log2-unit
increase in
PFOA
Rapid growth: 0.99 (0.86,
1.14)
Weight gain z-score:
0.04 (-0.04, 0.12)
Females: -0.03 (-0.14, 0.08)
Males: 0.13 (0.01,0.26)
p-value for sex interaction =
0.28
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
Meng et al.
(2018, 4829851)
Medium
Denmark, Cohort
1996-2002
Pregnant women Maternal serum BW (g), GA
and their infants Early pregnancy, (days), low
fromDNBC Later pregnancy LBW, PTB
N= 3,507 4.6 (3.3-6.0)
BW and GA:
Regression
coefficient per
doubling of
PFOA and by
quartiles
BW: -35.6 (-66.3, -5)
Q2: -20.4 (-70, 29.2)
Q3:-25.9 (-77.7, 25.9)
Q4:-117 (-172.3,-61.6)
Females:-25 (-71.4,21.5)
Males:-41.5 (-82.1,-0.9)
LBW and PTB: GA: -0.4 (-1, 0.3)
OR per
doubling of
PFOA and by
quartiles
Q2:-1.4 (-2.4,-0.3)
Q3: -1.2 (-2.2, -0.1)
Q4:-1.7 (-2.9,-0.6)
Females: -0.1 (-1.1, 0.9)
Males: -0.6 (-1.4, 0.3)
LBW: 1 (0.7, 1.5)
Q2: 1.5 (0.8,3.1)
Q3: 1.2 (0.5,2.5)
Q4: 1.5 (0.7,3.3)
PTB: 1.1 (0.8, 1.5)
Q2: 3.2 (1.8,5.6)
D-44
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Q3: 1.7(0.9,3.2)
Q4: 1.9 (1, 3.6)
BW and GA: no statistically
significant associations by
sex
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,
China, Nested case- Pregnant women Maternal blood and Septal defects,
OR for >75th
Maternal PFOA
7493134)
2014-2018 control and their
cord blood at conotruncal
percentile vs.
Septal defects: 0.54 (0.18,
Medium
children with
delivery defects, and
<75th
1.62)
(cases) and
total CHD
percentile
Conotruncal defects: 0.28
without
Maternal blood
PFOA
(0.07, 1.10)
(controls) CHD
Cases:
Total CHD: 0.64 (0.34, 1.21)
N = 316
1.524(1.275-1.914)
Controls:
Cord PFOA
1.491 (1.178-2.230)
Septal defects: 0.58 (0.16,
2.10)
Cord blood
Conotruncal defects: 1.66
Cases:
(0.12, 22.1)
1.083 (0.778-1.379)
Total CHD: 0.66 (0.23, 1.88)
Control:
1.169 (0.895-1.397)
CHD = congenital heart defects
Outcome: Total congenital heart defects included septal defects and conotruncal defects, as well as individual congenital heart defect
subtypes with a large number of cases.
Confounding: Maternal age, parity, infant sex
Robledo et al.
United Cohort Couples and
Serum BW (g), HC (cm),
Regression
Maternal PFOA
(2015, 2851197)
States, their children
Early pregnancy BL (cm),
coefficient for
Girls:
Medium
2005-2009 from the LIFE
Girls: GM = ponderal index
mean change
BW:-61.64 (-159.15, 35.87)
study
3.16 (95% CI = (g/cm3)
per 1-SD
HC:-0.18 (-0.59, 0.23)
N = 234
2.92, 3.42)
increase in
BL:-0.17 (-0.74, 0.40)
D-45
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Boys: GM =
5.00 (95% CI =
4.70, 5.32)
ln(maternal Ponderal Index: -0.02 (-0.09,
PFOA) or in 0.04)
ln(paternal Boys:
PFOA) BW: 4.78 (-85.44, 95.01)
HC: 0.18 (-0.25, 0.60)
BL: -0.24 (-0.77, 0.29)
Ponderal Index: 0.04 (-0.02,
0.10)
Paternal PFOA
Girls:
BW: 19.82 (-69.37, 109.02)
HC: -0.03 (-0.42, 0.36)
BL: -0.27 (-0.79, 0.25)
Ponderal Index: 0.06 (0.00,
0.12)
Boys:
BW:-11.04 (-112.32, 90.23)
HC: -0.04 (-0.52, 0.43)
BL: -0.26 (-0.86, 0.34)
Ponderal Index: 0.03 (-0.04,
Old)
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
Savitz et al.
United Nested case-
Pregnant women Modeled
PTB, stillbirth,
PTB, stillbirth,
Study I:
(2012, 1424946)
States, control
and their infants,
term SGA,
LBW, low
PTB: 1.02 (0.94, 1.1)
Medium
1990-2004
Study II linked
Study I:
term LBW,
SGA:
Q2: 1.0 (0.8, 1.1)
to C8HP data
7.7 (4.9-17.2)
BW (g)
OR per 100-
Q3: 1.0 (0.9, 1.2)
Study II:
unit increase in
Q4: 1.0 (0.9, 1.2)
Study I:
13.4 (5.6-61.2)
PFOA, or by
Per IQR: 1.02 (0.96, 1.08)
N= 3,695
quartiles, or per
Study II:
IQR increase in
Stillbirth: 1.2 (0.86, 1.68)
N = 4,547
ln-PFOA
Q2: 0.9 (0.4, 2.0)
Q3: 1.0 (0.5, 1.7)
BW:
Q4: 0.8 (0.5, 1.5)
D-46
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Adjusted mean
Per IQR: 1.0(0.76, 1.32)
difference per
100-unit
Term SGA: 0.86 (0.67, 1.11)
increase in
Q2: 1.0 (0.7, 1.4)
PFOA, or by
Q3: 1.0 (0.7, 1.5)
quartiles, or per
Q4: 0.8 (0.6, 1.2)
IQR increase in
Per IQR: 0.91 (0.78, 1.06)
ln-PFOA
Term LBW: 1.0 (0.86, 1.15)
Q2: 0.9 (0.7, 1.2)
Q3: 1.0 (0.8, 1.3)
Q4: 1.0 (0.8, 1.3)
Per IQR: 1.02 (0.92, 1.13)
BW: -14.8 (-43.28, 13.68)
Q2: 22.8 (-32.9, 78.5)
Q3: 2.3 (-50.3, 54.8)
Q4: -9.5 (-58.4, 39.4)
Per IQR: -10.72 (-32.26,
10.82)
Study II:
PTB: 1.09 (1.0, 1.18)
Q2: 1.2 (0.9, 1.5)
Q3: 0.8 (0.6, 1.1)
Q4: 1.2 (0.9, 1.6)
Per IQR: 1.09 (0.91, 1.32)
Term SGA: 1.07 (0.98, 1.17)
Q2: 1.0 (0.7, 1.4)
Q3: 1.1 (0.8, 1.6)
Q4: 1.3 (0.9, 1.7)
Per IQR: 1.18 (0.97, 1.43)
Term LBW: 1.0 (0.82, 1.21)
Q2: 0.9 (0.5, 1.7)
D-47
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Vesterholm et
al. (2014,
2850926)
Medium
Q3: 1.6 (1.0.2.8)
Q4: 0.9 (0.5, 1.7)
Per IQR: 1.04 (0.75, 1.44)
BW:-9.14 (-20.3, 2.02)
Q2: -3.8 (-40.4, 32.8)
Q3: -25.4 (-63.7, 12.9)
Q4: -33.3 (-73.1,6.5)
Per IQR:-21.89 (-45.91,
2.13)
C8HP = C8 Health Project
Outcome: PTB defined as birth at <37 weeks gestation. Term SGA is defined as BW <10th percentile by GA and sex. LBW defined as
BW <2,500 g. Stillbirths are only reported for Study I.
Results: Lowest quartile used as reference.
Confounding: Maternal age, education, parity, smoking status, exposure year, state of residence. Additional confounding for term LBW
andBW: GA.
Nested case-
control
Boys with (107
cases) or without
(108 controls)
cryptorchidism
N = 215
Cord blood
Cryptorchidism
2.6 (5th-95th
percentile: 1.4-
4.4)
OR per ln-unit
increase in
PFOA or by
tertiles
Continuous: 0.51 (0.21, 1.2)
T2: 0.58 (0.28, 1.22)
T3: 0.46 (0.20, 1.02)
p-trend = 0.06
Denmark
and Finland
Recruitment
1997-2002,
follow-up 3
months after
birth
Outcome: Cryptorchidism defined as by Scorer (1964).
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 (511
- 95thpercentile: 1.0-3.9); controls: 2.3 (5th-95th percentile: 1.2-4.8)
Results: Lowest tertile used as reference.
Confounding: BW, GA, parity
Wang et al. China
Cross-sectional Pregnant women Cord blood
BL (cm), BW (g),
Regression
BL
(2019,5080598) 2013
and their Later pregnancy
BW z-score, HC
coefficient per
0.09 (-0.39, 0.58); p-value =
Medium
children at birth 1.99 (1.22-3.11)
cm), ponderal
loglO-unit
0.702
N= 340 (171
index (g/cm3)
increase in
Girls:-0.13 (-0.86, 0.59); p-
girls, 169 boys)
PFOA
value = 0.715
Boys: 0.06 (-0.59, 0.72); p-
value = 0.855
D-48
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
p-value for interaction by sex
= 0.913
BW
-33.42 (-149.6, 82,77); p-
value = 0.573
Girls: -84.07 (-260.42,
92.28); p-value = 0.35
Boys: -21.24 (-171.66,
129.17); p-value = 0.782
p-value for interaction by sex
= 0.959
BW z-score
-0.09 (-0.41, 0.23); p-value
= 0.589
HC
-0.37 (-0.70, -0.04); p-value
= 0.028
Girls:-0.57 (-1.07, -0.08);
p-value = 0.023
Boys: -0.35 (-7.89, -0.96);
p-value = 0.124
p-value for interaction by sex
= 0.992
Ponderal index
-0.05 (-0.10, 0.01); p-value
= 0.103
Girls:-0.05 (-0.13, 0.03); p-
value = 0.23
Boys: -0.03 (-0.10, 0.04); p-
value = 0.401
p-value for interaction by sex
= 0.980
D-49
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
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.
(2017,4183148)
Medium
United
States,
Recruitment
:2003-
2006;
outcome
assessed at
birth
Cohort
Pregnant women Maternal serum
and their Later pregnancy
children at birth 5.4(3.8-8.1)
from the HOME
study
N = 272
BW (g)
Regression
coefficient per
loglO-unit
increase
maternal PFOA
-13.1 (-53.2, 27.0)
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
Yang et al.
China Nested case- Infants from the Cord blood at birth PTB, GA
OR (PTB) and
PTB
(2022,
2018-2019 control KBCS, Term births (weeks)
regression
1.03 (0.89, 1.2), p-value =
10176806)
N = 768 0.455 (0.221-0.785)
coefficient
0.71
Medium
(384 term births, PTBs
(GA) per IQR
384 PTBs) 0.289(0.167-0.562)
increase in
GA
PFOA
Term births
-0.38 (-1.33, 0.57), p-value
= 0.44
PTBs
-1.04 (-3.72, 1.63), p-value
= 0.44
KBCS = Kashgar Birth Cohort Study
Outcome: PTBs defined as live born infants with GA at delivery 28-36 weeks.
Confounding: Maternal age, maternal ethnicity, maternal BMI, household income, maternal education level, maternal tobacco smoking
during pregnancy, maternal alcohol consumption during pregnancy, parity, living near factory, periconceptional folic acid intake,
gestational diabetes, gestational hypertension, infant's sex.
Confounding: Gravida and mode of delivery
Callan et al.
(2016, 3858524)
Low
Australia,
2008-2011
Cross-sectional
Mother-infant
pairs enrolled in
AMETS,
Ages 19-44,
Maternal blood BW (g), BL Regression BW
(cm), coefficient per -48 (-203, 108)
0.86 (0.21-3.1) Proportion of ln-unit increase
optimal B W in PFOA BL
D-50
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
N = 98
(POBW), HC
(cm), ponderal
index (g/cm3 x
100),
proportion of
optimal BL
(POBL),
proportion of
optimal HC
(POHC)
0.06 (-0.7, 0.81)
POBW
0.83 (-3.6, 5.3)
HC
-0.4 (-0.96, 0.16)
Ponderal Index
-0.06 (-0.16, 0.05)
POBL
0.42 (-1, 1.9)
POHC
-0.66 (-2.3,
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,
5080197)
Low
China,
2013-2015
Cohort
Infants from
Zhoukou City,
China,
N= 337 (183
males, 154
females)
Postnatal
weight, postnatal
length, postnatal
head
circumference
N = 282 (157
males, 125
females)
Cord blood
1.25 (0.87-1.82)
BW (g), BL
(cm), ponderal
index (g/cm3),
postnatal
weight (g),
postnatal
length (cm),
postnatal HC,
birth defects
Regression
coefficient and
OR by tertiles
BW
T2: -42.3 (-165.6, 81)
T3: -26.3 (-149.1, 96.4)
Males
T2: -121.7 (-293.3, 49.8)
T3: -15.4 (-181.9, 151.2)
Females
T2: 41.3 (-135.1,217.7)
T3: -65.3 (-247.1, 116.6)
BL
T2:-0.21 (-0.56, 0.14)
T3: -0.45 (-0.79,-0.1)
Males
T2: -0.22 (-0.68, 0.23)
T3:-0.36 (-0.8, 0.09)
D-51
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Females
T2:-0.16 (-0.68, 0.37)
T3:-0.58 (-1.12,-0.04)
Ponderal Index
T2: -0.01 (-0.09, 0.09)
T3: 0.06 (-0.03,0.15)
Males
T2: -0.07 (-0.21, 0.08)
T3: 0.06 (-0.08,0.2)
Females
T2: 0.07 (-0.04,0.17)
T3: 0.05 (-0.07,0.16)
Postnatal Weight
T2:-429.2 (-858.4,-0.121)
T3:-114.9 (-562, 332.1)
Males
T2: -661.1 (-1193.8,
-128.4)
T3:-284.6 (-830.9, 261.7)
Females
T2: -103.3 (-825.5, 618.8)
T3: 8.1 (-757.5,773.6)
Postnatal Length
T2: -0.47 (-2.3, 1.37)
T3: 1.37 (-0.5, 3.28)
Males
T2: -1.95 (-4.3, 0.4)
T3: 0.58 (-1.82,2.99)
Females
T2: 1.4 (-1.51, 4.31)
T3: 2.13 (-0.95,5.21)
Postnatal HC
T2: 0.12 (-0.8, 1.03)
D-52
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
T3: -0.04 (-0.09,0.92)
Males
T2: 0.2 (-0.99, 1.4)
T3: 0.72 (-0.51, 1.94)
Females
T2: -0.23 (-1.65, 1.19)
T3: -1.46 (-2.96,0.05)
Birth Defects
T2 OR: 0.87 (0.38, 1.96)
T3 OR: 1.24 (0.57, 2.61)
Comparison: Tertiles were defined as follows: T2 = 0.99-1.59 vs. <0.99. T3 = >1.59 vs. <0.99. T2 OR = 0.99-1.59 vs. <0.99. T3 OR =
>1.52 vs. <0.74
Results: Lowest tertile used as reference
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 B W, birth defects, ponderal index, postnatal weight, postnatal length,
POHC: maternal education. Additional confounding for postnatal weight, postnatal length, and POHC: infant's age.
Espindola
Santos et al.
(2021, 8442216)
Low
Brazil Cross-sectional
Recruitment
2017
Mother-child
pairs of women
enrolled in the
PIP A project
BW:
N = 63
BL, weight for
length
N = 56
HC
N = 53
Cord blood from
newborns
0.44 (0.21-1.02)
BW, BL, Regression BW: 0.38 (-0.18, 0.93)
weight for coefficient per
length, and HC loglO-unit
(z-scores) increase in
PFOA
BL: 0.26 (-0.21, 0.73)
Weight for length: 0.50 (-
0.17, 1.16)
HC: 0.62 (-0.096, 1.269)
PIPA = Rio Birth Cohort Study
Population: Mothers were recruited between 28th - 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
D-53
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
Gross et al.
(2020, 7014743)
Low
United
States
2012-2014
Nested case-
control
Healthy and Newborn blood B W (z-score), Regression BW (z-score)
overweight 18-
month-old
Hispanic
children from
StEP,
N = 98
Mean (SD) = 0.376
(0.249)
overweight
coefficient
(BW) and OR
(overweight)
for PFOA
>mean level vs.
PFOA < mean
level
-0.26 (-0.63,0.11)
Overweight
0.91 (0.36, 2.29)
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, and intervention
status.
Nolan et al.
(2010, 1290813)
Low
United
States
2003-2005
Cross-sectional
Mother-child
pairs
N = 1,548
Drinking Water
LHWA
5.7 (Range = 1.7-
17.1)
Congenital
anomalies
OR by LHWA
exposure level
Congenital abnormalities
LHWA vs. no LHWA
1.1 (0.34,3.3)
LHWA vs. partial LHWA
1.1 (0.4,3.1)
Non-LHWA
0.0049 (Range =
0.0-0.017)
LHWA = Little Hocking Water Association (water service area with high PFOA)
Population: No LHWA was defined as residing in zip codes served by Marietta and Warren Water Service. Partial LHWA was defined
as zip codes served in part by the LHWA and in part by Belpre Water.
Confounding: Maternal age, PTB, parity, sec, race, maternal education, diabetic status, alcohol and tobacco use.
Wu et al. (2012, China, 2007 Cross-sectional Pregnant women Maternal serum BW (g), Apgar Apgar score, BW
2919186)
Low
residing in e-
waste recycling
(Guiyu) and
non-e-waste
recycling
(Chaonan) areas,
Guiyi: 16.95 (5.5-
58.5)
Chaonan: 8.70 (4.4-
30.0)
score (5-
minute), BL
(cm), GA
(weeks),
ponderal
index
(g/cmA3 x
100),
BL, BW, GA,
ponderal index:
regression
coefficient per
loglO-unit
increase in
PFOA
-267.3 (-573.27, -37.18), p-
value < 0.05
Apgar score
-1.37 (-2.42,
<0.05
BL
-0.32), p-value
D-54
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Sample Timing,
Levels"
Outcome Comparison
Resultsb
N = 167(108
Guiyu, 59
Chaonan)
Still births
N = 146
LBW
N = 150
Premature
delivery
N = 146
premature Premature -1.91 (-3.31,-0.52), p-value
delivery, still delivery, still <0.01
birth, term birth, term
LBW LBW: GA
comparison of -2.28 (-3.96, -0.61), p-value
meanloglO <0.01
unit PFOA
concentrations Ponderal index
0.095 (-0.2, 0.389)
Premature delivery
No: Mean = 1.1 (SD = 0.22)
Yes: Mean = 1.34 (SD =
0.18)
p-value = 0.003
Still birth
No: Mean= 1.11 (SD = 0.22)
Yes: Mean = 1.42 (SD =
0.14)
p-value <0.001
Term LBW
Low: Mean =1.10 (SD =
0.22)
Normal: Mean = 1.25 (SD =
0.24)
p-value = 0.025
Comparison: Logarithm base not specified.
Confounding: Apgar score, BL, BW, GA, ponderal index: Maternal age, educational level, smoking, husband smoking, catching cold
during pregnancy, parity, premature delivery history, spontaneous abortion history. Additional confounding for Apgar score, BL, BW,
ponderal index: baby sex, GA.
Notes: BL = Birth Length; BMt = 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; CI = Confidence Interval; GM = Geometric Mean; SD = Standard Deviation; SE = Standard Error; OR :
Odds Ratio; PTB = Preterm Birth; T2 = Tertile 2; T3 = Tertile 3
a Exposure reported as median (25th-75th percentile) in ng/mL unless otherwise specified.
D-55
-------
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.2 Reproductive
D.2.1 Male
Table D-2. Associations Between PFOA Exposure and Male Reproductive Effects in Recent Epidemiologic Studies
Reference, Location,
Design
Population,
Exposure Outcome
Comparison
Resultsb
Confidence Years
Ages,
Matrix,
N
Levels3
Children and Adolescents
Jensen et al. Denmark
Cohort
Infants from
Maternal serum Levels of FSH (IU/L), Regression
FSH: 10% (-0.4, 21.4);
(2020,6311643) 2010-2012
Odense Child
1.64 testosterone (nmol/L), coefficient
p-value = 0.06
High
Cohort
LH (IU/L),
(testosterone), or
N = 208 boys
testosterone/LH ratio,
percent change (%)
Testosterone, LH,
DHEAS (nmol/L),
per doubling of
testosterone/LH, DHEAS,
DHEA (nmol/L),
PFOA
DHEA, androstenedione, 17-
Androstenedione
OHP: no statistically significant
(nmol/L), 17-OHP
associations
(nmol/L)
Confounding: Age of the child at examination time, maternal parity0
Lind et al. Denmark
Cohort
Infants from
Maternal serum Penile width (mm),
Regression
AGDap
(2017,3858512) 2010-2012
Odense child
Total cohort: 1.7 Anogenital distance
coefficient per ln-
Continuous: 0.1 (-1.1, 1.3)
High
cohort
(AGD) (mm); penile
unit increase in
p-trend by quartiles = 0.71
N = 649 (296
(AGDap), scrotal
PFOA or by
boys)
(AGDas)
quartiles
AGDas
Continuous: -0.3 (-1.6, 1.0)
p-trend by quartiles = 0.58
Penile width: no statistically
significant associations; p-trend
by quartiles = 0.86
D-56
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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
1.60
In cord blood, loglO-
transformed levels of
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 loglO-
transformed, nmol/L),
testosterone (pg/mL)
Regression
coefficient per
loglO-unit increase
in PFOA, least
squares mean
(LSM) by quartiles
Inhibin B
0.197 (0.009,0.384);
p-value = 0.04
Ql: 36.9 (29.1,46.7)
Q2: 44.3 (36.0, 55.3)
Q3: 48.5 (39.0, 60.7)
Q4: 50.3 (39.2, 64.2)
E2, FSH, insulin-like 3, LH,
progesterone, prolactin, SHBG,
testosterone, testosterone/E2,
testosterone/SHBG: No
statistically significant
associations
Testosterone/E2 ratio,
testosterone/SHBG
ratio
Confounding: 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
Lopez-Espinosa
et al. (2016,
3859832)
Medium
United
States
2005-2006
Cross-
Sectional
Male children
ages 6-9 years
N = 1,169
Serum
34.8
Total testosterone (ln-
ng/dL)
Percent difference
between 75 th and
25th percentile of
ln-unit PFOA or by
quartiles
Total testosterone:
-4.9 (-8.7, -0.8)
Q2
Q3
Q4
-3.2 (-10.6, 4.7)
-10.4 (-17.6, -2.6)
-10 (-17, -2.4)
p-trend = 0.030
Results: Results by quartile used lowest quartile as reference.
Confounding: Age, month, time of sampling
Goudarzi et al.
(2017, 3981462)
Medium
Japan Cohort Children from the Serum
2002-2005 Hokkaido Study Total cohort:
N = 185 (81 1.40
males)
Levels (loglO-ng/mL) Regression
ofDHEA, coefficient per
androstenedione loglO-unit increase
in PFOA or by
quartiles
DHEA: -0.312 (-0.642, -0.043);
p-value = 0.025
Androstenedione: -0.23 (-0.49,
0.038); p-value = 0.093
Confounding: Gestational age, maternal age, parity, smoking and caffeine intake during pregnancy, maternal educational level, blood
sampling period
Liu et al. (2020,
6569227)
Medium
China
2013-2014
Cross-
sectional
Neonates
N = 374 (183
males)
Serum
Total cohort:
1.65
Cord blood levels
(ng/mL) of 17-OHP,
progesterone
Percent change per
interquartile ratio
increase in PFOA
17-OHP: 7.82 (-0.22, 16.51);
p-value = 0.57
D-57
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Progesterone: 9.45 (3.23, 16.05);
p-value < 0.01
Confounding: Maternal age at delivery, pre-pregnancy BMI, maternal education status, passive smoking during pregnancy, parity, gestational
weeks, sample collecting time.
Ernst et al.
(2019, 5080529)
Medium
Denmark Cohort
1999-2017
Children from the Maternal blood
Puberty Cohort of Sample 1: 5.1
the Danish Sample 2: 4.3
National Birth
Cohort
N = 565 boys
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
Confounding: Highest social class of parents, maternal age at menarche, maternal age
daily number of cigarettes smoked in first trimester
No statistically significant
associations
Regression
coefficient per
log2-unit increase
in first trimester
maternal serum
PFOA
Puberty indicator:
mean difference in
age at puberty by
tertiles
at delivery, parity, prepregnancy body mass index,
Tian et al. China Cohort
Male infants at
Maternal plasma Anopenile distance
Regression
AGDap
(2019, 5390052) 2012-2013
birth, 6 months,
20.13 (AGDap) (mm),
coefficient per ln-
Birth: 0.28 (-0.62, 1.18);
Medium
and 12 months
anoscrotal distance
unit increase in
p-value = 0.533
N = 500
(AGDas) (mm)
maternal PFOA or
6 mo.:-1.82 (-4.25,0.62);
by quartiles
p-value = 0.147
Q2: -3.57 (-6.73, -0.41);
p-value < 0.05
Q3: -1.44 (-4.70, 1.81)
Q4: -3.05 (-6.19,0.10)
12 mo.:-1.55 (-4.76, 1.66);
p-value = 0.342
AGDas
Birth:-0.16 (-0.92, 0.61);
p-value = 0.686
6 mo.:-2.17 (-4.58, 0.24);
p-value = 0.079
Q2:-3.36 (-6.51,-0.21);
p-value < 0.05
Q3:-2.39 (-5.62, 0.84)
Q4: -2.58 (-5.71, 0.54)
D-58
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
12 mo.: 1.12 (-1.56,3.79);
p-value = 0.411
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)
Arbuckle et al. Canada Cohort
(2020,6356900) 2008-2011
Medium
Newborns from Maternal plasma Anopenile distance Regression
the MIREC
cohort
N = 205 boys
1.7
(AGDap) (mm),
anoscrotal distance
(AGDas) (mm)
coefficient per ln-
unit increase in
maternal PFOA, or
by quartiles
AGDap
Per In increase: 0.1 (-0.94, 1.14)
Q2
Q3
Q4
-0.76 (-2.65, 1.12)
-0.02 (-1.91, 1.88)
-0.51 (-2.50, 1.48)
p-value for trend = 0.807
AGDas
Per In increase: 1.36 (0.30,2.41);
p-value < 0.05
Q2: 0.23 (-1.67,2.13)
Q3: -0.43 (-2.34, 1.47)
Q4: 1.77 (-0.23, 3.77)
p-value for trend = 0.148
Results: Lowest quartile used as reference.
Confounding: AGDap: recruitment site, education, active smoking status, gestational age; AGDas: household income, active smoking status,
gestational age
Di Nisio et al. Italy Cross-
(2019,5080655) 2017-2018 sectional
Low
Male high school Serum
students
N = 100 (50
unexposed
controls, 50
exposed)
Unexposed
controls: 4.70
Exposed: 7.35
Semen
Unexposed
controls: 0.1
Exposed: 0.24
AGD (cm), crown-to-
pubis distance (cm),
pubis-to-floor
distance (cm), crown-
to-pubis/pubis to floor
ratio, penis
circumference (cm),
penis length (cm),
testicular volume
(mL), normal
morphology (%),
semen pH, immotile
sperm (%),
nonprogressive
motility (%),
progressive motility
(%), total sperm count
Mann-Whitney test
(Exposed r.v.
Unexposed
controls)
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)
Adjusted p-value for comparison
of medians < 0.001
D-59
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(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 (nmol/L)
of medians < 0.001
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
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).
D-60
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Confounding: Age
|Gene|ral Population
Cui et al. (2020,
6833614)
Medium
China
2015-2016
Cross-
sectional
Chinese adult
men
N = 651
Serum
8.57
Semen
0.23
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
PFOA or by
quartiles
Free testosterone
Serum PFOA: -2.7 (-4.83,
-0.53); p-value = 0.015
p-trend by quartiles = 0.036
Semen PFOA: -4.42 (-7.12,
-1.55); p-value = 0.003
p-trend by quartiles = 0.001
Total testosterone
Serum PFOA: -3.1 (-5.32,
-0.84); p-value = 0.008
p-trend by quartiles = 0.012
Semen PFOA: -5.56 (-8.4,
-2.62); p-value < 0.000
p-trend by quartiles < 0.001
E2, semen PFOA: -5.49 (-10.6,
-0.17); p-value = 0.044
p-trend by quartiles = 0.031
Total testosterone/LH, semen
PFOA:-4.83 (-9.12,-0.35);
p-value = 0.035
p-trend by quartiles = 0.018
Results: Lowest quartile used as reference.
Confounding: Age, BMI, smoking status, blood sampling time, fasting status
No other statistically significant
associations or trends by quartile
Petersen et al. Denmark
Cross-
Faroese men born Serum
Levels (log-
Regression
Free testosterone: -0.28 (-0.56,
(2018, 5080277) 2007-2009
sectional
between 1981 2.8
transformed) of E2
coefficient per log-
0.002)
Medium
and 1984
(nmol/L), FSH
unit increase PFOA
N = 263
(IU/L), free
Free testosterone/E2: -0.12
testosterone (pmol/L),
(-0.21, 0.02)d; p-value = 0.02
inhibin B (pg/mL),
LH (IU/L), SHBG
No other statistically significant
associations
D-61
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(nmol/L), testosterone
(nmol/L)
Ratios of free
testosterone/E2, free
testosterone/LH,
Inhibin B/FSH,
testosterone/E2,
testosterone/LH
Normal morphology
(%), motile sperm
(logit-%), total sperm
count ((106)1/3) semen
volume (mL1/3),
sperm concentration
((106/mL)1/3)
Outcome: Logarithm base not specified.
Comparison: Logarithm base not specified.
Confounding: Age, BMI groups, current smoking, time of sampling
Linear regression 0.013; p-value = 0.05
adjusted r2
Kvist et al. Greenland, Cross-
(2012, 2919170) Poland, and sectional
Medium Ukraine
2002-2004
Serum
Mean
Greenland: 4.84
Poland: 5.19
Ukraine: 1.91
Confounding: Age, abstinence time, alcohol intake, CB-153
Male partners of
pregnant women
from INUENDO
N = 359
Y :X chromosome
ratio of sperm
Leter et al. Greenland, Cross-
(2014, 2967406) Poland, and sectional
Medium Ukraine
2002-2004
Male partners of
pregnant women
from INUENDO
N = 262
Serum
Mean = 4.0
Sperm DNA
methylation level (%
5-mC) at LINE-1,
Alu, or Sat-alpha;
global DNA
methylation level
(FCM DGML channel
no.)
Regression
coefficient per ln-
unit increase PFOA
LINE-1: 1.1 (-0.3,2.5)
Ukraine: 2.6 (0.3, 5.0);
p-value < 0.05
Greenland: -1.7 (-4.2, 0.7)
Poland: 1.7 (-1.4, 4.8)
Alu, Sat-alpha, or global
methylation levels: No
statistically significant
associations
Confounding: Site, age (ln-transformed), smoking status
D-62
-------
DRAFT FOR PUBLIC COMMENT
Pan etal. (2019, China Cross-
6315783) 2015-2016 sectional
Medium
Adult men in Serum
Nanjing 8.567
N = 664
Semen
0.229
MARCH 2023
Sperm normal
morphology (%),
count ((106)1/3),
concentration
((106/mL)1/3),
progressive motility
(%), curvilinear
velocity (VCL)
(|im/s): straight-line
velocity (VSL)
(|im/s). DNA
fragmentation index
(DFI) (ln-%), high
DNA stainability
(HDS) (ln-%); semen
volume (ln-mL)
Regression
coefficient per ln-
unit increase PFOA
in serum or in
semen, or by
quartiles
No statistically significant
associations by serum PFOA
levels; following results are by
semen PFOA
Sperm count
0.247 (0.061, 0.432)
p-value = 0.05
Q2: 0.37 (0.02, 0.71)
Q3: -0.08 (-0.43,0.27)
Q4: 0.42 (0.06, 0.78)
p-trend = 0.2
Sperm concentration
0.193 (0.075,0.311)
p-value = 0.02
Q2: 0.3 (0.08, 0.52)
Q3: 0.06 (-0.16,0.28)
Q4: 0.36 (0.13, 0.59)
p-trend = 0.2
Progressive motility
-2.377 (-3.94,-0.815)
p-value = 0.03
Q2: 0.31 (-2.65,3.27)
Q3: -1.49 (-4.48, 1.50)
Q4: -4.26 (7.30, 1.22)
p-trend = 0.02
Sperm VCL
-1.155 (-2.064,-0.245)
p-value = 0.06
Q2: -1.65 (-3.38, 0.07)
Q3: -1.61, (-3.35,0.12)
Q4: -2.64 (-4.41, -0.87)
p-trend = 0.08
Sperm VSL
-0.92 (-1.676,-0.165)
p-value = 0.08
-63
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Q2
Q3
Q4
-1.68 (-3.11,-0.24)
-0.87 (-2.32, 0.57)
-2.13 (-3.60,-0.66)
p-trend = 0.1
Sperm DFI
0.136(0.064,0.209)
p-value = 0.01
Q2: 0.05 (-0.09,0.19)
Q3: 0.14 (0,0.28)
Q4: 0.21 (0.07,0.35)
p-trend = 0.03
Sperm morphology, sperm HDS,
semen volume: no statistically
significant associations or trends
Results: Lowest quartile used as reference.
Confounding: Age, BMI, BMI2, smoking, alcohol intake, abstinence time
Notes: 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; E2 = estradiol; FSH = follicle stimulating hormone; HDS = high DNA
stainability; LH = luteinizing hormone; LSM = least squares mean; MIREC = Maternal-Infant Research on Environmental Chemicals; PFOA = perfluorooctanoic 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.
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 PFOA Exposure and Female Reproductive Health Effects in Female Children and
Adolescents
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure Matrix,
Levels" (ng/mL)
Outcome
Comparison
Resultsb
Jensen et al. Demnark,
(2020,6311643) 2010-2012
High
Cohort Female infants Maternal serum. Levels of 17-
from the Odense 1.70 (5th-95th OHP (nM),
Child Cohort, percentile = 0.67, DHEA (nM),
Age 4 months, 3.70)
Percent change per 17-OHP
doubling in PFOA 3 (-7.9, 15.2)
DHEA
-4.7 (-15.5, 7.4)
D-64
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure Matrix,
Levels3 (ng/mL)
Outcome
Comparison
Resultsb
N = 165
Confounding: Age of the child at examination time, maternal parity0
FSH (IU/L), LH
(IU/L)
FSH
3.8 (-6.4, 15)
LH
13.3 (-4.8, 34.9)
Lind et al. Denmark,
(2017,3858512) 2010-2012
High
Cohort
Infants from
Odense child
cohort
N = 649 (353
girls)
Maternal serum
Total cohort: 1.7
Anogenital
distance (AGD)
(mm); clitoral
(AGDac),
fourchette
(AGDaf)
Regression
coefficient per ln-
unit increase in
PFOA or by
quartiles
AGDac
Continuous: -0.5 (-1.8, 0.8)
p-trend by quartiles = 0.71
AGDaf
Continuous: 0.1 (-0.9, 1.1)
p-trend by quartiles = 0.94
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.
Yao et al. (2019,
5187556)
High
Regression
coefficient per
loglO-unit
increase in PFOA
China, Cross- Pregnant Cord blood, Levels of
2010-2013 sectional women (aged 34.67(20.48, estradiol (loglO
>18 years) and 57.84) pg/mL),
female infants, testosterone
N=171 (loglO-ng/mL),
testosterone to
estradiol ratio
Confounding: Maternal age, pre-pregnancy BMI, parity, mode of delivery, passive smoking during pregnancy, gestational age, household
income level among female infants separately
Estradiol
0.03 (-0.01, 0.07)
Testosterone
0.07 (-0.03,0.17)
Testosterone to estradiol ratio
0.04 (-0.05,0.13)
Ernst et al. Denmark,
(2019, 5080529) 1999-2017
Medium
Cohort
Female
adolescents
from the Danish
National Birth
Cohort,
N = 555
Maternal blood,
Sample 1:
4.8 (10th-90th
percentile = 2.7,
8.2)
Breast
development,
pubic hair
development,
age at attainment
of axillary hair
(months), age at
menarche
Regression
coefficient per
loglO-unit
increase in PFOA
Breast development
-1.37 (-6.14, 3.4)
Pubic hair development
3.05 (-0.94,7.04)
Axillary hair
-1.49 (-4.56, 1.58)
Menarche
-1.09 (-3.25, 1.07)
D-65
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure Matrix,
Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Exposure Levels: For Sample 2, median = 4.1 (10th-90th percentile = 2.3, 6.4). Samples 1 and 2 combined for analysis.
Outcome: Age in months at Tanner stage 5 used to measure breast development and pubic hair development.
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 trimester
Donley et al.
(2019, 5381537)
Medium
United Kingdom, Case- Mothers and Maternal serum, AMH (loglO-
Recruitment control their daughters 3.7(2.8,4.8) ng/mL)
1991-1992, fromALSPAC,
outcome assessed N = 446
at adolescence
Results: N for complete data = 173; N for imputation model = 446
Confounding: Maternal age at delivery, pre-pregnancy BMI, maternal education
Regression Complete AMH data
coefficient per unit 0.05 (0.01, 0.09)
increase in PFOA Multiple imputation model
0.04 (-0.01,0.09)
Goudarzi et al.
(2017, 3981462)
Medium
Levels of
androstenedione
(loglO-ng/mL),
DHEA (loglO-
ng/mL)
Regression
coefficient per
loglO-unit
increase in PFOA
Androstenedione
-0.17 (-0.46, 0.07)
DHEA
-0.10 (-0.27, 0.11)
Japan, Cohort Pregnant Maternal serum,
2002-2005 women and their 1.40 ( < LOD-
infants from the 5.30)
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. (2016,
3981465)
Medium
Japan, Cohort Female infants Maternal serum, Levels of
2002-2005 from the 1.35 (0.80,2.00) estradiol
Sapporo Cohort (logio- ng/mL),
of the Hokkaido progesterone
Study, (loglO-ng/mL),
N = 106 prolactin (loglO-
ng/mL), SHBG
(nmol/L),
testosterone
(loglO-pg/mL)
Confounding: Age, parity, BMI before pregnancy, annual income, smoking during pregnancy,
gestational weeks of blood sampling for PFOS/PFOA measurement
Regression Estradiol
coefficient per -0.04 (- 0.19, 0.11)
log 10 increase in Progesterone
PFOA 0.04 (-0.22, 0.29)
Prolactin
-0.16 (-0.36, 0.05)
SHBG
-0.12 (-0.29, 0.05)
Testosterone
-0.03 (-0.27, 0.20)
caffeine consumption during pregnancy,
D-66
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure Matrix,
Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Liu et al. (2020,
6569227)
Medium
Lopez-Espinosa
et al. (2016,
3859832)
Medium
China,
2013-2014
Cross-
sectional
Female
neonates,
N = 191
Cord blood,
1.65(1.31,2.11)
Levels of Percent change per Progesterone
progesterone IQR-unit increase -0.03 (-5.64, 5.9)
(ng/mL), 17-OHP in PFOA 17-OHP
(ng/mL) 0.69 (-5.98, 7.84)
Confounding: Maternal age at delivery, pre-pregnancy BMI, maternal education status, passive smoking during pregnancy, parity, gestational
weeks, sample collecting time
United States, Cross- Females from Serum,
2005-2006 sectional the C8 Health 30.1 (13.5,74.0)
Project,
Ages 6-9,
N = 1,123
Results: Lowest quartile used as the reference group.
Confounding: Age, month of sampling
Levels of
estradiol (ln-
pg/mL)
Percent difference
by quartiles of
PFOA
Q2: 12.6 (3,23.1)
Q3: 6.2 (-3, 16.4)
Q4: 8.1 (-1.2, 18.4)
Maisonet et al.
(2015, 3859841)
Medium
United Kingdom,
1991-1992
Cohort
Female
adolescents
from ALSPAC,
Age 15,
N = 72
Maternal serum,
3.6 (2.7, 4.7)
Levels of serum
total testosterone
(nmol/L), SHBG
(nmol/L)
Regression
coefficient by
tertiles of PFOA
Testosterone
T2: 0.15 (-0.02,0.32)
T3: 0.24 (0.05,0.43)
SHBG
T2: 0.32 (-15.97, 16.61)
T3: 5.02 (-13.07, 23.11)
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.
Tsaietal. Taiwan, Cross-
(2015, 2850160) 2006-2008 sectional
Medium
Female
adolescents,
Ages 12-17,
N = 95
Serum,
GM = 2.74
(GSD = 2.95)
Levels of serum
FSH (ln-
mlU/mL), serum
SHBG (ln-
nmol/L)
Means by quartile
of PFOA
FSH
Ql: 1.47 (SE = 0.2)
Q2: 1.38 (SE = 0.21)
Q3: 1.23 (SE = 0.25)
Q4: 1.35 (SE = 0.29)
SHBG
Ql: 3.5 (SE = 0.24),
p-value <0.05
Q2: 3.5 (SE = 0.25),
p-value <0.05
Q3: 3.45 (SE = 0.29),
p-value <0.05
D-67
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Population,
Location, Study .
Year(s) Design ^S'
Exposure Matrix,
Levels3 (ng/mL)
Outcome
Comparison Resultsb
Q4: 2.96 (SE = 0.34),
p-value <0.05
Confounding: Age, gender, BMI, high fat diet
Wang et al. China.
(2019, 5080598) 2013
Medium
Cross-
sectional
Pregnant Cord blood,
women and their 1.99 (1.22-3.11)
children,
N = 171
Levels of estrone
(loglO-ng/mL),
b-estradiol
(loglO-ng/mL),
estriol (loglO-
ng/mL)
Regression
coefficient per ln-
unit increase in
PFOA
Estrone
0.07 (-0.07, 0.21)
b-estradiol
0.14 (-0.04, 0.32)
Estriol
0.29 (0.02, 0.56),
p-value = 0.034
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 body mass index, gestational diabetes mellitus, infant sex, delivery mode, gestational
weight gain
Notes: 17-OHP = 17-hydroxyprogesterone; ALSPAC = Avon Longitudinal Study of Parents and Children; AMH = anti-Mullerian hormone; BMI = body mass index;
DHEA = dehydroepiandrosterone; DNBC = Danish National Birth Cohort; FSH = follicle stimulating hormone; LH = luteinizing hormone; GM = geometric mean;
GSD = geometric standard deviation; Q1 = quartile one; Q2 = quartile two; Q3 = quartile three; Q4 = quartile four; SD = standard deviation; SE = standard error; SHBG = sex
hormone binding globulin; T1 = tertile one; T2 = tertile two; 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.
Table D-4. Associations between PFOA Exposure and Female Reproductive Health Effects in Pregnant Women
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure Matrix,
Levels3 (ng/mL)
Outcome
Comparison
Resultsb
Huo et al. China,
(2020, 6505752) 2013-2016
High
Cohort
Females from Plasma,
the Shanghai
Birth Cohort
Study,
Ages > 20,
N= 3,220
11.85 (9.18,
15.29)
Gestational
hypertension,
hypertensive
disorders of
pregnancy,
preeclampsia
OR per ln-unit Gestational hypertension
increase in PFOA
1.37 (0.76, 2.48)
Hypertensive disorders
1.09 (0.72, 1.66)
Preeclampsia
0.89 (0.5, 1.57)
Confounding: Maternal age, pre-pregnancy BMI, parity, parental educational levels, gestational age of blood drawn, fetal sex0
D-68
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure Matrix,
Levels" (ng/mL)
Outcome
Comparison
Resultsb
Mitro et al.
(2020, 6833625)
High
United States,
1999-2005
Cohort
Females from
Project Viva,
N = 812
Plasma,
5.6 (4.0, 7.6)
Levels of SHBG
(nmol/L)
Percent difference
per log2-unit
increase in PFOA
-1.5 (-9.3, 7)
Women under 35 years during
pregnancy
-0.9 (-11.4, 10.8)
Women over 35 years during
pregnancy
-1.8 (-13.7, 11.6)
Confounding: Age, pre-pregnancy BMI, marital status, race/ethnicity, education, income, smoking, parity
Borghese et al. Canada, Cohort Females from Plasma,
Gestational
OR (GH, PE) or
Gestational hypertension
(2020,6833656) 2008-2011 the MIREC GM = =1.65
hypertension,
regression
1.06 (0.84, 1.35)
Medium study, (95% CI: 1.61,
preeclampsia,
coefficient (DBP,
Preeclampsia
Ages >18, 1.70)
DBP (mmHg),
SBP)
1.36 (0.9, 2.08)
N = 1,739
SBP (mmHg)
per log2-unit
DBP
increase in PFOA
0.64 (0.24, 1.05),
p-value = = = 0.002
SBP
0.82 (0.23, 1.42),
p-value = 0.006
Confounding: Maternal age, education, smoking status, pre-pregnancy BMI, parity
Huang et al. China, Cross- Females from Cord blood
Gestational
OR per increase in
Gestational hypertension
(2019,5083564) 2011-2012 sectional mother-infant plasma,
hypertension,
standardized
0.95 (0.61, 1.48)
Medium pairs, 6.98 (4.95,9.54)
hypertensive
PFOA
Hypertensive disorders of
N = 687
disorders of
pregnancy
pregnancy,
1.02 (0.73, 1.44)
preeclampsia
Preeclampsia
1.12(0.68, 1.84)
Results: Standardized PFOA calculated by subtracting PFOA concentration from mean PFOA concentration and dividing by the SD.
Confounding: Age, pre-pregnancy BMI, parity, education level
Lyngso et al. Greenland, Cross- Pregnant Serum,
Menstrual cycle
OR per log-unit
Length
(2014,2850920) 2002-2004 sectional women from the 1.5 (10th-90th
length (long),
increase in PFOA
1.5 (1.0,2.1)
Medium INUENDO percentile = 0.7,
irregularity
and by tertile
T2: 1.4(0.8,2.3)
cohort, 3.1)
T3: 1.8(1.0,3.3)
N = 1,623
Irregularity
1.3 (0.8, 1.9)
T2: 1.3 (0.8,2.3)
D-69
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Population, Exposure Matrix,
Location, Study . „ , „
„ . Ages, Outcome Comparison
Year(s) Design s
N Levels3 (ng/mL)
Resultsb
T3: 1.3 (0.7,2.3)
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.
(2016, 3981728)
Medium
United States,
2003-2006
Cohort
Females from
the HOME
study,
Ages >18,
N = 336
Serum,
5.5 (3.8, 7.7)
Breastfeeding
termination at
3 months and at
6 months
RR by quartiles of Breastfeeding termination
PFOA
At 3 months
Q2: 1.32 (0.92, 1.88)
Q3: 1.63 (1.16,2.28)
Q4: 1.77 (1.23,2.54)
p-value = 0.003
At 6 months
Q2: 1.25 (0.96, 1.62)
Q3: 1.38 (1.06, 1.79)
Q4: 1.41 (1.06, 1.87)
p-value for trend = 0.038
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
Rylander et al.
(2020, 6833607)
Medium
OR by quartiles of Q2: 0.94 (0.56, 1.57)
PFOA Q3: 1.42 (0.87, 2.31)
Q4: 1.13 (0.68, 1.87)
Sweden, 1989 Case- Females with or Serum, Preeclampsia
control without pre- Primiparous cases:
eclampsia, 2.82 (Minimum,
Ages 15-44, Maximum = 0.55,
N = 876 10.9)
Exposure Levels: [Muciparous cases] Median = 1.96 ng/mL (Minimum, Maximum = 0.42, 6.93 ng/mL); [Primiparous controls]
Median = 2.83 ng/mL (Minimum, Maximum = 0.39, 9.38 ng/mL); [Multiparous controls] Median = 1.81 ng/mL (Minimum,
Maximum = 0.40, 9.34ng/mL).
Confounding: Maternal age, BMI in early pregnancy, maternal smoking in early pregnancy, parity
Starling et al.
(2014, 2446669)
Medium
Plasma,
2.78 (2.14,3.57)
Preeclampsia
onset
HR per ln-unit
increase in PFOA
Norway, Nested Females from
2003-2007 case-control MoBa,
Ages 16-44,
N = 976
Confounding: Maternal age, pre-pregnancy BMI, education completed, smoking during pregnancy
0.89 (0.65, 1.22)
Timmermann et
al. (2017,
3981439)
Denmark,
1997-2000,
2007-2009
Cohort
Pregnant and
postpartum
females,
Serum,
2.40 (1.45,3.59)
Total Regression Total breastfeeding duration
breastfeeding coefficient per -1.3 (-1.9,-0.7)
duration doubling of PFOA Exclusive breastfeeding duration
D-70
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Year(s)
Study
Design
Population,
Exposure Matrix,
Ages,
N
Levels" (ng/mL)
Outcome
Comparison
Resultsb
Medium
N = 987
(months),
exclusive
breastfeeding
duration
(months)
-0.5 (-0.7, -0.3)
Confounding: Cohort, maternal age, pre-pregnancy BMI, pregnancy alcohol intake, pregnancy smoking, education, employment, parity
Wikstrom et al. Sweden, 2007-
(2019, 5387145) 2010
Medium
Cohort Females from Serum,
theSELMA 1.61(1.12,2.31)
study,
Ages 28-35,
N = 1,773
Population: N for nulliparous women = 812
Confounding: Parity, women's age, body weight, smoke exposure
Preeclampsia
OR per log2-unit
increase in PFOA
PE
All women: 1.31 (0.93, 1.87)
Nulliparous women: 1.38 (0.90,
2.21)
Notes: BMI = body mass index; CI = confidence interval; DBP = diastolic blood pressure; GM = geometric mean; GSD = geometric standard deviation; 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; MoBa = Norwegian Mother and Child Cohort Study; OR = odds ratio;
Q1 = quartile one; Q2 = quartile two; Q3 = quartile three; Q4 = quartile four; RR = relative risk ratio; SBP = systolic blood pressure; SD = standard deviation; SE = standard error;
SELMA = Swedish Environmental Longitudinal, Mother and child, Asthma and allergy study; SHBG = sex hormone binding globulin.
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-5. Associations between PFOA 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,
(2020, 6833612) 1999-2017
High
Cohort Pre-menopausal Serum,
women from the 4.0(2.8,5.7)
Study of
Women's
Health Across
the Nation,
Ages 42-52,
N = 1,120
Natural
menopause
HR per doubling
of PFOA and by
tertiles
1.11 (0.99, 1.24)
T2: 1.12 (0.9, 1.4)
T3: 1.31 (1.04, 1.65)
p-value for trend = 0.01
D-71
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure Matrix,
Levels" (ng/mL)
Outcome
Comparison
Resultsb
Results: Lowest tertile used as the reference group.
Confounding: 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
Times to
Cycle-specific time to pregnancy
(2017,3859813) 2008-2009 the Time to 2.79 (2.48,3.16)
time to
pregnancy:
1.15(0.66, 2.01)
Medium Conceive Study,
pregnancy, day-
FR per ln-unit
Day-specific time to pregnancy
Ages 30-44,
specific time to
increase in PFOA
0.96 (0.31, 1.94)
N = 99
pregnancy;
AMH
levels of AMH
AMH:
-0.56 (p-value = 0.75)
(ln-ng/mL)
Regression
coefficient per ln-
unit increase in
PFOA
Confounding: Age, mean cycle length (for cycle-specific outcome)
Dhingra et al. United States,
Cohort
Females from
Serum, measured Natural
OR per ln-unit
Measured
(2016, 3981432) 2005-2006,
the C8 Science
and modeled menopause
increase in PFOA,
1.09(1.002, 1.18),
Medium 2008-2011
Panel,
Measured:
or by quintiles, or
p-value = 0.04
Age > 40,
Mean = 69.2 |ig/m
by deciles
Quintile 2: 1.68 (1.21,2.35),
N= 9,192
L (SD = 195.6)
p-value = 0.002
Modeled:
Quintile 3: 1.45 (1.04,2.02),
Mean= 81.8 |ig/m
p-value = 0.03
L (SD = 175.0)
Quintile 4: 1.39 (1, 1.93),
p-value = 0.05
Quintile 5: 1.58(1.14,2.19),
p-value = 0.006
Modeled
0.98 (0.7, 1.37)
Quintile 2: 0.98 (0.7, 1.37)
Quintile 3: 1.05 (0.75, 1.45)
Quintile 4: 0.78 (0.56, 1.08)
Quintile 5: 0.92 (0.65, 1.3)
Dose-response by deciles:
increased up to the 4th
D-72
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure Matrix,
Levels" (ng/mL)
Outcome
Comparison
Resultsb
decile and then, except for a drop
at the 5th decile, remained
approximately level
thereafter
Results: Lowest quintile used as the reference group.
Confounding: Age, parous/nulliparous status, smoking status, education, BMI, birth year
Kim et al.
(2020, 6833596)
Medium
Australia,
2006-2011
Cross-
sectional
Females
undergoing
fertility
treatment,
Ages 23-42,
N = 97
Follicular fluid Fertilization rate Regression 0.71 (-2.22, 3.63)
Mean =2.4 coefficient per unit
(Minimum- increase in PFOA
Maximum = 0.3,
14.5)
Confounding: Age
Lum et al.
(2017, 3858516)
Medium
United States,
2005-2009
Cohort
Females from
the LIFE study,
Ages 18-40,
N = 483
Serum
Women with < 24-
day cycle: 3.1
(2.5, 4.0)
Women with 25 to
31-day cycle:
3.5 (2.3, 5.0)
Day-specific
probability of
pregnancy,
menstrual cycle
length
Regression
coefficient by
tertiles of PFOA
All women:
Day-specific probability of
pregnancy
T2: 1 (0.7, 1.5)
T3: 0.7 (0.5, 1.5)
Menstrual cycle length
T2: 0.98 (0.95, 1.01)
T3: 0.98 (0.96, 1)
Women with > 32-
day cycle: 3.1
(2.0, 4.7)
Results: Lowest tertile used as the reference group.
Exposure Levels: Presented for females with 25-31-day cycles. The study also present exposure levels for females with 24-day cycles or
shorter and females with cycles longer than 31 days.
Results: Lowest tertile used as the reference group.
Confounding: For menstrual cycle length: female age, BMI, active smoking at enrollment; For day-specific probability of pregnancy: couple
intercourse pattern, female menstrual cycle length, age, BMI, active smoking at enrollment
Tsai et al.
(2015, 2850160)
Medium
Taiwan, Cross- Females, Serum, Levels of FSH in Means by quartile
2006-2008 sectional Ages 18-30, GM = 2.74 serum (In- of PFOA
N = 265 (GSD = 2.95) mlU/mL),
SHBG in serum
(ln-nmol/L)
FSH
Ql: 1.69(SE = 0.24)
Q2: 1.65 (SE = 0.24)
Q3: 1.64 (SE = 0.25)
Q4: 1.79 (SE = 0.26)
D-73
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Confidence
Location,
Year(s)
Study
Design
Population,
Ages,
N
Exposure Matrix,
Levels" (ng/mL)
Outcome
Comparison
Resultsb
SHBG
Ql: 3.83 (SE = 0.21)
Q2: 3.86 (SE = 0.2)
Q3: 3.81 (SE = 0.22)
Q4: 3.78 (SE = 0.23)
Confounding: Age, BMI, high fat diet
Wang et al. China, Case-
(2017, 3856459) 2014-2015 control
Medium
Females of
reproductive
age,
N = 335
Plasma,
Cases:
14.67
(7.32, 23.73)
Endometriosis-
related infertility
OR by tertiles of
PFOA
T2: 0.89 (0.5, 1.59)
T3: 1.05 (0.58, 1.91)
Population: [Cases] Females with endometriosis; [Controls] Females from couples seeking treatment for male infertility
Exposure Levels: [Control] Median = 12.09 (25th-75th percentile = 7.33, 22.59)
Results: Lowest tertile used as the reference group.
Confounding: Age, BMI, household income, and education
Notes: 17-OHP = 17-hydroxyprogesterone; ALSPAC = Avon Longitudinal Study of Parents and Children; BMI = body mass index; DHEA = dehydroepiandrosterone;
DNBC = Danish National Birth Cohort; FR = fecundability ratio; FSH = follicle stimulating hormone; HR = hazard ratio; LH = luteinizing hormone; GM = geometric mean;
GSD = geometric standard deviation; OR = odds ratio; Ql = quartile one; Q2 = quartile two; Q3 = quartile three; Q4 = quartile four; SD = standard deviation; SE = standard error;
SHBG = sex hormone binding globulin; T1 = tertile one; T2 = tertile two; 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.3 Hepatic
Table D-6. Associations Between PFOA Exposure and Hepatic Effects in Epidemiology 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
3.20 (20th-80th
percentile = 1.82-
5.50)
Levels of iron
in serum,
bilirubin, and
albumin
Percent change Iron concentration in serum
per one percent 0.10 (0.07, 0.12), p-value < 0.05
increase in
PFOA Bilirubin
0.06 (0.04, 0.08), p-value < 0.05
D-74
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure Matrix
Levels (ng/mL)a
Outcome Comparison
Select Resultsb
Albumin
0.03 (0.03, 0.04), p-value < 0.05
Confounding: Age, sex, race, education, poverty income ratio, serum cotinine, BMI
Jain (2019,
5381541)
Medium
United States Cross- Adults from Serum Levels of ALT Regression ALT
2003-2014 sectional NHANES (loglO-IU/L), coefficient per Non-obese,
Age >20, AST (loglO- loglO-unit GF-1:0.009
N = 108-3,562 IU/L) increase in GF-2:0.047, p-value = 0.02
PFOA GF-3A: 0.001
GF-3B/4: -0.001
Obese,
GF-1: 0.077, p-value <0.01
GF-2: 0.035
GF-3A: 0.057
GF-3B/4: 0.164, p-value < 0.01
AST
Non-obese,
GF-1: 0.014
GF-2: 0.028
GF-3A: 0.002
GF-3B/4: 0.055, p-value = 0.03
Obese,
GF-1: 0.039, p-value <0.01
GF-2: 0.029
GF-3A : 0.036, p-value = 0.03
GF-3B/4: 0.050, p-value < 0.01
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. (2018,
4238396)
Medium
United Controlled
States, 2004- trial
2007
Overweight and Plasma
Obese patients from Males
the POUNDS-Lost, 5.2(3.9-8.6)
Age 30-70 study, Females
N = 150 4.1 (2.8-5.6)
Hepatic fat
mass
Partial
Spearman
correlation
coefficient
among baseline
PFOA (ng/ml)
Hepatic fat mass: 0.12
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
Liu et al. (2018,
4238514)
and hepatic fat
mass
Confounding: age, sex, race, education, smoking status, alcohol consumption, physical activity, menopausal status (women only), hormone
replacement therapy (women only), and dietary intervention groups
United Cross-
States, 2013- sectional
2014
Adults from Serum Levels of Regression Albumin
NHANES, GM=1.86 albumin (g/dL) coefficient per 0.09, SE = 0.02, p-value < 0.005
Age >18, (SE=1.02) ln-unit increase
N = 1871 inPFOA
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.
(2018, 5083555)
Medium
Sweden
2001-2014
Cohort
Plasma
Age 70
3.31 (2.52-4.39)
Age 75
3.81 (2.71-5.41)
Age 80
2.53 (1.82-3.61)
Levels of ALT Regression 0.04 (0.03, 0.06), p-value < 0.0016
(|ikat/L) coefficient per
ln-unit increase
in PFOA
Elderly adults in
Sweden,
Age 70
N = 1.002
Age 75
N = 817
Age 80
N = 603
Confounding: Sex, LDL and HDL cholesterol, serum triglycerides, BMI, fasting glucose levels, statins use, and smoking
Nian et al. (2019,
5080307)
Medium
China
2015-2016
Cross-
sectional
Adults in high
exposure area in
China,
Ages 22-96,
N = 1,605
Serum
6.19(4.08-9.31)
Levels of ALT
(ln-U/L), AST
(ln-U/L)
Percent change
per ln-unit
increase in
PFOA
ALT
7.4(3.9,
11.0)
AST
2.9 (0.7, 5.2)
Confounding: Age, sex, career, income, education, drink, smoke, giblet, seafood consumption, exercise, BMI
Yamaguchi et al.
Japan Cross-
Participants from
Blood
Levels of GGT
Spearman rank
GGT
(2013, 2850970)
2008-2010 sectional
the "Survey on the
2.1 (1.5-3.3)
(IU/L), AST
correlation
0.06, p-value = 0.120
Medium
Accumulation of
(IU/L), ALT
Dioxins and Other
(IU/L)
AST
Chemical
0.13, p-value = 0.002
Compounds"
project from urban,
ALT
agricultural and
0.09, p-value = 0.040
fishing areas,
Ages 15-76,
D-76
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location, _ . Population, Exposure Matrix
Design Outcome Comparison
Years Ages,N Levels (ng/mL)a
Select Resultsb
N = 590
Confounding: Age, sex, BMI, regional block, smoking habits, frequency of alcohol intake
Gallo et al. United States Cross-
(2012, 1276142) 2005-2006 sectional
Medium
Adults from the C8
Health Project,
Ages >18 years,
N = 46,452
Serum
28.0 (13.5-70.8)
Levels of GGT
(ln-IU/L), ALT
(ln-IU/L),
Direct bilirubin
(ln-mg/dL),
ALT (IU/L,
elevated)
GGT, ALT,
direct bilirubin:
Regression
coefficient per
ln-unit increase
in PFOA
Elevated ALT:
OR per ln-unit
increase in
PFOA, or by
deciles
GGT
0.015 (0.01,0.019), p-
value < 0.001
ALT
0.022 (0.018, 0.025), p-
value < 0.001
ALT, elevated (OR):
Decile 2: 1.09 (0.94, 1.26)
Decile 3: 1.19 (1.03, 1.37)
Decile 4: 1.26(1.09, 1.45)
Decile 5: 1.40 (1.22, 1.62)
Decile 6: 1.39 (1.21, 1.60)
Decile 7: 1.31 (1.14, 1.52)
Decile 8: 1.42 (1.23, 1.64)
Decile 9: 1.40 (1.21, 1.62)
Decile 10: 1.54 (1.33, 1.78)
p-trend <0.001
Per ln-unit increase: 1.1 (1.07,
1.13), p-value < 0.001
Direct bilirubin: 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 GGT, ALT, and direct bilirubin regression analyses: Race. Additional
confounding for OR analyses: increased serum iron.
Linetal. (2010,
United States
Cross-
Adults from
Serum
Levels of ALT
Regression
ALT
1291111)
1999-2000,
sectional
NHANES,
Total:
(U/L), GGT
coefficient per
Total:
Medium
2003-2004
Ages >18 years,
4.20 (2.90-5.95)
(log-U/L),
log-unit
Separate analysis: 1.86
Total
bilirubin (nM)
increase in
(SE = 0.62), p-value = 0.005
N = 2,216,
Mean (SE):
PFOA
Composite analysis: 2.19
Men
Men: 5.05 (1.03)
(SE = 0.79), p-value = 0.009
D-77
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location, _ . Population, Exposure Matrix .
„ ~Design Outcome Comparison Select Results
Confidence Years Ages,N Levels (ng/mL)a
N = 1,063,
Women: 4.06 (1.04)
Men:
Women
Ages 18-39: 4.48
1.55 (SE = 0.84), p-value = 0.076
N = 1,134,
(1.03)
Women:
Ages 18-39
Ages 40-59: 4.71
1.87 (SE = 1.13), p-value = 0.109
N = 944,
(1.04)
Ages 18-39:
Ages 40-59
Ages > 60: 4.22
1.02 (SE = 0.84), p-value = 0.234
N = 534,
(1.04)
Ages 40-59:
Ages > 60
1.83 (SE = 1.84), p-value = 0.329
N = 719
Ages > 60:
1.93 (SE = 1.10), p-value = 0.089
GGT
Total:
Separate analysis: 0.08
(SE = 0.03), p-value = 0.019
Composite analysis: 0.15
(SE = 0.04), p-value = 0.001
Men:
(SE = 0.03), p-value = 0.766
Women:
0.09 (SE = 0.05), p-value = 0.087
Ages 18-39:
0.06 (SE = 0.04), p-value = 0.078
Ages 40-59:
0.04 (SE = 0.08), p-value = 0.641
Ages > 60:
0.06 (SE = 0.04), p-value = 0.146
Bilirubin, total
Separate analysis: -0.09
(SE = 0.20), p-value = 0.645
Composite analysis: -0.20
(SE = 0.22), p-value = 0.378
Population: Stratified population counts do not include 19 individuals who were excluded due to their iron saturation being above 50%.
Comparison: Logarithm base not specified.
D-78
-------
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: Age, gender, race/ethnicity, smoking status, drinking status, education level, BMI, HOMA-IR, metabolic syndrome, iron
saturation status. Additional confounding for bilirubin, GGT and ALT composite analyses: PFHxS exposure, PFNA exposure, PFOS
exposure.
Costa et al.
(2009, 1429922)
Medium
Italy
2007
Cross-
sectional
Current and former
Serum
Levels of AST
Comparison of
male employees of
Production workers
(U/L), ALT
mean outcome
an Italian chemical
(2007): 3.89 ng/mL
(U/L), GGT
(Exposed vs
production plant,
(2.18-18.66 ng/mL)
(U/L), ALP
unexposed
Comparison of
(U/L),
workers)
means analysis
Albumins (%)
N = 68,
Regression
Exposed vs
coefficient
Unexposed analysis
(exposed
N = 141,
workers vs all
Continuous
workers)
regression analysis
N = 56
Regression
coefficient per
unit increase in
PFOA
No significant differences in
comparison of mean hepatic
outcomes
ALT
Exposed vs Unexposed: -5.18
(-13.7, 3.32)
Continuous: 0.116 (0.054, 0.177),
p-value < 0.01
ALP
Exposed vs Unexposed: -0.78
(-8.51,6.95)
Continuous: 0.057 (0.007, 0.107),
p-value < 0.05
AST
Exposed vs Unexposed: 1.35
(-2.72, 5.41)
Continuous: 0.038 (-0.003, 0.08)
GGT
Exposed vs Unexposed: 0.32
(-17.5, 18.1)
Continuous: 0.177 (0.076, 0.278),
p-value < 0.01
Albumins
Exposed vs Unexposed: -0.73
(-3.44, 1.97)
Continuous: -0.009 (-0.017,
0.001)
Confounding: Age, job seniority, body mass index, smoking and alcohol consumption. Additional confounding for continuous regression
analyses: year of observation
D-79
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure Matrix
Levels (ng/mL)a
Outcome Comparison
Select Resultsb
Sakr et al. (2007,
1291103)
Medium
United States Cross-
2004 sectional
Active employees
at a Washington
Works site where
APFO is used,
AST analysis
N = 1016,
ALT analysis
N = 1018,
GGT & bilirubin
analysis
N = 1019,
AST analysis,
workers not on
lipid-lowering
medications
N = 838,
ALT & GGT
analysis, workers
not on lipid-
lowering
medications
N = 840
Serum
Mean (SD) = 0.428
(0.86) ppm
Levels of AST Regression
(ln-IU/L), ALT coefficient per
(ln-IU/L), GGT unit increase in
(ln-IU/L), PFOA
Bilirubin (ln-
IU/L)
AST
0.012 (SE = 0.012), p-
value = 0.317
ALT
0.023 (SE = 0.015), p-
value = 0.124
GGT
0.048 (SE = 0.020), p-
value = 0.016
Bilirubin
0.008 (SE = 0.014), p-value = 0.59
AST, workers not on lipid-
lowering medication
0.023 (SE = 0.013), p-
value = 0.079
ALT, workers not on lipid-
lowering medication
0.031 (SE = 0.017), p-
value = 0.071
Confounding: Age, gender, BMI
GGT, workers not on lipid-
lowering medication
0.05 (SE = 0.023), p-value = 0.03
Bilirubin, workers not on lipid-
lowering medication
0.008 (SE = 0.017), p-
value = 0.637
Sakr et al. (2007,
1430761)
Medium
United States Cohort
1979-2007
Fluoropolymer
manufacturing site
workers,
Serum
Mean (SD) = 1.13
(2.1) ppm
Levels of total
bilirubin
(mg/dL), GGT
Regression Bilirubin, total
coefficient per -0.008 (-0.0139, -0.0021)
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
N = 454
(IU/L), AP
(IU/L), AST
(IU/L), ALT
(IU/L)
unit increase in GGT
PFOA 1.24 (-1.09, 3.57)
AP
-0.21 (-0.60,0.18)
AST
0.35 (0.10,0.60)
ALT
-0.54 (-0.46, 1.54)
Confounding: Age, BMI, gender, decade of hire. Additional confounding for total bilirubin regression analysis: age-squared.
Olsen et al. United
(2001, 10228462) States,
Medium Belgium
1994-2000
Cohort Male 3M
fluorochemical
plant workers in
Antwerp, Belgium
and Decatur,
Alabama
N = 175
Serum
Antwerp (2000)
Mean (SD): 1.43
ppm (1.21)
Decatur (2000):
1.83 ppm (1.53)
Levels of ALT
(ln-IU/L)
(ln-IU/L)
(ln-IU/L)
(ln-IU/L)
Regression
ALP coefficient per
AST unit increase in
GGT PFOA
ALT
0.015 (SE = 0.02), p-value = 0.46
PFOA x Years of observation
interaction p-value = 0.19
AST
0.027 (SE = 0.015), p-value = 0.06
PFOA x Years of observation
interaction p-value = 0.41
ALP
0.005 (SE = 0.012), p-value = 0.69
PFOA x Years of observation
interaction p-value = 0.62
GGT
-0.009 (SE = 0.025), p-
value = 0.72
PFOA x Years of observation
interaction p-value = 0.29
Confounding: Years of observation, PFOA x Years of observation, age, BMI, drinks/day, cigarettes/day, location, entry period, baseline
years worked, triglycerides
Olsen et al.
(2000, 1424954)
United States Cross-
1993-1997 sectional
Male workers
involved in
Serum
Levels of ALT
(IU/L),
Regression
coefficient per
ALT
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
Medium
ammonium
perfluorooctanoate
production,
N = 265
1993:1.1 Cholecystokini unit increase in 1993: 0.89 (SE = 2.88), p-
(Range = 0.0- n (pg/mL) PFOA value = 0.76
80.0) ppm 1995: 0.81 (SE = 2.62), p-
1995:1.2 value = 0.75
(Range = 0.0- 1997: 2.77 (SE = 1.27), p-
114.1) ppm value = 0.03
1997: 1.3
(Range = 0.1- Cholecystokinin
81.3) ppm -0.008 (SE = 0.004), p-
value = 0.07
Confounding: Age, alcohol use, cigarette use. Additional confounding for cholecystokinin regression analysis: Body mass index (BMI)
Olsen et al.
United States Cohort
3M Fluorochemical
Serum
Levels of ALT
Regression
ALT
(2012, 2919185)
2008-2010
plant employees
Mean change from
(IU/L), AST
coefficient per
-0.0097 (SD = 0.005), p-value =
Low
and contractors,
baseline,
(IU/L)
unit increase in
0.00495
N = 179
Employees:
PFOA
-218.3; Contractors:
AST
32.1
-0.0032 (SD = 0.003)
Confounding: Sex, age at baseline, BMI at baseline, alcohol consumption at baseline
Wang et al.
China Cross-
Male
Serum
Levels of ALT
Regression
ALT
(2012, 2919184)
2010-2011 sectional
fluorochemical
Residents: 284.34
(ln-IU/L), AST
coefficient per
Residents: -0.1 (-0.19, 0.00), p-
Low
plant workers and
(Range = 10.20-
(ln-IU/L)
ln-unit increase
value = 0.05
near-by residents,
2436.91);
in PFOA
Workers: 0.04 (-0.06, 0.15)
N= 55-132
Workers: 1635.96
(Range = 84.98-
AST
7737.13)
Residents: -0.04 (-0.10, 0.02)
Workers: -0.12 (-0.22, -0.02), p-
value = 0.02
Confounding: None
Darrow et al.
United States Cohort and
Adults from the C8
Modeled cumulative Levels of ALT
Regression
ALT
(2016, 3749173)
2005-2006 Cross-
Health Project,
PFOA,
(IU/L), Liver
coefficient
Modeled, 0.012 (0.008, 0.016)
Medium
sectional
Ages >18 years,
20th-80th percentile:
(enlarged,
per ln-unit
Quintile 2: 0.023 (0.006, 0.040)
N = 30,723
191-3998 y-ng/mL;
fatty, or
increase in
Quintile 3: 0.035 (0.018, 0.052)
Estimated in serum
cirrhosis),
PFOA or by
Quintile 4: 0.039 (0.022, 0.056)
16.5 (range = 2.6-
Liver disease
quintiles
Quintile 5: 0.058 (0.040, 0.076)
3,559)
(any)
p-trend < 0.0001
D-82
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location, _ . Population, Exposure Matrix .
„ ~Design Outcome Comparison Select Results
Confidence Years Ages,N Levels (ng/mL)a
Liver Estimated, 0.012 (0.009, 0.016)
(enlarged, Quintile 2: 0.001 (-0.016, 0.018)
fatty, or Quintile 3: 0.023 (0.007, 0.040)
cirrhosis) and Quintile 4: 0.036 (0.019, 0.053)
disease (any): Quintile 5: 0.048 (0.031, 0.066)
HR per 1-lny- p-trend < 0.001
ng/mL increase
in PFOA Liver (enlarged, fatty, or cirrhosis)
or by quintiles No lag, 0.97 (0.91, 1.04)
Quintile 2: 0.90 (0.65, 1.25)
Quintile 3: 0.83 (0.60, 1.15)
Quintile 4: 0.75 (0.54, 1.03)
Quintile 5: 0.83 (0.60, 1.16)
10-year lag, 1.00 (0.94, 1.07)
Quintile 2: 1.04 (0.72, 1.50)
Quintile 3: 0.91 (0.64, 1.31)
Quintile 4: 0.84 (0.59, 1.21)
Quintile 5: 0.87 (0.61, 1.25)
Liver disease (any)
No lag, 0.97 (0.92, 1.03)
Quintile 2: 1.19 (0.88, 1.59)
Quintile 3: 1.08 (0.81, 1.45)
Quintile 4: 1.04 (0.78, 1.40)
Quintile 5: 0.95 (0.70, 1.27)
10-year lag, 0.98 (0.93, 1.04)
Quintile 2: 1.15 (0.81, 1.63)
Quintile 3: 1.08 (0.76, 1.54)
Quintile 4: 0.90 (0.63, 1.28)
Quintile 5: 0.99 (0.70, 1.42)
Results: Regression coefficient for modeled continuous PFOA is per In y-ng/mL increase. Lowest quintile used as the reference group.
Confounding: Age, sex, BMI, alcohol consumption, regular exercise, smoking status, education, insulin resistance, fasting status, history of
working at DuPont plant, race
Adults - Other Hepatic Outcomes
D-83
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure Matrix
Levels (ng/mL)a
Outcome Comparison
Select Resultsb
Girardi and
Merler (2019,
6315730)
Low
Italy
1960-2018
Cohort
Male workers at a Serum
PFAS production T2:
Liver cancer or SMR by tertiles Liver cancer or cirrhosis mortality
plant
N = 462
GM= 13,051 ng/m
L-years
T3:
GM= 81,934 ng/m
L-years
cirrhosis
mortality
Liver cirrhosis
mortality
Mortality risk
ratio by tertiles
SMRs:
Tl: 0.44 (0.06,3.15)
T2: 2.76 (1.15, 6.63)
T3: 2.86 (1.36. 6.00)
RRs:
Tl: 1.17(0.15,9.42)
T2: 7.26 (2.37, 22.3)
T3: 6.68 (2.41, 18.5)
Liver cirrhosis mortality
SMRs:
T2: 2.76 (0.89, 8.56)
T3: 2.63 (0.85,8.14)
RRs:
T2: 6.59 (1.57, 27.7)
T3: 5.04(1.19,21.3)
Results: Workers at nearby non-chemical factory used as reference. Tertile 1 used as the reference group.
Confounding: For mortality risk ratio: age at risk, calendar period
Rantakokko et al. Finland
(2015, 3351439)
Medium
2005-2011
Cross-
sectional
Morbidly obese Serum
adults undergoing
bariatric surgery,
N = 160
2.56 (S^S4
percentile: 1.04-
4.66)
Lobular
inflammation
ORperloglO < 2 foci vs. none: 0.71 (0.10, 5.18)
umt increase in
PFOA by level
of lobular
inflammation
2-4 foci vs. none: 0.02 (< 0.01,
0.66), p-value = 0.027
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 from Serum
Levels of ALT
Regression
ALT
(2015,2966740) 2007-2010 sectional
NHANES, 3.7 (2.5-5.2)
(ln-U/L), AST
coefficient per
0.038(0.014,0.062), p-
Medium
Ages > 12,
(ln-U/L), GGT
ln-unit increase
value < 0.001
N = 4,333
(ln-U/L), ALP
in PFOA
ALT, elevated, OR:
(ln-U/L);
Q2: 1.43 (1.11, 1.86)
elevated ALT,
Elevated ALT,
Q3: 1.56(1.15,2.12)
GGT, or AST
GGT, or AST:
Q4: 1.52 (1.18, 1.96)
OR by quartile
p-trend=0.07
D-84
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location, _ . Population, Exposure Matrix .
„ ~Design Outcome Comparison Select Results
Confidence Years Ages,N Levels (ng/mL)a
GGT
0.058(0.021,0.096), p-
value < 0.01
GGT elevated, OR:
Q2: 1.10 (0.80, 1.53)
Q3: 1.12 (0.80, 1.53)
Q4: 1.36 (1.00, 1.82)
p-trend=0.042
AST
0.025 (0.007, 0.043), p-
value < 0.01
AST elevated, OR
Q2: 1.32 (1.03, 1.67)
Q3: 1.27 (0.98, 1.66)
Q4: 1.40 (1.07, 1.83)
p-trend=0.058
ALP
-0.003 (-0.023,
0.016)
Outcome: Elevated clinical biomarkers defined based on the 75th percentile value in the 2007-2010 NHANES.
Results: Lowest quartile used as reference group.
Confounding: Age, gender, race/ethnicity; and BMI group, poverty, smoking, alcohol consumption "if statistically significant associated
with both the exposure and outcome in univariate analysis."
Khalil et al.
(2018, 4238547)
Low
United States Cross-
2016 sectional
Obese children,
Ages 8-12,
N = 48
Serum
0.99 (IQR = 0.45)
Levels of ALT
(u/L), AST
(u/L)
Regression
coefficient per
unit increase in
PFOA
ALT
1.64 (-8.68, 12.00)
AST
0.14 (-4.73,
5.00)
Confounding: Age, sex, race
Attanasio (2019,
United States Cross-
Adolescents from
Serum
Levels of ALT Regression
ALT
5412069)
2013-2016 sectional
NHANES,
Boys:
(ln-IU/L), AST coefficient per
Boys,
Medium
Ages 12-19,
GM = 1.5
(ln-IU/L) ln-unit increase
-0.07 (-0.13,-0.01)
N = 305-354
(SE = 0.06)
in PFOA or by
Q2: 0.02 (-0.16,0.19)
Girls:
quartiles
Q3:-0.01 (-0.13, 0.10)
D-85
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location, _ . Population, Exposure Matrix .
„ ~Design Outcome Comparison Select Results
Confidence Years Ages,N Levels (ng/mL)a
GM = 1.22 Q4:-0.11 (-0.21,-0.01), p-
(SE = 0.06) value = 0.03
p-trend=0.09
Girls,
0.09 (0.02,0.17)
Q2: 0.09 (0.01,0.18)
Q3: 0.16 (0.05, 0.28)
Q4: 0.17 (0.05, 0.28)
p-trend=0.02
AST
Boys,
-0.06 (-0.12, 0)
Q2:-0.01 (-0.14, 0.12)
Q3: 0.00 (-0.08, 0.08)
Q4: -0.05 (-0.15,0.04)
Girls,
0.06 (0.00,0.13)
Q2: 0.04 (-0.02,0.11)
Q3: 0.10 (0.01, 0.19)
Q4: 0.11 (0.01,0.20)
Results: Lowest quartile used as reference group.
Confounding: Age, race/ethnicity, body weight status, education, poverty income ratio, exposure to smoking
United States Cohort Children from Plasma Levels of ALT Regression Prenatal exposure:
1999-2010 Project Viva, Prenatal: 5.4 (3.9- (U/L) coefficient per -0.5 (-1.3,0.2)
N = 508-630 7.6); IQR increase in Mid-childhood exposure:
Mid-childhood: 4.3 PFOA -0.7 (-1.4,0)
(3.0-6.0)
Confounding: For prenatal exposure maternal education, prenatal smoking, gestational age at blood draw, and child's sex, race/ethnicity, age
at lipids/ALT measurements; For mid-childhood exposure maternal education, prenatal smoking, and child's sex, race/ethnicity, age at
lipids/ALT measurements
Children and Adolescents - Other hepatic outcomes
Jin et al. (2020,
United States Cross-
Children and
Plasma
Ballooning,
OR per IQR
Ballooning
6315720)
2007-2015 sectional
adolescents
3.42 (2.5-4.1)
Grade of
increase in
Few balloon cells: 0.99 (0.52,
Medium
diagnosed with
steatosis,
Liver fibrosis,
PFOA
1.86)
Mora et al. (2018,
4239224)
Medium
D-86
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure Matrix
Levels (ng/mL)a
Outcome Comparison
Select Resultsb
nonalcoholic fatty
liver disease,
Ages 7-19,
N = 74
Lobular
inflammation,
Nonalcoholic
steatohepatitis,
Portal
inflammation
Many cells/prominent ballooning:
0.42 (0.07, 2.60)
Grade of steatosis
34%-66% steatosis: 1.41 (0.61,
3.23)
> 66% steatosis: 1.21 (0.60, 2.47)
Liver fibrosis
Mild (Stage 1): 1.68 (0.75, 3.73)
Significant (Stages 2-4): 0.97
(0.33, 2.82)
Lobular inflammation
<2 foci: 0.90 (0.45, 1.81)
2-4 foci: 1.32 (0.52,3.39)
Nonalcoholic steatohepatitis
1.21 (0.67,2.18)
Portal inflammation
Mild: 1.26 (0.65,2.43)
Moderate-to-severe: 0.65 (0.18,
2.39)
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, BMI z-score
Notes: ALT = alanine aminotransferase; APFO = ammonium perfluorooctanoate; AST = aspartate aminotransferase; BMI = body mass index; GF = glomerular filtration;
GGT = y-glutamyltransferase; GM = geometric mean; HOMA-IR = homeostasis model assessment of insulin resistance; HR = hazard ratio; IQR = interquartile range; LDL = low-
density lipoprotein; HDL = high-density lipoprotein; NHANES = National Health and Nutrition Examination Survey; OR = odds ratio; POUNDS = Preventing Overweight Using
Novel Dietary Strategies; 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 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-87
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.4 Immune
Table D-7. Associations between PFOA 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, Cohort
Children
Maternal serum
Antibody
Percent change
Child serum
(2012, 1248827) Denmark
followed from
(prenatal)
concentrations
per doubling in
Anti-diphtheria, prebooster, age 5
Medium Recruitment
birth to age 7
Geometric
(log-IU/mL) for
age 5 and
-6.8 (-28.3,21.0)
1997-2000,
Birth and
mean = 3.20
tetanus and
maternal serum
Anti-diphtheria, postbooster, age 5
Follow-up
infancy:
(2.56,4.01)
diphtheria
PFOA
-6.1 (-23.6, 15.5)
through 2008
N = 587
Anti-diphtheria, age 7
Prebooster
Child serum
-25.2 (-42.9, -2.0)
(mean age 5.0)
(5 years)
Anti-diphtheria, age 7 adjusted for
examination:
Geometric
age 5 Ab
N = 532
mean = 4.06
-23.4 (-39.3,-3.4)
Postbooster
(3.33-4.96)
(mean age 5.2)
Maternal serum
examination:
Anti-diphtheria, prebooster, age 5
N = 456
-16.2 (-34.2, 6.7)
Age 7 (mean
Anti-diphtheria, postbooster, age 5
age 7.5)
-6.2 (-22.4, 13.3)
examination:
Anti-diphtheria, age 7
N = 464
-22.8 (-39.4, -1.7)
Anti-diphtheria, age 7 adjusted for
age 5 Ab
-16.8 (-32.9, 3.3)
Child serum
Anti-tetanus, prebooster, age 5
-13.3 (-31.6,9.9)
Anti-tetanus, postbooster, age 5
-9.7 (-30.7, 17.7)
Anti-tetanus, age 7
-35.8 (-51.9, 14.2)
Anti-tetanus, age 7 adjusted for age
5 Ab
D-88
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location,
Confidence Years
Design
Exposure
Population, Matrix,
Ages, N Levels
(ng/mL)a
Outcome Comparison
Resultsb
-28.2 (-42.7,-10.1)
Maternal serum
Anti-tetanus, prebooster, age 5
-10.5 (-28.2, 11.7)
Anti-tetanus, postbooster, age 5
14.5 (-10.4, 46.4)
Anti-tetanus, age 7
7.4 (-17.1, 39.0)
Anti-tetanus, age 7 adjusted for age
5 Ab
12.3 (-8.6, 38.1)
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.
(2013, 1937228)
Medium
Norway Cohort Mother-infant Maternal serum
1999-2008 pairs from with three days
MoBaat3-year of delivery
follow-up 1.1 (0.8-1.4)
N = 56
Confounding: Maternal allergy, paternal allergy, maternal education,
Levels (OD) of
rubella anti-
vaccine
antibodies
Regression Rubella antibody
coefficient per -0.4 (-0.64, -17)
unit increase in p-value = 0.001
PFOA
: at
3-year follow-up.
Mogensen et al. Faroe Islands,
Cohort Children aged
Serum
Antibody
Percent change
Anti-diphtheria, age 7
(2015, 3981889) Denmark
5-7 years
4.4 (3.5-5.7)
concentrations
per doubling of
-25.4 (-40.9, -5.8)
Medium 2002-2007
N = 443
(log2-IU/mL)
PFOA
(7 years)
for diphtheria or
Anti-tetanus, age 7
tetanus
-20.5 (-38.2,2.1)
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:
diphtheria
per doubling of
Age 7: -4.1 (-25.4, 23.3)
Medium Enrollment:
7 years and
2.0 (1.6-2.5)
antibody (log2-
PFOA
p-value = 0.742
1997-2000
13 years
IU/mL), tetanus
Age 13: -17.5 (-35.6, 5.8)
7 years:
antibody (log2-
p-value = 0.129
N = 505
4.4 (3.5-5.7)
IU/mL)
(13 years)
Tetanus antibody
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
N = 427
(7 years)
Age 7: 9.4 (-24.7, 58.9)
p-value = 0.637
Age 13: 3.3 (-27.3,46.9)
p-value = 0.856
Confounding: Sex, age at antibody assessment, booster type at age 5
Grandjean et al. Faroe Islands,
Cohort and Infants 2 weeks
Serum
Age 5 levels of
Percent change
2007-2009 cohort
(2017, 4239492) Denmark
Cross-sectional after expected
tetanus antibody per doubling of
Tetanus antibody
Medium 1997-2000 and
term date,
18 months:
(IU/mL),
PFOA
Birth: -22.25 (-35.25, -6.63)
2007-2009
followed up at
median =2.8
diphtheria
p-value = 0.007
(year of birth)
18 months and
(2.0-4.5)
antibody
18 mo:-16.31 (-29.04,-1.31)
5 years
(IU/mL)
p-value = 0.034
5 years:
5 yr: -25.26 (-42.63, -2.64)
All: N = 490,
median = 2.2
p-value = 0.031
18 months:
(1.8-2.8)
N = 275,
Diphtheria antibody:
5 years:
Birth: -18.93 (-33.16,-1.66)
N = 349
p-value = 0.033
18 months: 4.19 (-11.76, 23.02)
p-value = 0.63
5 yr: 18.31 (-10.72, 56.78)
p-value = 0.24
Combined cohort
Tetanus antibody:
Birth: -17.59 (-28.38, -5.17)
p-value = 0.007
18 mo: -16.47 (-28.84, -1.96)
p-value = 0.028
5 yr:-18.75 (-31.79, -3.21)
p-value = 0.020
Diphtheria antibody:
Birth: -17.82 (-29.11, -4.74)
p-value = 0.009
18 mo: 5.44 (-10.28, 23.92)
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
p-value = 0.52
5 yr: 3.38 (-14.16, 24.50)
p-value = 0.73
Confounding: Age, sex
Abraham et al. Berlin,
(2020,6506041) Germany
Medium Enrollment:
1997-1999
Cross-sectional
Children, 1 year Plasma
old
All: N = 101,
formula fed:
N = 21,
breastfed:
N = 80
Formula fed:
mean = 3.8 ± 1.
1 (range = 1.6-
6.4)
Breastfed:
mean = 16.8 ± 6
.6 (range = 2.6-
36.7)
Levels of Hib Spearman
antibody, correlation
tetanus antibody coefficient
IgG,
tetanus antibody
IgGl,
diphtheria
antibody
Hib antibody: -0.32
p-value < 0.05
Tetanus antibody IgG: -0.25
p-value < 0.05
Tetanus antibody IgG: -0.22
p-value < 0.05
Diphtheria antibody: -0.23
p-value < 0.05
Confounding: Time since last vaccination
Timmermann et
al. (2020,
6833710)
Medium
Guinea-Bissau
2012-2015
Cohort
Infants enrolled
at 4-7 months
old (inclusion),
followed up at
9 months and
2 years
Inclusion:
N = 236
9-month
Unvaccinated
controls:
N = 100
Intervention:
N = 134
2-year
Maternal blood Measles Percent
0.68 (0.53-0.92) antibody difference per
concentration doubling of
(mlU/mL) PFOA
Inclusion (no measles vaccination):
-12 (-28, 7)
9-month visit
Control (no measles vaccination):
-11 (-36, 22)
Intervention (1 measles
vaccination): 7 (-15, 35)
2-year visit
Control (1 measles vaccination): -9
(-30, 18)
Intervention (2 measles
vaccinations): 12 (-11, 40)
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
Unvaccinated
controls:
N = 102
Intervention:
N = 92
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 Cohort and Vaccinated Maternal serum Levels (IU/mL)
al. (2022, Recruitment: cross-sectional children ages 7- from pregnancy of diphtheria
9416315) 1999-2005, 12 years and 2.28 (1.89-2.88) and tetanus
Medium Examination: their mothers at antibody
2012-2015 pregnancy Child serum
2.13 (1.68-2.54)
Maternal serum
N = 57
Child serum
N = 169
Percent
difference per
unit increase in
PFOA
OR per log 10-
unit increase in
PFOA
Diphtheria antibody
Child serum
Percent difference: -22 (-48, 16)
OR: 1.41 (0.91,2.19)
Maternal serum
Percent difference: 44 (-15, 145)
Tetanus antibody
Child serum
Percent difference: -8 (-30,
Maternal serum
Percent difference: -7 (-44,
21)
56)
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
Cohort
Infants from
Cord blood HFMD antibody Percent change
CA16
(2019, 5081554) 2013
Guangzhou
1.22 (0.86-1.74) titers (CA16 or or OR (below
Cord blood: -16.3 (-25.3, 6.1)
Low
Birth Cohort
EV71) in serum clinical
Girls: -8.7 (-22.6, 7.6)
Study at birth
of cord blood or protection) per
Boys: -22.0 (-33.1,-8.9)
and 3 months
at 3 months doubling of
p months: -6.9
Birth N= 194
PFOA
(-13.2, 0)
(91 girls, 103
Girls: -3.2 (-11.2, 5.5)
boys)
Boys: -11.1 (-20.7,-0.3)
3-month
N = 180 (89
CA16 below clinical protection
girls, 91 boys)
Cord blood: 1.56(1.13,2.14);
p-value = 0.007
Girls: 1.16(0.72, 1.87)
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
Boys: 1.95 (1.16, 3.27)
p-value for interaction by
sex = 0.181
q
Girls: 1.31 (0.71.
Boys: 2.49 (1.23
p-value for interaction by
sex = 0.263
months: 1.73
(1.08,2.75);
p-value = 0.022
2.44)
5.04)
EV71
Cord blood: -18.7 (-28.6, -7.4)
Girls: -14.6 (-30.4, 4.6)
Boys: -20.6 (-32.5, -6.6)
r months: -7.2
(-13.2, -0.8)
Girls: -4.9 (-13.7, 4.8)
Boys: -8.2 (-16.2, 0.5)
EV71 below clinical protection
Cord blood: 1.49 (1.09,2.05);
p-value = 0.014
Girls: 1.27 (0.84, 1.93)
Boys: 1.76 (1.07, 2.87)
p-value for interaction by
sex = 0.282
s months: 2.11
(1.27, 3.48);
p-value = 0.004
Girls: 1.52 (0.81,2.85)
Boys: 2.90 (1.34, 6.29)
p-value for interaction by
sex = 0.202
Outcome: Clinical protection threshold defined as titers > 1:8 in modified cytopathogenic effect assay.
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
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 Serum Influenza Regression Influenza type B titer rise
water districts GM(95% antibodies (titer coefficient per Per loglO-unit:-0.2 (-0.13, 0.09),
of Ohio and CI) = 33.74 ratio and titer loglO-unit p-value = 0.73
West Virginia (29.78-38.22) rise, loglO- increase in Q2:-0.03 (-0.19, 0.13), p-
with transformed): PFOA, or by value = 0.69
contaminated A/H1N1, quartiles Q3:-0.02 (-0.19,0.15), p-
drinking water A/H3N2, type value = 0.82
N = 403 B; influenza Influenza Q4:-0.07 (-0.24,0.10), p-
A/H3N2 A/H3N2 value = 0.42
seroconversion seroconversion: Influenza type B titer ratio
ORperloglO- Per loglO-unit: -0.02 (-0.11, 0.08),
unit increase in p-value = 0.73
PFOA, or by Q2: 0.05 (-0.09, 0.19), p-
quartiles value = 0.53
Q3: 0.07 (-0.07, 0.22), p-
value = 0.32
Q4: -0.03 (-0.17,0.12), p-
value = 0.71
Influenza A/H3N2 titer rise
Per loglO-unit: -0.01 (-0.17, 0.14),
p-value = 0.86
Q2:-0.28 (-0.51,-0.06), p-
value = 0.02
Q3:-0.37 (-0.60,-0.13), p-
value = 0.002
Q4:-0.12 (-0.36, 0.13), p-
value = 0.35
Influenza A/H3N2 titer ratio
Per loglO-unit: -0.12 (-0.25, 0.02),
p-value = 0.09
Q2:-0.10 (-0.30, 0.10), p-
value = 0.31
D-94
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Q3:-0.07 (-0.28, 0.14), p-
value = 0.49
Q4:-0.22 (-0.43,-0.01), p-
value = 0.04
Influenza A/H1N1 titer rise
Per loglO-unit: -0.30 (-0.14, 0.09),
p-value = 0.63
Q2: -0.09 (-0.27, 0.08), p-
value = 0.31
Q3:-0.10 (-0.28,-0.09), p-
value = 0.30
Q4:-0.12 (-0.30, 0.06), p-
value = 0.19
Influenza A/ H1N1 titer ratio
Per loglO-unit: 0.07 (-0.06, 0.21),
p-value = 0.30
Q2:-0.08 (-0.29, 0.12), p-
value = 0.43
Q3:-0.04 (-0.25, 0.18), p-
value = 0.72
Q4: 0.07 (-0.14, 0.29, p-
value = 0.51
Results: Lowest quartile used as reference group
Confounding: Age (cubic spline), gender, mobility, and history of previous influenza vaccination.
Influenza A/H3N2 seroconversion
not statistically significant
Pilkerton et al. United States
Cross-sectional Adults and
Serum
Rubella IgA
Regression
Adolescents:
(2018, 5080265) 1999-2000
adolescents
titers (log-IU)
coefficient by
Per quartile increase:
Medium for
12 years and
Women:
quartiles or per
No associations. F-value = 0.34,
youth
older
mean = 4.3,
quartile increase
p-value = 0.80
Low for adult
SE = 0.2
in PFOA
Youths:
Adults:
N = 1,012
Per quartile increase:
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
Adults: N = 542 Men:
women, 613 mean = 6.0
men SE = 0.3
F-value = 6.60, p-value = 0.002
Women
Q2: -0.25 (-0.52, 0.02)
p-value = 0.064
Q3: -0.15 (-0.9,0.6)
p-value = 0.686
Q4:-0.17 (-0.97, 0.64)
p-value = 0.677
Men
Q2:-0.14 (-0.43, 0.15)
p-value = 0.339
Q3: -0.55 (-0.81,-0.28)
p-value = 0.0002
Q4: -0.45 (-0.84, -0.05)
p-value = 0.028
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: 0.87 (0.70,
1.08), p-
adults aged 12-
GM (SE) = 2.54
cytomegalovirus
Prevalence ratio
value = 0.24
49 years
(0.06)
, Epstein-Barr
per doubling in
20-49 years: 0.98 (0.91,
1.05), p-
12-19 years:
20-49 years:
virus, hepatitis
PFOA
value = 0.57
N= 3,189
GM (SE) = 2.68
C, hepatitis E,
20-49 years:
(0.03)
herpes simplex
Pathogen
Epstein-Barr virus
N= 5,589
virus 1, herpes
burden: Relative
12-19 years: 0.99 (0.94,
1.05), p-
simplex virus 2,
difference per
value = 0.0.83
Toxoplasma
log2-unit
gondii, and
increase in
Hepatitis C virus
Toxocara
PFOA
20-49 years: 0.89 (0.62,
1.29), p-
species;
value = 0.54
pathogen burden
Hepatitis E virus
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
20-49 years: 1.01 (0.78, 1.31), p-
value = 0.92
Herpes simplex virus 1
12-19 years: 1.02 (0.93, 1.11), p-
value = 0.75
20-49 years: 1.03 (1.01, 1.06), p-
value = 0.01
Herpes simplex virus 2
20-49 years: 1.11 (1.05, 1.17), p-
value < 0.01
Toxoplasma gondii
12-19 years: 0.99 (0.68, 1.42), p-
value = 0.94
20-49 years: 1.03 (0.89, 1.18), p-
value = 0.70
Toxocara species
12-19 years: 1.21 (0.56, 2.65), p-
value = 0.63
20-49 years: 1.23 (1.00, 1.51), p-
value = 0.08
Pathogen burden
12-19 years: 1.36 (1.27, 1.45)
20-49 years: 1.09 (1.06, 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 Cohort and Adults from Serum Levels (In- Levels: Relative
etal. (2021, 2005-2006, cross-sectional C8HP 2005-2006: cells/|iL or difference per 1-
7751049) 2010 26.9 (13.2-69.2) percentage of
White blood cells, total
2005-2006: -0.27 (-0.62, 0.08)
2010: 0.84 (-2.20, 3.97)
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
Medium
2005-2006:
N = 42,782
2010: N = 526
2010: 35.7
(15.0-93.7)
white blood
In IQR increase
cells/lymphocyt
in PFOA
es) of white
blood cells,
Percentages:
neutrophils,
Difference per
monocytes,
1-ln IQR
eosinophils,
increase in
lymphocytes,
PFOA
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
Likelihood ratio test p-
value < 0.001 for the comparison
between the two time points
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: -4.34 (-30.69, 32.02)
Medium Recruitment:
birth, 7, 14, 22,
1.11
antibody,
increase in
7-year serum: -9.39 (-43.4, 45.04)
1986-1987,
and 28 years
(IQR = 0.62)
hepatitis B
PFOA
14-year serum: -7.4 (-47.65,
Follow-up
N = 399
antibody,
63.81)
through 2015
Serum
diphtheria
22-year serum: -21.24 (-42.2,
7 year: 5.11
antibody,
7.34)
(IQR = 2.45)
tetanus
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
14 year: 4.98
(IQR = 2.11)
22 year: 2.96
(IQR = 1.69)
28 year: 1.28
(IQR = 0.90)
antibody;
Hepatitis A
antibody signal-
to-cutoff ratio
Confounding: Sex
28-year serum:
7.35)
-16.77 (-35.47,
Hepatitis Type A
Cord blood: 0.05 (-0.36, 0.46)
7-year serum: 0.1 (-0.52, 0.72)
14-year serum: -0.71 (-1.52, 0.09)
22-year serum: -0.06 (-0.48, 0.35)
28-year serum: -0.24 (-0.59, 0.1)
Diphtheria
Cord blood: 28.14 (-0.38, 64.82)
7-year serum: -4.89 (-37.24,
44.11)
14-year serum: -11.6 (-47.55,
48.97)
22-year serum: 9.8 (-12.62, 37.96)
28-year serum: 23.56 (3.65, 47.29)
Tetanus
Cord blood: -2.57 (-20.38, 19.22)
7-year serum: 4.68 (-23.9, 43.99)
14-year serum: -0.77 (-36.35,
54.7)
22-year serum: -0.39 (-17.12,
19.72)
28-year serum: 3.1 (-10.42, 18.66)
Zeng et al. China
(2020,6315718) 2015-2016
Low
Cross-sectional Adults from the Serum Hepatitis B Regression
Isomers of C8 5.12 (3.82-8.11) surface antibody coefficient or
Health Project (HbsAb) (log- OR(HbsAb
N = 605 mlU/mL) or seronegative)
surface antigen per loglO-unit
(HbsAg) (mlU- increase in
mL); HbsAb PFOA
HbsAb concentration
-0.38 (-0.79, 0.02);
p-value = 0.061
HbsAb seronegative
1.89 (1.23, 2.90); p-value = 0.004
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
seronegative HbsAg concentration
( < 10 mlU/mL) 0.41 (-0.42, 1.25); p-value = 0.33
Confounding: Age, gender, BMI, career, income, alcohol drinking, smoking, regular exercise; education for HbsAb concentration alone
Notes: Ab = antibody; BMI = body mass index; C8HP = C8 Health Project; CI = confidence interval; HAI = hemagglutinin inhibition; ICH = immunohistochemistry;
MoBa = Norwegian Mother and Child Cohort Study; NHANES = National Health and Nutrition Examination Survey; 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.
c Confounding indicates factors the models presented adjusted for.
Table D-8. Associations between PFOA Exposure and Infectious Disease in Recent Epidemiological Studies
Reference, Location,
Confidence Years
Design
Exposure
Population, Matrix,
Ages, N Levels
(ng/mL)a
Outcome Comparison
Resultsb
Denmark,
Recruitment:
1996-2003;
Follow up: 2008
Children
Cross-sectional Mother infant Maternal plasma Infectious
and cohort pairs with Mean disease
follow-up to 11 (range) = 5.6 hospitalizations
years (DNBC) ( < LLOQ-
N= 1,400 41.5);
LLOQ = 1.0 ng/
mL
Fei et al. (2010;
1290805)
Medium
IRR by quartiles Girls
or per quartile Q2: 1.20(0.76, 1.89)
increase in Q3: 1.63 (1.03, 2.58)
PFOA Q4: 1.74(1.06,2.87)
Per quartile increase: 1.21 (1.04,
1.42)
Boys
Q2: 0.58 (0.40, 0.83)
Q3: 0.53 (0.36,0.76)
Q4: 0.57 (0.38,0.86)
Per quartile increase: 0.83 (0.73,
0.95)
All children
Q2: 0.71 (0.53,0.94)
Q3: 0.77 (0.59, 1.03)
D-100
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Q4: 0.84 (0.62, 1.13)
Perquartile increase: 0.96 (0.87,
1.06)
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
Gourdazi et al. Hokkaido,
(2017, 3859808) Japan
Medium 2003-2009
Children, early
pregnancy
followed up at
4 years
N = 1,558 (793
boys, 765 girls)
Maternal blood
2.01 (1.31-3.35)
Infectious
diseases, total
(including Otitis
media,
Pneumonia, RS
virus, Varicella)
Cohort Children, early Maternal blood Infectious OR by quartiles Girls
Q2: 1.45 (0.92,2.3)
Q3: 1.37 (0.87,2.19)
Q4: 1.37(0.86,2.21)
p-value for trend = 0.242
Boys
Q2: 1.02 (0.67, 1.56)
Q3: 1.34 (0.87,2.11)
Q4: 0.952 (0.61, 1.49)
p-value for trend = 0.854
All
Q2: 1.17(0.87, 1.6)
Q3: 1.32 (0.97, 1.82)
Q4: 1.11 (0.81, 1.54)
p-value for trend = 0.393
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 aged
Maternal blood LRTI
OR or RR per
OR
Salgado et al.
2003-2008
1.5, 4, or 7 years
2.35 (1.63-
log2-unit
1.5 years: 0.92 (0.79, 1.07)
(2019, 5412076)
Age 1.5:
3.30)
increase in
4 years: 1.11 (0.94, 1.31)
Medium
N = 1,188
PFOA
7 years: 0.69 (0.47, 1.01)
D-101
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Age 4:
N = 1,184
Age 7:
N = 1,068
RR, 1.5-7 years
All: 0.96 (0.85, 1.08)
Boys: 0.97 (0.82, 1.14)
Girls: 0.99 (0.83, 1.18)
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
Ait Bamai et al.
(2020, 6833636)
Medium
Hokkaido,
Japan
Enrollment:
2003-2012
Children, early
pregnancy
followed up at
7 years
N = 2,689
Maternal blood Chicken pox,
1.94 (1.30-2.95) RSV, otitis
media,
pneumonia,
wheeze, eczema
OR or RR per
ln-unit increase
in PFOA
Cohort Children, early Maternal blood Chicken pox, OR or RR per Pneumonia: OR: 1.17 (1.01, 1.37);
p-value = 0.043
Otitis media: OR: 1.06 (0.92, 1.22);
p-value = 0.45
Chicken pox: OR: 0.94 (0.81,
1.09); p-value = 0.381
RSV: OR: 0.96 (0.8, 1.17);
p-value = 0.694
Wheeze: RR: 0.92 (0.84, 1.01);
p-value = 0.089
Eczema: RR: 0.85 (0.77, 0.94);
p-value = 0.001
Confounding: Sex, maternal age, parity, maternal smoking during pregnancy, BMI pre-pregnancy, annual household income during
pregnancy, duration nursing, and presence of siblings
Grandjean et al. Denmark
(2020, 7403067) 2020
Medium
Cross-sectional Adults, ages 30-Plasma Covid-19 OR per unit
70 years, with 0.77 (0.43-1.18) severity increase in
known SARS- PFOA
CoV-2 infection
N = 323
Covid-19 severity
0.83 (0.57, 1.20)
Covid-19 severity (hospitalization
vs. no hospitalization)
1.11 (0.37,3.32)
D-102
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Covid-19 severity (intensive care
unit and/or deceased vs.
hospitalization)
0.90 (0.29, 2.80)
Confounding: Age, sex, kidney disease, other chronic disease, national origin, place of testing, and days between blood sampling and
diagnosis
Huang et al.
(2020, 6988475)
Medium
China
Recruitment:
2011-2013,
Follow-up at
5 years
Cohort
Children ages Cord blood Respiratory tract Recurrent Total respiratory tract infections
1-5 years 6.68 (4.82-9.13) infections (total respiratory tract 0.37 (-3.63, 4.38), p-value = 0.854
N = 344 (182 and recurrent) infections: OR
boys, 162 girls) for > 75th Recurrent respiratory tract
percentile vs. infections
< 75th percentile 0.90 (0.49, 1.64), p-value = 0.73
PFOA
Results stratified by age and sex not
statistically significant
Confounding: Infant sex, maternal age, maternal education level, birth weight
Dalsager et al.
(2021, 7405343)
Medium
Denmark
Recruitment:
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
1.68(0.27-12.5) from infection log2-unit
(any infection,
upper
respiratory tract,
lower
respiratory tract,
gastrointestinal,
other)
increase in
PFOA
1.13 (0.97, 1.29)
Upper respiratory infection
1.18(0.93, 1.5)
Lower respiratory infection
1.27 (1.01, 1.59)
Gastrointestinal infection
0.55 (0.32,0.95)
Other infection
1.12 (0.93, 1.35)
Results stratified by sex not
statistically significant
Confounding: Maternal age, parity, maternal educational level, child sex, child age
D-103
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Exposure
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Matrix,
Levels
(ng/mL)a
Outcome
Comparison Resultsb
Jietal. (2021,
China
Case-control
Adults
Urine
COVID-19
OR per log2-SD COVID-19
7491706)
2020
N = 160
infection
change inPFOA 2.73 (1.71, 4.55)
Medium
Controls: 24.8
(16.9-36.3) ng/g
creatinine
Cases: 39.6
(27.5-48.9) ng/g
creatinine
Confounding: Age, gender, body mass index, diabetes, cardiovascular diseases, and urine albumin-to-creatinine ratio
Wang et al.
China
Cohort
Pregnant
Maternal serum
Common cold,
OR per log 10- Common cold
(2022,
Recruitment:
women and
at delivery
bronchitis/pneu
unit increase in OR: 1.36 (0.60, 3.09), p-
10176501)
2010-2013,
their children at
45.82 (28.72-
monia, diarrhea
PFOA value = 0.469
Medium
Follow-up after
1 year
1 year from
LWBC
N = 235
77.34)
IRR: 1.18 (0.85, 1.63), p-
IRR per loglO- value = 0.329
unit increase in
PFOA Bronchitis/pneumonia
OR: 1.14 (0.37,3.54), p-
value = 0.822
IRR: 0.68 (0.30, 1.53), p-
value = 0.350
Diarrhea
OR: 4.99 (1.86, 13.39), p-
value = 0.001
IRR: 1.97 (1.32, 2.94), p-
value = 0.001
Confounding: Maternal age, pre-pregnancy BMI, smoking during pregnancy, maternal education level, and parity
Dalsager et al. Odense,
Cohort
Children,
Maternal serum
Fever, cough,
OR (of
Fever
(2016, 3858505) Denmark
pregnancy
1.68
nasal discharge,
proportion of
T2: 1.55 (0.90,2.95)
Low 2010-2012
followed up at
(range = 0.32-
diarrhea,
days with
T3: 1.97 (1.07,3.62);
1-4 years
10.12)
vomiting
symptoms) by
tertiles
p-value < 0.05
N = 346
Cough
T2: 0.72 (0.42, 1.24)
D-104
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location,
Confidence Years
Design
Exposure
Population, Matrix,
Ages, N Levels
(ng/mL)a
Outcome Comparison
Resultsb
T3: 1.01 (0.42, 1.24)
Nasal discharge
T2: 1.19(0.70,2.04)
T3: 1.37 (0.75,2.51)
Diarrhea
T2: 1.10(0.64, 1.89)
T3: 0.94 (0.51, 1.74)
Vomiting
T2: 1.05 (0.62, 1.78)
T3: 0.95 (0.52, 1.72)
Results: Lowest tertile used as reference group
Confounding: Maternal age, maternal educational level, parity, and child age.
Impinen et al.
Oslo, Norway Cohort, Nested
Infants followed
Cord blood
Common cold
Regression
Common cold 0-2 years
(2018, 4238440) Recruited 1992- case-control
up at 2 and
1.6(1.2-2.1)
episodes from
coefficient per
-0.04 (-0.08, 0.01)
Low
1993, followed
10 years of age
0-2 years, LRTI log2-unit
p-value = 0.089
up for 10 years
N = 641
episodes from
increase in
0-10 years
PFOA
LRTI 0-10 years
0.28 (0.22, 0.35)
p-value < 0.0001
Confounding: Child sex
Impinen et al.
Oslo, Norway, Cohort
Pregnant
Maternal blood
Common cold,
OR per 1-IQR
Common cold, 0-3 years: 0.96
(2019, 5080609) Enrollment:
women and
2.54 (1.86-3.30)
bronchitis/pneu
increase in
(0.94, 0.99); p-value < 0.05
Low
1999-2008
their infants
monia, throat
PFOA
followed up at 3
infection with
Bronchitis/pneumonia
and 7 years
strep,
0-3 years: 1.27 (1.12, 1.43);
pseudocroup,
p-value < 0.05
0-3 years:
ear infection,
6-7 years: 0.75 (0.45, 1.23)
N = 1,207
diarrhea/gastric
6-7 years:
flu, urinary tract
Throat infection with strep, 0-
N = 921
infection
3 years: 1.14 (0.96, 1.35)
D-105
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Exposure
Design Population, at , Outcome Comparison Resultsb
Ages, N Levels
(ng/mL)a
Other throat infections, 0-3 years:
0.91 (0.80, 1.04)
Pseudocroup, 0-3 years: 1.22 (1.07,
1.38); p-value < 0.05
Ear infection
0-3 years: 1.00 (0.92, 1.08)
6-7 years: 1.12(0.88, 1.41)
Diarrhea/gastric flu
0-3 years: 1.00 (0.94, 1.06)
6-7 years: 1.48 (1.31, 1.67);
p-value < 0.05
Urinary tract infection
0-3 years: 0.78 (0.69, 0.88);
p-value < 0.05
6-7 years: 0.66 (0.43, 1.01)
Confounding: Maternal age, maternal BMI, maternal education, parity, smoking during pregnancy
Kvalem et al. Norway
Cohort and Children,
Serum
Common cold, Colds: OR
Colds, 10-16 years
(2020,6316210) Enrollment:
cross-sectional 10 years
LRTI (reference: 1-2
3-5 colds
Low 1992-1993
N = 378 (193
All: 4.36 (IQR:
colds)
All: 1.23 (0.33,4.58)
Follow-up:
boys, 185 girls)
1.77)
p-value = 0.76
2002-2009
Boys: 4.53
LRTI: RR per
Boys: 1.41 (0.29, 6.89)
Children, 10-
(IQR: 1.86)
IQR-unit
p-value = 0.67
16 years
Girls: 4.13
increase in
Girls: 1.32(0.19,9.21)
N = 375 (191
(IQR: 1.63)
PFOA
p-value = 0.78
boys, 184 girls)
> 5 colds:
All: 1.29 (0.36, 4.64)
Children,
p-value = 0.7
16 years
Boys: 1.38 (0.29, 6.54)
N = 375 (191
p-value = 0.69
boys, 184 girls)
Girls: 1.67 (0.26, 1.09)
p-value = 0.59
Reference, Location,
Confidence Years
D-106
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Exposure
Reference, Location, _ . Population, Matrix,
„ j-. Design ,r ,T Outcome Comparison Results
Commence Years Ages, N Levels
(ng/mL)a
LRTI
10-16 years
All: 1.1 (1.02, 1.19)
p-value = 0.01
Boys: 1.11 (0.97, 1.26)
p-value = 0.12
Girls: 1.49 (1.15, 1.92)
p-value = 0.002
16 years
All: 1.14(0.81, 1.59)
p-value = 0.45
Boys: 1 (0.64, 1.59)
p-value = 0.99
Girls: 1.61 (0.72, 3.58)
p-value = 0.25
Confounding: Puberty status at 16 years, mother's education, physical activity level at 16 years
Occupational
Costa et al. Italy
Cross-sectional Current and
Serum
Concentration of Comparison of
No significant difference in
(2009, 1429922) 2007
former male
Production
WBC (x 109/L) mean outcome
comparison of mean WBC count
Medium
employees of
workers (2007):
(Exposed vs
an Italian
3.89 ng/mL
unexposed
WBC
chemical
(2.18-18.66
workers)
Exposed vs Unexposed: 0.58 (-
production
Hg/mL)
0.19, 1.35)
plant,
Regression
Continuous: 0.029 (-0.011, 0.071)
Comparison of
coefficient
means analysis
(exposed
N = 68,
workers vs all
Exposed vs
workers)
Unexposed
analysis
Regression
N = 141,
coefficient per
Continuous
unit increase in
regression
PFOA
analysis
D-107
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Exposure
Design Population, at , Outcome Comparison Resultsb
Ages, N Levels
(ng/mL)a
N = 56
Confounding: Age, job seniority, body mass index, smoking and alcohol consumption. Additional confounding for continuous regression
analyses: year of observation
Notes: CI = confidence interval; DNBC = Danish National Birth Cohort; IQR = interquartile range; IRR = incidence rate ratio; LLOQ = lower limit of quantitation;
LWBC = Laizhou Wan Birth Cohort; SE = standard error; BMt = body mass index; LRTI = lower respiratory tract infection; OCC = Odense Child Cohort; OR = odds ratio;
Q2 = quartile 2; Q3 = quartile 3; Q4 = quartile 4; RR = risk ratio; RSV = respiratory syncytial virus; T2 = tertile 2; T3 = tertile 3; WBC = white blood cell.
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.
Reference, Location,
Confidence Years
Table D-9. Associations Between PFOA Exposure and Asthma in Recent Epidemiologic Studies
Exposure
Reference, Location, , _ . Population, Matrix,
„ j-. Study Design f -T Outcome Comparison Results
Confidence Years Ages, N Levels
(ng/mL)a
Children
Dong et al. Taiwan, 2009-
Case control
Children from
Serum
Asthma,
Asthma: OR by
Asthma
(2013, 1937230) 2010
and cross-
GBCA with
Cases: 1.2
Asthma Control
quartiles of
Q2: 1.58 (0.89,2.8)
Medium
sectional
(cases) or
(0.50-2.2)
Test score,
PFOA
Q3: 2.67 (1.49, 4.79)
without
Controls: 0.5
asthma severity
Q4: 4.05 (2.21,7.42)
(controls)
(0.4-1.3)
score, IgE in
Asthma Control
p-trend <0.001
asthma, ages
serum (IU/mL),
Test score,
10-15 years,
AEC (106/L),
asthma severity
IgE
N = 231 (cases),
ECP in serum
score, IgE,
Q1:512.1 (329.4, 694.8)
N = 225
(^g/L)
AEC, ECP:
Q2: 604.6 (422.1,787.1)
(controls)
mean values by
Q3: 788.2 (607.1, 969.2)
quartiles
Q4: 836.4 (652, 1,020.8)
p-trend = 0.05
AEC
Ql: 325.9 (253.7, 398.1)
Q2: 339.7 (266.8, 412.6)
Q3: 422.1 (349.9, 494.2)
D-108
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Exposure
Reference, Location, , _ . Population, Matrix,
„ j-. Study Design ' .. Outcome Comparison Results
Confidence Years Ages, N Levels
(ng/mL)a
Q4: 498 (423.7, 572.3)
p-trend <0.001
ECP
Ql: 30.3 (14.3,46.3)
Q2: 34.8 (18.9, 50.7)
Q3: 44.3 (28.4, 60.2)
Q4: 57.8 (42.2, 73.4)
p-trend = 0.010
Asthma Control Test score, asthma
severity 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
4.0 (2.8-5.4)
Ever asthma
4.3 (3.1-5.7)
No current
asthma
4.0 (2.8-5.4)
Current asthma
4.2 (2.9-5.6)
No wheezing
4.0 (2.9-5.5)
Wheezing
4.4 (2.9-5.6)
Asthma, wheeze OR per
doubling in
PFOA or per
unit increase in
PFOA
Ever asthma
Per doubling: 1.18(1.01, 1.39), p-
value = 0.04
Per unit increase: 1.06(1.00, 1.11),
p-value =1.11
Current asthma
Per doubling: 1.12 (0.92, 1.36), p-
value = 0.26
Per unit increase: 1.03 (0.97, 1.10),
p-value = 0.30
Wheeze
Per doubling: 1.0 (0.80, 1.23), p-
value = 0.98
Per unit increase: 1.01 (0.94, 1.07),
p-value = 0.87
Exposure: No wheezing defined as no wheezing in the past 12 months. Wheezing defined as history of wheezing in the past 12 months.
Confounding: Sex, smoking, age, race/ethnicity, survey cycle, poverty income ratio, health insurance
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
Smit et al.
(2015, 2823268)
Medium
Ukraine and
Greenland,
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
Asthma
OR per SD
increase in
PFOA
Asthma ever (combined): 0.8 (0.62,
1.04)
Ukraine: 0.93 (0.47, 1.84)
Greenland: 0.79 (0.60, 1.03)
Maternal blood
Ukraine:
GM = 0.97 (P5-
P95: 0.45-2.34)
Greenland:
GM = 1.79 (P5-
P95: 0.80-3.66)
Confounding: Maternal allergy, smoking during pregnancy, education level, maternal age, child sex, child age at follow-up, gestational age at
blood sample, parity, breastfeeding, and birthweight0
Impinen et al.
(2018, 4238440)
Medium
Oslo, Norway,
1992-2002
Cohort, Nested
case-control
Infants followed Cord blood
up at 2 and 1.6(1.2-2.1)
10 years of age,
N = 641
Asthma
OR per log2-
unit increase
PFOA
Current asthma (10y):
1.06 (0.82, 1.37); p-value = 0.649
Asthma ever (lOy):
1.1 (0.78, 1.54); p-value = 0.589
Confounding: Sex
Beck et al. Denmark,
(2019, 5922599) Enrollment:
Medium 2010-2012
Children, early
pregnancy to
5 years
N = 970 (507
boys, 363 girls)
Maternal blood Wheeze, self-
1.68 (1.13-2.35) reported asthma,
doctor-
diagnosed
asthma
OR per
doubling in
maternal serum
PFOA
Cohort Children, early Maternal blood Wheeze, self- OR per Wheeze
All: 0.98 (0.78, 1.23)
Boys 0.94 (0.71, 1.23)
Girls: 1.08 (0.75, 1.55)
Self-reported asthma
All: 1.57 (0.93,2.68)
Boys: 2.17 (1.07,4.42)
Girls: 1.06 (0.49,2.30)
Doctor-diagnosed asthma
All: 0.81 (0.53, 1.22)
Boys: 0.72 (0.46, 1.12)
Girls: 1.70 (0.63,4.56)
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
without
(controls)
Asthma OR per log-unit 1.34 (0.55,3.29)
increase in
PFOA
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
asthma aged
13-22,
N = 118 (cases),
N = 169
(controls)
Cases: 1.80
(Range: 0.56-
5.03)
Controls: 1.38
(Range: 0.36-
4.28)
Comparison: Logarithm base not specified.
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 2.54 increase in
Total: 1.11 (0.69, 1.79);
Medium 1999-2008 their infants (1.86-3.30) PFOA
p-value = 0.657
(followed to age
Boys: 1.34 (0.70, 2.60);
7),
p-value = 0.38
N = 921
Girls: 0.91 (0.46, 1.82);
p-value = 0.799
Ever asthma:
Total: 0.99 (0.70, 1.39);
p-value = 0.933
Boys: 0.98 (0.63, 1.54);
p-value = 0.945
Girls: 0.99 (0.58, 1.70);
p-value = 0.982
Confounding: Maternal age, maternal BMI, maternal education, parity, smoking during pregnancy
Manzano-
Salgado et al.
(2019, 5412076)
Medium
Spain,
2003-2008
Cohort
Children,
4 years,
N = 1,184
7 years,
N = 1,068
Maternal blood Asthma
2.35 (1.63-3.30)
OR or RR per 4-year follow-up: OR = 0.77 (0.50,
log2-unit 1.17)
increase in 7-year follow-up: OR = 0.77 (0.54,
maternal PFOA 1.10)
4 and 7 years
Girls: RR= 1.01 (0.61, 1.68)
Boys: RR = 0.74 (0.49, 1.13)
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
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
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
Asthma
OR per log 10-
unit increase in
PFOA
Boys: 7.13
(5.15-9.97)
Girls: 6.51
(4.57-8.73)
Confounding: Child weight at age 5, gestational age, breastfeeding during the first 6 months, maternal education, maternal pre-pregnancy
BMI, and annual household income
All: 0.98 (0.22, 4.49),
p-value = 0.98
Boys: 0.32 (0.04, 2.36),
p-value = 0.26
Girls: 5.6 (0.22, 145.87),
p-value = 0.30
Huang et al. China
Cohort Children ages Cord blood IgG, IgE levels
Regression
IgG
(2020, 6988475) Recruitment:
1-5 years 6.68 (4.82-9.13)
coefficient per
0.01 (-0.05, 0.06), p-value = 0.856
Medium 2011-2013,
N= 344 (182
loglO-unit
Follow-up at
boys, 162 girls)
increase in
IgE
5 years
PFOA
-0.30 (-0.64, 0.04), p-
value = 0.084
Results stratified by age and sex not
statistically significant
Confounding: Infant sex, maternal age, maternal education level, birth weight
Jackson-Browne NHANES,
Cross-sectional Children, ages Serum Asthma
OR per ln-SD
p.l (0.9,1.4)
et al. (2020, United States,
3-11 years, GM =1.9 (1.4-
increase in
6833598) 2013-2014
N = 607 2.7)
PFOA
By age:
Medium
3-5 y: 1.6(1.0,2.7)
6-11 y: 1.0 (0.7, 1.3)
p-value for interaction by
age = 0.47
By sex:
Females: 1.1 (0.6, 1.7)
Males: 1.1 (0.9, 1.4)
p-value for interaction by
sex = 0.65
By race/ethnicity:
White, non-Hispanic: 1.3 (0.9,2.0)
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
Black, non-Hispanic: 0.9 (0.7, 1.3)
Hispanic: 1.3 (0.9, 1.9)
Other: 1.1 (0.6, 1.7)
p-value for interaction by
race = 0.41
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: 1.06 (0.93, 1.21)
Medium 1992-1993;
N= 378 (193
All: 4.36 (IQR: PFOA
Boys: 0.99 (0.84, 1.16)
Follow-up:
boys, 185 girls)
1.77)
2002-2009
Boys: 4.53
10-16 years
Children, 10-
(IQR: 1.86)
All: 1.04 (0.88, 1.23)
16 years
Girls: 4.13
Boys: 0.95 (0.72, 1.26)
N= 375 (191
(IQR: 1.63)
Girls: 1.36 (0.98, 1.89)
boys, 184 girls)
16 years
Children,
All: 1.04 (0.87, 1.24)
16 years
Boys: 0.99 (0.76, 1.27)
N= 375 (191
Girls: 1.21 (0.81, 1.82)
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
Okada et al.
(2012, 1332477)
Medium
Japan
2002-2005
Cohort
Pregnant
women and
children from
the Hokkaido
Study on
Environment
and Children's
Health; follow
up at 18 months
N = 128
Maternal serum IgE levels Regression Linear regression
1.3(0.8-1.7) (loglO-IU/mL) coefficients per 0.766(0.104,1.428)
loglO-unit
increase in Quadratic regression
PFOA -1.429 (-2.416, -0.422)
Cubic regression
-3.078 (-5.431,-0.726)
Results stratified by gender not
statistically significant for boys and
combined
D-113
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Exposure
Reference, Location, , _ . Population, Matrix,
„ j-. Study Design f -T Outcome Comparison Results
Confidence Years Ages, N Levels
(ng/mL)a
Confounding: Maternal age, maternal allergic history, distance from home to highway, parity, birth season, and blood sampling period
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) = 3.87
(3.13) (ig/L
Fractional
exhaled nitric
oxide (ppb)
Percent change
per doubling in
PFOA, or by
tertile
Results: Lowest tertile used as reference group
Confounding: Age, sex, race/ethnicity, BMI, annual family income,
smoking status
Fractional exhaled nitric oxide
2.64 (0.38, 4.96), p-value < 0.05
T2: 5.29 (1.88,8.81), p-
value < 0.01
T3: 6.34 (2.81, 10.01), p-
value < 0.001
p-trend <0.001
education level, serum cotinine, recent respiratory symptom, and
Zhou et al. Taiwan
Case-control
Children with
Serum Asthma
Asthma:
Asthma: Increased PFOA among
(2016,3981296) 2009-2010
(cases) or
Case boys: 1.3
Comparison of
asthmatics, p-value < 0.001
Low
without
(0.5-2.3)
PFOA
(controls)
Case girls: 0.8
distributions
asthma ages
(0.5-1.8)
(Wilcoxon rank-
10-15 from the
Control boys:
sum test)
GBCA
0.5 (0.4-1.4)
N = 456
Control girls:
Case boys: 158
0.5 (0.4-1.2)
Case girls: 73
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: 1.26
vs. lowest
trend = 0.001
Low
without
Case girls: 0.81
quartiles of
Girls: 3.68 (1.43, 9.48); p-value for
(controls)
Control boys:
PFOA exposure
trend = 0.005
asthma ages
0.52
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
10-15 from the
GBCA
N = 456
Case boys: 158
Case girls: 73
Control boys:
102
Control girls:
123
Control girls:
0.54
Confounding: Age, BMI, parental education, ETS, parental asthma, month of survey
Zhou et al. Taiwan
(2017,3858488) 2009-2010
Low
Case-control
Children with
Serum Asthma
OR per unit
Females with high testosterone:
(cases) or
Cases: 1.16
increase in
3.16 (1.47,6.78)
without
(0.48-2.16)
PFOA
Females with low testosterone: 2.8
(controls)
Controls: 0.52
(1.39, 5.97)
asthma ages
(0.44-1.27)
Males with high testosterone: 2.42
10-15 from the
(1.47, 3.99)
GBCA
Males with low testosterone: 2.82
N = 456
(1.60, 4.97)
Case boys: 158
Case girls: 73
Females with high estradiol: 2.56
Control boys:
(1.27, 5.12)
102
Females with low estradiol: 3.54
Control girls:
(1.61,7.79)
123
Males with high estradiol: 2.93
Sexes evenly
(1.64, 5.24)
divided into
Males with low estradiol: 1.85
high/low
(1.12, 3.06)
hormone
classifications
No statistically significant
interactions for low/high hormone
levels in either sex
Confounding: Age, sex, BMI, parental education, environmental tobacco smoke exposure, physical activity, month of survey
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
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
3.3 (2.5-1.0)
OR per
doubling of
maternal PFOA
Asthma (age 5): Total: 1.37 (0.81,
2.32)
No MMR vaccine before age 5:
10.37 (1.06, 101.93)
Yes MMR vaccine before age 5:
0.76 (0.41, 1.39)
Asthma (age 13):
Total: 1.12(0.67, 1.88)
No MMR vaccine before age 5:
9.92 (1.06, 93.22)
Yes MMR vaccine before age 5:
0.65 (0.35, 1.20)
Confounding: 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, daycare attendance at age 5, birth weight, and family
history of chronic bronchitis/asthma
Averina et al. Norway
(2019,5080647) 2010-2011
Low
Cohort
Adolescents in
their first year
of high school
fromTFFl and
TFF2
N = 675
Serum
Girls: GM = 2.1
(IQR = 1.2)
Boys: GM= 1.9
(IQR = 0.7)
Asthma, self-
reported doctor
diagnosed
OR by quartiles TFF1 Q4 vs. Ql: 2.07 (1.01, 4.23);
of PFOA p-value = 0.046
No other statistically significant
associations
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 Maternal plasma Recurrent
their infants 0.89 (Range: wheezing
N= 85 0.16-7.1) episodes
Difference in
prenatal PFOA
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 = Genetics and Biomarkers Study for
Childhood Autism; ETS = environmental tobacco smoke; GM = geometric mean; IQR = interquartile range; MMR = measles, mumps, rubella; NHANES = National Health and
Nutrition Examination Survey; OR = odds ratio; RR = risk ratio; SD = standard deviation; TFF1 = Tromse Fit Futures.
D-116
-------
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-10. Associations Between PFOA Exposure and Allergies in Recent Epidemiologic Studies
Reference,
Confidence
Location, years Study Design
Population,
Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Wang et al. Taiwan
(2011, 1424977) 2004
Medium
Cohort and
cross-sectional
Pregnant
women and
their children at
age 2
N = 244 (133
boys, 111 girls)
Cord blood
1.71 (0.75-
17.40)
Atopic
dermatitis, IgE
levels (log-
KU/L)
Atopic
dermatitis:
OR by quartiles
of PFOA
exposure
IgE:
Regression
coefficient per
ln-unit change
in PFOA
Atopic dermatitis
Q2: 0.84 (0.28, 2.48)
Q3: 1.03 (0.42,2.56)
Q4: 0.58 (0.22, 1.58)
IgE in cord blood at birth
All: 0.134 (SE = 0.115),
p-value = 0.047
Boys: 0.206 (SE = 0.164),
p-value = 0.025
Girls: 0.067 (SE = 0.231),
p-value = 0.823
IgE in serum at age 2
All: 0.027 (SE = 0.244),
p-value = 0.870
Boys: 0.097 (SE = 0.345),
p-value = 0.710
Girls: 0.001 (SE= 0.452),
p-value = 0.998
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
(2012, 1332477) 2002-2005
Medium
Cohort
Pregnant
women and
children from
the Hokkaido
Study on
Maternal serum
1.3 (0.8-1.7)
Food allergy,
eczema, otitis
media, and
wheezing
OR per loglO-
unit increase in
PFOA
Food allergy
1.67 (0.52, 5.37)
Eczema
0.96 (0.23, 4.02)
D-117
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location, years Study Design
Population,
Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Environment
and Children's
Health; follow
up at 18 months
N = 343
Otitis media
1.51 (0.45, 5.12)
Wheezing
1.27 (0.27, 6.05)
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
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 = 3.59
(2.46-5.36)
2007-2010:
GM = 3.27
(2.43-4.47)
Food allergy or
sensitization
OR by quartiles
ofPFOA
exposure
Food allergy, 2007-2010 cycle
Q2: 2.84 (0.83, 9.73)
Q3: 1.70 (0.51,5.65)
Q4: 9.09 (3.32, 24.9)
p-value for trend <0.001
Food sensitization, 2005-2006
cycle
Q2: 0.91 (0.47, 1.76)
Q3: 1.28 (0.59,2.76)
Q4: 1.23 (0.57,2.65)
p-value for trend = 0.74
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
2.01 (1.31-3.35)
Allergic
diseases, total
OR by quartiles
ofPFOA
exposure
Q2: 1.07 (0.79, 1.47)
Q3: 0.95 (0.70, 1.31)
Q4: 0.83 (0.59, 1.16)
p-value for trend = 0.208
No statistically significant
associations, trends, or interactions
by sex
Results: Lowest quartile used as reference.
Confounding: Maternal age, maternal educational level, sex, parental allergic history, number of older siblings, breast-feeding, day-care
attendance, ETS exposure
D-118
-------
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
3.3 (2.5-1.0)
Allergy, allergic
rhino-
conjunctivitis in
past 12 months,
positive skin
prick test, IgE
OR per
doubling of
PFOA
IgE: Percent
change per
doubling of
PFOA
Allergy at age 5
0.92 (0.53, 1.57)
Allergic rhino-conjunctivitis in past
12 months, at age 13
1.18(0.65,2.15)
Positive skin prick test, age 13
1.16(0.76, 1.77)
IgE, age 7: -5.15 (-31.92, 32.14)
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 1.6 (1.2-2.1)
and 10 years of
age,
N = 641
Rhinitis, rhino-
conjunctivitis,
SPT
OR per log2-
unit increase in
PFOA
Rhinitis, current, 10 y
1.30 (0.97, 1.74); p-value = 0.083
Rhinitis, ever, 10 y
1.29 (0.95, 1.74); p-value = 0.098
Rhino-conjunctivitis, ever, 10 y
1.32 (0.97, 1.79); p-value = 0.079
Rhinitis, ever, spes IgE > 0.35, 10 y
1.24 (0.90, 1.71); p-value = 0.185
SPT, any pos, 10 y
0.97 (0.75, 1.24); p-value = 0.788
SPT + and/pr slgE > 0.35, 10 y
1.03 (0.81, 1.30); p-value = 0.815
Confounding: Sex
D-119
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Exposure
Reference. Poihi Lit ion. Matrix, ,
„ .. . Location, years Study Design ' Outcome Comparison Results
Confidence J Ages, N Levels
(ng/mL)a
Pregnant Maternal blood Allergy, food or OR per IQR- Allergy, food, current
women and 2.54 (1.86-3.30) inhaled unit increase in All: 1.32 (0.92, 1.90);
their infants PFOA p-value = 0.136
(followed to age Boys: 1.49 (0.89, 2.50);
7), p-value = 0.131
N = 921 Girls: 1.15 (0.68, 1.94);
p-value = 0.602
Allergy, food, ever
All: 1.10 (0.77, 1.57);
p-value = 0.613
Boys: 1.04 (0.63, 1.73);
p-value = 0.867
Girls: 1.14 (0.68, 1.91);
p-value = 0.626
Allergy, inhaled, current
All: 0.96 (0.55, 1.67);
p-value = 0.887
Boys: 1.0(0.46,2.15);
p-value = 0.994
Girls: 0.88 (0.39, 2.01);
p-value = 0.765
Allergy, inhaled, ever
All: 1.25 (0.88, 1.78);
p-value = 0.213
Boys: 1.13 (0.71, 1.80);
p-value = 0.597
Girls: 1.44 (0.84, 2.47);
p-value = 0.189
Confounding: Maternal age, maternal BMI, maternal education, parity, smoking during pregnancy, nursey attendance
Ait Bamai et al. Hokkaido, Cohort Early pregnancy Maternal blood Rhino- RRperln-unit 0.95 (0.83, 1.09); p-value = 0.487
(2020,6833636) Japan, to 7 years, 1.94 (1.30-2.95) conjunctivitis increase in
Medium 2003-2012 N = 2,689 PFOA, from
Impinen et al. Oslo, Norway, Cohort
(2019, 5080609) Enrollment:
Medium 1999-2008
D-120
-------
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 Serum
Rhinitis, skin Change in RR Rhinitis
10 years:
N = 377
Age 16 years:
N = 375
All: 4.36 (IQR:
1.77)
Boys: 4.53
(IQR: 1.86)
Girls: 4.13
(IQR: 1.63)
prick test (SPT)
per IQR 10 years
increase in All: 0.84 (0.61, 1.15);
PFOA p-value = 0.28
Boys: 0.77 (0.53, 1.11);
p-value = 0.16
Girls: 0.84 (0.48, 1.49);
p-value = 0.56
16 years
All: 1.08 (1.01, 1.14);
p-value = 0.02
Boys: 1.06 (0.84, 1.32);
p-value = 0.63
Girls: 1.16 (0.90, 1.50);
p-value = 0.25
SPT
10 years
All: 1.11 (1.07, 1.15);
p-value < 0.0001
Boys: 1.02 (0.82, 1.27);
p-value = 0.84
Girls: 1.19 (0.79, 1.80);
p-value = 0.39
16 years
All: 1.07 (1.05, 1.08);
p-value < 0.0001
Boys: 1.05 (1.03, 1.06);
p-value < 0.0001
D-121
-------
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: 1.13 (0.86, 1.47);
p-value = 0.38
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; ETS = environmental tobacco smoke; 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 PFOA Exposure and Eczema in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
General Population
Goudarzi et al. Japan
(2016, 3859523) 2003-2013
Medium
Cohort
Children at age
4 from the
Hokkaido Study
N = 1,558 (765
girls, 793 boys)
Maternal blood Eczema
2.01 (1.31-3.35)
OR by quartiles
of PFOA
Q2: 1.10(0.76, 1.59)
Q3: 0.92 (0.623, 1.34)
Q4: 0.84 (0.56, 1.27)
p-value for trend = 0.287
Girls
Q2: 0.88 (0.50, 1.55)
Q3: 1.16 (0.67,2.03)
Q4: 1.21 (0.68,2.17)
p-value for trend = 0.356
Boys
Q2: 1.31 (0.80,2.18)
Q3: 0.74 (0.43, 1.27)
Q4: 0.59 (0.32, 1.08)
p-value for trend = 0.022
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
p-value for interaction by
sex = 0.039
Results: Lowest quartile used as reference.
Confounding: Maternal age, maternal educational level, sex, parental allergic history, number of older siblings, breast-feeding, day-care
attendance, ETS exposure0
Timmermann et
al. (2017,
3858497)
Medium
Denmark
1997-2000
Cohort
Serum Atopic eczema
Prenatal at birth: at age 13
3.3 (2.5-1.0)
Age 5/7: 4.0
(3.3-5.0)
OR per
doubling of
PFOA at age 13
Pregnant
women and
infants from the
CHEF study at
ages 5, 7, and
13 years
N = 559
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
Age 5: 0.72 (0.42, 1.25)
Age 13: 1.36 (0.85,2.19)
MMR vaccination before age 5
Yes: 4.48 (0.42, 47.69)
No: 0.82 (0.49, 1.36)
Chen et al.
(2018, 4238372)
Medium
Infants followed
up at 6, 12, and
24 months
N = 687
children (328
female and 359
male)
China Cohort Infants followed Cord blood
2012-2015 up at 6, 12, and All: 6.98
(Range = < 0.09
-29.97)
Female: 7
(Range = 0.70-
29.97)
Male: 6.89
(Range = < 0.09
-25.99)
Comparison: Logarithm base not specified.
Results: Lowest quartile used as reference group
Confounding: Maternal age, maternal pre-pregnancy BMI, gestational
parity, mode of delivery, family history of allergic disorders, infant sex.
Atopic
dermatitis
OR per log-unit
increase in
PFOA, or by
quartiles
All: 1.35 (0.93, 1.97)
Q2: 1.48 (0.87,2.52)
Q3:1.16 (0.67, 2)
Q4:1.74 (1.02, 2.95)
Female: 2.07 (1.13, 3.8)
Q2: 1.23 (0.52,2.93)
Q3:1.81 (0.79, 4.14)
Q4: 2.52 (1.12, 5.68)
Male: 0.98 (0.58, 1.64)
Q2: 1.57 (0.76, 3.23)
Q3: 0.81 (0.37, 1.78)
Q4: 1.34(0.64,2.82)
week at delivery, birth weight, maternal education, paternal education,
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
Cord blood
1.6 (Ql-
Q3 = 1.2-2.1)
Atopic
dermatitis
diagnosed
OR per log2-
unit increase
PFOA
Ages 0-2: 1.18 (0.94, 1.5)
Ages 0-10:0.99 (0.59, 1.67)
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
at 0, 2, and
10 years
N = 641
anytime
between 0-
2 years old, or
between 0-
10 years old
Confounding: Sex
Manzano-
Salgado et al.
(2019, 5412076)
Medium
Spain
2003-2015
Cohort
Pregnant
women and
children
followed up at
ages 1.5, 4, and
7 from the
INMA study
N = 1,188 at 1.5
and 4 years,
N = 1,071 at
7 years
Maternal plasma Eczema
2.35 (1.63-3.30)
OR or RR per
log2-unit
increase in
PFOA
Age 1.5: 1.1 (0.91, 1.31)
Age 7: 0.96 (0.81, 1,14)
Follow up at age 4: 0.97 (0.81,
1.17)
Boys at ages 1.5, 4, and 7: 0.98
(0.81, 1.18)
Girls at ages 1.5, 4, and 7: 0.9
(0.75, 1.07)
From ages 1.5 to 7 years: 0.96
(0.85, 1.08)
No statistically significant
associations
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 Cord blood Atopic
2 years 0.65 (0.23-1.96) dermatitis
N = 839
Results: Lowest tertile used as reference.
Confounding: Sex, family income, maternal atopy, breast feeding, and maternal age at childbirth
OR by tertiles of T2: 0.75 (0.26, 1.89)
PFOA T3: 2.58 (1.27, 5.32);
p-value < 0.01
Wen et al.
(2019, 5387152)
Medium
Taiwan Cohort
2001-2005
Infants followed Cord blood Atopic
from birth up to 0.65 (0.23-1.96) dermatitis
5 years of age
N = 863
Hazard ratio for 1.89 (1.1, 3.16); p-value < 0.05
PFOA
>1.96 ng/mL
vs.
<1.96 ng/mL
Confounding: Sex, parental education, parental atopy, breast feeding, and maternal age at childbirth
Notes: CHEF = Children's Health and Environment in the Faroe Islands; ECA = Environment and Childhood Asthma; ETS = environmental tobacco smoke; INMA = Spanish
Environment and Childhood (Infancia y Medio Ambiente); Q2 = Quartile 2, Q3 = Quartile 3, Q4 = Quartile 4.
D-124
-------
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-12. Associations Between PFOA Exposure and Autoimmune Health Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Population,
Study Design Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Steenland et al. West Virginia Cohort
(2013, 1937218) 1952-2011
Medium
Males and
females from
C8 Health
Project,
Ages > 20,
N= 32,254
Serum
26 (13-68)
Occurrence of
conditions with
and without a
10-year lag:
rheumatoid
arthritis (RA),
lupus, multiple
sclerosis (MS),
ulcerative colitis
(UC), Crohn's
disease (CD)
RR by quartiles
of PFOA
RA, no lag
Q2: 1.24 (0.85, 1.79)
Q3: 1.40 (0.96,2.03)
Q4: 0.99 (0.68, 1.43)
p-trend = 0.84
RA, with lag
Q2: 1.53 (0.61,2.58)
Q3: 1.73 (1.10,2.71)
Q4: 1.35 (0.87,2.11)
p-trend = 0.73
Lupus, no lag
Q2: 1.49 90.68, 3.34)
Q3: 1.01 (0.44,2.30)
Q4: 0.71 (0.31, 1.65)
p-trend = 0.94
Lupus, with lag
Q2: 0.79 (0.27, 2.34)
Q3: 1.26 (0.40,4.03)
Q4: 0.61 (0.19, 1.91)
p-trend = 0.93
MS, no lag
Q2: 0.85 (0.44, 1.63)
Q3: 1.56 (0.81,3.00)
Q4: 1.26 (0.65,2.42)
p-trend = 0.22
D-125
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Location,
Population,
Exposure
Reference,
Confidence
Matrix,
Resultsb
Years
Study Design Ages,
N
Levels
(ng/mL)a
Outcome
Comparison
MS, with lag
Q2: 1.16(0.54,2.47)
Q3: 1.62 (0.74,3.52)
Q4: 1.32(0.61,2.84)
p-trend = 0.59
UC, no lag
Q2: 1.76(1.04,2.00)
Q3 2.63 (1.56,4.43)
Q4: 2.86 (1.65, 4.96)
p-trend < 0.0001
UC, with lag
Q2: 1.71 (0.89, 3.27)
Q3: 2.05 (1.07, 3.91)
Q4: 3.05 (1.56, 5.96)
p-trend < 0.0001
CD, no lag
Q2: 1.25 (0.61,2.58)
Q3: 1.15 (0.55 (2.41)
Q4: 1.00 (0.48,2.09
p-trend = 0.73
CD, with lag
Q2: 0.80 (0.32, 1.99)
Q3: 0.97 (0.36,2.60)
Q4: 0.69 (0.26, 1.82)
p-trend = 0.79
Results: Lowest quartile used as reference.
Confounding: Sex, race/ethnicity, smoking, BMI, alcohol consumption0
Gaylord et al. United States Case-control
Children and
Serum
Celiac disease OR per ln-unit 3.85 (0.71, 21.1)
(2020,6833754)
adolescents
Cases: 1.26
change inPFOA Girls: 20.6 (1.13, 375);
Medium
younger than
(IQR = 0.76)
p-value < 0.05
21 years with
Boys: 1.05 (0.11, 9.59)
D-126
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
(cases) and
without
(controls) celiac
disease
N = 88 (42 girls,
46 boys)
Controls: 0.99
(IQR = 0.51)
Confounding: Genetic susceptibility score, albumin, BMI, age, race (non-Hispanic white vs. other race/ethnicity) and sex
Steenland et al. United States
(2018, 5079806) 1999-2012
Low
Case-control
Patients with
UC, CD, or
healthy controls
N = 114 UC, 60
CD, 75 neither
Serum
UC: 2.93
CD: 1.78
Controls: 1.33
UC
OR of UC vs.
CD and/or
neither per ln-
unit increase in
PFOA, or by
quintiles
UCvs. CD: 1.68 (1.07,2.23)
UC vs. neither: 2.00 (1.08, 3.67)
UCvs. CD & neither: 1.60 (1.14,
2.24)
Q2: 0.81 (0.22,2.93)
Q3: 40.98 (11.67, 150.34)
Q4: 33.36 (11.32, 119.36)
Q5: 2.86 (0.94, 8.75)
Results: Lowest quintile used as reference.
Confounding: Age, sex, ethnic group (white or non-white)
Sinisalu et al. Finland
Cohort
Pregnant
Cord blood Celiac disease
Comparison of
No significant differences in
(2020,7211554) 1999-2005
women and
Case: 2.32
mean PFOA
exposure between cases and control
Low
infants at birth
(min-max:
exposure levels
at birth or 3 months
and 3 months
1.31-4.80)
from the Type 1
Control: 2.43
Diabetes
(min-max:
Prediction and
1.23-4.46)
Prevention
Study in Finland
3-month serum
(DIPP)
Case: 4.34
N = 33 (17
(min-max:
celiac disease,
1.23-9.17)
16 controls)
Control: 4.05
(min-max:
0.98-6.25)
D-127
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Location,
Population,
Exposure
Reference,
Confidence
Matrix,
Resultsb
Years
Study Design Ages,
N
Levels
(ng/mL)a
Outcome
Comparison
Xu et al. (2020,
Sweden Cohort Residents of
Serum
CD, UC
HR for exposure CD: 1.58 (1.00-2.49) for early
6315709)
2014-2016 Ronneby
Ronneby panel
period r.v. not
(1985-94) exposure period
Low
municipality
study: 20 (11-
exposed (1980-
No associations for the later years
29)
1984)
UC: No associations any exposure
Ronneby panel
Ronneby
periods
study: N = 57
resampling: 16
Ronneby
(9-23)
resampling:
Karlshamn: 2
N = 113
(1-2)
Karlshamn:
N = 19
Confounding: Age, gender, calendar year
Ammitzboll et
Denmark Case-control Adults with
Serum
Relapsing
Percent change
-12 (-24, 2); p-value = 0.099
al. (2019,
2019 (cases) or
Cases: 1.88
remitting
in PFOA
Females: 7 (-13, 32);
5080379)
without
(1.34-2.32)
multiple
comparing MS
p-value = 0.526
Low
(controls)
Controls: 1.94
sclerosis
cases r.v, healthy Males: -28 (-42, -9);
RRMS or CIS
(1.38-3.01)
(RRMS)
controls
p-value = 0.006
N = 162 (92
women, 70
men)
Confounding: Age, sex, breastfeeding
Notes: BMI = body mass index; CD = Crohn's disease; CIS = clinically isolated serum syndrome; HR = hazard ratio;
arthritis; RR = risk ratio; RRMS = relapsing remitting multiple sclerosis; UC = ulcerative colitis.
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.
MS = multiple sclerosis; OR = odds ratio; RA = rheumatoid
D-128
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.5 Cardiovascular
D.S.I Cardiovascular Endpoints
Table D-13. Associations Between PFOA Exposure and Cardiovascular Effects in Recent Epidemiological Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
Children and Adolescents
Li etal. (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 Pregnant women
2003-2006 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 SBP (z-score), Regression
Gestation: 5.3
(3.7-7.2)
score)
mean of SBP coefficient per
and DBP (z- log2-unit IQR
Cord serum
At birth: 3.2
(2.4-1.7)
increase in
PFOA
Confounding0: visit, visit*PFAS, maternal age
and parity; and child age, sex, race, and pubertal
duration.
Serum
At age 3: 5.4
(3.7-7.4)
At age 8: 2.4
(1.8-3.2)
At age 12: 1.3
(1.0-1.6)
maternal education, maternal pre-pregnancy BMI,
stage. Additional confounding for analyses at age
SBP (z-score)
Gestation: 0.1 (-0.1, 0.2)
At birth: 0.1 (-0.1, 0.3)
Age 3:0 (-0.2, 0.3)
Age 8: 0 (-0.4, 0.5)
Age 12:0.2 (-0.1,0.6)
Mean of SBP and DBP (z-score)
Gestation: 0 (-0.1, 0.2)
At birth: 0.1 (-0.1, 0.2)
Age 3: 0.1 (-0.1, 0.3)
Age 8: 0.1 (-0.2, 0.4)
Age 12: 0.2 (0.0, 0.5)
gestational serum cotinine concentrations,
3, age 8, and age 12: Breastfeeding
Ma et al. (2019, United States Cross- Adolescents Serum
5413104) 2003-2012 sectional aged 12-20 Levels not
Medium from NHANES provided
N = 2,251
(1,048 female,
1,203 male)
DBP, SBP
Regression
coefficient per
loglO-unit
increase in
PFOA
DBP
Total cohort: 0.008 (-0.009, 0.026)
Females: -0.005 (-0.027, 0.016)
Males: 0.018 (-0.01, 0.046)
SBP
Total cohort: -0.003 (-0.01, 0.004)
Females: -0.005 (-0.015, 0.004)
Males: -0.004 (-0.014, 0.007)
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
Confounding: Age, gender, race, BMI, cotinine, dietary calcium, caloric intake, sodium consumption, potassium consumption, sampling year
Warembourg et
al. (2019,
5881345)
Medium
Cohort Pregnant women
and their
children at ages
6 and 11 from
the HELIX
Project
N = 1,277
Prenatal
exposure
Postnatal
exposure
Confounding: Cohort of inclusion, maternal age,
child age, child sex, child height
France, Spain,
Lithuania,
Norway,
Greece,
United
Kingdom
1999-2015
Maternal blood DBP, SBP
2.3 (1.4-3.3)
Plasma
1.5 (1.2-2.0)
Regression
coefficient per
log2-unit IQR
increase PFOA
DBP
Maternal PFOA: 0.29 (-0.55, 1.13)
Childhood PFOA: 0.23 (-0.45, 0.91)
SBP
Maternal PFOA: -0.1 (-1, 0.8)
Childhood PFOA: 0.39 (-0.34, 1.12)
maternal education level, maternal pre-pregnancy BMI, parity, parental country of birth,
Canova et al.
Italy
Cross-
Adolescents
Serum DBP, SBP
Regression
DBP
(2021,
2017-2019
sectional
aged 14 to
Adolescents:
coefficient per
Adolescents
10176518)
19 years and
38.9 (20.1-68.8)
ln-unit increase
Per ln-unit increase: -0.11 (-0.37, 0.15)
Medium
children aged 8
in PFOA, or by
Q2: -0.23 (-0.84, 0.39)
to 11 years from Children: 20.9
quartiles
Q3: -0.28 (-0.93,0.36)
health
(12.9-33.5)
Q4: -0.08 (-0.77, 0.61)
surveillance
Children
program in
Per ln-unit increase: 0.16 (-0.23, 0.54)
Veneto Region
Q2: 0.58 (-0.28, 1.44)
Adolescents:
Q3: 0.37 (-0.50, 1.24)
N = 6,669
Q4: 0.68 (-0.21, 1.57)
Children:
N = 2,693
SBP
Adolescents
Per ln-unit increase: -0.16 (-0.53, 0.20)
Q2: -0.44 (-1.31,0.43)
Q3:-1.01 (-1.92,-0.10)
Q4: -0.44 (-1.42, 0.54)
Children
Per ln-unit increase: -0.51 (-1.02,
-0.01)
D-130
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
Q2
Q3
Q4
-0.08 (-1.20, 1.05)
-0.22 (-1.35,0.91)
-0.98 (-2.14,0.18)
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
Cohort
al. (2021,
9960593)
Medium
Kingdom,
France, Spain,
Lithuania,
Norway,
Greece
Recruitment
1999-2010,
Follow-up:
2013-2015
Mother-child
pairs from the
HELIX Project,
children
followed up
around age 8
(range 6-12)
N = 1,101
Maternal
plasma
(prenatal)
2.22 (1.34-3.29)
Plasma
(childhood)
1.53 (1.17-1.96)
DBP (z-score),
SBP (z-score)
Regression
coefficient
per doubling
in PFOA, or
by quartiles
DBP
Maternal PFOA: -0.01 (0.10, 0.09)
Q2: 0.04 (-0.14,0.21)
Q3: 0.00 (-0.22,0.21)
Q4: 0.08 (-0.17,0.33)
p-trend = 0.614
Childhood PFOA: 0.01 (-0.11, 0.13)
Q2: -0.01 (-0.16,0.14)
Q3: 0.00 (-0.16,0.16)
Q4: 0.09 (-0.09, 0.27)
p-trend = 0.390
SBP
Maternal PFOA: 0.03 (-0.08, 0.14)
Q2: 0.08 (-0.11,0.28)
Q3: 0.04 (-0.19,0.28)
Q4: 0.06 (-0.22, 0.33)
p-trend = 0.910
Childhood PFOA: 0.03 (-0.11, 0.16)
Q2
Q3
Q4
-0.05 (-0.22,0.11)
-0.05 (-0.23,0.13)
0.10 (-0.10, 0.30)
p-trend = 0.388
Comparison: Maternal quartiles are defined as follows (in |ig/L PFOA): Ql: 0.02-1.33; Q2: 1.34-2.22; Q3: 2.22-3.29; Q4: 3.29-31.64;
childhood quartiles are defined as follows (in |ig/L PFOA): Ql: 0.21-1.17; Q2: 1.17-1.53; Q3: 1.53-1.96; Q4: 1.96-6.66.
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, PFOS
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
Manzano-
Salgado et al.
(2017, 4238509)
Medium
Spain
2003-2008
Cohort
Pregnant women Maternal blood Blood Pressure Regression BP
and their
GM = 2.32
(BP) (z-
score)
Cardiometaboli
c Risk Score
(CMR)
coefficient per All age 4: -0.06 (-0.16, 0.04)
log2-unit
increase in
PFOA
children at ages (1.63-3.31)
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
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
Girls:-0.04 (-0.18,0.1)
Boys: -0.08 (-0.23, 0.07)
All age 7: -0.02 (-0.11,0.07)
Girls: -0.08 (-0.21, 0.04)
Boys: 0.04 (-0.08,0.16)
CMR
All age 4: 0.27 (-0.35, 0.89)
Girls:-0.22 (-1.1,0.66)
Boys: 0.72 (-0.17, 1.62)
Linetal. (2013,
2850967)
Medium for
CIMT
Low for Systolic
BP
Taiwan
2006-2008
Cross-
sectional
Adolescents and
young adults
ages 12-30
N = 637
Serum
3.49 (75th
percentile = 6.5
4)
SBP, CIMT Mean by PFOA SBP: No associations
exposure group
CIMT: No associations
Comparison: Groups were defined as follows: (1) up to 50th percentile; (2) 50th-75th percentile; (3) 75th-90th percentile; (4) above 90th
percentile
Confounding: Age, gender, smoking status, alcohol drinking, body mass index; for CIMT, also includes SBP, low density lipoprotein
cholesterol, triglyceride, high sensitivity CRP, homeostasis model assessment of insulin resistance
Geiger et al.
(2014, 2851286)
Medium
United States Cross-
Children ages Serum
Hypertension OR per ln-unit Hypertension
1999-2000,
2003-2008
sectional
<18 years from
NHANES
N = 1,655
Mean
(SE) = 4.4 (0.1)
increase in Per ln-unit increase: 0.76 (0.53, 1.10)
PFOA, orby Q2: 0.89 (0.53, 1.49)
quartile Q3: 0.96 (0.53,1.73)
Q4: 0.69 (0.41, 1.17)
p-trend = 0.2477
Results: Lowest quartile used as the reference group.
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
Confounding: Age, sex, race-ethnicity, BMI categories, annual household income categories, moderate activity, total cholesterol, and serum
cotinine
Averina et al.
(2021,7410155)
Medium
Linetal. (2016,
3981457)
Medium
Norway
2010-2011
Cross-
sectional
First level high
school students
ages 15-
19 years from
TFF1
N = 940
Hypertension
OR by quartiles Hypertension
Q2: 1.28 (0.74, 2.22), p-value = 0.37
Q3: 1.45 (0.85, 2.49), p-value = 0.175
Q4: 2.08 (1.17, 3.69), p-value = 0.013
Serum
Girls: GM
(IQR) = 2.14
(1.26)
Boys: GM
(IQR) = 1.86
(0.67)
Outcome: Hypertension defined as systolic blood pressure > 130 mmHg and/or diastolic blood pressure > 80 mmHg.
Comparison: Quartiles are defined as follows (inng/mL PFOA): Ql: 0.28-1.56; Q2: 1.57-1.92; Q3: 1.93-2.44; Q4: 2.45-13.97.
Results: Lowest quartile used as the reference group.
Confounding: Sex, age, BMI and physical activity outside school
Taiwan
1992-2000
Cross-
sectional
Adolescents and
young adults
ages 12-30
N = 848
Serum
GM = 3.21
(95% CI: 3.00-
3.46)
8-OHDG (log-
l-ig/g
creatinine)
CIMT
CD31+ /
CD42a-
(log coiint/|iL)
CD31+ /
CD42a+
(log coiint/|iL)
CD62E
(log coiint/|iL)
CD62P
(log coiint/|iL)
Mean by PFOA 8-OHDG: Borderline statistically
exposure significant increase across exposure
level group groups, 7.55-7.68 (Group 3);
p-trend = 0.059
CIMT: No associations across exposure
groups; p-trend = 0.2868
CD31+ / CD42a-: Statistically significant
decrease across exposure groups, 5.14—
4.77; p-trend = 0.036
CD31+ / CD42a+, CD62E, CD62P: No
statistically significant associations
across exposure groups
Comparison: Groups were defined as follows: (1) up to 50th percentile; (2) 50th-75th percentile; (3) 75th-90th percentile; (4) above 90th
percentile.
Confounding: Age, gender, smoking status, BMI, systolic blood pressure, low density lipoprotein, triglyceride, homeostasis model
assessment of insulin resistance, and high sensitivity CRP
Khalil et al.
(2018, 4238547)
United States Cross-
2016 sectional
Obese children Serum
ages 8-12
DBP, SBP Regression DBP: 7.75 (-0.25, 15.7)
coefficient per SBP: 7.99 (-2.29, 18.3)
D-133
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Design
Years
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Low
00
-t
II
z;
0.99
(IQR = 0.45)
unit increase in
PFOA
Confounding: Age, race, sex
Koshy et al.
(2017, 4238478)
Low
United States Cross-
2011-2012 sectional
Children and Serum
adolescents
from the World
Trade Center
Health Registry
(WTCHR)
N = 308
1.81
(IQR = 0.90)
Comparison:
1.39
(IQR = 0.75)
Augmentation Regression
Index (AI) coefficient per
Brachial Artery ln-unit increase
Distensibility in PFOA
(BAD)
Pulse Wave
Velocity
(PWV)
AI: -1.41 (-4.59, 1.78)
BAD: 0.45 (0.04, 0.87)
PWV: 0.05 (-0.17,0.28)
Confounding: BMI category, caloric intake, cotinine concentration, physical activity, race, sex
Pregnant Women
Matilla-
Santander et al.
(2017, 4238432)
Medium
CRP
2.86 (-8.12, 14.3)
By quartile:
Spain Cohort Pregnant women Plasma CRP Percent median
2003-2008 from INMA 2.35 (1.63-3.30) (loglO mg/dL change by
study ) quartiles and
N= 1,240 per loglO-unit
increase in
PFOA
Results: Lowest quartile 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
Q2
Q3
Q4
-12.19 (-27.3,6.18)
-3.92 (-22.1, 17.3)
3.05 (-17.3,28.4)
Liao et al.
(2020, 6356903)
High
United States Cross- Adults ages 20+ Serum DBP, SBP,
2003-2012 sectional fromNHANES 3.33(2.13-5.10) hypertension
N = 6,967
(3,439 females,
3,528 males)
DBP and SBP: DBP:-0.34 (-1.43, 7.55)
Regression SBP: 1.83 (0.40, 3.25)
coefficient per
loglO-unit Hypertension
increase in T2: 1.03 (0.89, 1.2)
PFOA T3: 1.32 (1.13, 1.54), p-value < 0.01,
p-trend < 0.001
Hypertension: No significant interactions by age
OR by tertiles Females
or regression T2: 0.96 (0.77, 1.19)
coefficient T3: 1.42 (1.12, 1.79), p-value < 0.001,
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
around
p-trend = 0.003
inflection point
Males: No statistically significant
(1.80 ng/mL)
associations, or trends
Ages > 60 years
T2: 0.84 (0.66, 1.06)
T3: 1.32 (1.03, 1.68)
p-trend = 0.003
Ages < 60 years: No statistically
significant associations or trends
Levels < 1.80 ng/mL: 0.56 (0.32, 0.99)
Levels > 1.80 ng/mL: 1.32 (1.03, 1.68)
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 PFOA): T1 < 2.5; 2.5< T2 < 4.4; 4.4< 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-control
Rural men Serum
CHD
OR by quartiles CHD
(2015, 3859607)
1990-1991,
N = 462 Cases: 4.2
Q2: 0.79 (0.44, 1.43)
High
2002-2003
(IQR = 1.8)
Controls: 4.0
(IQR = 2.0)
Q3: 1.18(0.67,2.06)
Q4: 0.88 (0.5, 1.55)
Results: Lowest quartile used as reference group.
Confounding: BMI, systolic blood pressure, total cholesterol, HDL, tobacco use
Mobacke et al.
Sweden Cross-
Adults aged 70 Serum
Left
Regression
LVEDD: 0.58 (-0.03, 1.18)
(2018, 4354163)
Years not sectional
from the Mean
Ventricular
coefficient per
LVMI: -0.65 (-1.94,0.65)
High
reported
Prospective (SD) = 3.59
End-Diastolic ln-unit increase
RWT: -0.12 (-0.22, -0.001)
Investigation of (1.69)
Diameter
in PFOA
the Vasculature
(LVEDD)
in Uppsala
(mm)
Seniors
Left
(PIVUS) study
Ventricular
N = 801
Mass Index
D-135
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
(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,
3860099)
Medium
China
2015-2016
Cross-
sectional
Adults aged 22-
96
N = 1,612(408
females, 1,204
males)
Serum
6.19(4.08-9.31)
DBP, SBP,
hypertension
Regression
coefficient
per ln-unit
increase in
PFOA
Hypertension:
OR per ln-
unit increase
PFOA
DBP
Total: 2.18 (1.35,2.98)
SBP
Total: 1.69 (0.25,3.13)
Females: 2.91 (0.1, 5.72)
Males: No association
Hypertension: No statistically significant
associations
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, United States Controlled Overweight and Plasma DBP, SBP Partial
4238396) 2004-2007 trial obese adults Females: 4.1 Spearman
Medium ages 30-70 in (2.8-5.6) correlation
Males: 5.2 (3.9- coefficient
6.8)
DBP: 0.1; p-value < 0.05
SBP: 0.04
ages 30-70 in
the POUNDS
Lost Study
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
Lin et al. (2020, United States Cohort
Adults from the
Serum
DBP, SBP,
Regression
DBP: No statistically significant
6311641) 1996-2014
Diabetes
Baseline: 4.9
pulse
coefficient
associations by timepoint, by quartiles,
Medium
Prevention
(3.5-6.7)
pressure
per log2-unit
or by sex (p-value for interaction by
Program (DPP)
increase in
sex = 0.81)
D-136
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
T Population,
Location, '
Design Ages,
Years ^
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
and Outcomes
Year 2: 5.7
(mmHg), and
PFOA or by
Study (DPPOS)
(4.0-8.0)
hypertension
quartiles
SBP:
N = 957 at
Year 14: 2.8
Baseline: 1.49 (0.29, 2.70)
baseline, 956
(2.0-3.8)
Hypertension:
Baseline males: 2.36 (0.13, 4.60);
at year 2, and
HR or RR per
p-value for interaction by sex = 0.28
346 at year 14
log2-unit
No statistically significant associations
increase
by follow-up timepoint or by quartiles
PFOA or by
quartiles
Pulse Pressure: No statistically
significant associations by timepoint, by
quartiles, or by sex (p-value for
interaction by sex = 0.24)
Hypertension
Baseline males: 1.27 (1.06, 1.53);
p-value for interaction by sex = 0.07165
No statistically significant associations
by timepoint or by quartiles
Outcome: Hypertension defined as SBP > 140 mmHg and DBP > 90 mmHg in those without diabetes, SBP > 30 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
Mi et al. (2020, China Cross-
Shenyang Serum
DBP, SBP,
DBP, SBP:
DBP
6833736) 2015-2016 sectional
residents ages 4.8 (3.6-7.4)
hypertension
egression
1.49 (0.34, 2.64)
Medium
23-94
coefficient per
Females: 0.38 (-0.75, 1.51)
N = 1238 (559
ln-unit
Males: 1.82 (-0.04, 3.67)
women, 679
increases in
p-interaction = 0.05
men)
PFOA
Ages >60: 1.96 (0.62,3.31)
Ages 23-60: No associations
Hypertension:
No statistically significant sex
OR per ln-unit
interactions within age groups
increase in
PFOA
SBP: No statistically significant
associations or interactions by sex or age
D-137
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Hypertension
1.72(1.27, 2.31)
Females: 2.32 (1.38, 3.91)
p-interaction = 0.22
Ages >60: 3.58 (2.14,5.98)
Ages 23-60: No associations
No statistically significant sex
interactions within age groups
Outcome: Hypertension defined as mean SBP > 140 mmHg or DBP > 90 mmHg, or use of antihypertensive medicines for previous two
weeks.
Confounding: Age, sex, ethnicity, career, education, smoking, alcohol drinking, physical activity, annual household income, and seafood
consumption
Mitro et al.
(2020, 6833625)
Medium
DBP, SBP, CRP: No statistically
significant associations
United States Cohort Pregnant women Plasma DBP, SBP, Percent
1999-2005 and their 5.6(4.0-7.6) CRP (mg/L) difference per
children at age 3 log2-unit
from Project increase in
Viva PFOA
N = 761 mothers Regression
(496 ages < 35, coefficient per
265 ages >35) log2-unit
increase in
PFOA
Population: For measurements of CRP, 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
Pitter et al. Italy Cross-
Adults aged 20-
Serum DBP, SBP,
DBP, SBP:
DBP
(2020,6988479) 2017-2019 sectional
39 years from
35.8 (13.7-78.9) hypertension
Regression
0.34 (0.21, 0.47)
Medium
Veneto Region
Male: 58.3 risk
coefficient per
Q2: 0.24 (-0.16,0.64)
with PFAS
(25.1-114.7)
ln-unit increase
Q3: 0.78 (0.36, 1.20)
contaminated
Female: 22.6
in PFOA, or by
Q4: 0.97 (0.53, 1.42)
drinking water
(8.8-19.4)
quartiles
Males: 0.23 (0.04, 0.42)
DBP and SBP:
Females: 0.39 (0.21,0.57)
N = 15,380
Hypertension
risk: OR per ln-
SBP
D-138
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
(7,428 males,
7,952 females)
Hypertension
risk: N = 15,786
(7,667 males,
8,119 females)
unit increase in 0.37 (0.19, 0.54)
PFOA, or by
quartiles
Q2: 0.26 (-0.29,0.81)
Q3: 0.74 (0.16, 1.31)
Q4: 1.07 (0.46, 1.68)
Males: 0.46 (0.19,0.73)
Females: 0.31 (0.08,0.55)
Hypertension risk
1.06(1.01, 1.12)
Q2: 1.00 (0.85, 1.16)
Q3: 1.02 (0.87, 1.20)
Q4: 1.16 (0.99, 1.37)
Males: 1.08 (1.02, 1.15)
Females: 1.06 (0.97, 1.15)
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
Minetal. (2012,
2919181)
Medium
United States
2003-2006
Cross-
sectional
Hypertension OR by quartile
Hypertension
Q4: 1.84 (1.07,3.18)
Adults ages 20+ Serum
from NHANES GM = 4.0
N= 1,415 (3.86-4.13)
Outcome: Hypertension defined as SBP >140 mmHg or DBP > 90 mmHg or as a self-reported medical diagnosis of hypertension.
Results: Lowest quartile used as the reference group.
Confounding: Age, sex, race/ethnicity, education, income, smoking habits, alcohol use, obesity status, total saturated fatty acid intake,
physical activity, serum folate, total cholesterol, and poor kidney function
Winquist and United States Cohort Workers at a
Steenland (2014, 2008-2011 Mid-Ohio
2851142) Valley chemical
Medium plant and
residents of the
surrounding
community from
Serum Hypertension HR by quintiles Hypertension
26.1 (12.8-68.1) Q2: 1.10 (1.02, 1.19)
Q3: 1.10(1.02, 1.18)
Q4: 1.05 (0.97, 1.12)
Q5: 0.98 (0.91, 1.06)
D-139
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
C8 Health
Project
N = 32,254
Liu et al. (2018,
4238514)
Medium
CMstensen et
al. (2019,
5080398)
Medium
Outcome: Hypertension cases were identified based on self-reported diagnosis.
Results: Lowest quintile used as the reference group.
Confounding: Age, sex, years of schooling, race, smoking, smoking duration, smoking pack-years, regular alcohol consumption, BMI, self-
reported type-2 diabetes
United States Cross- Adults ages 18+ Serum Hypertension ORperln-unit Hypertension: 1.13 (0.81, 1.58)
2013-2014 sectional fromNHANES GM(SE)=1.86 increase in
N= 1,871 (1.02) PFOA
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
United States Cross- Adults ages 20+ Serum
2007-2014 sectional fromNHANES 2.8(1.8-4.3)
N = 2,975
Hypertension OR by quartiles Hypertension
No statistically significant associations
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, PFOS, PFUnDA, race/ethnicity, smoking status, survey cycle
Donat-Vargas et Sweden Cohort Adults aged 30- Plasma Hypertension ORbytertiles
al. (2019, 1990-2013 60 at baseline Baseline: 2.9 orperSD-unit
5080588) N = 187 (2.2-L2) increase in
Medium Follow-up: 2.7 PFOA
(1.9-3.6)
Hypertension
Baseline:
OR per increase: 1.12(0.78, 1.59)
Follow-up:
OR for T3: 1.14(0.51,2.58)
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: Results by tertile use lowest tertile as the reference group.
Confounding: Gender, age, education, sample year, body mass index, smoking habit, alcohol consumption, physical activity, healthy diet
score
D-140
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Jeddi et al.
(2021, 7404065)
Medium
Elevated blood pressure: 1.05(1.01,
1.08), p-value < 0.05
Italy Cross- Residents aged Serum Elevated blood OR per ln-unit
2017-2019 sectional 20-39 from the GM (range): pressure increase in
PFAS- 67.66 (0.70- PFOA
contaminated 1400.0)
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
Shankar et al.
(2012, 2919176)
Medium
United States
1990-2000,
2003-2004
Cross-
sectional
Adults ages 40+
Serum
CVD,
OR by quartiles CVD
from NHANES
Female: 3.9
cardiovascula
Q3: 1.77 (1.04,3.02)
N = 1,216(623
(2.9, 5.6)
r heart
Q4: 2.01 (1.12,3.60)
females, 593
Male: 4.3 (3.0,
disease
Increasing trend by quartiles;
males)
6.1)
(CVHD),
p-trend = 0.01
peripheral
arterial
CVHD
disease
Q4: 2.24 (1.02, 4.94)
(PAD),
Increasing trend by quartiles;
stroke, CVD
p-trend = 0.007
or PAD
Cardiovascular
PAD
Disease
Q4: 1.78 (1.03,3.08)
(CVD)
Increasing trend by quartiles;
p-trend = 0.04
Stroke
Q2: 4.39 (1.44, 13.37)
Q3: 3.94 (1.48, 10.05
Q4: 4.26 (1.84, 9.89)
p-trend = 0.02
CVD or PAD:
Q3: 1.72 (1.13,2.64)
Q4: 2.28 (1.40, 3.71)
D-141
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
Increasing trend by quartiles;
p trend < 0.001
Females:
Q4: 2.99 (1.53, 5.81)
Increasing trend by quartiles;
p-trend = 0.004
Males:
Q3: 1.75 (1.04,2.96)
Q4: 1.83 (1.02, 3.28)
Increasing trend by quartiles;
p-trend = 0.04
Results: Lowest quartile used as reference.
Confounding: Age, sex, race/ethnicity, educational level, smoking status, alcohol intake, body mass index, hypertension, diabetes mellitus,
serum total cholesterol level; serum high-sensitivity CRP level and serum uric acid level for CVD and PAD outcomes only
Fry and Power
(2017, 4181820)
Medium
Lind et al.
(2017, 3858504)
Medium
United States
2003-2006
Cohort
Adults ages 60+
from NHANES
N = 1,023
Serum
23.7 ng/g
(SE = 0.7 ng/g)
Mortality by HR per SD-unit Mortality
cerebrovascul increase in 0.98(0.81,1.17)
ar or heart PFOA
diseases
Confounding: Age, education, gender, race/ethnicity, smoking status
Sweden
2001-2004
Cross-
sectional
Adults ages 70+
in Uppsala,
Sweden
N = 1,016(509
females and 507
males)
Plasma
3.3 (2.52-4.39)
CIMT, carotid
artery intima-
media
complex grey
scale median
(CIM-GSM),
carotid artery
atheroscleroti
c plaque
CIMT, CIM-
GSM:
Regression
coefficient
per ln-unit
increase in
PFOA
CIMT, CIM-GSM, atherosclerotic
plaque: no statistically significant
associations; all p-values > 0.25
Plaque: OR per
ln-unit increase
in PFOA
Confounding: Sex, HDL- and LDL cholesterol and serum triglycerides, BMI, blood pressure, smoking exercise habits, energy and alcohol
intake, diabetes, educational level
D-142
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Huang et al.
(2018, 5024212)
Medium
United States
1999-2014
Cross-
sectional
Adults from
NHANES ages
18+
N = 10,859
Serum
3.17 (1.97-4.90)
CVD, angina
pectoris,
congestive
heart disease,
CHD, heart
attack, stroke,
CRP (mg/L)
OR by quartiles CVD: No association by quartiles, no
significant trend; p-trend = 0.703
CRP: No associations, trend, or interaction by
Spearman age groups
correlation Females
coefficient Q2: 0.76 (0.49, 1.18)
Q3: 1.04 (0.66, 1.66)
Q4: 1.14 (0.75, 1.75)
Males
Q2: 1.49 (0.98,2.26)
Q3: 1.56 (1.02,2.40)
Q4: 1.45 (0.92,2.28)
No trend or interaction by sex
Angina pectoris: No association by
quartiles, no significant trend;
p-trend = 0.391
Congestive heart disease: No association
by quartiles, no significant trend;
p-trend = 0.670
CHD: No association by quartiles, no
significant trend; p-trend = 0.097
Heart attack
Q2: 1.57 (1.06,2.34)
Q3: 1.62 (1.04,2.53)
Q4: 1.47 (0.91,2.37)
p-trend = 0.231
Stroke
Q2: 1.01 (0.70, 1.44)
Q3: 1.42 (0.94,2.13)
Q4: 1.37 (0.92,2.05)
D-143
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
p-trend = 0.045
CRP: -0.068; p-value < 0.001
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
OR per log2- MVD, nephropathy, neuropathy,
unit increase in retinopathy: No statistically significant
baseline PFOA associations
United States Controlled Prediabetic Plasma MVD,
1996-2014 trial adults ages 25+ GM nephropathy,
from DPP and (IQR) = 4.82 neuropathy,
DPPOS (3.20) retinopathy
N = 877
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
Hutcheson et al.
(2020, 6320195)
Medium
United States
2005-2006
Cross-
sectional
Adults from C8
Health Project
N = 48,206
Stroke
OR per ln-unit
increase in
PFOA
Serum
With diabetes:
28.7 (12.9-73.6)
Without
diabetes: 27.6
(13.4-70.4)
Confounding: Age, BMI, CRP, diabetes duration, eGFR, HDL, LDL, history of smoking, race, sex
0.96 (0.91, 1.01)
Osorio-Yanez et
al. (2021,
7542684)
Medium
United States
1999
Cohort
CAC
(Agastston
score)
OR per
doubling in
PFOA
CAC (11-400): 1.17 (0.91, 1.50)
CAC (>400): 1.05 (0.71, 1.57)
Prediabetic Plasma
adults ages 25+ 5.35
enrolled in the (IQR = 3.60)
DPP trial
N = 666
Results: CAC <11 used as reference group.
Confounding: Sex, age, body mass index, race/ethnicity, cigarette smoking, education, treatment assignment, statin use.
He et al. (2018, United States Cross-
Adults ages 20+
Serum DBP, SBP
Percent
DBP: No associations in men or women.
4238388) 2003-2012 sectional
from NHANES
Female Mean
difference per
No significant trend (p-trend = 0.390 and
Low
N = 3,948
(SE) = 3.46
interquartile
0.167 among females and males,
(females) and
(0.04)
ratio increase
respectively)
3,956 (males)
D-144
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
Male Mean
(SE) = 4.50
(0.06)
in PFOA by SBP: No associations in men or women,
quartiles No significant trend (p-trend = 0.096 and
0.642 among females and males,
respectively)
Results: Lowest quartile used as the reference group. Interquartile ratio = 75th/25th percentiles of serum PFOA: 2.43 ng/mL.
Confounding: None listed
Yang et al. China Cross- Adult men Serum DBP, SBP,
(2018,4238462) Years not sectional N= 148 1.90 (Range: hypertension
Low reported 0.6-5.0)
Regression
coefficient per
log-unit
increase in n-
PFOA
Hypertension:
OR for
elevated
pressure (DBP
> 90 or SBP
>140 mmHg)
comparing
above or below
median
DBP: No statistically significant
associations
SBP: 12.94 (-1.46, 27.35)
OR: 10.8(1.31,90)
Outcome: Hypertension evaluated by individual BP components.
Comparison: Logarithm base not specified.
Confounding: Age
Chen et al. Croatia Cross-
(2019,5387400) 2007-2008 sectional
Low
Adults aged 44- Plasma
56 GM
N = 122 (range) = 2.87
(1.03-8.02)
DBP, SBP
Regression
coefficient per
ln-unit increase
in PFOA
DBP:-1.00 (-4.11,2.11)
SBP: -2.15 (-8.49, 4.18)
Confounding: Age, sex, education, socioeconomic status, smoking, dietary pattern, physical activity
Graber et al. United States
Cross-
Members of Serum
Cardiovascular
OR per unit
Any condition
(2019,5080653) 2016-2017
sectional
community with 2.98(1.94-4.69)
conditions,
increase in
0.97 (0.9, 1.05)
Low
exposed water
self-reported
PFOA
supply
D-145
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Design
Years
Population,
Ages,
N
Exposure
Matrix,
Outcome
Levels
Comparison
Resultsb
(ng/mL)a
(Paulsboro, NJ)
ages 12+
N = 105
Confounding: Age, BMI
Honda-Kohmo
United States Cross-
Adults ages 20+
Serum CHD
OR per ln-unit
CHD
etal. (2019,
2005-2006 sectional
from C8 Health
28.4 (12.6-74.9)
increase in
Diabetic adults: 0.9 (0.85, 0.96)
5080551)
Low
Project
N = 5,270 with
diabetes and
49,191 without
diabetes
PFOA or by
quintiles
Q2: 0.92 (0.71, 1.18)
Q3: 0.86 (0.67, 1.11)
Q4: 0.74 (0.58, 0.96)
Q5: 0.73 (0.57,0.94)
Diabetic females: 0.88 (0.80, 0.96)
Diabetic males: 0.93 (0.85, 1.00)
Nondiabetic adults: 0.95 (0.92, 0.98)
Results: Results by quintile use lowest quintile as the reference group.
Confounding: Age, BMI, CRP, diabetes duration, eGFR, HDL, LDL, hemoglobin, iron, sex, smoking history, uric acid, white blood cell
count
Occupational Populations
Steenland et al.
United States Cohort
Current and
Serum CHD,
Incidence rate
CHD: No associations with or without
(2015,2851015) 2008-2011
Low
former workers
at a chemical
plant
Cumulative hypertension,
exposure IQR stroke
with or without
ratio (RR) by
quartiles
lag; RRs ranging from 0.93 to 1.23. No
significant trend.
N = 3,713
10-year lag:
0.8-7.04 or
3.03-
11.42 ng/mL-
Hypertension: No association with or
without lag; RRs ranging from 0.91 to
1.04
No significant trend.
year
Stroke
No lag Q2: 2.63 (1.06, 6.56)
No associations with lag; RRs ranging
from 2.63 to 2.07. No significant trend.
Outcome: Hypertension was self-reported and only analyzed if participants reported taking medication for it.
Results: Lowest quartiles used as the reference group.
Confounding: Gender, race, education, BMI, smoking, alcohol consumption
D-146
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
CMstensen et United States Cross-
Male anglers Serum
Cardiovascular OR per unit
al. (2016,
3858533)
Low
2012-2013
sectional
ages 50+
N = 154
2.50 (1.80-3.30)
condition
(any), CHD,
hypertension
increase of
PFOA
Any condition: 0.96 (0.72, 1.29)
CHD: 0.97 (0.61, 1.45)
Hypertension: 0.74 (0.52, 1.01)
Outcome: Hypertension was self-reported
Confounding: Age, BMI, work status, and alcohol consumption
Girardi and
Italy Cohort
Male workers Serum
Mortality by
Standardized
Merler (2019,
1960-2018
N = 154 GMby
circulatory
Mortality
6315730)
tertiles = 1,700;
disease,
Ratio by
Low
13,051; and
ischemic
tertiles
81,934 ng/mL-
heart disease,
Mortality Risk
years
or stroke
(ictus)
Ratio (for
PFAS plant
workers r.v.
nearby metal
factory
workers)
Mortality: No statistically significant
associations
Exposure: Tertiles of cumulative serum PFOA were defined as follows (in ng/mL-years): T1 < 4,034; 4,034< T2 < 16,956; 16,956< T3
Confounding: Age at risk, calendar period
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; CVHD = cardiovascular heart disease; DBP = diastolic blood pressure (mmHg); DPPOS = Diabetes Prevention Program
Outcomes Study; DPP = Diabetes Prevention Program; GM = geometric mean; HDL = high density lipoprotein cholesterol; HELIX = Human Early-Life Exposome;
HOME = Heath Outcomes and Measures of the Environment; IQR = Interquartile Range; LDL = low-density lipoprotein cholesterol; LVEDD = left ventricular end-diastolic
diameter (mm); LVMI = left ventricular mass index (g/m2); MP AH = 2-(N-methyl-PFOSA) acetate; MVD = microvascular disease; NHANES = National Health and Nutrition
Examination Survey; OR = odds ratio; PAD = peripheral arterial disease; PFOA = perfluorooctanoic acid; PFDE = perfluorodecanoic acid; PFOS = perfluorooctane sulfonate;
PFHxS = perfluorohexane sulfonic acid; PFNA = perfluorononanoic acid; PFUnDA = perfluoroundecanoic acid; PWV = pulse wave velocity; 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-147
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.5.2 Serum Lipids
D.5.2.1 Forest Plots
Effort Estrrale
Confidence Exposure Study -1 0 1 2 3 4 6
Rating Reference Matrix Design Exposure LevtSs Subpapulitfjari Comparison iEE
Low Chen el al.. 2019 Pbsma dlttLnla* Coefficient (per 1-ta Q.22
confidence secOonal
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure D-l. Overall Levels of Total Cholesterol in Adults from Epidemiology Studies Following Exposure to PFOA
Interactive figure and additional study details available on Tableau.
D.5.2.2 Tables
Table D-14. Associations Between PFOA Exposure and Serum Lipid Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Children
Li etal. (2021,
7404102)
High for
gestation, birth,
and childhood
exposures (3-year
and 8-year)
Medium for
exposure at 12-
year follow-up
United Cohort Pregnant women
States and their children
2003-2006 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
Maternal
serum
Gestation: 5.3
(3.7-7.2)
Cord serum
At birth: 3.2
(2.4-1.7)
Serum
At age 3: 5.4
(3.7-7.4)
At age 8: 2.4
(1.8-3.2)
At age 12: 1.3
(1.0-1.6)
Levels (mg/dL) of
triglycerides and HDL;
triglycerides to HDL
ratio
Regression
coefficient per
log2-unit IQR
increase in PFOA
Triglycerides
Gestation: 0.0 (-0.2, 0.2)
At birth: 0.1 (-0.1, 0.3)
Age 3:-0.2 (-0.4, 0.1)
Age 8: 0 (-0.3, 0.2)
Age 12: 0.1 (-0.2,0.3)
HDL
Gestation: -1.5 (-4.7,
1.7)
At birth: -2.1 (-5.6, 1.4)
Age 3: 0.4 (-3.5,4.4)
Age 8: 2.1 (-3.0, 7.3)
Age 12: 3.1 (-1.6,7.8)
Triglycerides to HDL
ratio
Gestation: 0.0 (-0.2, 0.3)
At birth: 0.2 (-0.1,0.4)
Age 3:-0.2 (-0.5, 0.0)
Age 8:-0.1 (-0.4, 0.2)
Age 12: 0.0 (-0.3, 0.3)
D-149
-------
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: 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
1290820) States
Medium 1999-2000
and 2003-
2004
Cross-sectional Adolescents ages Serum
Metabolic syndrome OR per loglO-unit Metabolic syndrome
12-20 years from
NHANES
N = 474
Mean
(SEM) = 1.51
(0.05) loglO-
ng/mL
HDL cholesterol and
metabolic syndrome
triglycerides
increase in PFOA
HDL cholesterol
Model 4: 1.20 (0.60,
2.39)
Model 5: 1.50 (0.67,
3.36)
Metabolic syndrome
triglycerides
Model 4: 1.64 (0.72,
3.73)
Model 5: 1.15 (0.54,
2.47)
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
Level (mg/dL) of HDL Regression
coefficient by
quartiles
HDL
Q4: 4.3 (0.1,
*•5)
United Cross-sectional Adolescent girls Serum
States ages 12-19 years Level not
2003-2004 from NHANES reported
N not reported
Results: Results by quartiles use lowest quartile as the reference group. Quartile analyses discussed in-text only and values provided for
Q4 only.
Confounding: Not reported.
Geiger et al.
(2014, 2850925)
Medium
United
States
1999-2008
Cross-sectional
Adolescents ages Plasma
12-18 years from Mean
NHANES (SE) = 4.2
N = 815 (0.2)
Levels (mg/dL) of TC,
LDL, HDL, and
triglycerides; elevated
TC; elevated LDL;
depressed HDL;
Lipid levels:
Regression
coefficient per ln-
unit increase in
PFOA, Mean
TC: 4.55 (0.90, 8.20)
T2: 4.72 (-1.23, 10.67)
T3: 7.0 (1.40, 12.60)
p-trend = 0.170
elevated triglycerides change by tertiles
D-150
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Elevated or
depressed: OR per
ln-unit increase in
PFOA, or by
tertiles
HDL: -1.52 (-3.02,
-0.03)
T2: 0.53 (-1.23, 2.30)
T3:-1.19 (-2.94, 0.56)
p-trend = 0.177
LDL: 5.75 (2.16, 9.33)
T2: 3.61 (-1.13, 8.36)
T3: 8.18 (3.04, 13.32)
p-trend = 0.0027
TG: 1.74 (-2.88, 6.36)
T2: 3.0 (-5.68, 11.68)
T3: 0.09 (-6.11,6.30)
p-trend = 0.994
Elevated TC: 1.44(1.11,
1.88)
T2: 1.49 (0.97,2.29)
T3: 1.49 (1.05,2.12)
p-trend = 0.025
Depressed HDL: 1.32
(0.82,2.13)
T2: 1.06 (0.65, 1.73)
T3: 1.45 (0.87,2.41)
p-trend = 0.149
Elevated LDL: 1.61
(1.14,2.27)
T2: 1.26 (0.74,2.15)
T3: 1.56 (0.98,2.48)
p-trend = 0.054
D-151
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Elevated TG: 1.10(0.64,
1.89)
T2: 1.35 (0.60, 3.01)
T3: 0.86 (0.46, 1.64)
p-trend = 0.598
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.
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 Abnormal TC, OR by quintiles Abnormal TC
Mean abnormal HDL, Q2: 1.1 (1.0, 1.3)
(SD) = 69.2 abnormal LDL, and Q3: 1.2 (1.0, 1.4)
(111.9) abnormal fasting Q4: 1.2 (1.1, 1.4)
triglycerides Q5: 1.2 (1.1, 1.4)
Abnormal HDL
Q2: 1.0 (0.8, 1.2)
Q3: 1.0 (0.8, 1.2)
Q4: 1.0 (0.9, 1.2)
Q5: 0.9 (0.8, 1.1)
Abnormal LDL
Q2: 1.2 (1.0, 1.5)
Q3: 1.2 (1.0, 1.4)
Q4: 1.2 (1.0, 1.4)
Q5: 1.4 (1.2, 1.7)
Abnormal fasting
triglycerides
Q2: 1.0 (0.7, 1.5)
Q3: 1.3 (0.9, 1.9)
Q4: 1.6 (1.1,2.3)
Q5: 1.0 (0.7, 1.6)
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
D-152
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
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 Plasma
ages 8-10 from 9.3
Danish (Range = 0.8-
component of 35.2)
EYHS
N = 400 normal
weight, N = 59
overweight
Triglycerides (mmol/L) Percent change per
10-unit increase
PFOA
Normal weight: 1.4
(-9.0, 13.0), p-
value = 0.79
Overweight: 76.2 (22.8,
153), p-value = 0.002
p-value for PFOA-BMI
interaction = 0.004
Confounding: Sex, age, ethnicity, paternal income, fast-food consumption, and fitness
Maisonet et al. United
Case-control Pregnant women Serum
(2015, 3981585)
Medium for TC
and HDL at age 7
and all lipids at
age 15
Low for
Triglycerides and
LDL at age 7
Kingdom
1991-1992
and their
daughters
followed-up at
ages 7 and 15
from ALSPAC
Age 7: N = 111
Age 15: N = 88
Levels (mg/dL) of TC, Regression
3.6 LDL, HDL, and
(Range =1.2- triglycerides (ln-
16.4) mg/dL)
coefficient per unit
increase in PFOA
by tertiles
TC
Age 7
Tl: 13.75 (0.05,27.45
T2: -0.53 (-15.39,
14.33)
T3: -1.53 (-4.61, 1.54)
Age 15
Tl: 17.19 (0.45, 33.93)
T2: -1.22 (-16.45,
14.01)
T3: -2.09 (-5.59, 1.40)
LDL
Age 7
Tl: 14.01 (3.26, 24.76)
T2: -5.56 (-17.22,6.10)
T3: 0.03 (-2.38,2.45)
Age 15
Tl: 14.26 (0.25,28.26)
D-153
-------
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: Previous live births, maternal education, and maternal age at delivery
T2: -1.29 (-14.03,
11.45)
T3: -1.41 (-4.33, 1.51)
HDL
Age 7
Tl: 0.50 (-5.78,6.79)
T2: 4.49 (-2.33, 11.30)
T3: -0.40 (-1.82, 1.01)
Age 15
Tl: 0.56 (-7.02,8.15)
T2: 1.04 (-5.87,7.94)
T3: -0.52 (-2.10, 1.06)
Triglycerides
Age 7
Tl: -0.063 (-0.278,
0.153)
T2:-0.150 (-0.384,
0.084)
T3:-0.020 (-0.068,
0.029)
Age 15
Tl: 0.135 (-0.049, 0.319)
T2: -0.047 (-0.215,
0.120)
T3: -0.013 (-0.051,
0.025)
Zeng et al. (2015, Taiwan Cross-sectional Children Serum
2851005) 2009-2010 ages 12-15 Median = 05
Medium N = 225
Levels (ng/dL) of TC,
LDL, HDL, and
triglycerides
Regression TC: 6.57 (2.72, 10.42)
coefficient per In- p-value = 0.001
unit increase PFOA LDL: 4.66 (1.67, 7.65)
p-value = 0.002
D-154
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
HDL: -1.56 (-3.20,
0.08)
p-value = 0.06
Triglycerides: 19.63
(14.82, 24.34)
p-value <0.001
Confidence: Results for triglycerides and LDL considered low confidence because of a lack of fasting prior to blood sample collection.
Confounding: Age, gender, BMI, parental education level, exercise, ETS exposure0
Domazet et al.
(2016, 3981435)
Medium
Denmark Cohort Members of the
1997-2009 European Youth
Study (EYHS)
evaluated at ages
9 and 15
(N = 260), 9 and
21 (N= 175), or
15 and 21
(N = 171)
Plasma Levels (mmol/L) of
Median at age triglycerides
9 = 9.7 (male)
Percent change in
triglycerides at
ages 9 and 15, or
age 9 and 21, or
age 15 and 21 per
10 ng/mL increase
in PFOA at age 9
or 15
Age 9 and 15: -1.46
(-17.84, 18.22)
Age 9 and 21: -8.07
(-30.3, 20.9)
Age 15 and 21: 2.54
(-31.18, 84.56)
or 9.0
(female)
Median at age
15 = 3.7
(male) or 3.4
(female)
Median at age
21 = 3.1
(male) or 2.7
(female)
Confounding: Sex, age, and triglycerides 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
Spain
2003-2008
Cohort
Pregnant women
and their children
(age 4) from
INMA study
N = 627
Maternal
plasma during
1st trimester
GM = 2.32
Levels (z-score) of TC,
LDL, HDL, and
triglycerides
Regression
coefficient per
log2-unit increase
PFOA
TC: 0.02 (-0.10,0.15)
LDL: 0.03 (-0.08,0.15)
HDL: -0.04 (-0.15,
0.08)
Boys: -0.20 (-0.37,
-0.03)
Girls: No association
Triglycerides: 0.04
(-0.07,0.15)
Confidence: Results for triglycerides 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
D-155
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Jain et al. (2018, United
Cross-sectional Children ages 6- Serum
Levels (loglO-mg/dL) Regression
TC: -0.0085
5079656)
Medium
States
2013-2014
11
N = 458
GM = 1.78
of TC, HDL, and non-
HDL
coefficient per
loglO-unit increase
linear PFOA
p-value = 0.46
Non-HDL: -0.0016
p-value = 0.61
HDL: 0.0223
p-value = 0.45
Confounding: Gender, race/ethnicity, age, poverty income ratio, body mass index percentiles, fasting time, and exposure to secondhand
smoke
Kangetal. (2018,
4937567)
Medium
Korea Cross-sectional Children ages 3- Serum
2012-2014 18 from KorEHS- Median = 1.8
C 8
N = 147
Levels of TC (mg/dL),
LDL (mg/dL), and
triglycerides (ln-
mg/dL)
Regression TC:-2.26 (-11.49, 6.98)
coefficient per In- LDL: 3.90 (-4.81, 12.61)
unit increase PFOA Triglycerides: 0.02
(-0.13,0.18)
Results: LDL and triglycerides evaluated at ages 7-18 only (N = 117).
Confounding: Age, sex, BMI z-score, household income, second-hand smoking
Mora et al. (2018,
4239224)
Medium
Levels (mg/dL) of TC,
HDL, LDL, and
triglycerides
Regression
coefficient
increase in
United Cohort and Pregnant women Prenatal
States cross-sectional and their children maternal
1999-2010 from Project Viva plasma
N = 512 prenatal, Median =5.4
596 mid-
childhood Mid-
childhood
plasma
Median =4.3
Confounding: maternal education, prenatal smoking, gestational age at blood draw (for prenatal data)
age at lipids/ALT measurements
No statistically
per IQR significant prenatal
PFOA exposure associations
Mid-childhood:
TC: 2.6 (-0.5, 5.7)
Boys: 1.2 (-3.0, 5.4)
Girls: 5.2 (0.4, 9.9)
HDL: 1.5(0.1,2.9)
, and child's sex, race/ethnicity, and
Jensen et al. Denmark Cohort
Pregnant women
Maternal
Levels (standard
Regression
Regression coefficients
(2020,6833719) 2010-2012
and their children
serum
deviation score) of TC,
coefficient per unit
for all children were
Medium
assessed at
Median = 1.6
LDL, HDL, and
increase in PFOA
between -0.07 and 0.1,
3 months and
2
triglycerides
all with p-values > 0.05
18 months
N = 260 at
LDL at 18 months
3 months, 83 at
Boys: -0.29 (-0.58,
18 months
-0.003)
D-156
-------
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 for interaction
with sex = 0.01
Spratlen et al.
(2020, 5915332)
Medium
Triglycerides at
18 months Boys: 0.43
(0.16,0.70)
p-value for interaction
with sex < 0.01
Confounding: Maternal age, parity, pre-pregnancy BMI, pre-pregnancy BMI2, education, smoking, sex, and lipid outcome at 3 months
United
States
2001-2002
Cross-sectional Pregnant women Cord blood Levels (mg/dL) of TC, Percent change per
and their children
from the
Columbia
University World
Trade Center
birth cohort
N = 222
Median = 2.4
6
total lipids, and
triglycerides in cord
blood
1% increase in
PFOA
Geometric mean
ratios (GMRs) by
quartiles
Confounding: Maternal age, child sex, maternal education, maternal race, parity, pre-pregnancy BMI, marital
and gestational age
TC: 0.038 (-0.032,
0.109)
GMR p-trend = 0.39
Total lipids: 0.087
(0.021,0.153)
GMR p-trend = 0.04
Triglycerides: 0.256
(0.129, 0.383)
GMR p-trend = 0.001
status, family smoking,
Blomberg et al.
Faroe
Cohort and
Children from the
Serum Levels (mmol/L) of
Regression
TC: Regression
(2021, 8442228)
Islands
cross-sectional
Faroese Birth
PFAS at birth: TC, HDL
coefficient per
coefficients were
Medium for HDL
Recruitment:
Cohort 5 at birth,
0.9 (0.63-
log2-unit increase
between -0.14 and 0.18,
and TC
2007-2009
18 months, and
1.34)
in PFOA
all with p-values > 0.05
Low for LDL and
9 years
Female: 0.93
TG
Birth: N = 459
(0.65-1.42)
HDL: Regression
(219 female, 240
Male: 0.87
coefficients were
male)
(0.61-1.22)
between-0.031 and
18 months:
0.041, all with p-values
N = 334
PFAS at
>0.05
9 years: N = 366
18 months:
D-157
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
2.74(1.19-
1.74)
PFAS at
9 years: 1.43
(1.19-1.74)
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, BP; 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:
38.9 (20.1-
68.8)
Children: 20.9
(12.9-33.5)
HDL, LDL,
triglycerides
coefficient per In- Adolescents: 1.05(0.31,
unit increase in 1.80)
PFOA Children: 0.85 (-0.44,
2.14)
HDL
Adolescents: -0.17
(-0.47,0.14)
Children: 0.64 (0.09,
1.19)
LDL
Adolescents: 1.03 (0.39,
1.66)
Children: 0.17 (-0.98,
1.32)
Triglycerides
D-158
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
„ , Exposure
Population, ' .
Re ere nee, Location, Design Ages, a ' Outcome Comparison Select Resultsb
Commence Years Levels
^ (ng/mL)a
Adolescents: 0.01 (0.00,
0.03)
Children: 0.00 (-0.02,
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.
Papadopoulou et
United Cohort
Mother-child
Maternal
Levels (z-scores) of
Regression
HDL
al. (2021,
Kingdom,
pairs from the
plasma
HDL, LDL, and
coefficient per
Maternal PFOA: -0.01
9960593)
France,
HELIX Project,
(prenatal)
triglycerides
doubling in PFOA,
(-0.13,0.10)
Medium
Spain,
children
2.22 (1.34-
or by quartiles
Q2: -0.11 (-0.32,0.10)
Lithuania,
followed-up
3.29)
Q3: -0.06 (-0.31,0.19)
Norway,
around age 8
Q4: -0.05 (-0.35, 0.24)
Greece
(range 6-12)
Plasma
p-trend = 0.821
Recruitment
N = 1,101
(childhood)
Childhood PFOA: 0.17
1999-2010,
1.53 (1.17-
(0.03, 0.32)
Follow-up:
1.96)
Q2: 0.04 (-0.14,0.22)
2013-2015
Q3: 0.11 (-0.08,0.31)
Q4: 0.18 (-0.03, 0.40)
p-trend = 0.160
LDL
Maternal PFOA: -0.04
(-0.08,0.15)
Q2: -0.07 (-0.28,0.14)
Q3:-0.14 (-0.40, 0.11)
Q4:-0.14 (-0.44, 0.16)
p-trend = 0.394
Childhood PFOA: -0.17
(-0.32, -0.03)
Q2: 0.03 (-0.15,0.21)
Q3: -0.03 (-0.23,0.16)
Q4:-0.10 (-0.32, 0.12)
p-trend = 0.195
D-159
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Triglycerides
Maternal PFOA: 0.09
(-0.03, 0.21)
Q2: 0.20 (-0.01,0.41)
Q3: 0.17 (-0.08, 0.43)
Q4: 0.28 (-0.02, 0.58)
p-trend = 0.244
Childhood PFOA: -0.06
(-0.21, 0.08)
Q2
Q3
Q4
-0.15 (-0.33,0.03)
-0.21 (-0.41, -0.01)
-0.13 (-0.35,0.09)
p-trend = 0.345
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, PFOS
Tian et al. (2021, China
7026251) 2012
Medium
Cohort Pregnant women Maternal
and their newborn plasma
children from the 19.6 (14.6-
S-MBCS 27.2)
N = 306
Levels (ln-mg/dL) of
TC, LDL, HDL, and
triglycerides
Regression
coefficient per ln-
unit increase in
PFOA, or by tertile
TC:
Per ln-unit: -0.06 (-0.17,
0.05), p-value = 0.259
T2:-0.10 (-0.22, 0.02)
T3: -0.07 (-0.18,0.05)
LDL:
Per ln-unit: 0.0 (-0.14,
0.14), p-value = 0.982
T2: -0.06 (-0.22, 0.09)
T3: -0.02 (-0.17,0.13)
HDL:
Per ln-unit: -0.09 (-0.22,
0.03), p-value = 0.153
T2:-0.14 (-0.28, 0.01)
T3:-0.11 (-0.25,0.03)
Triglycerides:
D-160
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Select Resultsb
(ng/mL)a
Per ln-unit: 0.03 (-0.09,
0.16), p-value = 0.586
T2: -0.03 (-0.17,0.11)
T3: 0.04 (-0.09,0.18)
Results: Lowest tertile used as reference group.
Confounding: Maternal age, pre-pregnancy BMI, household income, infant sex, gestational age.
Pregnant Women
Starling et al.
Norway
Cross-sectional Women in mid Plasma
Levels (mg/dL) of TC,
Regression
TC
(2014, 2850928)
2003-2004
pregnancy 2.25 (1.66-
HDL, LDL, and
coefficient per ln-
Per ln-unit: 2.58 (-4.32,
Medium for TC,
(median= 18 wee 3.03)
triglycerides (ln-
unit or IQR
9.47)
HDL, and LDL
ks of gestation)
mg/dL)
increase in PFOA,
Per IQR: 1.55 (-2.60,
Low for
from MoBa
or by quartiles
5.69)
Triglycerides
N = 891
Q2: 1.49 (-6.49, 9.48)
Q3: 3.54 (-4.51, 11.59)
Q4: 3.90 (-5.00, 12.80)
HDL
Per ln-unit: 2.13 (-0.26,
4.51)
Per IQR: 1.28 (-0.15,
2.71)
Q2: 0.22 (-2.38, 2.83)
Q3: 2.31 (-0.59,5.20)
Q4: 3.42 (0.56,6.28)
LDL
Per ln-unit: 2.25 (-3.97,
8.48)
Per IQR: 1.36 (-2.38,
5.10)
Q2: 0.94 (-6.08, 7.96)
Q3: 4.16 (-3.19, 11.50)
Q4: 3.35 (-4.35, 11.06)
D-161
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Triglycerides
Per ln-unit: 0 (-0.07,
0.06)
Per IQR: 0 (-0.04, 0.04)
Q2: 0.03 (-0.04,0.11)
Q3: 0.01 (-0.08,0.09)
Q4: —0.04 (-0.12,0.04)
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.
Skuladottir et al.
(2015, 3749113)
Medium
Denmark
1988-1989
Cross-sectional Pregnant women Serum
N = 854 Mean = 4.1
Levels (mmol/L) of TC Regression
coefficient by
quintile
TC:
Q2: 0.10 (-0.19, 0.39)
Q3: 0.39 (0.10, 0.67)
Q4: 0.24 (-0.05, 0.54)
Q5: 0.45 (0.15,0.75)
p-trend = 0.003
Results: Lowest quintile used as the reference group.
Confounding: Age, parity, education, smoking and pre-pregnancy BMI, total caloric intake, and intake of vegetables, meat, and meat
products
Matilla-Santander
et al. (2017,
4238432)
Medium
Spain Cohort
2003-2008
Pregnant women
from the Spanish
INMA birth
cohort
N = 1240
Plasma Levels of TC (mg/dL), Percent change in
Median = 2.3 triglycerides (loglO- median lipid level
5 mg/dL), and C-reactive per loglO-unit
protein (loglO-mg/dL) increase in PFOA
TC: 1.26 (0.01, 2.54)
Triglycerides: -2.78
(-6.15, 1.42) with
inverted U-shaped dose-
response
Confidence: Triglycerides 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
Starling et al.
(2017, 3858473)
Medium
Serum
Median = 1.1
Regression HDL: 1.90 (0.22, 3.59)
coefficient per In- Triglycerides: -0.006
unit increase PFOA (-0.049, 0.036)
United Cohort Pregnant women
States ages 16-45 from
2009-2014 the Healthy Start
study
N = 598
Confounding: Maternal age, race/ethnicity, pre-pregnancy body mass index, education, gravidity, smoking, and gestational age at
Levels of HDL
(mg/dL) and
triglycerides (ln-
mg/dL)
D-162
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location,
Confidence Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
blood draw
Yang et al. (2020, China
7021246) 2013-2014
Medium
Cohort
Pregnant women
ages 20-40 years
in early
pregnancy
N = 436
Serum
5.41 (3.40-
9.08)
Levels (ln-mmol/L) of
TC, triglycerides,
HDL, and LDL;
LDL/HDL ratio
Regression
coefficient per ln-
unit increase in
PFOA, or by
quartiles
TC
Per ln-unit: -0.013
(-0.156,0.131)
Q2: 0.41 (0.11,0.71)
Q3: 0.26 (-0.12,0.64)
Q4: -0.20 (-0.59,0.19)
p-trend = 0.523
Triglycerides
Per ln-unit: 0.044
(-0.131,0.217)
Q2: 0.33 (-0.10,0.76)
Q3: 0.23 (-0.22,0.68)
Q4: 0.07 (-0.40, 0.54)
p-trend = 0.484
HDL
Per ln-unit: 0.018
(-0.025, 0.062)
Q2: 0.06 (-0.02,0.14)
Q3: 0.05 (-0.01,0.11)
Q4: 0.01 (-0.10,0.12)
p-trend = 0.837
LDL
Per ln-unit: -0.046
(-0.143,0.051)
Q2: 0.23 (-0.01,0.47)
Q3: 0.07 (-0.18,0.32)
Q4: -0.24 (-0.50, 0.02)
p-trend = 0.090
LDL/HDL ratio
D-163
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Exposure
Matrix,
Outcome Comparison Select Results
Levels
(ng/mL)a
Per ln-unit: -0.042
(-0.075, -0.009)
p-value <0.05
Q2: 0.01 (-0.06,0.07)
Q3: -0.02 (-0.10,0.06)
Q4:-0.11 (-0.21,-0.01)
p-trend = 0.019
Results: Results by quartiles use 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
Italy
Cross-sectional Pregnant women
Serum
Levels (mg/dL) of TC,
Regression
TC
al. (2021,
2017-2020
ages 18-44 from
16.0 (6.7-
HDL, and LDL
coefficient per ln-
Per ln-unit: -4.25 (-8.26,
7277682)
an area exposed
35.5)
unit increase in
-0.23), p-value< 0.05
Medium for TC
to PFAS through
I Trimester:
PFOA, or by
Q2:-1.12 (-13.24,
HDL; low for
drinking water
17.7 (8.9-
quartiles
-11.00)
LDL
N = 319
35.9)
Q3: -12.65 (-25.25,
I Trimester:
II Trimester:
-0.06), p-value< 0.05
N = 101
15.4 (4.7-
Q4: -13.76 (-26.68,
II Trimester:
35.5)
-0.83), p-value< 0.05
N = 88
III Trimester:
III Trimester: N=
14.5 (6.5-
HDL
=130
34.4)
Per ln-unit: 2.01 (0.53,
3.48), p-value< 0.05
Q2: 4.56 (0.13,9.00), p-
value< 0.05
Q3: 3.74 (-0.88,8.37)
Q4: 6.88 (2.14, 11.62), p-
value< 0.05
Population,
Reference, Location, A
Confidence Years 8
N
LDL
Per ln-unit: -6.74
(-10.15,-3.34), p-
value< 0.05
Q2: -4.70 (-15.02, 5.62)
D-164
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Q3: -15.81 (-26.55,
-5.07), p-value< 0.05
Q4: -21.17 (-32.22,
-10.12), p-value< 0.05
First Trimester
TC: 7.62 (-1.33, 16.57)
HDL: 2.88 (-1.03, 6.80)
LDL: 3.45 (-3.30, 10.22)
Second Trimester
TC: -0.55 (-7.20, 6.08)
HDL: 1.34 (-1.85, 4.54)
LDL:-1.80 (-6.93,3.31)
Third Trimester
TC:-11.02 (-18.07,
-3.96), p-value < 0.05
HDL: 1.98 (-0.15,4.13)
LDL:-13.92 (-20.31,
-7.52), p-value < 0.05
Results: Results by quartile use 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,
United
Cross-sectional Adults ages
Serum
Metabolic syndrome
OR per loglO-unit
Metabolic syndrome
1290820)
States
20+ years from
Mean
HDL cholesterol and
increase in PFOA
HDL cholesterol
Medium
1999-2000
NHANES
(SEM) = 1.48
metabolic syndrome
Model 4: 1.14 (0.84,
and 2003-
N = 969
(0.04) loglO-
triglycerides
1.55)
2004
ng/mL
Model 5: 1.22 (0.86,
1.71)
Metabolic syndrome
triglycerides
Model 4: 0.91 (0.69,
1.20)
D-165
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Model 5: 0.86 (0.65,
1.13)
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.
Nelson et al.
(2010, 1291110)
Medium
United
States
2003-2004
Cross-sectional
Adults ages 20-
80 years from
NHANES
N = 860
Serum
3.9
(Range = 0.1-
37.3)
Levels (mg/dL) of TC,
HDL, non-HDL, LDL
Regression
coefficient per unit
increase in PFOA,
or by quartiles
TC
Per unit increase: 1.22
(0.04, 2.40)
Q4: 9.8 (-0.2, 19.7)
p-trend by
quartiles = 0.07
HDL
20-80 years
Per unit increase: -0.12
(-0.41,0.16)
60-80 years
Q4:-8.7 (-16.3,-1.1)
Non-HDL
Per unit increase: 1.38
(0.12,2.65)
LDL
Per unit increase: -0.21
(-1.91, 1.49)
Results: Results by quartile use lowest quartile 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.
Liu et al. (2018, United Cross-sectional Adults ages 18+ Serum
4238514) States fromNHANES GM=1.86
Medium 2013-2014 N=1871
Levels of TC (mg/dL),
LDL (mg/dL), HDL
(mg/dL), triglycerides
(ln-mg/dL)
Regression TC: 5.58 (2.03)
coefficient (SE) per p-value < 0.05
ln-unit increase in LDL: 4.47 (2.47)
PFOA HDL: 1.93 (0.64)
D-166
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Jain et al. (2019,
5080642)
Medium
p-value <0.01
Triglycerides: -0.08
(0.04)
Confounding: Age, gender, ethnicity, smoking status, alcohol intake, household income, waist circumference, and medications (anti-
hypertensive, anti-hyperglycemic, and anti-hyperlipidemic agents)
Dong et al. (2019,
5080195)
Medium
United
States
2003-2014
Cross-sectional
Adults aged 20-
80 from
NHANES
N = 8849
Serum
Mean =3.7
Levels (mg/dL) of TC,
LDL, HDL
Regression
coefficient per unit
increase PFOA
TC all cycles: 1.48(0.2,
2.8)
Inconsistent associations
with LDL or HDL across
NHANES cycles.
Confounding: Age, gender, race, family income index, BMI, waist circumference, physical activities, diabetes status, smoking status,
number of alcoholic drinks per day
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 = 2.5
Male = 3.4
Levels (mg/dL) of TC,
LDL, HDL,
triglycerides
Regression
coefficient per
loglO-unit increase
PFOA
TC
OM: 0.0519(0.0128,
0.0911)
p-value = 0.01
No clear associations in
NF, NM, or OF
LDL
OM: 0.0822 (0.0098,
0.1546)
p-value = 0.03
No clear associations in
NF, NM, or OF
HDL: No clear
associations
Triglycerides: No clear
associations
Confounding: race/ethnicity, smoking status, age, poverty income ratio (PIR), fasting time, use of lipid lowering medicine, physical
exercise, survey year, daily dietary intake of total cholesterol, daily intake of total saturated fat, calories, caffeine, alcohol, protein intake
Fan et al. (2020, United Cross-sectional Adults age 20+ Serum
7102734) States from NHANES Median =2.0
Medium 2011-2014 N= 1067 5 ng/mL
Levels (mg/dL) of TC,
LDL, HDL, and
triglycerides
Regression
coefficient per
loglO-unit increase
in PFOA
TC: 6.74 (3.23, 10.2)
p-value <0.001
LDL: 4.67 (1.57, 7.77)
p-value = 0.003
D-167
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
HDL: 2.23 (0.97, 3.49)
p-value = 0.001
Triglycerides: 0 (-0.05,
0.04)
p-value = 0.891
Confounding: Age, gender, race, education level, PIR, BMI, smoking status, alcohol use, energy intake levels, screen time
Jain and
Ducatman (2020,
6988488)
Medium
Apolipoprotein B
(loglO-mg/dL)
Regression
coefficient per
loglO-unit increase
in PFOA
United Cross-sectional Adults age 20+ Serum
States from NHANES Levels not
2007-2014 Non-diabetic non- reported
LLM users:
N = 2,872
Diabetic non-
LLM users:
N = 316
Non-diabetic
LLM users:
N = 519
Diabetic LLM
users: N = 293
Population: LLM = Lipid lowering medication
Confounding: Gender, age, age squared, race/ethnicity, poverty income ratio, fasting time in hours, log 10-transformed BMI, smoking
status, survey year, daily intake of cholesterol, caffeine, alcohol, total calories, total protein, and total fat
Apolipoprotein B
Non-diabetic non-LLM
users: 0.03878, p-
value < 0.01
Diabetic non-LLM users:
-0.02055, p-value = 0.52
Non-diabetic LLM users:
-0.01042, p-value = 0.59
Diabetic LLM users:
-0.00058, p-value = 0.98
Steenland et al.
United
Cross-sectional Adults ages 18+
Serum Levels (ln-mg/dL) of
Lipid levels, ratios:
TC
(2009, 1291109)
States
from the C8
26.6 (Range: TC, LDL, HDL, non-
Regression
Per ln-unit: 0.01112
Medium for TC,
2005-2006
Health Project,
0.25-17556.6) HDL cholesterol, and
coefficient per ln-
(SD = 0.00076)
HDL
current or former
triglycerides; TC/HDL
unit increase in
D2: 0.01 (SE = 0.004), p-
Low for TG, LDL
residents from
ratio; high TC
PFOA, or by
value = 0.0026
areas supplied
deciles
D3: 0.02 (SE = 0.004), p-
with
value < 0.0001
contaminated
High TC:
D4: 0.03 (SE = 0.004), p-
water
OR by PFOA
value < 0.0001
N = 46494
quartiles
D5: 0.04 (SE = 0.004), p-
value < 0.0001
D-168
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
D6: 0.03 (SE = 0.004), p-
value < 0.0001
D7: 0.04 (SE = 0.004), p-
value < 0.0001
D8: 0.04 (SE = 0.004), p-
value < 0.0001
D9: 0.04 (SE = 0.004), p-
value < 0.0001
D10: 0.05 (SE = 0.004),
p-value < 0.0001
„ , Exposure
Population, Matrix
Location, Design Ages, a nx' Outcome Comparison Select Resultsb
Years s Levels P
(ng/mL)a
HDL
0.00276 (SD = 0.00094)
LDL
0.01499 (SD = 0.00121)
Triglycerides
0.00169 (SD = 0.00219)
TC/HDL ratio
0.00831 (SD = 0.0011)
Non-HDL
0.01406 (SD = 0.03476)
High TC
Q2: 1.21 (1.12, 1.31)
Q3: 1.33 (1.23, 1.43)
Q4: 1.38 (1.28, 1.50)
p-trend < 0.0001
Outcome: High TC defined as > 240 mg/dL.
Results: Results by quartile use lowest quartile as the reference group; results by decile use lowest decile as the reference group.
Confounding: Age, male gender, smoking status, education level, drinks alcohol, currently exercises, and BMI
D-169
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Eriksen et al.
(2013,2919150)
Medium
Denmark Cross-sectional Adults ages 50- Plasma Levels of TC (mg/dL) Regression 4.4(1.1,7.8)
1993-1997 65fromDCH Mean = 7.1 coefficient per IQR p-value = 0.01
N = 753 increase in PFOA
Confounding: Sex, education, age, BMI, smoking status, intake of alcohol, egg, and animal fat and physical activity
Fisher etal. Canada Cross-sectional Adults ages 18-
Plasma Levels (ln-mmol/L) of
Lipid levels,
TC
(2013,2919156) 2007-2009 74 years from
GM TC, HDL, LDL, non-
TC/HDL ratio:
0.03 (-0.017,0.07), p-
Medium CHMS, cycle 1
(SD) = 2.46 HDL, triglycerides;
Regression
value = 0.22
N = 2,700
(1.83) TC/HDL ratio (In-
coefficient per ln-
TC, HDL, Non-
transformed); high
unit increase in
HDL
HDL, TC/HDL
cholesterol
PFOA
0.0009 (-0.04, 0.04), p-
ratio: N = 2,345
value = 0.96
LDL,
High cholesterol:
triglycerides:
OR per ln-unit
LDL
N = 1,168
increase in PFOA,
0.02 (-0.06,0.091), p-
High cholesterol:
or by quartiles
value = 0.63
N = 1,042
Non-HDL
0.036 (-0.01, 0.08), p-
value = 0.13
Triglycerides
-0.003 (-0.13,0.12), p-
value = 0.94
TC/HDL ratio
0.02 (-0.016,0.0), p-
value = 0.22
High cholesterol
per ln-unit increase: 1.22
(0.89, 1.67)
Q2: 1.61 (1.02,2.53)
Q3: 1.26 (0.76,2.07)
Q4: 1.50 (0.86,2.62)
D-170
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location,
Confidence Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Select Resultsb
(ng/mL)a
p-trend = 0.10
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
(2013,2850962) States
Medium for TC, Baseline:
HDL 2005-2006;
Low for TG, LDL Follow-up:
2010
Cohort
Baseline GM LDL, HDL, and
(SD) = 74.8 triglycerides
Adults ages 20- Serum
60 from C8
Short-Term
Follow-up Study (208.7)
living in West Follow-up
Virginia and Ohio GM
with PFOA- (SD) = 30.8
contaminated (143.9)
drinking water
N = 560 (N = 521
for LDL analysis)
Confounding: Age, sex, interval between measurements, and fasting status
Levels (mg/dL) of TC, Percentage
decrease (loglO of
final and initial
ratio change per
log 10 of ratio
change in PFOA)
TC: 1.65 (0.32, 2.97)
R2 = 0.03
LDL: 3.58 (1.47, 5.66)
R2 = 0.06
HDL: 1.33 (-0.21, 2.85)
R2 = 0.04
Triglycerides: -0.78
(-5.34, 3.58)
R2 = 0.08
Winquist and
Steenland (2014,
2851142)
Medium
United
States
2008-2011
Workers at a
Mid-Ohio Valley
chemical plant
and residents of
the surrounding
community from
C8 Health Project
N= 32,254
Cohort Workers at a Serum Hypercholesterolemia HR by quintiles Hypercholesterolemia
26.1(12.8- Whole cohort
68.1) Q2: 1.24 (1.15, 1.33)
Q3: 1.17 (1.09, 1.26)
Q4: 1.19(1.11, 1.27)
Q5: 1.19(1.11, 1.28)
p-trend = 0.005
Men 40-59 years of age
Q2: 1.38 (1.21, 1.56)
Q3: 1.32 (1.17, 1.50)
Q4: 1.31 (1.16, 1.48)
Q5: 1.44(1.28, 1.62)
p-trend < 0.001
Outcome: Hypercholesterolemia cases were identified based on self-reported diagnosis.
Results: Lowest quintile used as the reference group.
Confounding: Age, sex, years of schooling, race, smoking, smoking duration, smoking pack-years, regular alcohol consumption, BMI,
self-reported type-2 diabetes
D-171
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Donat-Vargas et
al. (2019,
5080588)
Medium
Sweden Cohort
1990-2013
Non-diabetic Plasma
Levels (mmol/L) of TC
Regression
Per change in PFOA
adults ages 30-60 Baseline
and triglycerides
coefficient per 1-
TC
at baseline in median = 2.9
SD increase in
Baseline: -0.19 (-0.36,
Vasterbotten Median at 10-
PFOA, or by
-0.02)
Intervention year follow-
tertiles
Follow-up: -0.03 (-0.21,
Programme (VIP) up = 2.7
0.15)
N = 187
Prospective: -0.12
(-0.23, 0)
Triglycerides
Baseline: -0.03 (-0.14,
0.07)
Follow-up: -0.08 (-0.20,
0.04)
Prospective: -0.07
(-0.13,-0.01)
Overall non-significant
inverse association using
tertiles
Confounding: Gender, age, education, sample year, body mass index, smoking habit, alcohol consumption, physical activity and healthy
diet score
Linetal. (2019,
United
Cohort and Prediabetic adults Plasma
Levels (mg/dL) of TC,
Regression
Cross-sectional
5187597)
States
cross-sectional age 25+ from the Median = 4.9
LDL, HDL,
coefficient per
TC: 6.09 (3.14, 9.04);
Medium
1996-2014
Diabetes
triglycerides, non-
doubling PFOA
p<0.01
Prevention
HDL, and very low
LDL: 2.93 (0.22, 5.63);
Program (DPP)
density lipids (VLDL);
HR or OR for
p-value <0.05
and Outcomes
hypercholesterolemia,
hypercholesterolem
HDL: -0.49 (-1.38,
Study (DPPOS)
hypertriglyceridemia
ia or
0.40)
N = 940 (888 not
hypertriglyceridemi
Triglycerides: 17.75
on metformin)
a per doubling of
(9.77, 25.74); p-
PFOA
value < 0.01
VLDL: 3.66 (2.18, 5.15);
p-value <0.01
D-172
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Hypercholesterolemia at
baseline OR: 1.29 (1.05,
1.57)
Hypertriglyceridemia at
baseline OR: 1.48(1.21,
1.81)
Prospective
Hypercholesterolemia
HR: 1.06 (0.94, 1.19)
Greater effect in the
placebo group
Hypertriglyeridemia HR:
1.23 (1.04, 1.45)
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 = 35.
8
Female = 22.6
5
Male = 58.3
Levels (mg/dL) of TC,
LDL, HDL, non-HDL,
and triglycerides
Regression
coefficient per ln-
unit increase
PFOA, or by decile
TC
1.94(1.48, 2.41)
p-value for interaction
with sex = 0.15
Associations for deciles
2-10 consistently
increase from 2.83 to
9.10
LDL
1.12(0.71, 1.52)
p-value for interaction
with sex = 0.577
Associations for deciles
2-10 moderately increase
from 1.4 to 5.3
D-173
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Exposure
Matrix,
Outcome Comparison Select Results
Levels
(ng/mL)a
HDL
0.49 (0.32, 0.67)
Male: 0.13 (-0.11, 0.37)
Female: 0.83 (0.57, 1.1)
p-value for interaction
with sex <0.001
Associations for deciles
2-10 moderately increase
from 0.45 to 2.07
Triglycerides
0.02 (0.01, 0.03)
p-value for interaction
with sex = 0.815
Associations for deciles
2-10 increase from 0.04
to 0.09
Results: Lowest decile used as the 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,
Taiwan Cross-sectional Adults aged 55 to
Serum
Levels (mg/dL) of TC,
Regression
TC
6988476)
2016-2017 75 that resided in
8.6(6.2-11.6)
HDL, LDL, and
coefficient by
Q2: 2.48 (-8.00, 12.96)
Medium
the study area for
triglycerides
quartiles
Q3: 2.88 (-7.64, 13.40)
more than
Q4: 4.04 (-6.65, 14.73)
10 years and not
p-trend = 0.47
taking lipid-
lowering
HDL
medication
Q2: 0.45 (-3.57,4.48)
N = 352
Q3:-3.36 (-7.40, 0.68)
Q4: -1.72 (-5.82,2.38)
p-trend = 0.18
LDL
Q2: 4.79 (-4.65, 14.23)
Q3: 8.72 (-0.76, 18.20)
Q4: 8.06 (-1.57, 17.69)
Population,
Reference, Location, A
Confidence Years 8
N
D-174
-------
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-trend = 0.07
Triglycerides
Q2: 0.55 (-17.93, 19.03)
Q3: 14.43 (-4.13, 32.98)
Q4: 15.45 (-3.40, 34.30)
p-trend = 0.05
Results: Lowest quartile used as the reference group.
Confounding: Age, sex, smoking status, and drinking status
Liu et al. (2020,
6318644)
Medium
United Randomized Adults from Plasma Levels (mg/dL) of TC, Least-squared TC
States clinical trial POUNDS Lost Median =4.6 triglycerides, and means (LSM) by Tl: 189.1 (7.9)
2004-2007 study ages 20+ apolipoproteins loglO- tertile PFOA T2: 189.3 (7.6)
N = 326 ApoB, ApoE, and T3: 188.4 (7.7)
ApoC-III p-trend = 0.67
Triglycerides
Tl: 111.1 (11.2)
T2: 137.3 (10.8)
T3: 131.8 (10.9)
p-trend = 0.06
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, China Case-control
Adults ages 25 to
Serum
Levels (loglO-mmol/L) Regression
TC: 0.01 (-0.05,0.07)
7762348) 2016-2017
74 including type
Cases: 10.05
of TC, HDL, LDL, and coefficient per
HDL: -0.03 (-0.09,
Medium
2 diabetes cases
(6.75-17.05)
triglycerides
loglO-unit increase
0.04)
and healthy
Controls:
in PFOA
LDL: 0.02 (-0.07,0.10)
controls
11.40 (9.20-
Triglycerides: 0.09
N = 304
17.40)
(-0.06, 0.23)
Confounding: Age, sex, BMI.
Jeddi et al. (2021, Italy Cross-sectional
Residents aged
Serum
Reduced HDL,
OR per ln-unit
Reduced HDL: 0.93
7404065) 2017-2019
20-39 from the
GM (range):
elevated triglycerides
increase in PFOA
(0.89, 0.97), p-
Medium
PFAS-
67.66 (0.70-
value < 0.05
contaminated
1400.0)
Veneto region
D-175
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Select Resultsb
(ng/mL)a
N= 15,876 Elevated triglycerides:
1.10(1.05, 1.16), p-
value < 0.05
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 Cross-sectional
Current and
Serum Levels of cholesterol
Regression
Cholesterol
(2003, 1290020)
States,
former workers at Antwerp (ln-mg/dL)
coefficient per unit
0.032 (0.013,0.051)
Medium
Belgium
two
Mean
increase in PFOA
1994-2000
fluorochemical
(SD) = 1.03
production plants
ppm (1.09);
Male
Decatur =1.9
N = 421,
0 ppm (1.59)
Female
N = 97,
Regression
analysis
N = 174
Confounding: Age, BMI, drinks/day, cigarettes/day, location, entry period, baseline years worked
Costa et al.
Italy Cross-sectional
Current and
Serum Levels of TC and HDL
Comparison of
No significant
(2009, 1429922)
2007
former male
Production (mg/dL)
mean outcome
differences in
Medium
employees of an
workers
(Exposed vs
comparison of mean
Italian chemical
(2007): 3.89
unexposed
HDL
production plant,
Hg/mL (2.18-
workers)
Comparison of
18.66 ng/mL)
Comparison of mean TC
means analysis
Regression
p-value = 0.003
N = 68,
coefficient
Exposed vs
(exposed workers
TC
Unexposed
vs all workers)
Exposed vs Unexposed:
analysis
21.7 (6.83,36.6), p-
N = 141,
value = 0.005
D-176
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Continuous
regression
analysis
N = 56
Regression
coefficient per unit
increase in PFOA
Continuous: 0.028
(0.002, 0.055), p-
value < 0.05
HDL
Exposed vs Unexposed:
2.42 (-2.30, 7.13)
Continuous: -0.018
(-0.047, 0.012)
Confounding: Age, job seniority, body mass index, smoking and alcohol consumption. Additional confounding for continuous
regression analyses: year of observation
Sakr et al. (2007, United
1291103) States
Medium 2004
Cross-sectional
Active employees
at a Washington
Works site where
APFO is used
N = 1,019
Workers not on
lipid-lowering
medications
N = 840
Serum
Mean
(SD) = 0.428
(0.860) ppm,
Range = 0.00
5-9.550 ppm
Levels (mg/dL) of TC,
LDL, HDL and levels
(ln-mg/dL) of VLDL
and triglycerides
Regression
coefficient per unit
increase PFOA
TC
4.036 (1.284)
p-value = 0.002
Workers not on lipid-
lowering medications:
5.519 (1.467)
p-value <0.001
LDL
2.834 (1.062)
p-value = 0.008
Workers not on lipid-
lowering medications:
3.561 (1.213)
p-value = 0.003
HDL
-0.178 (0.432)
p-value = 0.680
Workers not on lipid-
lowering medications:
0.023 (0.508)
p-value = 0.964
D-177
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Select Resultsb
(ng/mL)a
VLDL
0.045 (0.021)
p-value = 0.031
Workers not on lipid-
lowering medications:
0.055 (0.025)
p-value = 0.026
Triglycerides
0.018(0.021)
p-value = 0.384
Workers not on lipid-
lowering medications:
0.030 (0.024)
p-value = 0.207
Results: Reported as effect estimate (standard error).
Confounding: Age, gender, BMI
Olsen et al.
United
Cross-sectional Male workers
Serum
Level (mg/dL) of HDL Regression
HDL
(2000, 1424954)
States
involved in
1993: 1.1
coefficient per
1993:-0.14 (SD = 0.33),
Low
1993-1997
ammonium
(Range = 0.0-
1 ppm increase in
p-value = 0.67
perfluorooctanoat
80.0) ppm
PFOA
1995:-0.10 (SD = 0.08),
e production
1995: 1.2
p-value = 0.18
N = 265
(Range = 0.0-
114.1) ppm
1997: 1.3
(Range = 0.1-
81.3) ppm
1997:-0.19 (SD = 0.13),
p-value = 0.16
Confounding: Age, alcohol and cigarette use, BMI, testosterone
Notes: APFO = ammonium perfluorooctanoate; ALSPAC = Avon Longitudinal Study of Parents and Children; CHMS = Canadian Health Measures Survey; DCH = Diet, Cancer
and Health; EYHS = European Youth Study; HDL = high density lipids; HELIX = Human Early-Life Exposome; HR = hazard ratio; IQR = interquartile range; LDL = low
density lipids; HOME = Heath Outcomes and Measures of the Environment; KorEHS-C = Korea Environmental Health Survey in Children and Adolescents; MoBa = Norwegian
Mother and Child Cohort Study; NHANES = National Health and Nutrition Examination Survey; OR = odds ratio; S-MBCS = Shanghai-Minhang Birth Cohort Study;
SE = standard error; TC = total cholesterol; 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.
D-178
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
c Confounding indicates factors the models presented adjusted for.
D.6 Endocrine
Table D-15. Associations Between PFOA Exposure and Endocrine Effects in Recent Epidemiologic Studies
„ , Exposure
Population, M^atr.x
Ages, ' Outcome Comparison Select Resultsb
„ Levels
(ng/mL)a
Reference, Location, _ .
Confidence Years Design
General Population
Lebeaux et al.
(2020, 6356361)
High for Cord
serum thyroid
hormones;
Medium for
maternal serum
thyroid
hormones
United States
2003-2007
Cohort
Mother-infant
pairs from
Health Outcome
Measures of the
Environment
(HOME) Study
N = 256 for
cord serum
N = 185 for
maternal serum
Cord serum
5.6
Maternal serum
5.5
Levels of TSH
(HIU/L), TT4
(Hg/dL), TT3
(ng/dL), FT4
(ng/dL), and FT3
(pg/mL)
Regression
coefficient per
Cord serum
TSH: 0.06 (-0.08, 0.19)
log2-unit increase TT4
in PFOA TT3
FT4
FT3
0.03 (-0.02, 0.08)
-0.01 (-0.09,0.06)
-0.01 (-0.04, 0.03)
-0.01 (-0.06, 0.03)
Maternal serum
TSH: 0.09 (-0.14,
0.33)
TT4
TT3
FT4
FT3
-0.03 (-0.10,0.04)
-0.01 (-0.05,0.04)
-0.01 (-0.06, 0.03)
-0.01 (-0.04, 0.01)
Confounding: Individual PFAS, maternal age at delivery, race/ethnicity, marital status at baseline, maternal education level, household
income, mean loglO 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.
(2018, 5080657)
Medium
Fernand, Ohio,
USA
1991-2008
Cohort
Fernald
Drinking water
Levels of TSH
Percent change
TSH
Community
Serum
(ln-|iIU/mL) and per IQR increase
-0.48 (-9.68, 9.65)
Cohort,
12.7
TT4 (ln-ng/dL)
in PFOA
p-value = 0.92
Median age
Males: 9.38 (-7.47, 29.3)
38 years at
p-value = 0.47
enrollment,
Females:-6.64 (-17.8, 5.97);
N = 122 for
p-value = 0.31
TSH
measurements;
TT4
47 male and 75
-1.18 (-5.12, 2.92); p-value =
female
Males: -2,71 (-9.05,4.08);
p-value = 0.43
D-179
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
N = 144 for
TT4
measurements;
63 males and 81
females
Females: -1.62 (-6.88, 3.94);
p-value = 0.56
Confounding: Age, year of measurement, sex, education, income, marital status, BMP
Jain and
Ducatman
(2019, 6315816)
Medium
United States
2007-2012
Cross-sectional
Adults from
NHANES aged
20+
Glomerular
filtration (GF)
status:
GF-1 = 1,653
GF-2 = 720
GF-3A = 114
GF-3B/4 = 62
Serum
Levels not
reported
Levels of TSH
(log-|iIU/mL).
TGN (log-
ng/mL),
TT4 (log-ng/dL),
FT4 (log-ng/dL),
TT3 (log-ng/dL),
FT3 (log-pg/mL)
Regression
coefficient per
loglO-unit
increase in PFOA
TSH
GF-1: -0.004, p-value = 0.89
GF-2: 0.085, p-value <0.01
GF-3A: -0.229, p-value = 0.04
GF-3B/4: 0.012, p-value = 0.88
FT4
GF-1: -0.010, p-value = 0.17
GF-2: -0.020, p-value = 0.08
GF-3A: 0.038, p-value = 0.07
GF-3B/4: -0.040, p-value = 0.15
GF Stages: GF-1: GFR> 90 mL/min/1.73m2; GF-2: GFRbetween60 and 90 mL/min/1.73m2; GF- 3A: GFRbetween45 and
60 mL/min/1.73m2; GF- 3B/4: GFR between 15 and 45 mL/min/1.73m2
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: Significantly increased levels
2168068)
2007-2008 children from
Total cohort
TSH (nIU/L),
coefficient per
(T3 vs. Tl), p-value < 0.01
Low
NHANES aged
Lowest tertile
FT3 (pg/L),
loglO-unit
12+
T1 <3.3
TT3 (fg/dL),
increase in
TT3: 0.032, p-value = 0.01
N = 1,540
Highest tertile
FT4 (pg/L),
PFOA, or by
including
T3 >5.1
TT4 (pg/L),
tertiles
FT3, FT4, TT4, TGN: No
children
TGN
statistically significant associations
Results: Lowest tertile used as the reference group
Confounding: Gender, race, age, iodine deficiency, iodine replete
Lewis et al.
United States Cross-sectional Children and
Serum
Levels of
Percent change
TSH
(2015,3749030) 2011-2012 adults from
Females 12-20:
TSH (nIU/mL),
per doubling of
Females
Low
NHANES, aged
1.53
TT3 (ng/dL),
PFOA
12-20: 16.6 (2.6, 28.6)
12-80
FT3 (pg/mL),
20-80: No associations
D-180
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Select Resultsb
(ng/mL)a
145 females 12
to <20
680 females 20-
80
158 males 12
to <20
699 men
Females 20-40: TT4 (|ig/mL).
1.49 FT4 (ng/dL)
Females 40-60:
1.62
Females 60-80:
2.55
Males 12-20:
1.85
Males 20-40:
2.35
Males 40-60:
2.31
Males 60-80:
2.48
Males, all age groups: No
associations
TT3
Females
60-80: 3.3 (0.6, 6)
Younger than 60: No associations
Males, all age groups: No
associations
FT3
Females
60-80: 1.8 (0.2, 3.4)
Younger than 60: No associations
Males, all age groups: No
associations
TT4
Females
12-20: 4.1 (0.6, 8.9), p-value < 0.1
20-80: No associations
Males
40-60: -3.1 (-6.2,0.1),
p-value < 0.10
12-40 or 60-80: No associations
FT4
Females
20-40:2.0(0,4.1)
12-20 or 40-80: no associations
Males, all age groups: No
associations
Confounding: Age, BMI, poverty income ratio, serum cotinine, and race/ethnicity
D-181
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Byrne et al.
St. Lawrence Cross-sectional Alaska Natives, Serum
(2018, 5079678) Island, Alaska,
Low USA
2013-2014
aged 18-45
N = 85
38 men
47 women
1.01
Male: 1.47
Female: 0.772
Levels of
TSH (ln-
|iIU/mL).
TT3 (pg/mL),
FT3 (ng/dL),
TT4 ((ig/dL),
FT4 (ng/dL)
Regression
coefficient per ln-
unit increase in
PFOA
TSH
Total cohort: 0.63 (0.22, 1.03),
p-value < 0.005
TT3
Total cohort: -7.67 (-18.61, 3.27),
p-value = 0.17
Males: -14.24 (-26.24, -2.24),
p-value = 0.02
Females: 11.29 (-5.25, 27.83)
p-value for sex interaction = 0.18
FT3, TT4, FT4: No statistically
significant associations
Confounding: Age, sex, smoking status
Convertino et al. Scotland
(2018,5080342) 2008-2011
Low
Controlled trial
Adults, Solid- Serum Levels of FT4
tumor cancer Median PFOA (mmol/L)
patients ranging from
49 9-
1,530 nmol/mL
Regression
coefficient per
unit increase in
PFOA
Median and mean
FT4 levels by
exposure
categories
0.003, p-value = 0.21
Increasing trend in FT4 by
exposure categories
Confounding: None given
Heffernan et al. United
(2018, 5079713) Kingdom
Low 2015
Cross-sectional
Women aged
20-45 years,
with (cases) or
without
(controls)
polycystic
ovarian
syndrome
(PCOS)
Serum Levels of
Geometric TSH (mU/L),
mean = 2.49 for FT3 (ln-pmol/L),
both cases and FT4 (ln-pmol/L)
controls
Regression
coefficient per ln-
unit increase in
PFOA
TSH
PCOS cases: 0.86, p-value < 0.01
PCOS controls: -0.13,
p-value = 0.75
FT3, FT4: No statistically
significant associations
D-182
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
N = 59
Confounding: Serum albumin
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
Cases:11.10
Controls: 8.35
Levels (ng/mL)
of TSH, FT3,
FT4
Regression
coefficient per
log-unit increase
in PFOA
TSH
POI cases: 1.39 (0.18,2.59)
POI controls: 1.65 (0.86, 2.44)
FT4:
POI cases: -3.42 (-5.39, -1.46)
POI controls: No association
Comparison: Logarithm base not specified.
Confounding: Age, BMI, education, income, sleep, and parity
FT3
No statistically significant
associations
Children
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
Maternal blood
Geometric
mean = 2.37 ng
Cord serum
levels of
TSH (log-IU/L),
T4 (log-pmol/L),
FT3 (log-
pmol/L),
FT4, (log-
pmol/L)
Regression TSH :23.1 (1.9, 48.6)
coefficient per T4: 1.9 (-4.1, 8.3)
log2-unit increase FT3: 0.5 (-5.6, 6.9)
in PFOA FT4: 1.9 (-11.5, 17.2)
FT3 resin
uptake, FT4
index (FTI) (log-
IU/L)
Confounding: Child sex (in detailed results), parity, maternal BMI, maternal height, maternal education, maternal age, smoking and drinking
alcohol during pregnancy, total PCB, mercury
Kim et al. South Korea
(2020,6833758) 2012-2017
Cohort Children, aged Serum Levels of TSH, Regression FT4 at age 6
2, 4, 6 years Age 2: 4.39 FT4 (ng/dL), and coefficient per In- All: 0.07, p-value < 0.05
D-183
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
N= 181-660 Age 4: 3.65
Age 6: 3.83
T3 (ng/dL) at
age 6
Subclinical
hypothyroidism
High
unit increase in
PFOA
Subclinical
hypothyroidism:
OR per increase
in PFOA
Boys: 0.04, p-value < 0.05
No interaction with sex
TSH, T3: No statistically
significant associations between or
within age groups
Confounding: Age, sex, dietary iodine intake
Kang et al. Korea
(2018, 4937567) 2012-2014
Medium
Cross-sectional Children from Serum
1.88
Levels (ng/dL) Regression
of TSH, FT4
Seoul and
Gyeonggi aged
3-18
N = 147
Confounding: Age, sex, BMI z-score, household income, second-hand smoking
coefficient per ln-
unit increase in
PFOA
TSH: -0.14 (-0.62,0.34),
p-value = 0.341
FT4: 0.04 (-0.01, 0.09),
p-value = 0.075
Aimuzi et al. China
(2019, 5387078) 2012-2013
Medium
Cross-sectional
Levels of TSH
(ln-mlU/L), FT3
(pmol/L), FT4
(pmol/L)
Regression
coefficient per ln-
unit increase in
PFOA
Pregnant Cord blood
women and 7.57
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
FT4
All children: 0.14 (0.02, 0.26)
Boys: 0.25 (0.08, 0.42)
Girls: 0.01 (-0.16,0.18)
Itoh et al. (2019,
5915990)
Medium
Japan
2003-2005
Cohort
Pregnant
women and
their children
259 male
children
240 female
children
Plasma
2.00
Levels of
TSH (ln-
|iU/mL).
FT3 (ln-pg/mL),
FT4 (ln-pg/mL),
TPOAb (ln-
IU/mL),
TgAb (ln-
IU/mL)
Regression
coefficient per ln-
unit increase in
PFOA
TgAb
Boys, maternal TA negative: -0.13
(-0.27, -0.002), p-value = 0.047
All boys or maternal TA positive:
no association
Girls, maternal TA positive: 0.27
(0.95, 0.44), p-value = 0.007
All girls or maternal TA negative:
no association
D-184
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
TSH, FT3, FT4, TPOAb: No
statistically significant associations
Confounding: Age at delivery, parity, educational level, alcohol consumption, smoking during pregnancy, pre-pregnancy BMI
Pregnant Women
Dreyer et al. Denmark
(2020,6833676) 2010-2012
High
Cohort
Pregnant
women from
Odense Child
Cohort (OCC)
N = 1,048
Serum
1.64
Levels of diurnal Percent change
urinary (dU)
Cortisol
(nmol/24-hours),
dU-cortisone
(nmol/24-hours),
dU-
cortisol/cortisone
, serum Cortisol
(nmol/L)
per 2-fold
increase in PFOA
Serum Cortisol: -15.8 (-33.1, 1.5)
T2:-19.6 (-51.0, 11.8)
T3: -35.1 (-69.4,-0.7), p-
value < 0.05
p-trend = 0.05
dU-cortisol, dU-cortisone, dU-
cortisol/cortisone: No statistically
significant associations
Confounding: Age, parity, and offspring sex
Xiao et al.
(2020, 5918609)
High
Faroe Islands,
Denmark
1994-1995
Cross-sectional
Pregnant
women and
their children,
Maternal age 28
(SD = 5.6)
Maternal blood
Geometric
mean = 2.37 ng
Regression TSH: 12.6 (-4.5, 32.8)
coefficient per T4: 0.7 (-5.5, 7.3)
log2-unit increase FT3: 3.1 (-1.2, 7.6)
in PFOA FT4: -0.4 (-5.4, 4.8)
N = 172 and
153 for
measurements
in maternal and
cord serum,
respectively
Confounding: Child sex (in detailed results), parity, maternal BMI,
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
maternal height, maternal education, maternal age, smoking and drinking
Preston et al.
(2018, 4241056)
Medium
United States Cross-sectional Pregnant Maternal Levels of Percent
1999-2002 women and plasma TSH (mlU/mL), difference in
their children 5.6 FT4 |ig/dL). hormone level
TT4 ([ig/dL)
FT4: -1.87 (3.4,-0.31)
TSH: 0.28 (-9.26, 10.8)
D-185
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
N = 726 and
718 for free T4
and TSH
measures,
respectively
per IQR increase
in PFOA
TSH TPOAb negative: 0.88 (-9.22,
12.1)
TSH TPOAb positive: -19 (-35.1,
1.15)
p-value for effect modification by
TPOAb status = 0.08
Confounding: Maternal age, race/ethnicity, smoking status, fish intake, parity, and gestational week at blood draw
Itoh et al. (2019, Japan
5915990)
Medium
2003-2005
Cross-sectional Pregnant
women and
their children
N = 499
Regression
coefficient per ln-
unit increase in
PFOA
Plasma Levels of
2.00 TSH (ln-
|iU/mL).
FT3 (ln-pg/mL),
FT4 (ln-pg/mL),
TPOAb (ln-
IU/mL),
TgAb (ln-
IU/mL)
Confounding: Age at delivery, parity, pre-pregnancy BMI, educational level, alcohol consumption, and smoking habits
TPOAb: -0.23 (-0.44, -0.02),
p-value = 0.033
TgAb:-0.01 (-0.21, 0.19),
p-value = 0.929
Aimuzi et al. Shanghai,
(2020,6512125) China
Cross-sectional
Medium
2013-2016
Pregnant
women prior to
16 weeks of
gestation
N = 1877
1615 TPOAb-
negative
222 TPOAb-
positive
Serum
Total cohort:
12.32
TPOAb-
negative: 12.32
TPOAb-
positive: 12.3
Levels of
TSH (ln-mlU/L),
FT3 (pmol/L),
FT4 (pmol/L)
Regression
coefficient per ln-
unit increase in
PFOA
FT4
Total cohort: 0.12 (0.02, 0.23)
TPOAb-negative: 0.11 (-0.01,
0.22)
TPOAb-positive: 0.14 (-0.20,
0.48)
TSH, FT3: All associations not
statistically significant
Confounding: Pre-pregnancy BMI, gestational age at thyroid hormone (TH) measurement, fish intake, maternal age, hospital indicators,
maternal education, difference between PFAS and THs measured gestational weeks
Notes: BMI = body mass index; FT3 = free triiodothyronine; FT4 = free thyroxine; GF = glomerular filtration; GFR = glomerular filtration rate; POI = premature ovarian
insufficiency; T3 = triiodothyronine; T4 = thyroxine; TgAb = thyroglobulin antibody; TGN = thyroglobulin; TPOAb = thyroid peroxidase antibody; TSH = thyroid stimulating
hormone; TT3 = total triiodothyronine; TT4 = total thyroxine.
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-186
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.7 Metabolic/Systemic
Table D-16. Associations Between PFOA Exposure and Metabolic Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Exposure
Population, Matrfx Outcome
Ages, N
Levels3
Comparison
Resultsb
Children and Adolescents
Ashley-Martin
Canada, Cohort Pregnant Maternal blood Adiponectin,
Regression
Adiponectin, leptin: No statistically
etal. (2017,
Recruitment women and 1.7 leptin
coefficient per
significant associations
3981371)
2008-2011 their children,
loglO-unit
High
from the
increase in
MIREC Study
PFOA
N = 1,176
Confounding: Maternal age, pre-pregnancy body mass index, sex, and parity0
Buck et al.
United States, Cohort Pregnant Maternal serum Adiponectin,
Percent change
Adiponectin, leptin: No statistically
(2019,5080288) 2003-2006 women and 5.6 leptin
per doubling of
significant associations
High
their children in
PFOA
the HOME
study
N = 230
Confounding: Maternal age, race, education, income, parity, maternal body mass index, serum cotinine, delivery mode, and infant sex
Chen et al.
China, Cohort Infants followed Cord blood BMI, WC, body Regression
BMI, WC, body fat, waist to height
(2019,5080578) 2012-2017 up at age 5, 6.74 fat, waist-to-
coefficient per
ratio: No statistically significant
High
N = 404 height ratio
ln-unit increase
association
in PFOA, or by
tertile
Confounding: Maternal age, maternal pre-pregnancy BMI, gestational week at delivery, maternal education, paternal smoking during
pregnancy, and parity
Jensen et al.
Denmark, Cohort Pregnant Maternal serum BMI z-score,
Regression
BMI z-score, WC: No statistically
(2020,6833719) 2010-2012 women and 1.62 WC
coefficient per
significant associations
High
their infants
unit increase in
assessed at
PFOA
birth, 3 months,
and 18 months,
Odense Child
Cohort
N = 593
D-187
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
Confounding: Maternal age, parity, pre-pregnancy BMI, pre-pregnancy BMI2, education, smoking, sex, visit, adiposity marker at birth
Minatoya et al.
(2017, 3981691)
High
Japan,
2002-2005
Cohort
Pregnant
women and
their children
N = 168
Serum
1.4
Adiponectin,
leptin
Regression
coefficient per
loglO-unit
increase in
maternal serum
PFOA
Confounding: Maternal BMI, parity, smoking during pregnancy, blood sampling period, gestational age, infant sex
Adiponectin, leptin: No statistically
significant associations
Alderete et al.
(2019, Medium
5080614)
United States
2001-2012
Cohort
Obese Hispanic
children, 8-
14 years
N= 39
Plasma
GM = 2.78
Regression
coefficient per
ln-unit increase
in PFOA
Glucose
(fasting, 2 hour.
AUC), Insulin
(fasting, 2 hour.
AUC), HOMA-
IR
Confounding: sex, baseline social position (categorical), baseline outcome, baseline and change in age at
(categorical), baseline and change in body fat percent at follow-up.
Glucose (2 hour): 30.6 (8.8, 52.4),
p-value < 0.05
Glucose (fasting, AUC), insulin,
HOMA-IR: No statistically
significant association
follow-up, pubertal status
Braun et al.
(2016, 3859836)
Medium
United States,
recruitment
2003-2006
Cohort
Pregnant
women and
their children
N = 285
Serum
5.3
Overweight/obe BMI z-score:
se, BMI z-score, Regression
WC, body fat coefficient by
percentage, Terciles
weight-for-age
Other outcomes:
Mean change
between 2 and
8 years by
tercile
BMI z-score: 0.44 (0.13, 0.74) T2:
0.44 (0.23, 0.64)
T3: 0.37 (0.14, 0.6)
WC: 4.3 (1.7, 6.9)
Body fat percent: 3.6 (1.8, 5.5)
Weight-for-age
T2: 0.49 (0.31,0.67)
T3: 0.43 (0.23,0.64)
Overweight/obese: No statistically
significant association
Results: Lowest tercile used as the reference group. Tercile 1 (0.5-4.3 ng/mL), tercile 2 (4.4-6.7 ng/mL), tercile 3 (6.8-26 ng/mL) maternal
PFOA.
Confounding: Maternal age, race, education, income, parity, employment, marital status, depressive symptoms, BMI at 16 weeks gestation,
fruit/vegetable consumption, fish consumption, prenatal vitamin use, maternal serum cotinine concentrations, child age in months.
D-188
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
Conway et al.
(2016, 3859824)
Medium
United States, Cross-Sectional Children living Serum
2005-2006 in six PFOA-
contaminated
water districts
with type 1
diabetes
N = 39
Confounding: Age, sex, race, BMI, eGFR, hemoglobin, iron
Mean = 68.4 ng/ uncategorized
L diabetes
T1D, T2D, and OR per ln-unit T1D: 0.52 (0.54, 0.97)
increase in
PFOA T2D and uncategorized diabetes:
No statistically significant
association
Domazet et al.
(2016, 3981435)
Medium
Denmark,
1997-2009
Cohort
Children
Blood, plasma,
WC (cm),
Percent change
followed
glucose
HOMA-B,
in WC at
through ages 9,
Age 15
HOMA-IR,
21 years old in
15, and 21,
Males: 9.7
insulin, glucose,
higher levels of
II
z;
Females: 9.0
skinfold
PFOA at age 21
Age 21
thickness, BMI
Males: 3.1
Percent change
Females: 2.7
in HOMA-B at
age 15 per 10-
unit increase in
PFOA exposure
at age 9
WC:-11.11 (-19.90,
p-value = 0.03
1.36),
HOMA-B: -10.93 (-19.67, -1.11)
HOMA-IR, insulin, glucose,
skinfold thickness, BMI: No
statistically significant association
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
the European Boys: 9.5
Youth Heart Girls: 9.5
Study aged
9 years
N = 242
Confounding: age, sex, parity, maternal income level
Leptin, fat mass, Percent change Body fat:-1.22 (-2.91, 0.5),
adiponectin per 10% p-value = 0.161
increase in Adiponectin: 1.7 (-0.15, 3.59),
PFOA p-value = 0.071
Leptin: -4.44 (-8.74, 0.06),
p-value = 0.053
Gyllenhammar
Sweden, Cohort
Mothers and Maternal serum BMI z-score
Regression
BMI z-score:
et al. (2018,
1996-2011,
their children 2.3
coefficient per
4238300)
children
from the
IQR increase in
Ages 36 and 48 months: Positive
Medium
followed up at
POPUP Study
maternal PFOA
statistically significant associations.
age 5
N = 193
D-189
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
Age 60 months: Non-significant
positive association (numeric
results not provided)
Confounding: Sampling year, maternal age, pre pregnancy BMI, maternal weight gain during pregnancy, maternal weight loss after delivery,
years of education, and total time of breastfeeding
Hartman, (2017,
3859812)
Medium
United
Kingdom,
recruitment
1991-1992
Cohort
Pregnant
women and
their daughters
N = 319
Maternal serum
3.7
WC (cm), Trunk Regression
fat (%), coefficient per
BMI (kg/m2) unit increase in
PFOA
WC: -0.54 (-0.9,0.11),
p-value = 0.01
Trunk fat:
-0.27 (-0.55, 0.0), p-value = 0.05
BMI:
-0.16 (-0.32, 0.0), p-value = 0.05
Body fat percentage: No
statistically significant associations
Confounding: sampling design, pre-pregnancy BMI (kg/m2) and maternal educational status
Kang et al. Korea,
(2018, 4937567) 2012-2014
Medium
Cross-sectional
Plasma
5.68
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
Fasting blood Regression
glucose (mg/dL) coefficient per
ln-unit increase
in PFOA
Blood glucose:
1.262 (-1.108,3.633),
p-value = 0.294
Kobayashi et al. Japan,
(2017, 3981430) 2002-2005
Medium
Cross-sectional Infants from Maternal serum Ponderal index
Hokkaido Study 1.4 at birth
on Environment
and Children's
Health
N = 177
Regression Ponderal index:
coefficient per -0.44 (-0.99, 0.12),
ln-unit increase p-value = 0.123
in PFOA
Confounding: Maternal age, pre-pregnancy BMI, parity, maternal education, maternal smoking during pregnancy, gestational age, infant sex,
and maternal blood sampling period
Karlsen et al. Faroe Islands,
(2017,3858520) recruited 2007-
Medium 2009 (at birth)
Cohort
Mother-child
pairs
N = 444
Serum
BMI z-score,
Overweight
Regression
coefficient or
RR per log 10-
BMI z-score at age 5:
-0.27 (-0.52, -0.02),
p-value < 0.05
D-190
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
follow up at
child ages
18 months and
5 years
Maternal 2-
week serum:
1.40
Child 5-year
serum: 2.20
unit increase in
child or
maternal PFOA,
or by tertiles
Overweight at age 5:
RR: 1.5 (1.01, 2.24), p-value < 0.05
T3: 1.88 (1.05,3.35),
p-value < 0.05
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
Lauritzen et al.
(2018, 4217244)
Medium
BMI, triceps skin fold, subscapular
skinfold, overweight: No
statistically significant associations
Norway and Cohort Pregnant Serum BMI, triceps Regression
Sweden, women and Norway: 1.64 skinfold, coefficient or
Recruitment their children at Sweden: 2.33 subscapular ORperln-unit
1986-1988 5-year follow up skinfold, increase in
N = 412 overweight maternal PFOA
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
et al. (2016,
3859832)
Medium
United States, Cohort Children ages Serum
2005-2006 6-9 years Girls: 30.1
N= 1123 (girls) Boys: 34.8
N= 1169 (boys)
Results: Lowest quartile used as the reference group.
Confounding: age and month of sampling
Insulin-like Percent
growth factor-1 difference by
(IGF-1) (In- quartiles.
ng/mL)
IGF-1
Girls:
Q3: -3.6 (-6.6, -0.5)
Boys
Q3: -7.4 (-12.8,-1.6)
No other statistically significant
associations
Manzano- Spain,
Salgado et al. Recruitment
(2017,4238509) 2003-2008
Medium
Cohort Mother-child Maternal blood BMI, WC,
pairs, followed GM = 2.32 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
PFOA
Martinsson et al. Sweden,
(2020,6311645) 2003-2008
Medium
Case-control
Pregnant
women and
their children at
Serum
3.1
Overweight OR by quartiles
OW: No statistically significant
associations
D-191
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
age 4, Southern
Sweden
Maternity
Cohort
N = 1,048
Results: Lowest quartile used as reference.
Confounding: Risk strata, difference from strata-specific mean, sex
Mora et al.
(2017, 3859823)
Medium
United States,
1999-2002
Cohort
Early childhood
N = 992
Maternal WC (cm), Regression WC
Plasma Skinfold coefficient per All: 0.31 (0.04, 0.57)
5.6 thickness, BMI, IQR-unit Boys: 0.5 (0.06,0.93)
increase in
PFOA Skinfold thickness, BMI, waist-to-
hip ratio, obesity, overweight, total
fat mass index, total fat-free mass
index: No statistically significant
association
Confounding: maternal age, race/ethnicity, education, parity, pre-pregnancy BMI, timing of blood draw, household income, child sex, age at
outcome assessment
WC (cm),
Skinfold
thickness, BMI,
waist-to-hip
ratio, obesity,
overweight,
total fat mass
index, total fat-
free mass index
Pinney et al. Greater
(2019, 6315819) Cincinnati and
Cohort Girls, age 6-8 Serum
N = 667 6.4
Medium
the San
Francisco Bay
Area,
Recruitment
2004-2007,
followed
annually or
semi-annually
until 2014
BMI, waist-
height ratio,
waist-hip ratio
Regression
coefficient by
quintiles or per
ln-unit increase
in PFOA
BMI:
Quintile 4 vs. Quintile 1: -0.248
(-0.489, 0.007), p-value = 0.044
Quintile 5 vs. Quintile 1: -0.436
(-0.685, 0.187), p-value = 0.001
Per ln-unit increase -0.264
(-0.416, 0.112), p-value = 0.001
Waist-height
Per ln-unit increase: -0.009
(-0.017, 0.002), p-value = 0.013
Waist-hip ratio: No statistically
significant association
Results: Lowest quintile used as the reference group.
D-192
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Exposure
Location, Design Population, Matrix, Outcome Comparison
Years Ages, N
Levels3
Resultsb
Confounding (BMI): Race, parental education, average kcal, physical activity
Confounding (Waist-height ratio): Age at exam, race, parental education, average kcal, physical activity
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: -0.19 (-0.5,0.12)
(BMIZ), height- coefficient per T2:-0.3 (-0.6, 0.01)
GM=1.95(SE for-age z-score ln-unit increase T3:-0.15 (-0.49, 0.2)
= 0.08) (HAZ), weight- inPFOAandby Females:-0.45 (-1, 0.1)
for-age z-score tertiles T2: -0.2 (-0.68, 0.29)
(WAZ) T3: -0.31 (-0.9, 0.28)
Males: -0.02 (-0.35, 0.3)
T2: -0.38 (-0.7, -0.05)
T3:-0.07 (-0.5, 0.37)
HAZ:-0.31 (-0.67,0.04)
T2: -0.17 (-0.38,0.03)
T3: -0.28 (-0.65,0.08)
Females: -0.36 (-0.87, 0.14)
T2: -0.25 (-0.45, -0.05)
T3:-0.35 (-0.88, 0.17)
Males: -0.28 (-0.7, 0.14)
T2:-0.2 (-0.53, 0.13)
T3: -0.23 (-0.64, 0.19)
WAZ: -0.34 (-0.68, -0.01)
T2: -0.33 (-0.63, -0.04)
T3: -0.28 (-0.65,0.08)
Females:-0.53 (-1.18,0.12)
T2:-0.28 (-0.73, 0.16)
T3: -0.43 (-1.08, 0.23)
Males: -0.22 (-0.51, 0.08)
T2: -0.42 (-0.77, -0.07)
T3: -0.21 (-0.56, 0.15)
No statistically significant
associations trends by sex
NHANES = National Health and Nutrition Examination Survey
Results: Lowest tertile used as reference
D-193
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
Confounding: Age, quadratic age, race/ethnicity, poverty income ratio, serum cotinine, birthweight, maternal smoking during pregnancy,
hematocrit, sex
Fleisch et al.
(2017, 3858513)
Medium for
metabolic
function
Low for
HOMA-IR
United States,
Pregnant
women
recruited 1999-
2002, outcome
assessed at mid-
childhood
follow-up
Cohort
Mid-childhood,
7.7 years
N = 584
Plasma
GM = 4.2
Leptin,
Adiponectin,
HOMA-IR
Percent change
by quartiles
Leptin
Q3: -23.3 (-37,-6.5)
Q4:-20.1 (-35.1,-1.6)
Adiponectin
Q2: 16.3 (1.8,32.9)
Q3: 22.7 (6.9, 40.8)
HOMA-IR: No statistically
significant association
Results: Lowest quartile used as reference
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
Mitro et al.
(2020, 6833625)
High
United States
Recruitment
1999-2002
Cohort
Pregnant
women
N = 786
Plasma
5.6
WC(cm), BMI, Percent change WC:
Adiponectin, (%) or 1.1% (0.1, 2.2), p-value < 0.05
skinfold Regression
thickness, arm coefficient per BMI:
circumference, log2-unit 0.3 (0.0, 0.6), p-value < 0.05
leptin increase in
PFOA Adiponectin, skinfold thickness,
arm circumference, hemoglobin,
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, Cohort Pregnant
1999-2002 women from the
Project Viva
cohort
N = 1533
Plasma
5.9
GDM, impaired
glucose
tolerance,
isolated
hyperglycemia,
blood glucose
levels
Regression Gestational diabetes, impaired
coefficient by glucose tolerance, isolated
quartiles hyperglycemia, blood glucose
levels: No statistically significant
OR by quartiles association
Confounding: Maternal age, pre-pregnancy BMI, prior history of gestational diabetes/parity, race/ethnicity, smoking, and education
D-194
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
Starling et al.
(2017, 3858473)
High
United States,
2009-2014
Cohort
Pregnant
women and
their children
N = 628
Maternal serum
1.1
Maternal
glucose
(ln(mg/dl))
Regression
coefficient by
tertiles
Maternal glucose:
T3: -0.025 (-0.046,0.004)
Confounding: Maternal age, pre-pregnancy body mass index (BMI).
gestational age at blood draw
Maternal glucose (continuous) and
T2: No statistically significant
association
race/ethnicity, education, smoking during pregnancy, gravidity, and
Ashley-Martin
Canada, Cohort
Pregnant
Serum
GWG (kg)
Regression
No statistically significant
et al. (2016,
Pregnant
women from
15.2
coefficient per
associations
3859831)
women
MIREC
log2-unit
Medium
recruited 2008-
N = 1,609
increase in
2011, outcome
PFOA
assessed at birth
Confounding: Age, income, parity
Jaacks et al.
United States, Cohort
Pregnant
Serum
GWG (kg)
Regression
GWG
(2016, 3981711) 2005-2007
women
Mean= 3.66
coefficient and
0.09 (-0.84 1.02)
Medium
N = 218
OR per SD-unit
OR for excessive GWG:
increase in
1.06 (0.76, 1.47)
PFOA
Confounding: Pre-pregnancy non-fasting serum lipids, BMI
Jensen et al.
Denmark, Cohort
Pregnant
Serum
Blood glucose,
Percent change
No statistically significant
(2018, 4354143) recruitment
women, Odense
1.67
insulin, c-
per log2-unit
associations
Medium
2010-2012,
Child Cohort
peptide, 2-hour
increase in
outcome
N = 158
glucose, insulin
PFOA
assessed 12-
resistance, beta
20 weeks later
cell function,
insulin
sensitivity
Confounding: Age, parity, education level, pre-pregnancy BMI
Liu et al. (2019,
China, Case-control
Pregnant
Serum
GDM, glucose
Regression
GDM:
5881135)
2013-2015
women without
2.25
homeostasis
coefficient per
Per ln-unit increase sum m-PFOA:
Medium
history or
ln-unit increase,
1.23 (0.92, 1.64)
family history
or by tertiles of
T2: 0.91 (0.4, 2.07)
of diabetes
T3: 2.01 (0.92,4.37)
D-195
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
N = 189
sum m-PFOA
or L-PFOA
Per ln-unit increase sum m-PFOA:
2.04 (0.99,4.21)
T2: 1.04 (0.47,2.34)
T3: 2.04 (0.94,4.46)
Per ln-unit increase sum L-PFOA:
Glucose homeostasis (1 hour): 0.55
(0.01, 1.1), p-value = 0.049
Glucose homeostasis (2 hours):
0.73 (0.27, 1.18), p-value = 0.002
Glucose homeostasis (fasting,
1 hour, 2 hour) for m-PFOA and
glucose homeostasis (fasting) for L-
PFOA: No statistically significant
association
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 PFOA
GWG: No statistically significant
associations
Serum
Mothers of
sons: 3.0
Mothers of
daughters: 3.7
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 = 1.99
GDM
Risk Ratio per
SD-unit
increase in
PFOA
GDM (family history of T2D): 1.27
(1.11, 1.45)
Overall cohort, no family history of
T2D, normal pre-pregnancy BMI,
overweight pre-pregnancy BMI: No
statistically significant association
Confounding: Maternal age, enrollment BMI, education, parity, race/ethnicity, serum cotinine
D-196
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
Ren et al. (2020,
6833646)
Medium
China, 2012
Cross-sectional
Pregnant
women enrolled
in the
Shanghai-
Minhang Birth
Cohort Study
N = 705
Plasma Glucose (1 hour, Regression
20.2 fasting) coefficient per
ln-unit increase
in PFOA
Glucose (1 hour tolerance test):
0.31 (0.03, 0.52), p-value = 0.031
Glucose after fasting, glucose after
1 hour tolerance test by gestational
weeks: No statistically significant
association
Confounding:
complication.
maternal age at enrollment, pre-pregnancy BMI, per capita household income, education level, passive smoking, pregnancy
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= 1.68
Gestational
impaired
glucose
tolerance
GM = 1.70
Women with
GDM
GM = 1.64
Confounding: Maternal age, race, pre-pregnancy BMI, and education
GDM,
gestational
impaired
glucose
tolerance
OR per quartile Gestational diabetes, gestational
PFOA impaired glucose tolerance: No
statistically significant association
Valvi et al.
(2017, 3983872)
Medium
Faroe Islands,
1997-2000
Cohort
Pregnant
women and
their children
N = 604
Maternal serum
3.31
Gestational
diabetes
OR per
doubling of
PFOA, or by
tertiles
Gestational diabetes:
Per doubling:
0.79 (0.44, 1.41)
T2: 1.01 (0.5, 2.06)
T3: 0.66 (0.3, 1.48)
Results: Lowest tertile used as the reference group.
Confounding: Maternal age at delivery, education, parity, pre-pregnancy BMI, smoking during pregnancy
Wang et al. China
(2018, 5079666) 2013
Medium
Case-control
Pregnant
women with
(cases) and
without
(controls) GDM
Serum Fasting blood Fasting blood Fasting blood glucose, GDM: No
Cases: 1.38 glucose, GDM glucose: OR by statistically significant associations
Controls: 1.30 tertiles of n-
PFOA
D-197
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location, . Population,
Confidence Years esign Ages, N
Exposure
Matrix, Outcome Comparison Resultsb
Levels3
N = 242 GDM: OR per
unit increase in
n-PFOA
Confounding: Fasting blood glucose: BMI, age, GDM status; GDM: BMI, GWG, ethnic groups, maternal education, parity, maternal
drinking during pregnancy, household income
Wangetal. China, Cohort Pregnant Serum Fasting blood LSMby tertiles No statistically significant
(2018,5080352) 2013-2014 women aged 7.3 glucose, fasting associations
Medium 20-40 insulin, HOMA-
N = 385 IR, gestational
diabetes, oral
glucose
tolerance
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
Gestational
OR per unit
Gestational diabetes mellitus
6833677)
2017-2019
control women
Cases: 8.19
diabetes
increase in
Q2: 1.05 (0.45,2.04)
Medium
N = 165 cases,
Controls: 7.91
mellitus
PFOA; OR per
Q3: 1.12(0.46,2.20)
330 controls
loglO-unit
Q4: 1.20 (0.28,2.21)
increase in
p-trend = 0.60
PFOA
log-PFOA: 1.51 (0.63, 3.84), p-
value = 0.33
Confounding: Maternal age, sampling time, parity,
BMI, educational level, and serum lipids
General Population
Cardenas et al.
United States,
Cohort Adults at high
Plasma
Adiponectin
Regression
Adiponectin: -0.29 (-0.54, -0.04),
(2017, 4167229) Recruitment
risk of Type-2
GM = 4.82
(Hg/mL),
coefficient per
p-value = 0.02
High
July 1996-May
diabetes
HbAlc (%),
doubling of
1999, outcome
N = 956
Insulin (fasting)
PFOA
HbAlc: 0.04 (0.001, 0.07),
assessed
(|iU/mL).
p-value = 0.05
annually until
Glucose
May 2001
(fasting)
Insulin (fasting): 2.26 (1.16, 3.35),
(|iU/mL).
p-value = 0.000056
HOMA-IR,
Insulin (30 min,
Glucose (fasting): 0.66 (0.07, 1.24),
|iU/mL).
p-value = 0.03
D-198
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
Proinsulin
(fasting, pM),
HOMA-B
(beta), Insulin
(corrected
response),
Insulinogenic
index, Diabetes,
HOMA-IR,
glucose (30
mins), glucose
(2 hours), BMI
HOMA-IR: 0.64 (0.34, 0.94),
p-value = 0.000031
Insulin (30 min): 7.85 (3.63, 12.07),
p-value = 0.00028
Proinsulin (fasting): 1.17 (0.72,
2.71), p-value = 0.00070
HOMA-B: 15.93 (6.78, 25.08),
p-value = 0.00066
Insulin (corrected): 0.04 (0.01,
0.07), p-value = 0.01
Insulinogenic index: 0.08 (0.01,
0.15), p-value = 0.02
Diabetes, HOMA-IR, glucose (30
mins), glucose (2 hours), BMI: No
statistically significant association
Confounding: Sex, race/ethnicity, BMI, age, marital status, education, smoking history.
Blake et al. United States
(2018, 5080657) 1991-2008
Medium
Cohort
Serum
12.7
BMI
Adults living in
a community
with water
supply from a
PFAS-
contaminated
aquifer
N = 192
Confounding: Age, year of measurement, sex, education, income, marital status, and BMI
Percent change
per IQR
increase in
PFOA
BMI: No statistically significant
associations
Cardenas et al. United States,
(2019, 5381549) 1996-2014
Medium
Controlled trial
Adults older
than 25 without
diabetes and
Plasma
GM = 4.82
T2D
Hazard ratio per Diabetes:
log2-unit HR: 1.05 (0.94, 1.18)
increase in
D-199
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
with elevated
fasting and
postload
glucose,
Diabetes
Prevention
Program
N = 956
baseline PFOA
and by PFOA
tertiles
T2: 0.94 (0.75, 1.17)
T3: 0.94 (0.75, 1.18)
Results: Lowest tertiles used as the reference group.
Confounding: Sex, race/ethnicity, baseline age, marital status, education, income, smoking history, BMI, maternal diabetes, paternal diabetes,
treatment assignment
Christensen et
al. (2019,
5080398)
Medium
United States,
2007-2014
Cross-sectional
Adults from
NHANES age
20+
N = 2,975
Serum Elevated waist
2.8 circumference
(Males: > 102
cm. Females:
> 88 cm),
metabolic
syndrome,
glucose
Confounding: PFDE, PFOS, PFHxS, MP AH, PFNA, PFUnDA, survey cycle, age, sex,
OR by quartiles
WC
Q2: 0.66 (0.46, 0.92),
p-value < 0.05
Q3: 0.62 (0.39,0.98),
p-value < 0.05
Metabolic syndrome, glucose level:
No statistically significant
association
race/ethnicity, family income, alcohol intake, and
smoking status
Conway et al.
(2016, 3859824)
Medium
United States,
2005-2006
Cross-sectional Adults working Serum
All participants
mean = 68.4
or living in six
PFOA-
contaminated
water districts
with diabetes
N = 6,460
Confounding: Age, sex, race, BMI, eGFR, hemoglobin, iron
T1D, ORperln-unit All
T2D, increase in T1D: 0.76 (0.71,0.8)
Uncategorized PFOA T2D: 0.94 (0.92, 0.97)
Diabetes Uncategorized DM: 0.94 (0.9, 0.99)
Adults
Type 1 DM: 0.74 (0.7, 0.79)
Type 2 DM: 0.91 (0.89,0.94)
Uncategorized DM: 0.92 (0.88,
0.96)
D-200
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
Donat-Vargas et Sweden, 1990- Case-control Adults with
al. (2019, 2003,2001- (cases) and
5083542) 2012 without
Medium
Plasma
Cases: 2.8
Controls: 3.0
Type 2
Diabetes,
HOMA-IR,
HOMA-Beta
OR per 1-loglO T2D: 0.65 (0.43, 0.97)
SD increase in
baseline PFOA
(controls) type-2
diabetes living
in Sweden
N = 248
Confounding: gender, age, sample year, red and processed meat intake, fish intake, BMI
HOMA-IR, HOMA-Beta: No
statistically significant association
Duan et al. China, 2017
(2020, 5918597)
Medium
Confounding:
diabetes
Cross-sectional
Adults, 19 to
87 years old
N = 252
Serum
14.83
Fasting glucose
(nmol/L),
HbAlc
Regression
coefficient per
1% increase in
PFOA
Glucose (fasting): 0.018 (0.004,
0.033), p-value = 0.014
HbAlc: No statistically significant
association
sex, age, body mass index, smoking and alcohol-drinking status, exercising status, education level, and family history of
Jain et al. (2019, United States, Cross-sectional Adults from Serum
Obesity
Comparison of Obesity: p-value = 0.02
5080621)
Medium
2011-2014
NHANES, age
20+
N = 2,883
GM = 2.2 (non-
obese); 2.0
(obese)
geometric mean
PFOA levels
non-obese vs.
obese
Confounding: Sex, nice, age, poverty income ratio, physical activity, BMI, and serum cotininc
Jeddy et al. England,
(2018, 5079850) mothers
Medium recruited 1991—
1002, outcome
assessed at age
17
Nested case- Pregnant Maternal serum Fat mass Regression
control studies mothers and 3.8 coefficient per
their 17-year old unit increase in
daughters, PFOA
ALSPAC
N = 221
105.88 (-621.59, 833.34)
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,
Controlled Trial Overweight and Plasma, glucose Leptin, HOMA- Partial
obese patients Males: 27.2 IR, insulin, Spearman
from the Females: 22.3 resting correlation
POUNDS metabolic rate, coefficient with
Spearman correlations
Leptin: 0.09, p-value < 0.05
HOMA-IR: 0.1, p-value < 0.05
change
Louisiana,
2004-2007
LOST Trial
body weight, baseline PFOA Resting metabolic rate, body
HbAlc, glucose,
weight, HbAlc, glucose, VAT fat
D-201
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
Uninformative
for insulin
resistance
Liu et al. (2018,
4238514)
Medium
, Ages 30-70
years
N = 621
VAT fat mass,
whole body fat,
BMI, waist
circumference
Regression
coefficient
loglO-unit
increase in
PFOA, or
by tertile
mass, whole body fat, BMI, waist
circumference: No statistically
significant association
Confounding: age, sex, race, education, smoking status, alcohol consumption, physical activity, menopausal status (women only), hormone
replacement therapy (women only), and dietary intervention groups.
United States,
2013-2014
Cross-sectional
Adults from
NHANES
N = 1,871
Serum
GM = 1.86
Fasting blood
glucose, 2-hour
glucose, HbAlc,
insulin levels,
HOMA-IR, beta
cell function,
metabolic
syndrome, WC
Regression HbAlc:-0.12 (0.05),
coefficient per p-value < 0.05
ln-unit increase Beta cell function: 0.12 (0.05);
in PFOA p-value < 0.05
Fasting blood glucose, 2-hour
glucose, insulin levels, HOMA-IR,
metabolic syndrome, WC: No
statistically significant associations
Results: Effect estimates are reported with SE in parentheses
Confounding: Age, gender, ethnicity, smoking status, alcohol intake, household income, WC, and medications (anti-hypertensive, anti-
hyperglycemic, and anti-hyperlipidemic agents)
Mancini et al.
(2018, 5079710)
Medium
France,
1990-2012
Cohort
Women, 40-60
N = 71,294
Dietary T2D
Mean =0.86 ng/
kg body
weight/day
Hazard ratio by
deciles
T2D
Decile 4: 1.21 (1.06,
p-value < 0.05
Decile 5: 1.35(1.15,
p-value < 0.05
Decile 6: 1.19(1.05,
p-value < 0.05
Results: Lowest decile used as the reference group.
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
1.46),
1.59),
1.41),
Su et al. (2016,
Taiwan,
Cross-sectional Adults aged 20- Plasma
Diabetes,
OR by quartiles, Diabetes:
3860116)
2009-2011
60 living in 8.0
Fasting blood
per doubling of Q2: 0.39 (0.16, 0.96)
Medium
Taiwan
glucose
PFOA Q3: 0.2 (0.07, 0.58)
N = 571
(ng/mL),
Q4: 0.16 (0.05, 0.5)
Total: 0.56 (0.43, 0.75)
D-202
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
blood glucose Geometric mean
(120 mins) (In)
(ng/mL),
glucose AUC
(ng/mL),
HbAlc (In) (%)
ratio (GMR) by
quartiles, or per
doubling of
PFOA
Glucose (Fasting):
Q2: 0.96 (0.93,0.99)
Q3: 0.95 (0.92,0.97)
Q4: 0.95 (0.92, 0.98)
Per doubling PFOA: 0.98 (0.97,
0.99)
Glucose (120 min)
Q2: 0.87 (0.82, 0.94)
Q3: 0.9 (0.94, 0.95)
Q4: 0.85 (0.79,0.91)
Per doubling PFOA: 0.96 (0.94,
0.98)
Glucose AUC:
Q2: 0.9 (0.85, 0.95)
Q3: 0.9 (0.86, 0.95)
Q4: 0.88 (0.84,0.93)
Per doubling PFOA: 0.97 (0.95,
0.99)
HbAlc:
Q2: 0.98 (0.96, 1.0)
Q4: 0.97 (0.95, 1.0)
Per doubling PFOA: 0.99 (0.98,
1-0)
Results: Lowest quartile used as reference group.
Confounding (Diabetes): age, sex, education, smoking (ever vs. never), alcohol (ever vs. never), BMI, hypertension, total cholesterol, regular
exercise
Confounding (Other): age, sex, education, smoking, alcohol, BMI, hypertension, total cholesterol, regular exercise
Sun etal. (2018, United States, Case-control
Female nurses
Plasma
Type 2
Regression
T2D
4241053) 1989-201ld
drawn from the
Cases: 4.96
Diabetes,
coefficient per
Per increase:
Medium
Nurses' Health
Controls: 4.57
HbAlc, fasting
SD loglO-unit
1.24 (1.06, 1.45), p-value = 0.009
Study II cohort
insulin,
increase in
study
adiponectin
PFOA
OR for T3: 1.54 (1.04,2.28)
D-203
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
N = 1586
OR by tertiles
HbAlc, fasting insulin,
adiponectin: No statistically
significant association
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.
Chen et al. Croatia
Cross-sectional Residents of
Plasma
BMI, fasting
Metabolic
(2019, 5387400) 2007-2008
Hvar ages 44-
GM = 2.87
insulin
syndrome: OR
Medium for
56 years
(Range: 1.03—
(nIU/mL),
per ln-unit
metabolic
N = 122
8.02)
fasting plasma
increase in
syndrome
glucose
PFOA
Low for all
(mmol/L),
other outcomes
glycated HbAlc
All other
(%), hip
outcomes:
circumference
regression
(cm),
coefficient per
homeostatic
ln-unit increase
model
in PFOA
assessment of
beta-cell
function
(HOMA-(3),
homeostatic
model
assessment of
insulin
resistance
(HOMA-IR),
metabolic
syndrome
defined by the
ATP III criteria,
waist
circumference
(cm)
Metabolic syndrome: 2.19 (0.88,
4.42); p-value = 0.09
All other outcomes: No statistically
significant associations
D-204
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages, N
Exposure
Matrix,
Levels3
Outcome Comparison
Resultsb
Confounding: Age, sex, education, socioeconomic status, smoking, dietary pattern, and physical activity
Occupational Populations
Steenland et al. United States Retrospective
(2013, 1937218) 2005-2006 Occupational
Medium Cohort
Adult residents
and workers
from the C8
Health Project
N= 32,254
Serum
26
Type 1 diabetes,
with and
without a 10-
year lag
RR by quartiles T1D, validated and self-reported
No lag: No statistically significant
associations or trends by quartiles
With lag
Q2: 0.42 (0.09, 2.00)
Q3: 0.70 (0.14,0.35)
Q4: 0.38 (0.08, 1.93)
p-trend = 0.84
T1D, validated cases only: No
statistically significant associations
or trends by quartiles
Confounding: Sex, race/ethnicity, smoking, BMI, alcohol consumption
Notes: AUC = area under the curve; BMI = body mass index; DM = diabetes mellitus; EYHS = European Youth Heart Study; HbAlc = hemoglobin Ale; HOMA = homeostatic
model assessment; HOME = Health Outcomes and Measures of the Environment; GDM = gestational diabetes mellitus; GM = geometric mean; GWG = gestational weight gain;
IGF = insulin-like growth factor; IR = insulin resistance; IQR = interquartile range; 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; RR = risk ratio; POPUP = Persistent Organic Pollutants in
Uppsala Primiparas; 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 Recruitment 1989, blood sample collection 1995-2000, outcome assessed during biennial follow up through June 2011.
D.8 Nervous
Table D-17. Associations Between PFOA Exposure and Neurological Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
Children and Adolescents
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
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: 5.6
(4.1-7.7)
Child: 4.4 (3.1-
6.0)
Both age groups:
Wide Range
Assessment of
Visual Motor
Abilities
(WRAVMA)
score
Early childhood
only: Peabody
Picture
Vocabulary Test
(PPVT-III) score
Mid-childhood
only: Kaufman
Brief Intelligence
Test Second
Edition (KBIT-2)
non-verbal and
verbal IQ,
(WRAML2)
design memory
and picture
memory
Mean Visual-motor
differences by Early childhood
quartiles of Q2: 1.0 (-1.0, 2.9)
PFOA exposure Q3: 0.5 (-1.6, 2.6)
Q4: 2.3 (0.1,4.5)
Mid-childhood (maternal plasma)
Mid-childhood (child plasma)
Q2
Q3
Q4
-4.1 (-8.0, -0.2)
-0.4 (-4.5, 3.7)
-6.1 (-10.5,-1.6)
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)
Mid-childhood (child plasma)
Q2
Q3
Q4
0.4 (-3.3,4.1)
-1.5 (-5.4, 2.3)
-3.2 (-7.4, 1.0)
Verbal IQ
Mid-childhood (maternal plasma)
Q2
Q3
Q4
-3.3 (-5.7, -1.0)
-2.7 (-5.2, -0.2)
-2.4 (-5.1, 0.2)
Mid-childhood (child plasma)
Q2
Q3
Q4
-1.0 (-3.9, 2.0)
-2.0 (-5.11, 1.1)
-2.8 (-6.2, -0.6)
Design memory
Mid-childhood (maternal plasma)
Q2: 0.2 (-0.3, 0.8)
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
Q3: 0.3 (-0.3,0.8)
Q4: 0.7 (0.1, 1.3)
Mid-childhood (child plasma)
Q2
Q3
Q4
0 (-0.6, 0.6)
-0.4 (-1.1, 0.2)
-0.4 (-1.1, 0.3)
Picture memory
Mid-childhood (maternal plasma)
Q2: -0.6 (-1.2, 0)
Q3: 0.1 (-0.5,0.7)
Q4:-0.1 (-0.7,0.5)
Mid-childhood (child plasma)
Q2: -0.3 (-1.0, 0.4)
Q3: 0.2 (-0.5, 1.0)
Q4: 0 (-0.8, 0.7)
PPVT-III: No statistically
significant associations
Niu et al. (2019, China,
5381527) Recruitment:
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
Cohort
High
2012; Follow-up
at age 4 years
Pregnant
women and
their children
from the
Shanghai-
Minhang Birth
Cohort
N = 533 (236
Females; 297
Males)
Maternal plasma ASQ-3 skill
19.9 (15.3-27.4) scales:
communication,
gross motor, fine
motor, problem
solving, personal-
social
RR per ln-unit Communication
increase in
PFOA and by
tertiles
0.84 (0.59, 1.19)
Females: 0.64 (0.34, 1.19)
T2: 0.86 (0.49, 1.50)
T3: 0.55 (0.28, 1.10)
p-trend <0.10
Males: 1.07 (0.70, 1.62)
T2: 1.02 (0.65, 1.6)
T3: 0.96 (0.61, 1.52)
p-value for interaction by
sex = 0.255
D-207
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
„ , Exposure
Population, M^trix
Ages, ' Outcome Comparison Resultsb
„ Levels
(ng/mL)a
Gross Motor
0.86 (0.47, 1.58)
Females: 2.31 (0.75,7.10)
T2: 1.08 (0.33, 3.57)
T3: 1.90 (0.66, 5.44)
Males: 0.47 (0.25, 0.89);
p-value <0.05
T2: 0.51 (0.23, 1.11)
T3: 0.45 (0.19, 1.04)
p-trend <0.10
p-value for interaction by
sex = 0.002
Fine Motor
0.99 (0.53, 1.84)
No statistically significant
associations, trends, or
interactions by sex
Problem Solving
1.26 (0.73,2.15)
No statistically significant
associations, trends, or
interactions by sex
Personal-Social Skills
1.67 (0.89,3.14)
Females: 9.00 (3.82,21.21);
p-value <0.05
Males: 1.03 (0.53,2.01)
T2: 1.60 (0.80, 3.19)
T3: 1.50 (0.77,2.93)
p-value for interaction by
sex = 0.002
Outcome: Neuropsychological problems defined as scores < 10th percentile.
Reference, Location, .
„ .. , Design
Confidence Years
D-208
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Results: Lowest tertile used as reference. For personal-social skills, no cases of neuropsychological problems were present among the lowest
tertile of PFOA exposure among girls; as a result, the Poisson regression model did not converge.
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 sex
Oulhote et al. Faroe Islands,
(2016, 3789517) Recruitment:
High 1997-2000,
Follow-up at
ages 5 and
7 years
Cohort
Children at
5 years
(n = 508) and
7 years
(n = 491)
Serum
Maternal: 3.34
(2.56-4.01)
5 years: 4.06
(3.33-4.98)
7 years: 4.37
(3.53-5.66)
Strengths and
Difficulties
Questionnaire
(SDQ) scores:
Total score
(hyperactivity/inatt
ention, conduct
problems, peer
relationship
problems,
emotional
symptoms),
prosocial behavior,
internalizing
problem,
externalizing
problems, autism
screening (peer-
problems minus
pro-social)
Mean difference
(autism,
internalizing,
externalizing,
total) or mean
ratio
(hyperactivity/in
attention,
conduct, peer
relationship,
emotional,
prosocial) per
doubling of
PFOA
SDQ total score
Prenatal exposure: -0.37 (-1.34,
0.61), p-value = 0.46
5-year serum: 1.03 (0.11, 1.95),
p-value = 0.03
7-year serum: 0.1 (-0.83, 1.03),
p-value = 0.84
Hyperactivity /Inattention
Prenatal exposure: 0.93 (0.76,
1.13), p-value = 0.43
5-year serum: 1.1 (0.91, 1.32),
p-value = 0.33
7-year serum: 0.97 (0.8, 1.16),
p-value = 0.71
Conduct
Prenatal exposure: 0.86 (0.71,
1.04), p-value = 0.12
5-year serum: 1.19(0.99, 1.44),
p-value = 0.06
7-year serum: 1.01 (0.84, 1.22),
p-value = 0.92
Peer Relationship
Prenatal exposure: 0.99 (0.71,
1.38), p-value = 0.96
5-year serum: 1.54 (1.16, 2.06),
p-value <0.01
7-year serum: 1.23 (0.92, 1.65),
p-value = 0.17
D-209
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Emotional
Prenatal exposure: 1.04 (0.84,
1.3), p-value = 0.7
5-year serum: 1.09 (0.88, 1.34),
p-value = 0.45
7-year serum: 0.98 (0.8, 1.21),
p-value = 0.85
Prosocial
Prenatal exposure: 1.02 (0.95,
1.1), p-value = 0.58
5-year serum: 0.97 (0.9, 1.04),
p-value = 0.4
7-year serum: 1 (0.93, 1.07),
p-value = 0.95
Internalizing
Prenatal exposure: 0 (-0.55,
0.55), p-value = 0.99
5-year serum: 0.59 (0.06, 1.13),
p-value = 0.03
7-year serum: 0.19 (-0.34, 0.72),
p-value = 0.49
Externalizing
Prenatal exposure: -0.37 (-0.99,
0.24), p-value = 0.24
5-year serum: -0.09 (-0.69, 0.5),
p-value = 0.15
7-year serum: -0.09 (-0.69, 0.5),
p-value = 0.76
Autism screening
D-210
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Prenatal exposure: -0.22 (-0.67,
0.23), p-value = 0.35
5-year serum: 0.68 (0.25, 1.11)
7-year serum: 0.18 (-0.25, 0.6),
p-value = 0.42
Confounding: Age, sex, maternal age, pre-pregnancy BMI, parity, socio-economic status, alcohol, and smoking during pregnancy
Braun et al.
(2014, 2345999)
Medium
United States,
Recruitment:
2003-2006;
Follow-up at
age 4-5 years
Cohort
Maternal Serum Social
5.5 (3.8-7.6) Responsiveness
Scale (SRS) total
score
Regression
coefficient per
loglO-unit/2SD
increase in
PFOA
Pregnant
women and
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
-0.9 (-3.1, 1.4)
Females: -1.8 (-4.6, 1.0)
Males: 0.7 (-2.5, 3.8)
p-value for interaction by
sex = 0.22
Chen et al.
(2013, 2850933)
Medium
Taiwan,
Recruitment:
2004-2005;
Follow-up at
age 2 years
Cohort
Pregnant
women and
their children
from the Taiwan
Birth Panel
Study
N = 239
Cord blood
Mean = 2.6
(SD = 2.5)
Comprehensive
Developmental
Inventory (CDI)
skill quotients:
cognitive, fine-
motor, gross-
motor, language,
self-help, social,
Regression
coefficient per
IQR increase in
ln-unit PFOA
Cognitive: -0.3 (-3.3, 2.7)
Fine-Motor: -0.1 (-3.1, 2.9)
Gross-Motor: -1.1 (-4.7, 2.3)
Language: 0.8 (-2.4, 3.9)
Self-Help: -1.7 (-5.6, 2.2)
Social: 0.8 (-3.2, 4.9)
Whole Test: -0.6 (-3.7, 2.4)
Confounding:
cotinine levels.
whole test
Maternal education, family income, infant sex and gestational age, breastfeeding, HOME score at 24 months of age, cord blood
postnatal environmental tobacco smoke exposure
Ghassabian et
United States, Cohort
Children aged
Blood
SDQ scores: total
Regression
Total Behavioral Difficulties ((3)
al. (2018,
2008-2010
7 years from
1.12
behavioral
coefficient (total
-0.01 (-0.06, 0.05)
5080189)
Upstate KIDS
(IQR = 0.96)
difficulties - total
behavioral
Q2: -0.05 (-0.19,0.10)
Medium
Study
score, borderline
difficulties,
Q3: 0.03 (-0.12,0.17)
N = 788
problems;
problem scores)
Q4: -0.05 (-0.21,0.12)
hyperactivity,
and OR
conduct, peer, or
(borderline
D-211
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
emotional
problems;
difficulties in
prosocial behavior
behavioral Difficulties in Prosocial Behavior
difficulties, (OR)
problem scores, 1.35 (1.03,1.75)
difficulties in Q2: 2.63 (0.97, 7.14)
prosocial Q3: 2.93 (1.03, 8.28)
behavior) per Q4: 3.23 (1.04, 10.07)
log-SD increase
in PFOA and by All other outcomes: No
quartiles statistically significant
associations
Outcome: Borderline behavioral difficulties were defined as having SDQ Total Difficulties Score within the borderline/abnormal range.
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,
(2016, 3981536) 2002-2005
Medium
Cohort
Pregnant
women and
their infants at 6
and 18 months
from the
Hokkaido Study
on Environment
and Children's
Health
N = 90 Females;
83 Males
Maternal serum
1.2 (0.8-1.7)
Bayley Scales of
Infant
Development,
Second Edition
(BSID-II) mental
Regression
coefficient
loglO-unit
increase in
PFOA and by
development index quartiles, least
(MDI),
psychomotor
development index
(PDI)
square means by
quartiles
MDI
Females (6 months)
-0.296 (-11.96, -0.682)
Ql: 89.25 (82.03, 96.47)
Q2: 89.68 (82.14, 97.23)
Q3: 89.03 (81.35, 96.72)
Q4: 84.19(76.11, 92.28),
p-trend = 0.045
PQ2
PQ3
PQ4
0.43 (-4.39, 5.25)
-0.21 (-5.29, 4.86)
-5.05 (-10.66, 0.55)
Males (6 months)
0.110 (-3.31, 7.14)
No statistically significant trend
by quartiles, p-trend = 0.615
PQ2
PQ3
PQ4
0.23 (-5.29, 5.77)
2.44 (-2.39, 7.29)
0.44 (-4.91,5.81)
D-212
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
„ , Exposure
Population, M^trix
Ages, ' Outcome Comparison Resultsb
„ Levels
(ng/mL)a
PDI
6 months: -0.006 (-5.93, 5.50)
Females: 0.055 (-8.37, 12.93)
Males: 0.068 (-5.56, 9.26)
18 months: 0.002 (-7.66, 7.85)
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
Jeddy et al. Great Britain. Cohort
Mothers and Maternal serum
MacArthur
Regression
Nonverbal, 15 mo.: 0.10 (-0.07,
(2017, 3859807) Recruitment:
daughters aged 3.7 (2.8-4.8)
Communicative
coefficient per
0.27)
Medium 1991-1992;
15 and
Development
unit change in
Follow-up at
38 months from
Inventories
PFOA
Social, 15 mo.: -0.06 (-0.36,
age 15 and
ALSPAC
(MCDI):
0.23)
18 months
N = 353
communicative,
intelligibility,
Verbal, 15 mo.: 0.24 (0.12, 0.36)
language,
Maternal age < 30: No
nonverbal
statistically significant
communication,
associations
social
Maternal age > 30: 0.35 (0.15,
development,
0.55)
verbal
comprehension,
Vocabulary, 15 mo.: 0.29 (-2.07,
and vocabulary
2.64)
comprehension
Maternal age < 25: -11.39
scores
(-22.76, -0.02)
Maternal age > 25: No
statistically significant
associations
Communicative, 38 mo.: -0.02
(-0.08, 0.04)
Maternal age < 25: 0.29 (0.03,
0.54)
Maternal age > 25: No
statistically significant
associations
Reference, Location, .
„ .. , Design
Confidence Years
D-213
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Intelligibility, 38 mo.: -0.04
(-0.08, -0.01)
Maternal age < 30: No
statistically significant
associations
Maternal age >30: -0.06 (-0.11,
-0.01)
Language, 38 mo.: -0.83 (-2.21,
0.54)
Nonverbal, social, 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 Factor
Liew et al. Denmark,
(2015,2851010) Recruitment:
Case-Control
Medium
1996-2002;
Follow-up at
average age
10.7 years
Mother-child
pairs from
Danish National
Birth Cohort
Maternal plasma ADHD, ASD
Cases: 4.06
(3.08-5.50)
Controls: 4.00
(3.01-5.42)
RR and OR
(stratified by
quartile or by
sex) per ln-unit
increase in
PFOA or by
quartiles
ADHD: 0.98 (0.82, 1.16)
ASD: 0.98 (0.73, 1.31)
No 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. Denmark,
(2018, 5079744)
Cohort Pregnant Maternal plasma Wechsler Primary Regression
women and 4.28 (3.15-5.49) and Preschool coefficient for
Full Scale IQ:-0.1 (-2.7, 2.4)
Performance IQ: 0.5 (-2.1, 3.0)
D-214
-------
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:
their children
Scales of mean difference
Verbal IQ: -1.1 (-3.7, 1.6)
1996-2002;
from the Danish
Intelligence- per ln-unit
Follow-up at
National Birth
Revised (WPPSI- increase in
No statistically significant
age 5 years
Cohort
R) full scale IQ, PFOA and by
associations or trends by
N = 1,592
performance IQ, quartiles
verbal IQ
quartiles
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.
Denmark,
Case-Control Pregnant
Amniotic fluid ASD OR per unit
0.164 (0.013,2.216),
(2019, 5080602) Recruitment:
women and
Cases: 0.29 increase in
p-value = 0.167
Medium
1982-1999;
their children
(Range: 0.10- PFOA
Females: 0.001 (0, 192.7),
Follow-Up:
from the
0.78)
p-value = 0.275
1993-2009
Historic Birth
Controls: 0.32
Males: 0.270 (0.020, 3.634),
Cohort at
(Range: 0.10-
p-value = 0.536
Statens Serum
1.86)
Institute
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's age at childbirth, birth weight, gestational week at sampling,
gestational age at birth, Apgar score, parity, congenital malformation
Lyall et al. United States, Case-Control
Children and
Maternal serum
ASD measured by
OR per ln-unit
ASD: 0.78 (0.60, 1.01)
(2018, 4239287) 2007-2009
adolescents
Cases:
Diagnostic and
increase in
Q2: 0.56 (0.39,0.81)
Medium
aged 4.5-
GM = 3.58
Statistical Manual
PFOA and by
Q3: 0.58 (0.40, 0.86)
9 years from
(95%
of Mental
quartiles
Q4: 0.62 (0.41,0.93),
EMA study
CI = 3.41-3.76)
Disorders, Fourth
p-trend = 0.05
N = 985 (553
Controls:
Edition (DSM-IV-
Cases; 432
GM = 3.67
TR), intellectual
Intellectual Disability: 0.63
Controls)
(95%
disability
(0.43, 0.92)
CI = 3.49-3.86)
Q2: 0.44 (0.26, 0.76)
D-215
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Q3: 0.67 (0.39, 1.14)
Q4: 0.48 (0.26, 0.88),
p-trend = 0.06
Results: Lowest quartile used as reference.
Confounding: Matching factors, parity, maternal age, race/ethnicity, weight at sample collection, and maternal birthplace
Oulhote et al. Faroe Islands,
(2019,6316905) Recruitment:
Medium 1997-2000;
Follow-up at
age 7 years
Cohort
Children Maternal blood Boston Naming Regression
N = 419 3.25 (2.54-3.99) Test with and coefficient per
without cues, SDQ IQR increase in
total score PFOA
Boston Naming Test
With Cues
Prenatal: -0.14 (-0.26, 0.05)
5-year serum: -0.01 (-0.07,
0.05)
Without Cues
Prenatal: -0.07 (-0.16, 0.00)
5-year serum: -0.01 (-0.07,
0.05)
SDQ
Prenatal: 0.11 (0.02, 0.26)
5-year serum: 0 (-0.06, 0.06)
Confounding: None reported
Quaak et al. Netherlands,
(2016, 3981464) Recruitment:
Medium 2011-2013;
Follow-up
through age
18 months
Cohort
Pregnant
women and
their children
from LINC
54 (20 Females;
34 Males)
Cord blood
870.0 ng/L
(Range: 200-
2,300 ng/L)
Child Behavior
Checklist 1.5-5
(CBCL 1.5-5)
measures of
ADHD,
externalizing
behavior
Regression ADHD
coefficient by Slightly negative, not statistically
tertiles significant associations for
overall population and males.
Slightly positive for females. No
interactions reported by sex.
Externalizing Behavior
T2: -3.33 (-7.65, 0.29),
p-value = 0.12
T3:-2.30 (-6.88, 1.55),
p-value = 0.31
Females
T2: -5.24 (-12.82, 0.00),
p-value = 0.10
D-216
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
T3: 0.71 (-3.83, 5.21),
p-value = 0.74
Males
T2: -5.87 (-10.76, -0.43),
p-value = 0.05
T3: -5.54 (-11.57,-0.29),
p-value = 0.09
Results: Lowest tertile used as reference.
Confounding: Alcohol use, smoking, family history of ADHD, education
Shin et al. United States,
(2020, 6507470) Recruitment:
Case-Control
Medium
2002-2009;
Follow-up:
2009-2017
Mother-child
pairs from the
CHARGE
study, with
children aged 2-
5 years
453 (239 Cases;
214 Controls;
88 Females; 365
Males)
Maternal serum
2.33 (1.59-3.32)
ASD measured by
Autism Diagnostic
Interview-Revised
(ADI-R)
OR per increase By modeled prenatal exposure
(ln-unit or linear Ln-unit: 0.94 (0.59, 1.49)
Linear: 1.01 (0.89, 1.14)
By measured postnatal levels
Ln-unit: 1.09 (0.71, 1.67)
Linear: 1.06 (0.84, 1.33)
No statistically significant
associations, trends, or
interactions by quartiles or by
sex
Results: Lowest quartile used as reference.
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
scale) in
modeled,
maternal,
prenatal PFOA
or measured,
maternal,
postnatal PFOA
and by quartiles
Skogheim et al. Norway, Cohort
Mother-child
Maternal plasma Nonverbal and
Regression
Nonverbal Working Memory
(2019, 5918847) Recruitment:
pairs from
2.50 (1.77-3.21) Verbal Working
coefficient per
Q2:-0.12 (-0.32, 0.09)
Medium 1999-2008;
MoBa
Memory measured
unit increase in
Q3:-0.19 (-0.41, 0.03)
Follow-up:
N = 943
by Stanford Binet
PFOA and by
Q4:-0.18 (-0.41, 0.05)
2007-2011
Intelligence Scales
quintiles
Q5:-0.38 (-0.61,-0.15),
p-value <0.01
Verbal Working Memory
Q2: 0.17 (-0.05, 0.40)
Q3: 0.32 (0.07, 0.56)
D-217
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
Q4: 0.24 (-0.01,0.49)
Q5: 0.24 (-0.01,0.50)
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. United States,
(2020, 6364693) Recruitment:
Medium 2001-2001;
Follow-up at
age 1, 2, and
3 years
Pregnant
women and
their children
from the
Columbia
University Birth
Cohort
N = 302 (150
Females; 152
Males)
BSID-II scores:
Mental and
Psychomotor
Development
Index (MDI and
PDI), Full IQ,
Performance IQ,
Verbal IQ
Cohort Pregnant Cord blood BSID-II scores: Regression MDI
GM = 2.31 Mental and coefficient of Year 1:-1.10 (-3.83, 1.63)
(Range: 0.18- Psychomotor mean difference Year 2: 1.26 (-2.64, 5.16)
>.14) Development per log-unit Year 3:3.93 (0.08,7.77)
increase in
maternal PFOA PDI
Year 1:-1.05 (-6.02,3.92)
Year 2: 0.23 (-3.27,3.74)
Year 3: 2.35 (-2.84,7.54)
Full IQ
Year 4: 2.50 (-1.15,6.15)
Year 6: 0.87 (-3.89, 5.63)
Performance IQ
Year 4: 0.64 (-4.12,5.4)
Year 6: -1.37 (-6.25, 3.51)
Verbal IQ
Year 4: 3.99 (-0.34,8.32)
Females: 5.97 (0.34, 11.6)
Males: 1.92 (-4.76, 8.60)
Interaction p-value = 0.29
Year 6: 3.02 (-2.49,8.53)
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
D-218
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Stein et al. United States,
(2013, 2850964) Recruitment:
Medium 2005-2006,
Follow-Up:
2009-2010
Cohort
Pregnant
mothers and
their children
aged 6-12 years
from the C8
Health Project
Modeled = 284
Measured = 319
Modeled in
utero exposure
43.7(11.7-
110.8)
NEPSY-II scores:
comprehension of
instructions,
design copying,
narrative memory
free and cued
Regression
coefficient per
ln-unit increase
in PFOA and by
quartiles
Serum
35.0 (15.3-93.2) recall, word
generation
semantic/initial
letter
Wechsler
Abbreviated Scale
of Intelligence:
Full-scale IQ,
performance IQ,
verbal IQ
Conners'
Continuous
Performance test
scores: clinical
confidence index,
commissions T-
score, hit reaction
time T-score,
omissions T-score
Wechsler
Individual
Assessment Test II
(WIAT-II) scores:
word
reading/pseudowor
d decoding,
Comprehension of instructions
Prenatal: 0.14 (-0.08, 0.36)
By serum: 0.03 (-0.22, 0.28)
Design copying
Prenatal: 0.21 (-0.06, 0.48)
Q4: 1.02 (0, 2.04)
By serum: 0.26 (-0.04, 0.55)
Narrative memory free and cued
Recall
Prenatal: -0.14 (-0.36, 0.08)
By serum: -0.07 (-0.31, 0.17)
Word generation semantic/initial
letter
Prenatal: 0.10 (-0.09, 0.30)
By serum: 0.03 (-0.19, 0.25)
Full-scale IQ
Prenatal: 0.83 (-0.13, 1.79)
Q4: 4.61 (0.68, 8.54)
By serum: 0.99 (-0.06, 2.04)
Performance IQ
Prenatal: 0.58 (-0.39, 1.55)
By serum: 0.94 (-0.14, 2.01)
Verbal IQ
Prenatal: 0.41 (-0.60, 1.42)
By serum: 0.29 (-0.83, 1.40)
Clinical confidence index
Prenatal: -2.37 (-4.24, -0.50)
D-219
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
numeral
operations
Q2
Q3
Q4
By
Q2
Q3
Q4
-2.14 (-9.86, 5.57)
-7.68 (-15.32, -0.04)
-8.49 (-16.14, -0.84)
serum:-2.15 (-4.19,-0.10)
-5.62 (-12.52, 1.27)
-3.23 (-10.37,3.91)
-6.90 (-14.04, 0.25)
Commissions t-score
Prenatal: -0.17 (-0.89, 0.55)
Q2: 1.52 (-1.46,4.51)
Q3: 0.16 (-2.79, 3.12)
Q4: 0.03 (-2.93, 2.99)
By serum: 0.12 (-0.66, 0.91)
Q2: 0.95 (-1.71,3.61)
Q3:-0.32 (-3.08, 2.44)
Q4: 0.60 (-2.16,3.36)
Hit reaction time t-score
Prenatal: -0.37 (-1.22, 0.49)
Q2: -1.69 (-5.24, 1.86)
Q3: -1.88 (-5.40, 1.63)
Q4:-1.38 (-4.90, 2.14)
By serum: -0.70 (-1.63, 0.24)
Q2: -1.67 (-4.84, 1.49)
Q3: -1.76 (-5.04, 1.52)
Q4: -1.73 (-5.01, 1.55)
Omissions t-score
Prenatal: -0.02 (-1.06, 1.03)
Q2: 0.10 (-4.21, 4.42)
Q3: -0.40 (-4.68,3.88)
Q4: 0.10 (-4.19, 4.38)
By serum: 0.12 (-0.66, 0.91)
Q2: -2.20 (-5.95, 1.55)
D-220
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
„ , Exposure
Population, M^trix
Ages, ' Outcome Comparison Resultsb
„ Levels
(ng/mL)a
Q3: 0.07 (-3.82,3.95)
Q4: -0.57 (-4.46,3.31)
Word reading
Prenatal: 0.50 (-0.40, 1.41)
Q2: 1.72 (-2.05, 5.48)
Q3: 0.61 (-3.07,4.30)
Q4: 2.27 (-1.43, 5.96)
By serum: -0.02 (-1.01, 0.98)
Q2:-1.32 (-4.70, 2.06)
Q3: -1.91 (-5.34, 1.52)
Q4: -1.09 (-4.54,2.36)
Numerical operations
Prenatal: 0.65 (-0.48, 1.78)
Q2: 4.45 (-0.25,9.14)
Q3: 4.75 (0.13,9.36)
Q4: 3.12 (-1.51, 7.76)
Females: -0.6 (-5.0, 3.9)
Males: 4.4 (0.4, 9.2)
p-value for interaction by
sex = 0.14
By serum: 0.15 (-1.17, 1.46)
Q2: 0.36 (-4.17, 4.88)
Q3: 1.11 (-3.51,5.73)
Q4: -0.41 (-5.06,4.25)
Females: -4.1 (-8.6, 0.3)
Males: 3.9 (0.2, 9.6)
p-value for interaction by
sex = 0.01
No other statistically significant
interactions by sex
Results: Lowest quartile used as reference. For brevity, only statistically significant associations by quartiles are included for NEPSY-II and
Wechsler Abbr.
Reference, Location, .
„ .. , Design
Confidence Years
D-221
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
Confounding: Child age at neuropsychological assessment, child sex, test examiner, HOME score, maternal Full-Scale IQ, child BMI at
neuropsychological assessment
Strom et al.
(2014, 2922190)
Medium
Denmark
Recruitment:
1988-1999
Follow-up: 2010
Cohort
Pregnant
women and
their children,
from the
DaF088 cohort
N = 876
Maternal serum
3.7 (IQR = 2.0)
Depression,
ADHD, scholastic
achievement
Depression, Depression
ADHD: Hazard T2: 1.37 (0.85, 2.21)
ratio (depression T3: 1.03 (0.61, 1.73)
and ADHD) by p-value for trend = 0.28
tertile
ADHD
T2: 0.48 (0.18, 1.28)
T3: 0.74 (0.29, 1.87)
p-value for trend = 0.45
Scholastic Achievement:
-0.07 (-0.15,0.001),
p-value = 0.18
T3: -0.25 (-0.64,0.14),
p-value = 0.21
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
Scholastic
achievement:
Regression
coefficient per
unit increase in
PFOA and by
tertiles
Vuong et al. United States, Cohort
Children ages 5
Maternal serum
Behavior Rating
All outcomes:
Behavioral Regulation: 1.11
(2016, 3352166) Recruitment:
and 8 years
5.4 (3.6-7.5)
Inventory of
OR for score
(-1.22, 3.44)
Medium 2003-2006;
from the HOME
Executive
> 60 per unit
Metacognition: 0.58 (-1.77,
Follow-up at
study
Function (BRIEF)
increase in
2.93)
ages 5 and
N = 218
scores for
PFOA
Global Executive Function: 1.06
8 years
behavioral
(-1.33, 3.45)
regulation index,
Index and
No statistically significant
metacognition
compositive
associations or interactions by
index, global
scores only:
age; no statistically significant
executive
Regression
associations or trends by
composite, inhibit,
coefficient per
quartiles
shift, emotional
ln-unit increase
control, working
in PFOA and by Inhibit: 1.45 (0.76, 2.77)
memory,
quartiles
Shift: 1.01 (0.51, 1.98)
plan/organize,
D-222
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
initiate,
organization of
materials, monitor
Emotional control: 1.33 (0.62,
2.84)
Working memory: 0.84 (0.47,
1.47)
Plan/organize: 1.43 (0.74, 2.76)
Initiate: 2.13 (0.89, 5.10)
Organization: 1.83 (0.81, 4.16)
Monitor: 1.80 (0.86, 3.78)
Confounding: Maternal age, race, education, income, maternal serum cotinine, maternal depression, HOME score, maternal IQ, marital
status, child sex
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: 5.4
(3.7-7.4)
8 years: 2.4
(1.8-3.2)
BRIEF measures
of behavioral
regulation,
metacognition,
global executive
composite indices
OR per ln-unit
increase in
PFOA
Behavioral Regulation
3 years: 1.01 (0.29, 3.53)
8 years: 1.56 (0.49,4.92)
Metacognition
3 years: 1.30 (0.47, 3.57)
8 years: 3.18 (1.17, 8.60)
Global Executive Function
3 years: 1.39 (0.45,4.24)
8 years: 2.69 (0.92,7.90)
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, Cohort
Mother-child
Serum
Conners'
Regression
Conners'
(2018, 5079693) Recruitment:
dyads from the
Prenatal: 5.2
Continuous
coefficient per
Commissions
Medium 2003-2006;
HOME study
(3.6-7.6)
Performance Test
ln-unit increase
Prenatal: -2.0 (-3.8, -0.3)
Follow-up at
N = 204
3 years: 5.4
II commissions t-
in PFOA
3 years: -0.1 (-2.3, 2.1)
age 3 and
(3.7-7.4)
score, omissions t-
8 years: -0.01 (-2.4, 2.4)
8 years
8 years: 2.5
score, hit reaction
Omissions
(1.7-13.2)
time, tau (ms)
Prenatal: -2.3 (-7.1,2.6)
3 years:-1.9 (-7.8, 3.9)
Virtual Morris
8 years: 1.0 (-5.8, 7.8)
Water Maze
Hit reaction time
(VMWM) scores
Prenatal: -0.7 (-3.5, 2.2)
D-223
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
„ , Exposure
Population, M^trix
Ages, ' Outcome Comparison Resultsb
„ Levels
(ng/mL)a
for visual-spatial
3 years: 0.2 (-3.5, 4.0)
learning distance
8 years: -2.3 (-6.8, 2.3
(pool units),
Tau
learning time (s),
Prenatal: -10.6 (-43.6, 22.3)
memory retention
3 years: 22 (-16.5, 60.6)
distance (%), and
8 years: 14.6 (-21.9,51.1)
memory retention
time (s)
Visual-spatial scores (VMWM)
Learning distance
Prenatal: -0.1 (-1.7, 1.5)
3 years: 0.5 (-1.2, 2.2)
8 years: 0.1 (-1.8, 2.0)
Learning time
Prenatal: 0.5 (-2.0, 3.0)
3 years: 1.4 (-1.4, 4.2)
8 years: -0.1 (-3.5, 3.3)
Memory retention distance
Prenatal: 2.8 (-1.7, 7.4)
3 years: -0.9 (-7.1, 5.4)
8 years: 1.1 (-5.8, 8.0)
Memory retention time
Prenatal: -0.3 (-2.0, 1.3)
3 years:-1.5 (-3.3,0.2)
8 years: -0.1 (-2.4, 2.1)
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, Cohort
Pregnant
Serum
Wechsler
Regression
Full Scale IQ
(2019, 5080218) Recruitment:
women and
Maternal:
Intelligence Scale
coefficient per
Prenatal: 3.3 (-0.4, 6.9)
Medium 2003-2006;
their children
GM = 5.2
for Children-
ln-unit increase
3 years: 2.4 (-1.5, 6.4)
Follow-up at
from the HOME
8 years:
Fourth Edition
in PFOA
8 years: 2.3 (-3.3, 7.9)
age 3 and
study
GM = 2.4
(WISC-IV): full
8 years
N = 221
scale IQ,
Perceptual Reasoning
perceptual
Prenatal: 0.7 (-3.2, 4.6)
reasoning,
3 years: 1.2 (-3.0,5.4)
processing speed,
8 years: 2.3 (-3.7, 8.2)
Reference, Location, .
„ , Design
Confidence Years
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
verbal
comprehension,
working memory
Processing Speed
Prenatal: 3.3 (-0.8, 7.5)
3 years: 1.7 (-2.6, 6)
8 years: 2.8 (-3.0, 8.5)
Verbal Comprehension
Prenatal: 2.3 (-1.1, 5.6)
3 years: 1.0 (-2.9, 4.8)
8 years:-1.8 (-6.9, 3.2)
Working Memory
Prenatal: 4.1 (0.3, 8.0)
3 years: 2.9 (-1.0, 6.7)
8 years: 4.3 (-0.7, 9.3)
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.
(2020, 6833684)
Medium
United States
Recruitment:
2003-2006;
Follow-up at
age 8 years
Cohort
Maternal serum
Mean = 6.1
(SD = 3.8)
Mother-child
pairs with
children aged
8 years from the
HOME study
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
Wide Range
Achievement Test
4 (WRAT-4)
reading composite
score
Regression
coefficient per
loglO-unit
increase in
PFOA
12.6 (3.0, 22.2)
Wang et al. Taiwan, Cohort
Pregnant
Serum
Full Scale IQ,
Regression
Full Scale IQ
(2015, 3860120) Recruitment:
women and
5 years: 2.50
Performance IQ,
coefficient per
5 years: 1.2 (-1.0, 3.5)
Medium 2000-2001;
their children
(1.54-3.35)
Verbal IQ
log2-unit
8 years: -0.4 (-2.5, 1.7)
Follow-up at
aged 5 and
8 years: 2.50
increase in
age 5 years
8 years from
(1.54-3.33)
PFOA
Performance IQ
TMICS
5 years: 1.0 (-1.4, 3.4)
N = 120
8 years: -1.1 (-3.2, 1.0)
Verbal IQ
D-225
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
5 years: 0.9 (-1.4, 3.3)
8 years: 0.5 (-1.5, 2.5)
Confounding: Maternal education, family annual income, children's age, sex, HOME score at IQ assessment
Zhang et al. United States, Cohort
Pregnant
Serum
Basic reading,
Basic Reading
(2018, 4238294) Recruitment:
women and
Maternal: 5.4
brief reading,
Maternal Serum: 0.7 (-4.9, 6.2)
Medium 2003-2006;
their children
(3.6-7.3)
letter word
Year 3 Serum: 6.4 (-1.6, 14.1)
Follow-up at
aged 3, 5, and
3 years: 5.5
identification,
age 3, 5, and
7 years from the
(3.9-7.7)
passage
Brief Reading
7 years
HOME study
8 years: 2.4
comprehension
Maternal Serum: 3.7 (-1.8, 9.3)
N = 167
(1.8-3.2)
measured by
Year 3 Serum: 10.4 (2.8, 18.1)
Woodcock
Johnson Test of
Letter Word Identification
Achievement-Ill
Maternal Serum: 2.0 (-3.1, 7.1)
(WJ-III)
Year 3 Serum: 9.2 (2.1, 16.3)
Reading
Passage Comprehension
composite, word
Maternal Serum: 3.8 (0.1, 7.7)
reading, sentence
Year 3 Serum: 8.5 (3.3, 13.7)
comprehension
measured by Wide
Word Attack
Range
Maternal Serum: 0.5 (-5.1, 6.1)
Achievement Test
Year 3 Serum: 4.9 (-2.0, 11.8)
4 (WRAT-4)
Reading Composite
Maternal Serum: 3.5 (-1.1, 8.2)
Year 3 Serum: 2.8 (-3.1, 8.8)
Year 8 Serum: 2.6 (-3.1, 8.2)
Word Reading
Maternal Serum: 2.3 (-2.1, 6.7)
Year 3 Serum: 1.0 (-4.7, 6.7)
Year 8 Serum: 6.1 (0.9, 11.3)
Sentence Comprehension
Maternal Serum: 3.7 (-1.6, 9.0)
D-226
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
Year 3 Serum: 3.1 (-4.1, 10.1)
Year 8 Serum: -0.1 (-6.6, 6.4)
Confounding: Maternal age, race, education, household-income, parity, smoking (serum cotinine concentration, ng/mL), maternal IQ,
breastfeeding duration (weeks), HOME score
General Population
Ding and Park
(2020, 6711603)
Medium
United States, Cross-sectional Adults aged 20- Serum
2003-2016 69 years from 2.0(1.3-2.9)
NHANES
N = 2,731
High and low OR per log2-
frequency hearing unit increase in
impairment (HFHI PFOA and
and LFHI) > 90th
percentile
vs. < 90th
percentile
HFHI
OR (per doubling): 0.97 (0.82,
1.14)
OR (90th percentiles): 1.05
(0.61, 1.81)
LFHI
OR (per doubling): 0.98 (0.73,
1.32)
OR (90th percentiles): 1.40
(0.48, 4.07)
Confounding: Age, age square, sex, race/ethnicity, education level, poverty-income ratio, smoking status, BMI, noise exposures
(occupational, recreational, firearm noise), NHANES cycles
Gallo et al.
(2013, 2272847)
Medium
United States,
2005-2006
Cross-sectional
Adults aged
50+ years from
the C8 Health
Project
N = 21,024
Serum
Range: 0.25-
22,412
Memory
impairment (self-
reported)
OR per
doubling of
PFOA and by
quartiles
OR: 0.96 (0.94, 0.98)
Q2: 0.88 (0.79, 0.97)
Q3: 0.83 (0.75,0.92)
Q4: 0.79 (0.71,0.88)
Q5: 0.79 (0.71,0.88)
p-trend < 0.001
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.
(2019, 5080366)
Medium
1.35 (0.87, 2.11), p-value = 0.183
Norway, Cohort Children and Breast milk ADHD OR per IQR
Recruitment: adults from 40.000 ng/L increase in ln-
2003-2009; HUMIS (26.809- unit PFOA
Follow-up: N= 1,199 61.256 ng/L)
2008-2016
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
D-227
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
Li (2020,
6833686)
Medium
United States,
1999-2006
Cross-sectional
Adults aged
20+ years from
NHANES
N = 2,525
Serum
2.25 (Range:
0.07-51.1)
Hearing threshold
> 25 dB by
frequency
OR by quartiles
2,000 Hz
Q2: 1.41 (0.95,2.10)
Q3: 1.26 (0.85, 1.87)
Q4: 1.76(1.20,2.60), p-
trend < 0.01
3,000 Hz
Q2: 1.39 (0.98, 1.98)
Q3: 1.38 (0.98, 1.96)
Q4: 1.64(1.16,2.34),
p-trend = 0.02
Results: Lowest quartile used as reference.
Confounding: Age, sex, BMI, education, ethnicity group, family income, sample weights
4,000 Hz
Q2: 1.31 (0.95,
Q3: 1.12(0.81,
Q4: 1.41 (1.01,
p-trend = 0.11
1.83)
1.56)
1.98),
Shrestha et al. United States,
Cross-sectional Residents aged
Serum
Regression
Depression:
(2017, 3981382) 2000-2002
55-74 years
8.1 (5.9-11.9)
Affective state:
coefficient per
0.08 (-0.85, 1.02),
Medium
who lived
Beck Depression
IQR increase in
p-value = 0.86
adjacent to
Inventory (BDI)
ln-unit PFOA
Hudson River
total score, State-
CVLT Total Score:
N = 126
Trait Anxiety
2.63 (0.20, 5.06), p-value = 0.03
Inventory state and
trait t-scores
Wisconsin card-sorting test
Perseverative Error:
Attention: Trail
-0.18 (-0.34,-0.01),
making test Part A
p-value = 0.04
(ln-transformed
Perseverative Response:
time to complete)
-0.20 (-0.38, -0.02),
p-value = 0.03
Executive
function: Stroop
Wechsler Memory Scale
D-228
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
color word test t-
Logical Memory
score, Trail
Immediate Recall: 0.28 (-0.85,
making test part B
1.42), p-value = 0.62
(ln-transformed
Delayed Recall: 0.09 (-0.98,
time to complete),
1.15), p-value = 0.87
Wisconsin card
Visual Reproduction
sorting test
Immediate Recall: -0.11 (-0.79,
preservative ln-
0.56), p-value = 0.74
transformed error
Delayed Recall: -0.12 (-0.83,
and response
0.59), p-value = 0.74
Memory and
No statistically significant
learning:
associations: State-Trait Anxiety
California Verbal
Inventory, Stroop color word
Learning Test total
test, trail-making tests, motor
and subscores,
function outcomes, visuospatial
Wechsler Memory
outcomes
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
D-229
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
test ln-transformed
total numbers of
contacts and times
touching,
dominant hand
reaction time
Visuospatial
function: Wechsler
Adult Intelligence
Scale-Revised
total scores for
block design and
digit symbol
coding
Confounding: Age, sex, education, serum total PCB
Pregnant Women
Vuong et al. United States
(2020, 6356876) Recruitment:
Cohort
Medium
2003-2006
Follow-up:
~20 weeks
gestation and
postpartum
(4 weeks, 1, 2,
3, 4, 5, and
8 years)
Pregnant
women from the
HOME study
N = 300
Maternal serum
5.4 (3.6-7.6)
Beck Depression
Inventory-II (BDI-
II)
Relative risk per Medium Score Trajectory: 1.3
ln-unit increase (0.8, 2.0)
in PFOA High Score Trajectory: 0.9 (0.5,
1.9)
OR for score >13 from
pregnancy to 8 years postpartum:
1.1 (0.7, 1.6)
Notes: ADHD = attention deficit hyperactivity disorder; ALSPAC = Avon Longitudinal Study of Parents and Children; ASD = autism spectrum disorder; ASQ = Ages and Stages
Questionnaire; BMI = body mass index; 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; HOME = Health Outcomes and Measures of the Environment; ID = intellectual
disability; HUMIS = Human Milk Study; IQR = interquartile range; LINC = Linking Maternal Nutrition to Child Health; MoBa = Mother, Father, and Child Cohort Study;
NHANES = National Health and Nutrition Examination Survey; OR = odds ratio; PFOA = perfluorooctanoic 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.
D-230
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
: Confounding indicates factors the models presented adjusted for.
D.9 Renal
Table D-18. Associations Between PFOA Exposure and Renal Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
General Population
Dhingra et al. United States,
(2016,3981521) 1951-2011
High
Cohort Adults from C8 Serum CKD HR by PFOA Main cohort
Health Cumulative quintiles, at 0-, 0-year lag:
Project/C8 PFOA exposure 5-, 10-, and Quintile 2: 1.26 (0.9, 1.75),
Science Panel, at failure or end 20- year lags p-value = 0.18
>20 years, of follow-up: Quintile 3: 1.12 (0.8, 1.55),
Mean = 3.32 ng/ p-value = 0.52
Main mL-yr Quintile 4: 1.12 (0.81, 1.56),
cohort = 28,240, (SD = 7.26) p-value = 0.49
prospective Quintile 5: 1.24 (0.88, 1.75),
cohort = 27,952 p-value = 0.21
p-value for trend = 0.80
5-, 10-, and 20-year lag: No
statistically significant associations
or trends
Prospective cohort
Quintile 2: 1.36 (0.89,2.09),
p-value = 0.16
Quintile 3: 0.94 (0.62, 1.45),
p-value = 0.79
Quintile 4: 1.12 (0.72, 1.75),
p-value = 0.6
Quintile 5: 1.08 (0.7, 1.66),
p-value = 0.74
p-value for trend = 0.77
Outcome: CKD was self-reported then confirmed by medical records or presence in United States Renal Data System renal failure registry
(non-neoplastic, non-genetic, and diagnosed after age 20).
D-231
-------
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 quintile used as reference group.
Confounding: Gender, time-varying self-reported hypertension, time-varying self-reported diabetes diagnosis, time-varying self-reported high
cholesterol diagnosis, time-varying smoking, category of BMI, and education category0
Dhingra et al. United States,
(2017, 3981432) 2005-2006
Medium
Women from
Serum eGFR
Regression
Modeled serum PFOA
C8 Science
Measured:
coefficient per
Per In increase: 0.05 (0.01),
Panel, 30-65
60th
ln-unit increase
p-value = 0.43
years,
percentile = 36.
in PFOA, or by
Quintile 2: -0.08 (0.27),
N = 29641
3 (ig/mL (20th-
quintiles, or by
p-value = 0.77
80th
deciles
Quintile 3: 0.37 (0.27),
percentile =11.
p-value = 0.17
1-88.0 (ig/mL)
Quintile 4: 0.21 (0.27),
p-value = 0.44
Modeled:
Quintile 5: 0.23 (0.27),
60th
p-value = 0.41
percentile = 26.
8 (ig/mL (20th-
Dose-response by deciles:
80th
decreased until the 4th decile and
percentile = 5.8
remained approximately flat
-82.4 (ig/mL)
thereafter
Measured serum PFOA
Per In increase: -0.14 (0.07),
p-value = 0.03
Quintile 2: -0.64 (0.27),
p-value = 0.018
Quintile 3: -1.03 (0.27),
p-value = 0.0001
Quintile 4: -0.84 (0.27),
p-value = 0.0019
Quintile 5: -0.98 (0.27),
p-value = 0.0003
Results: Lowest quintile used as reference group. Effect estimates are provided with standard deviation in parentheses.
Confounding: Smoking status, BMI, education level, race, sex, and birth year
D-232
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Uric acid
3.49 (75th (mg/dL)
percentile = 6.5
4)
Lin etal. (2013, Taiwan, 2006- Cross-sectional Adolescents and Serum
2850967) 2008 young adults
Low fromYOTA
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
PFOA
percentiles
< 50th percentile: 6.08 (0.1)
50th-75th: 6.08 (0.11)
75th-90th: 6.11 (0.14)
> 90th: 6.13 (0.17)
p-value for trend = 0.983
Blake et al.
(2018, 5080657)
Medium
United States,
1991-2008
Serum eGFR
12.7 (7.83-19.5)
Cohort Adults and
children,
Fernald
Community
Cohort (FCC)
N = 192(115
females, 77
males)
Confounding: Age, year of measurement, sex, education, income, marital status, and BMI
Percent change
per IQR
increase in
PFOA
All:
Repeated measures model: -0.83
(-2.44, 0.77); p-value = 0.31
Latent model: -0.74 (-2.45, 0.96);
p-value = 0.39
Female: -1.38 (-3.41, 0.65),
p-value = 0.18
Male: 0.95 (-3.08, 4.98),
p-value = 0.21
p-value for interaction by
sex = 0.38
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, HDL, LDL, white blood cell count, CRP, hemoglobin, and iron
Serum
Diabetic: 28.6
(12.6-72.7)
Non-diabetic:
28.0 (13.6-71.4)
CDK (eGFR OR per ln-unit
of < 60 mL/min/ increase in
1.73 m2) PFOA
Diabetics: 0.92 (0.86, 0.98)
Non-diabetic: 0.99 (0.96, 1.03)
Covertino et al. United Controlled trial
Adults, solid-
Plasma
Creatinine
Regression
No observable differences with
(2018, 5080342) Kingdom,
tumor cancer
Exposure levels
(|imol/L). urea
coefficient per
measured plasma PFOA
Low 2008-2011
patients
non reported
(|imol/L)
l-(iM increase
concentrations
N = 49
in PFOA
Confounding: None reported
D-233
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Arrebola et al. Spain, 2009-
(2019, 5080503) 2010
Low
Cross-sectional
Adults,
BIOAMBIENT.
ES study
N = 342
Serum Uric acid
1.83 (1.34-2.53) (mg/dL),
hyperuricemia
OR(hyperurice Uric acid
mia) or Wet-basis and lipid-basis models:
regression 0.04 (-0.06, 0.14); p-value = 0.425
coefficient per Wet-basis model with adjustment
log-unit increase for serum lipids:
in PFOA
0.04 (-0.06, 0.14); p-value = 0.459
Hyperuricemia (OR)
Wet-basis and lipid-basis models:
1.83 (0.93, 3.68); p-value = 0.083
Wet-basis model with adjustment
for serum lipids:
1.78 (0.90, 3.45); p-value = 0.095
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
Liu et al. (2018,
4238514)
Low
United States,
2013-2014
Cross-sectional
Adults from
NHANES,
18+ years,
N = 1,871
Serum
GM = 1.86
(SE= 1.02)
Total protein
(g/dL)
Regression
coefficient per
ln-unit increase
in PFOA
0.05 (SE = 0.03)
Confounding: Age, gender, ethnicity, smoking status, alcohol intake, household income, waist circumference, and medications (anti-
hypertensive, anti-hyperglycemic, and anti-hyperlipidemic agents)
Chen et al.
(2019, 5387400)
Low
Croatia,
2007-2008
Cross-sectional
Adults, 44-
56 years
N = 122
Plasma
GM = 2.87
(range = 1.03-
8.02)
Uric acid
(|imol/L),
creatinine
(|imol/L)
Regression
coefficient per
ln-unit increase
in PFOA
Uric acid: 5.02 (-22.09, 32.09)
Creatinine: 0.46 (-5.60, 6.52)
Confounding: Age, sex, education, socioeconomic status, smoking, dietary pattern, and physical activity
Jain and
United States,
Cross-sectional Adults from
Serum
Levels of Regression
Albumin in urine
Ducatman
2005-2014
NHANES,
Levels not
albumin in urine coefficient per
Per loglO-unit increase: -0.17
(2019, 5381566)
> 20 years,
reported
(loglO-ng/mL), loglO-unit
p-value < 0.01
Low
N = 8,220
creatinine in increase in
D-234
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Exposure
Matrix,
T , Outcome Comparison Select Resultsb
Levels
(ng/mL)a
urine (loglO-
PFOA, or
Negative associations across GF
mg/dL),
percent change
stages
albumin-to-
per 10%
Percent change per 10% increase:
creatinine ratio
increase in
-1.59
in urine (log 10-
PFOA
p-value < 0.05
mg/g), albumin
p-value for gender and
in serum (loglO-
race/ethnicity interaction = 0.15
mg/dL),
creatinine in
Creatinine in urine
serum (loglO-
Per loglO-unit increase: 0.02
mg/dL)
p-value = 0.2
No significant associations across
eGFR stages
Percent change per 10% increase:
0.22
p-value for gender and
race/ethnicity interaction = 0.02
Albumin- to-creatinine ratio in
urine
Per loglO-unit increase: -0.19
p-value < 0.01
Negative associations across GF
stages
Percent change per 10% increase:
-1.82
p-value < 0.05
p-value for gender and
race/ethnicity interaction = 0.88
Albumin in serum
Per loglO-unit increase: 0.02
p-value < 0.01
Positive associations across eGFR
stages
Population,
Reference, Location, A
Confidence Years 8
N
D-235
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Percent change per 10% increase:
0.17
p-value < 0.05
p-value for gender and
race/ethnicity interaction = 0.74
Creatinine in serum
Per loglO-unit increase: 0.01
p-value = 0.19
Positive associations in GF-1
Negative associations in GF-3B/4
Percent change per 10% increase:
0.07
p-value for gender and
race/ethnicity interaction < 0.01
GF Stages: GF-1: GFR> 90 mL/min/1.73m2; GF-2: GFRbetween60 and 90 mL/min/1.73m2; GF- 3A: GFRbetween45 and
60 mL/min/1.73m2; GF- 3B/4: GFR between 15 and 45 mL/min/1.73m2
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 = 3330,
females = 3506
Uric acid
(mg/dL) by
glomerular
filtration (GF)
stage
Regression
coefficient per
loglO-unit
increase in
PFOA
Males
GF-1: 0.04, p-value <0.01
GF-2: 0.05, p-value <0.01
GF-3A: 0.03, p-value = 0.27
GF-3B: -0.07, p-value < 0.01
Females
GF-1: 0.03, p-value = 0.01
GF-2: 0.02, p-value = 0.11
GF-3A: 0.09, p-value <0.01
GF-3B: 0.004, p-value = 0.91
GF Stages: GF-1: eGFR > 90 mL/min per 1.73 m2, GF-2: 60 < eGFR < 90 mL/min per 1.73 m2, GF-3A: 45< eGFR < 60 mL/min per 1.73 m2
GF-3B/4: 15< eGFR < 45 mL/min per 1.73 m2
Confounding: Gender, race/ethnicity, age, loglO(BMI), Iogl0(serum cotinine), poverty income ration, NHANES survey period
Serum
Males:
GM = 2.36
(2.24-2.48)
Females:
GM = 3.19
(3.06-3.32)
D-236
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location,
Confidence Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Select Resultsb
Wang et al. China, 2015-
Cross-sectional
Adults, Isomers
Serum
CKD, eGFR
OR(CKD), or
CKD (OR)
(2019, 5080583) 2016
of C8 Health
6.19(4.08-9.31)
regression
Per ln-unit increase: 0.73 (0.57,
Low
Project
coefficient per
0.95), p-value = 0.019
N = 1,612
ln-unit increase
Q2: 0.72 (0.45, 1.13)
(males = 1,204,
in PFOA, or by
Q3: 0.83 (0.52, 1.31)
females = 408)
quartiles
Q4: 0.60 (0.36, 1.01)
p-value for trend = 0.234
eGFR
Per ln-unit increase:
All: 1.23 (0.30,2.17),
p-value = 0.008
Males: 1.29 (0.21, 2.36),
p-value = 0.019
Females: 1.54 (-0.36, 3.44),
p-value = 0.111
p-value for interaction by
sex = 0.999
Q2: 1.00 (-0.8, 2.81)
Q3: 0.63 (-1.2,2.46)
Q4: 2.07 (0.22, 3.91)
p-value for trend = 0.050
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, 2015
(2019, 5918630) 2016
Low
PFOA
Uric acid
Cross-sectional
Adults, Isomers
of C8 Health
Project
N = 1,612
(males = 1,204,
females = 408)
Serum Uric acid
6.19(4.08-9.31) (mg/dL),
hyperuricemia
OR
(hyperuricemia)
or regression
coefficient (uric
acid) per loglO-
unit increase in
Hyperuricemia (OR)
All: 1.29 (1.08, 1.54)
Males: 1.21 (1, 1.46)
Females: 1.76 (1.06, 2.94)
p-value for interaction by
sex = 0.183
D-237
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
All: 0.18 (0.09,0.26),
p-value < 0.001
Males: 0.17 (0.06,0.27)
Females: 0.14 (0.01,0.27)
p-value for interaction by
sex = 0.988
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
Lee et al. (2020,
6833761)
Low
United States
1999-2016
Cross-sectional
Serum
Exposure levels
not reported
Albuminuria
Adults from
NHANES,
18+ years,
N = 46,748
Outcome: Albuminuria defined as urine albumin-to-creatinine ratio > 30 mg/g.
Confounding: Age, age-squared, sex, diabetes mellitus, hypertension, BMI, race/ethnicity, smoking, and socioeconomic status
OR per SD-unit Discovery data set: 0.69 (0.57,
increase in 0.83). FDR=0.006
loglO-PFOA Validation data set: 0.68 (0.58,
0.80), p-value = 0.029
Scinicariello et United States, Cross-sectional Adults from Serum Uric acid
al. (2020, 2009-2014 NHANES GM = 2.37 (mg/dL),
6833670) N = 4915 (no (SE = 0.06) hyperuricemia,
Low CKD = 4103; gout
CKD = 874)
OR
(hyperuricemia,
gout), or
regression
coefficient (uric
acid) by
quartiles
Uric acid
Overall population
Q2: 0.17 (0.06, 0.29)
Q3: 0.24 (0.11,0.37)
Q4: 0.42 (0.26, 0.57)
p-value for trend = 0.0001
Participants with CKD
Q2: 0.14 (-0.38,0.65)
Q3: -0.05 (-0.63,0.53)
Q4: 0.6 (-0.04, 1.24)
p-value for trend = 0.02
Participants without CKD
Q2: 0.08 (-0.03, 0.2)
Q3: 0.31(0.17,0.46)
Q4: 0.16 (0.01,0.31)
p-value for trend = 0.001
Hyperuricemia (OR)
Overall population
D-238
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Select Resultsb
(ng/mL)a
Q2: 1.05 (0.77, 1.44)
Q3: 1.21(0.87, 1.69)
Q4: 1.81 (1.29,2.55)
p-value for trend = 0.004
Participants with CKD
Q2: 1.15 (0.69, 1.92)
Q3: 0.95 (0.53, 1.69)
Q4: 1.82 (0.96, 3.47)
p-value for trend = 0.21
Participants without CKD
Q2: 0.96 (0.64, 1.44)
Q3: 1.19 (0.75, 1.88)
Q4: 1.65 (1.1,2.46)
p-value for trend = 0.02
Gout (OR)
Overall population
Q2: 1.75 (0.9,3.31)
Q3: 2.34 (1.32,4.15)
Q4: 3.17 (1.68, 5.98)
p-value for trend = 0.01
Participants with CKD
Q2: 1.83 (0.79,4.19)
Q3: 3.02 (1.28,7.15)
Q4: 2.73 (1.28, 5.84)
p-value for trend = 0.04
Participants without CKD
Q2: 2.11 (0.72,6.23)
Q3: 2.57 (1,6.59)
Q4: 3.88 (1.46, 10.33)
p-value for trend = 0.05
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
D-239
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Select Resultsb
(ng/mL)a
Children and Adolescents
Geiger et al.
(2013,2919148)
Low
United States,
1999-2000;
2003-2008
Cross-sectional
Children and
adolescents
from NHANES,
12-18 years,
N = 1,772
Serum
Mean= 4.3
(SE = 0.1)
Uric acid
(mg/dL),
hyperuricemia
OR
(hyperuricemia)
or regression
coefficient (uric
acid) per ln-unit
increase in
PFOA or by
quartiles
Hyperuricemia (OR)
Per In increase: 1.59(1.19,2.13)
Q2: 0.94 (0.58, 1.53)
Q3: 1.01 (0.62, 1.63)
Q4: 1.62(1.1,2.37)
p-value for trend = 0.007
Uric acid
Per In increase: 0.2 (0.11, 0.29)
Q2: 0.02 (-0.10,0.14)
Q3: 0.03 (-0.11,0.17)
Q4: 0.3 (0.17,0.43)
p-value for trend = 0.0001
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,
Cross-sectional Children and Serum
eGFR
Regression
eGFR
(2015, 3859835) 2003-2010
adolescents 3.5 (2.5-4.7)
(min/mL/1.73 m
coefficient by
Q2: -2.63 (-7.07, 1.81)
Low
from NHANES,
2), uric acid
quartiles
Q3: -5.42 (-11.46,0.61)
12-19 years,
(mg/dL),
Q4: -6.61 (-11.39,-1.83),
N = 1,962
creatinine
p-value < 0.01
(mg/dL)
Uric acid
Q2: 0.17 (-0.033, 0.37)
Q3: 0.13 (-0.03, 0.28)
Q4: 0.21 (0.056, 0.37),
p-value < 0.01
Creatinine
Q2: 0.007 (-0.012, 0.027)
Q3: 0.021 (-0.008,0.05)
Q4: 0.029 (0.004, 0.054),
p-value < 0.05
D-240
-------
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 reference group.
Confounding: Sex, poverty-income ratio, caregiver education, serum cotinine, prehypertension, insulin resistance, BMI,
hypercholesterolemia, race/ethnicity categories
Qinetal. (2016, Taiwan
3981721) 2009-2010
Low
Cross-sectional
Children from
GBCA Study,
12-15 years,
N = 225 (123
girls, 102 boys)
Serum
All: 0.5 (0.4-
1.3)
Boys: 0.5 (0.4-
1.4)
Girls: 0.5 (0.4-
1.2)
Uric acid
(mg/dL),
hyperuricemia
Regression
coefficient per
ln-unit increase
in PFOA (uric
acid), and by
quartiles; OR
scaled with
increasing
quartiles
(hyperuricemia)
Uric acid
All: 0.15 (0.01,0.28),
p-value = 0.032
Boys: 0.24 (0.06, 0.42),
p-value = 0.011
Increasing trend in mean uric acid
levels by quartiles; Q1 = 4.85 (4.53,
5.17) vs. Q4 = 5.65 (5.33,5.96);
p-value for trend = 0.033
Girls: 0.01 (-0.19,0.22),
p-value = 0.892
No trend in mean uric acid levels
by quartiles; Q1 = 4.64 (4.43, 4.94)
vs. Q4 = 4.73 (4.41,5.06);
p-value for trend = 0.756
Hyperuricemia (OR)
All: 2.16(1.29,3.61),
p-value < 0.05
Boys: 2.76 (1.37, 5.56),
p-value < 0.05
Girls: 1.64 (0.69, 3.85)
Outcome: Hyperuricemia defined as uric acid level > 6 mg/dL.
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 -0.02 (-0.15, 0.11)
(2018, 4238547) 2016
8-12 years
0.99
(mg/dL)
coefficient per
Low
N = 40
(IQR = 0.45)
unit increase in
PFOA
Confounding: Age, sex, race
Pregnant Women
D-241
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Gyllenhammar
etal. (2018,
4238300)
Medium
Maternal serum Cystatin C
(GFRcc)
Regression
coefficient per
(mL/min/1.73 m IQR increase in
2) maternal PFOA
0.004 (SD = 0.002),
p-value = 0.022
Sweden; 1996- Cohort Mothers and
2011 infants follow 2.3(1.6-3.0)
up to 5 years of
age, POPUP
study
N = 381
Confounding: Sampling year, maternal age, pre pregnancy BMI, maternal weight gain during pregnancy, maternal weight loss after delivery,
years of education, and total time of breastfeeding
Nielsen et al.
(2020, 6833687)
Low
Pregnant
Serum
eGFR:
women,
Early
LMrev, CKD-
PONCH study
pregnancy: 1.8
EPI(creatinine),
N = 73
(s^s4
CAPA, CKD-
percentile = 0.8
EPI(cystatin C),
-4.4)
mean of LMrev
Late pregnancy:
and CAPA,
1.5 (5^-95^
mean of CKD-
percentile = 0.7
EPIcreatinine and
-3.1)
CKD -EP Icystatin C
Glomerular pore
size
Sweden, 2009- Cohort Pregnant Serum eGFR: Spearman's Cross-sectional correlations
2014 women, Early LMrev, CKD- correlation consistently weak and
coefficient nonsignificant
Early to late pregnancy changes:
No significant associations
eGFR:
LMrev: 0.002, p-value = 0.99
CKD-EPI(creatinine): 0.03,
p-value = 0.83
CAPA: 0.06, p-value = 0.64
CKD-EPI(cystatin C): 0.03,
p-value = 0.83
mean of LMrev and CAPA: 0.04,
p-value = 0.76
mean of CKD-EPI(creatinine) and
CKD-EPI(cystatin C): 0.002,
p-value = 0.98
Glomerular pore size:
CAPA/LMrev: 0.09, p-value = 0.47
CKD-EPI(cy statin C)/CKD-
EPI(creatinine): -0.003,
p-value = 0.98
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-242
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Select Resultsb
(ng/mL)a
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 0.021 (SE = 0.032), p-value = 0.508
(2015, 3859842)
past exposure to 4.2 (|imol/L)
coefficient per
Low
AFFF, 17- (range = 0.3-
loglO-unit
66 years old 18)
increase in
N = 137 (97%
PFOA
male)
Confounding: Age, sex, BMI, smoking status, total serum protein, PFOS, PFHxS
Notes: FCC = Fernald Community Cohort; YOTA = Young Taiwanese Cohort Study; GBCA = Genetic Biomarkers Study for Childhood Asthma;; eGFR = estimated glomerular
filtration rate (mL/min per 1.73 m2); GF = glomerular filtration; CKD = chronic kidney disease; BMI = body mass index; GM = geometric mean; OR = odds ratio; SD = standard
deviation; SE = standard error; NHANES = National Health and Nutrition Examination Survey; POPUP = Persistent Organic Pollutants in Uppsala Primiparas;
PONCH = Pregnancy Obesity Nutrition and Child Health study; LMrev = Lund Malmo Revised; CKD-EPI = Chronic Kidney Disease Epidemiology Collaboration study;
CAPA = Caucasian Asian Pediatric Adult; AFFF = aqueous film-forming foam.
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 PFOA Exposure and Hematological Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Select Resultsb
(ng/mL)a
General Population
Etzel et al.
(2019,
5043582)
Medium
United States,
2003-2010
Cross-
sectional
Children and
adults from
NHANES,
>12 years of
age,
N = 7,040
Serum
Median =3.9
(2.6-5.5)
Vitamin D
deficiency
(< 50 ng/mL),
25-hydroxy
Vitamin D
Regression
coefficient or
prevalence OR
(POR) per
doubling of
Per doubling of PFOA:
Vitamin D deficiency
POR: 1.02 (0.93, 1.11)
25-hydroxy Vitamin D
-0.3 (-1.0,0.4)
D-243
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Exposure
Reference Location Population, Matm,
„ .... ' ' Design Ages, Outcome Comparison Select Resultsb
Commence Years Levels
N
(ng/mL)a
([25(OH)D], nm PFOA, or by
ol/L) quintiles No significant associations or trends
Jain (2020,
6333438)
Medium
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 period0
-0.02 (-0.07, 0.03)
Chen et al. Croatia Cross- Adults, 44- Plasma
(2019, 2007-2008 sectional 56 years of age, GM = 2.87
5387400) N = 122 (min= 1.03,
Medium max = 8.02)
Confounding: Age, sex, education, socioeconomic status, smoking, dietary pattern, and physical activity
Calcium in Regression
serum (mmol/L) coefficient per ln-
unit increase
PFOA
United States
2003-2016
Cross-
sectional
Adults from
Adult serum
NHANES,
non-anemic
> 20 years of
males:
age,
GM = 3.3 (95%
N = 11,251
CI: 3.2, 3.4)
non-anemic
females:
GM = 2.5 (95%
CI: 2.4, 2.6)
anemic males:
GM = 2.4
(95% CI: 2.1,
2.7)
anemic females:
GM = 1.6(95%
CI: 1.4, 1.7)
Whole blood
hemoglobin
(WBHGB)
(logl0-g/dL)
Regression
coefficient per
loglO-unit
increase in PFOA
Non-anemic males: 0.009, p-value < 0.01
Non-anemic females: 0.006,
p-value <0.01
Anemic males: 0.026, p-value <0.01
Anemic females: 0.034, p-value < 0.01
Outcome: Anemia defines as whole blood hemoglobin concentrations < 12 g/dL (females) and < 13 g/dL (males).
Confounding: Age, BMI, poverty income ratio, serum cotinine, survey year, daily alcohol intake
D-244
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Select Resultsb
(ng/mL)a
Convertino et
al. (2018,
5080342)
Low
"Almost no observable differences"
(statistical results not provided)
United Controlled Solid-tumor Plasma aPTT (s) Regression
Kingdom, trial cancer patients Range = 0- Fibrinogen (g/L) coefficient per
2008-2011 >18 years of -633,527 ^M PPT (s) unit increase in
age, N = 49 PFOA
Confounding: By design, randomly assigned exposure levels and excluded patients with life expectancy < 3 months, anticancer therapy within
the last 4 weeks, HIV infection, hepatitis B or hepatitis C, inadequate hematologic function, inadequate renal function, abnormal liver function
tests, lack of physical integrity of the gastrointestinal tract, uncontrolled cardiac disease, or use of warfarin, phenytoin, or tolbutamide.
Khalil et al.
United States, Cross- Children with
Serum,
25-hydroxy
Regression 1.90 (-5.49, 9.30)
(2018,
2016 sectional obesity, 8-
median = 0.99
Vitamin D
coefficient (per
4238547)
12 years of age,
(IQR = 0.45)
(ng/mL)
unit increase in
Low
N = 47
PFOA)
Confounding: Age, sex, race
Notes: aPTT = activated partial thromboplastin time. HIV = human immunodeficiency virus. PPT = prothrombin time; GM = geometric mean; BMt = body mass index;
IQR = interquartile range; NHANES = National Health and Nutrition Examination Survey; OR = odds ratio.
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.ll Respiratory
Table D-20. Associations Between PFOA Exposure and Respiratory Effects in Recent Epidemiologic Studies
Reference, Location,
Confidence Years
Design
Population,
Ages,
N
Exposure Matrix,
Outcome
Levels (ng/mL)a
Comparison
Resultsb
Agier et al. France,
Cohort
Pregnant women and
Maternal and child's FEV1
Regression
Prenatal: -1.4
(2019,5043613) Greece,
their children, ages 6-
serum, plasma, or whole
coefficient per
(-2.7, -0.1), p-value = 0.03
Medium Lithuania,
12 years,
blood
log2-unit
Norway,
N = 1,033
increase in PFOA Postnatal: 0.5
Spain, United
Prenatal (maternal)
(-0.6, 1.5), p-value = 0.33
Kingdom
Median = 2.4 (IQR = 2)
2019
Postnatal (child)
D-245
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Population,
Reference, Location, _ . . Exposure Matrix, „ n h
„ j... Design Ages, Outcome Comparison Results"
Confidence Years b * Levels (ng/mL)a
Median = 1.5
(IQR = 0.8)
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, and passive smoking status0
Gaylord et al.
New York, Cross- Adolescents and young Adolescents and young
FEV1
Regression
No statistically significant
(2019, 5080201)
US sectional adults ages 13- adults' serum
FVC
coefficient per
differences observed
Medium
2014-2016 22 years,
FEV1/FVC
log-unit increase
between exposure groups
N =287 Comparison group:
TLC
in PFOA
for the measured outcomes,
median = 1.38
RV
p-value > 0.05
(min = 0.36,
FRC
max = 4.28)
Resistance at
an oscillation
WTCHR group:
frequency of
median =1.80
5Hz, 5-20Hz,
(min = 0.56,
20Hz
max = 5.03)
Comparison: Logarithm base not specified.
Confounding: Sex, race/ethnicity, age, BMI, tobacco smoke exposure
Impinen et al.
Norway Cohort Infants followed up at Cord blood,
Oslo Severity
OR per log2-unit
1.43 (1.03, 1.98),
(2018, 4238440)
1992-2002 2 years and 10 years, Median =1.6 (1.2, 2.1)
Score (1-5 vs.
increase in PFOA
p-value = 0.033
Medium
N = 641
0)
1.25 (0.83, 1.89,
Oslo Severity
(p-value = 0.276
Score (6-12 vs.
0)
1.08 (0.56,2.07),
p-value = 0.819
Reduced lung
function at
birth
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
D-246
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Levels (ng/mL)a
Outcome Comparison
Resultsb
Spain Cohort Pregnant women and
2003-2015 children followed up at
ages 1.5, 4, and 7 years,
N = 503 (4 years)
N = 992 (7 years)
Maternal blood,
Median = 2.35 (1.63,
3.30)
Manzano-
Salgado et al.
(2019, 5412076)
Medium
FEV1,
FVC
FEV1/FVC,
FEF25-75%
Regression
coefficient per
log2-unit
increase PFOA
FVC (4 years): -0.17 (-
0.34, -0.01) p-value not
reported
FEV1, FEV1/FVC,
FEF25-75%: No
statistically significant
associations
Confounding: Maternal age at delivery, parity, previous breastfeeding, pre-pregnancy BMI, region of residence, and country of birth
Qinetal. (2017,
3869265)
Medium
Taiwan, Case-
2009-2010 control
Children with asthma
and without asthma,
aged 10-15,
N = 132 (with asthma)
N = 168 (without
asthma)
Serum,
Children with asthma:
Median = 1.02 (0.48,
2.13)
Children without
asthma:
Median =0.50 (0.43,
0.69)
FEV1
FVC
FEF25-75%
PEF
Regression
coefficient per
ln-unit increase
PFOA, or by
quartiles
Children with asthma:
FEV1:-0.10 (-0.19,-
0.02), p-value < 0.05
Quartile analysis:
p-value for trend=0.002
FEF25-75%: -0.22 (-0.40,
-0.05), p-value < 0.05
Quartile analysis
p-value for trend = 0.014
FVC, PEF: No statistically
significant associations
Children without asthma:
No statistically significant
associations for any
outcomes
Confounding: age, sex, BMI, parental education level, exercise, environmental tobacco smoke exposure, and month of survey
Steenland et al. United States, Cohort
Adult workers and
No lag cumulative
COPD no lag Rate ratio (RR)
No lag:
(2015,2851015) 2008-2011
former workers at a
exposure, 3.03-11.42
and 10-year lag by quartiles
Q2: 1.2(0.64,2.27)
Low
chemical plant,
ug/mL-year
Q3: 1.25 (0.65,2.37)
-t
II
z;
Q4: 1.13 (0.59,2.17)
10-year lag:
Q2: 0.75 (0.38, 1.48)
D-247
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design
Population,
Ages,
N
Exposure Matrix,
Levels (ng/mL)a
Outcome Comparison
Resultsb
10-year lag cumulative
exposure, 0.8-7.04
ug/mL-year
Results: Lowest quartile used as reference group
Confounding: Gender, race, education, BMI, smoking, alcohol consumption
Q3: 1.16(0.6,2.26)
Q4: 0.77 (0.38, 1.57)
Notes: 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; BMI = body mass index.
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.12 Musculoskeletal
Table D-21. Associations Between PFOA Exposure and Musculoskeletal Health Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Population, Exposure
Study Design Ages, Matrix, Outcome
N Levels" (ng/mL)
Comparison
Resultsb
Children and Adolescents
Jeddy et al. England,
(2018, 5079850) 1991-2009
Medium
Cohort
Females from
the ALSPAC
Study,
Age 17,
N = 221
Maternal serum
3.8 (2.9-1.9)
Area adjusted Regression
BMC (g), bone coefficient per
area (cm2), BMC unit increase in
(g), BMD, PFOA
cortical bone
area (cm2),
cortical BMC
(mg), cortical
BMD (mg/cm2),
cortical thickness
(mm), endosteal
circumference
(mm), height
(cm), periosteal
circumference
(mm), total
Height: -0.6 (-1.06, -0.14)
Bone area: -15.48 (-29.40, -1.55)
No other statistically significant
associations
D-248
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix,
Levels3 (ng/mL)
Outcome Comparison
Resultsb
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,
Overall N = 531
Male N = 296
Female N = 280
Plasma
Overall:
4.4 (IQR = 3.2)
aBMD z-score
-0.16 (-0.25,-0.06)
Males:-0.11 (-0.23, 0.00)
Females: -0.24 (-0.4, -0.07)
p-value for interaction by sex = 0.27
Confounding
race/ethnicity,
Areal bone Regression
mineral density coefficient per
(aBMD) z-score, log2-unit
bone mineral increase in
content (BMC) PFOA
z-score
BMC z-score: No statistically
significant associations
Maternal age, education, census tract median household income, individual household income, and child age, sex,
year of blood draw, dairy intake, physical activity
Khalil et al.
(2018, 4238547)
Low
United States
2016
Cross-sectional
Obese children,
ages 8-12
N = 23
Serum
0.99
(IQR = 0.45)
BMD measured
as broadband
ultrasound
attenuation
(dB/MHz) and
speed of sound
(m/s), stiffness
index (%)
Regression
coefficient per
unit increase in
PFOA
BMD (broadband ultrasound
attenuation)
-0.08 (-24.2, 24)
BMD (speed of sound)
-31.2 (-64, 1.54)
Stiffness index
-8.79 (-28.1, 10.5)
Confounding: Age, sex, race
Di Nisio et al.
(2019, 5080655)
Low
Italy Cross-sectional Male high school Serum Arm span (cm) Mann-Whitney
2017-2018 students Controls: 4.70 test (Exposed vs.
N = 100 (50 (3.5-6.6) Controls)
controls, 50 Exposed: 7.35
exposed) (4.7-14.9)
Arm span
Controls: 182.75 (178.0, 185.8)
Exposed: 179.00 (174.2, 187.0)
Adjusted p-value for comparison of
medians = 0.738
Semen
D-249
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix,
Levels3 (ng/mL)
Outcome Comparison
Resultsb
Controls: 0.1
(0.08-0.11)
Exposed: 0.24
(0.11-0.99)
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 Females from Serum
NHANES,
Ages 20-84,
N = 1,921
Ages 20-49
N = 1,104
(All adults
N= 3,809)
Females 20-84:
Weighted
mean = 4.22
Females 20-49:
Weighted
mean = 4.83
Osteoarthritis OR per ln-unit Females ages 20-84
increase in 1.35 (1.02, 1.79), p-value < 0.05
PFOA and by Q2: 1.44 (0.80, 2.62)
quartiles Q3: 1.18 (0.67, 2.08)
Q4: 1.98 (1.24, 3.19), p-value < 0.01
Females ages 20-49
2.23 (0.81,6.12)
Q2: 2.71 (0.93,7.91)
Q3: 1.52 (0.36,6.39)
Q4: 4.95 (1.27, 19.4), p-value < 0.05
All adults ages 20-49
Q4: 3.76 (1.25, 11.4)
No other statistically significant
associations
Results: Lowest quartiles used as the reference group.
Confounding: Age, race/ethnicity, SES, smoking, BMI, vigorous reactional activity, prior wrist, hip, or spine fracture
Lin et al. (2014,
United States,
Cross-sectional Adults from Serum
Total BMD
OR per ln-unit
All fracture types
5079772)
2005-2006,
NHANES Ages GM=3.96
(g/cm2) in hip or
increase in
Males: 0.84 (0.67, 1.07)
Medium
2007-2008
>20, (SD = 3.86)
lumbar spine;
PFOA
Females: 0.98 (0.75, 1.28)
Males N= 1,192,
fractures in hip,
Females in menopause: 1.53 (0.63,
Females
wrist, spine, or
3.74)
N = 842,
all types
Females in
Other outcomes: no statistically
menopause
significant associations
N = 305
D-250
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix,
Levels3 (ng/mL)
Outcome Comparison
Resultsb
Confounding: Age, race/ethnicity, BMI, smoking, drinking, treatment for osteoporosis, use of prednisone or cortisone daily
Khalil et al. United States,
(2016, 3229485) 2009-2010
Medium
Cross-sectional
Adolescents and
adults from
NHANES,
Ages 12-80,
N = 958 females,
956 males
Serum
Mean= 3.7
(SE = 0.18)
BMD (g/cm2) of BMD:
total femur, Regression
femoral neck, coefficient per
lumbar spine; ln-unit increase
Osteoporosis in PFOA and by
among females quartiles
Osteoporosis:
OR per ln-unit
increase in
PFOA and by
quartiles
Total femur
Females: -0.017 (-0.038, 0.003)
Q2: -0.02 (-0.04, -0.001),
p-value < 0.05
Q3:-0.002 (-0.038,0.034)
Q4: -0.03 (-0.063, 0.003)
Males: Not statistically significant
Femoral neck
Females: -0.017 (-0.033, -0.001)
No statistically significant
associations by quartiles
Males: Not statistically significant
Osteoporosis: 1.84 (1.17, 2.90), p-
value = 0.008
Q2: 1.25 (0.38,4.06)
Q3: 1.23 (0.37,4.05)
Q4: 2.59 (1.01, 6.67),
p-value = 0.049
Lumbar spine: 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 Plasma BMD and 2-yr
POUNDS-LOST Cross-sectional: ABMD (g/cm2)
study, Ages 30- Mean = 5.2 (3.5- of spine, total
70, 6.5) hip, femoral
N = 294 Cohort: neck, hip
Mean= 5.4 (3.7-trochanter, hip
6.6) intertrochanteric
area, and Ward's
triangle area
Regression Spine BMD analyses
coefficient per Cross-sectional: -0.021 (-0.038,
SD increase in -0.004)
PFOA 2-yr ABMD: -0.002 (-0.007, 0.004)
Total hip BMD analyses
Cross-sectional: -0.015 (-0.029,
-0.001)
2-yr ABMD: -0.004 (-0.008, 0.000)
D-251
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix,
Levels3 (ng/mL)
Outcome Comparison
Resultsb
Femoral neck BMD analyses
Cross-sectional: -0.016 (-0.03,
-0.002)
2-yr ABMD: -0.001 (-0.007, 0.004)
Hip trochanter BMD analyses
Cross-sectional: -0.015 (-0.029,
-0.002)
2-yr ABMD: -0.003 (-0.007, 0.001)
Hip intertrochanteric area BMD
analyses
Cross-sectional: -0.016
(-0.032, 0.000)
2-yr ABMD: -0.006 (-0.011,
-0.001), p-value < 0.05
Ward's triangle area BMD analyses
Cross-sectional: -0.015 (-0.033,
0.003)
2-yr ABMD: -0.004 (-0.012, 0.005)
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-yr weight change
Occupational Populations
Steenland et al. United States Retrospective
(2015,2851015) 2008-2011 occupational
Low cohort
DuPont plant Drinking water/ Osteoarthritis
workers from the occupational,
C8 Health
Project
N= 3,713
serum
Median =113;
Cumulative
exposure, 25th-
75 th percentiles
with or without
10-year lag: 0.8-
Incidence rate Osteoarthritis no lag
ratio by quartiles Q2: 0.88 (0.58, 1.34)
Q3: 0.97 (0.71, 1.54)
Q4: 0.97 (0.59, 1.59)
p-trend logPFOA cumulative
exposure = 0.92
p-trend via quartiles = 0.48
Osteoarthritis with lag
D-252
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix,
Levels3 (ng/mL)
Outcome Comparison
Resultsb
7.04 or 3.03-
11.42 ng/mL-
year
Results: Lowest quartile used as the reference group.
Confounding: Gender, race, education, BMI, smoking, alcohol consumption
Q2: 0.74 (0.49, 1.10)
Q3: 0.56 (0.34,0.93)
Q4: 0.67 (0.39, 1.14)
p-trend logPFOA cumulative
exposure = 0.13
p-trend via quartiles = 0.15
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 PFOA Exposure and Gastrointestinal Health Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix, „ ,
, , Outcome
Levels
Comparison
Resultsb
(ng/mL)a
Timmerman et
Guinea-Bissau
Cohort
Children
Serum Diarrhea
OR per
At inclusion: 1.09 (0.56, 2.09)
al. (2020,
6833710)
Medium
2012-2015
aged < 2 years
previously
enrolled in a
RCT for
measles
vaccination
N = 236 (113
girls, 123 boys)
0.68 (0.53-0.92)
doubling of
PFOA at
inclusion or 9-
month visit
At 9 months: 1.54 (0.72, 3.29)
No statistically significant
associations or interactions by sex
Confounding: Weight and age at inclusion, sex, maternal education, breastfeeding without solids0
Dalsager et al.
Denmark
Cohort
Pregnant
Serum Diarrhea,
Incidence rate
Diarrhea
(2016, 3858505) 2010-2015
Low
women and
their children
1.68 (Range: vomiting ratio (number of Number of days with symptom
0.32-10.12) (number of days days) or OR T2: 1.07 (0.61, 1.89)
D-253
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
T Population,
Location, „ .
v Study Design Ages,
Y curs jy
Exposure
Matrix,
Levels
(ng/mL)a
Outcome
Comparison
Resultsb
from the Odense
with symptom
(proportion of
T3: 1.08 (0.55,2.13)
Child Cohort,
or proportion of
days) by tertiles
Proportion of days under/above
Ages 1-4 years
days
ofPFOA
median
N = 346
under/above
exposure
T2: 1.10(0.64, 1.89)
median)
T3: 0.94 (0.51, 1.74)
Vomiting
Number of days with symptom
T2: 0.89 (0.61, 1.32)
T3: 0.95 (0.62, 1.47)
Proportion of days under/above
median
T2: 1.05 (0.62, 1.78)
T3: 0.95 (0.52, 1.72)
Results: Lowest tertile used as reference.
Confounding: Maternal age, maternal educational level, parity, and child age
Hammer et al.
Faroe Islands
Cohort
Children and
Blood
Inflammatory
Incidence rate
0.60 (0.23, 1.56)
(2019, 8776815) Enrollment:
adults from
Low exposure:
bowel disease
ratio for highest
Low
1986-2009;
CHEF
GM = 0.95
vs. lowest tertile
follow-up until
N = 2,843
(0.76-1.34)
ofPFOA
2017
High exposure:
exposure
GM = 4.42
(3.55-4.98)
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: -0.006 (-0.03, 0.02)
Low
municipality
study: 20 (11-
(ln-ng/mL levels unit increase in
Resampling: -0.01 (-0.03, 0.005)
29)
of calprotectin
PFOA
Karlshamn: -0.15 (-0.84, 0.55)
Ronneby panel
Ronneby
or zonulin)
study: N = 57
resampling: 16
Zonulin
Ronneby
(9-23)
Panel study: -0.002 (-0.02, 0.02)
resampling:
Karlshamn: 2
Resampling: -0.01 (-0.02, 0.01)
N = 113
(1-2)
Karlshamn: -0.29 (-0.85, 0.27)
Karlshamn:
N = 19
D-254
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location, „ .
Study Design
Years
Population,
Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
Confounding: Age, BMI, gender
Notes: PFOA = perfluorooctanoic acid; RR = risk ratio; BMt = body mass index; RCT = randomized controlled trial; CHEF = Children's Health and the Environment in the
Faroes.
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-23. Associations Between PFOA Exposure and Dental Health Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Population,
Study Design Ages,
N
Exposure
Matrix,
Levels
Outcome
Comparison
Resultsb
(ng/mL)a
Rameshetal. United States Cross-sectional Adolescents Serum Dental caries ORperlog2- 1.00(0.91,1.12)
(2018,5080517) 1999-2002 fromNHANES Median =3.5 unit increase in Q2: 0.95 (0.74, 1.20)
Medium aged 12- (2.3-4.9) PFOA and by Q3: 1.04 (0.82, 1.32)
19 years quartiles Q4:0.95 (0.74, 1.21)
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
United States Cross-sectional Children from Serum Dental caries ORperlQR 1.33 (0.70,2.53); p-value = 0.352
2013-2014 NHANES aged GM=1.92 experience increase in
3-11 years (95% CI: 1.74, PFOA
N = 629 2.11)
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: PFOA = perfluorooctanoic acid; NHANES = National Health and Nutrition Examination Survey; OR = odds ratio; CI = confidence interval; IQR = interquartile range.
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.
Wiener &
Waters (2019,
5386081)
Medium
D-255
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.140cular
Table D-24. Associations Between PFOA Exposure and Ocular Effects in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design
Population, Exposure Matrix,
Ages, N Levels (ng/mL)a
Outcome
Comparison
Resultsb
Zeeshan et al. China,
(2020,6315698) 2016
Medium
Cross-sectional Adults, from the Serum
Isomers of C8 Median = 6.06
Health Project, (3.97-9.12)
ages 22-
96 years,
N = 1,202
Visual impairment,
synechia, macula
disorder, corneal
pannus, shallow
anterior chamber,
vitreous disorder,
retinal disorder,
lens opacity,
conjunctival
disorder, combined
eye disease
Confounding: Age, sex, BMI, education, income, career, exercise time, drinking, smoking0
OR per ln-unit Visual impairment
increase in PFOA 1.8 (1.37,2.37);
p-value < 0.05
Eye disease, combined
<65 years: 1.25 (1.01,
1.56); p-value < 0.05
>65 years: 1.19(0.71,
1.98)
All other outcomes: No
statistically significant
associations
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.15 Dermal
Table D-25. Associations Between PFOA Exposure and Dermal Health Effects in Recent Epidemiologic Studies
Reference, Location,
Confidence 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: -5.16 (-8.50,-1.82)
(2019, 5080529) 1999-2017
women and
(1st trimester)
occurrence
coefficient per
T3: -6.09 (-12.10,-1.70)
Medium
their children
Girls Sample 1:
(months)
log2-unit
from the
4.8 (2.7-8.2)
increase in
Boys: -1.06 (-3.62, 1.49);
Puberty Cohort
Girls Sample 2:
PFOA, or by
p-value = 0.58
4.1 (2.3-6.4)
tertiles
D-256
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Study Design
Population,
Ages,
N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Resultsb
within the
DNBC
N = 555 girls,
565 boys
Boys Sample 1:
5.1 (2.8-8.3)
Boys Sample 2:
4.3 (2.2-6.7)
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.16 Cancer
Table D-26. Associations Between PFOA Exposure and Cancer in Recent Epidemiologic Studies
Reference,
Confidence
Location,
Years
Design Population, Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Eriksen et al.
(2009, 2919344)
Medium
Denmark Cohort
1993-2006
Adults with no
previous cancer
diagnosis,
Ages 50-65 at
enrollment,
Prostate cancer,
1,393;
Bladder cancer,
1,104;
Pancreatic cancer,
900;
Liver cancer, 839
Serum Cancers:
Mean (5th-95th prostate,
percentile):
Cases, men: 6.8
(3.1-14.0);
Controls, men:
6.9 (3.2-13.3);
Cases, women:
6.0 (2.6-11.0);
Controls,
women: 5.4
(2.2-11.6)
bladder,
pancreatic
, liver
IRR per unit Prostate cancer:
increase in PFOA, or Q2: 1.09 (0.78, 1.53)
by quartiles Q3: 0.94 (0.67, 1.32)
Q4: 1.18 (0.84, 1.65)
Per unit increase: 1.03 (0.99, 1.07)
Bladder cancer:
Q2: 0.71 (0.46, 1.07)
Q3: 0.92 (0.61, 1.39)
Q4: 0.81 (0.53, 1.24)
Per unit increase: 1.00 (0.95, 1.05)
Pancreatic cancer:
Q2: 0.88 (0.49, 1.57)
Q3: 1.33 (0.74,2.38)
D-257
-------
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: 1.55 (0.85, 2.80)
Per unit increase: 1.03 (0.98, 1.1)
Liver cancer:
Q2: 1.0 (0.44, 2.23)
Q3: 0.49 (0.22, 1.09)
Q4: 0.60 (0.26, 1.37)
Per unit increase: 0.95 (0.86, 1.06)
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 cancer0
Bonefeld-
Jorgensen et al.
(2011,2150988)
Medium
Greenland
2000-2003
Case-
control
Greenlandic Inuit
women with and
without breast
cancer, 76
Plasma Breast
Cases: 2.5 cancer
(Range = 0.2-
7.2)
Controls: 1.6
(Range = 0.2-
7.6)
Confounding: Age, BMI, pregnancy, cotinine, breastfeeding, and menopausal status
OR per ln-unit
increase in PFOA
1.2 (0.77, 1.88), p-value = 0.43
Barry et al. United States Cohort
Adults from C8
Modeled
Cancers
HR per unit increase
Kidney cancer (no lag):
(2013,2850946) 2005-2006
Health Project, Ages
Community:
(no-lag
in PFOA, or by
Q2: 1.23 (0.70,2.17)
Medium
> 20 years, 32,254
19.4
and 10-
quartiles
Q3: 1.48 (0.84,2.60)
(Range = 2.8-
year lag):
Q4: 1.58 (0.88,2.84)
9,217
kidney,
p-trend = 0.18
Worker: 174.4
testicular,
Per unit increase: 1.1 (0.98, 1.24), p-
(Range = 5.2-
thyroid,
value = 0.1
3,683)
breast,
lung
Kidney cancer (10-year lag):
Q2: 0.99 (0.53, 1.85)
Q3: 1.69 (0.93,3.07)
Q4: 1.43 (0.76,2.69)
D-258
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location,
Confidence Years
Design Population, Ages, N
Exposure
Matrix,
Levels
Outcome Comparison
Select Resultsb
(ng/mL)a
p-trend = 0.34
Per unit increase: 1.09(0.97, 1.21), p-
value = 0.15
Testicular cancer (no lag):
Q2: 1.04 (0.26,4.22)
Q3: 1.91 (0.47,7.75)
Q4: 3.17 (0.75, 13.45)
p-trend = 0.04
Per unit increase: 1.34 (1.00, 1.79), p-
value = 0.05
Testicular cancer (10-year lag):
Q2: 0.87 (0.15,4.88)
Q3: 1.08 (0.20,5.90)
Q4: 2.36 (0.41, 13.65)
p-trend = 0.02
Per unit increase: 1.28 (0.95, 1.73), p-
value = 0.10
Thyroid cancer (no lag):
Q2: 1.54 (0.77,3.12)
Q3: 1.48 (0.74,2.93)
Q4: 1.73 (0.85, 3.54)
p-trend = 0.25
Thyroid cancer (10-year lag):
Q2: 2.06 (0.93, 4.56)
Q3: 2.02 (0.90,4.52)
Q4: 1.51 (0.67,3.39)
p-trend = 0.65
Breast cancer (no lag):
Per unit increase: 0.94 (0.89, 1.00), p-
value = 0.05
D-259
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Exposure
Matrix,
Levels
(ng/mL)a
Reference, Location, Desjgn Population, Ages, N ' Outcome Comparison Select Resultsb
Confidence Years b r Levels
Breast cancer (10-year lag):
Per unit increase: 0.93 (0.88, 0.99), p-
value = 0.03
Lung cancer (no lag):
Per-unit increase: 0.88 (0.78, 1.00), p-
value = 0.05
Lung cancer (10-year lag):
Per unit increase: 0.92 (0.81, 1.04), p-
value = 0.17
Results: Lowest quartile used as the reference group
Confounding: Time-varying smoking, time-varying alcohol consumption, sex, education, and stratified by 5-year period of birth year
United States Cohort Exposed DuPont Serum Mortality: SMR by quartiles, or Bladder cancer mortality (no lag):
1948-2009 chemical plant 4.3 ng/mL-years bladder for all quartiles Ql: 1.24 (0.15,4.47)
Steenland and
Woskie (2012,
2919168)
Medium
workers in West
Virginia,
5,791
cancer,
kidney
cancer,
mesotheli
oma
Q2: 2.49 (0.97, 5.78)
Q3: 0.39 (0.01, 2.17)
Q4: 0.36 (0.10, 2.01)
All quartiles: 1.08 (0.52,
1.99)
Kidney cancer mortality (no lag):
Ql: 1.07 (0.02, 3.62), p-value < 0.05
Q2: 1.37 (0.28, 3.99), p-value < 0.05
Q3: 0.00 (0.00, 1.42), p-value < 0.05
Q4: 2.66 (1.15, 5.24), p-value < 0.05
All quartiles: 1.28 (0.66, 2.24)
Kidney cancer mortality (10-year lag):
Ql: 1.05 (0.13, 3.79), p-value < 0.05
Q2: 0.87 (0.11, 3.15), p-value < 0.05
Q3: 0.44 (0.01, 2.44), p-value < 0.05
Q4: 2.82 (1.13, 5.81), p-value < 0.05
D-260
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design Population, Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Kidney cancer mortality (20-year lag)
Ql: 1.34 (0.28, 3.91), p-value < 0.05
Q2: 0.46 (0.01, 2.55), p-value < 0.05
Q3: 0.00 (0.00, 2.03), p-value < 0.05
Q4: 3.67 (1.48, 7.57), p-value < 0.05
Mesothelioma mortality (no lag):
Ql: 0.00 (0.00, 15.4), p-value < 0.05
Q2: 0.00 (0.00, 7.51), p-value < 0.05
Q3: 1.73 (0.04, 9.65), p-value < 0.05
Q4: 6.27 (2.04, 14.63), p-value < 0.05
All quartiles: 2.85 (1.05, 6.20), p-
value < 0.05
Results: Other DuPont workers from the region were used as the reference group
Confounding: Not reported
Mesothelioma mortality (10-year lag):
Ql: 0.00 (0.00, 17.80)
Q2: 0.00 (0.00, 9.55)
Q3: 3.08 (0.37, 11.12)
Q4: 4.66 (1.27, 11.93)
Mesothelioma mortality (20-year lag):
Ql: 9.09 (0.23,50.60)
Q2: 0.00 (0.00, 15.24)
Q3: 2.60 (0.31,9.39)
Q4: 3.44 (0.71, 10.05)
Vieira et al. United States Case-
Adults living near
Modeled
Cancers:
OR by exposure
Kidney cancer:
(2013,2919154) 1996-2005 control
the Dupont Teflon-
Low:
kidney,
category
Low: 0.8 (0.4, 1.5)
Medium
manufacturing plant,
Range = 3.7-
prostate
Medium: 1.2 (0.7, 2.0)
7,869
12.8 ng/L
High: 2.0 (1.3, 3.2)
Medium:
Very high: 2.0 (1.0, 3.9)
Range = 12.9-
30.7 ng/L
Prostate cancer:
High:
Low: 1.1 (0.8, 1.5)
D-261
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design Population, Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Range = 30.8-
109 (ig/L
Very high:
Range = 110—
655 ng/L
Medium: 0.8 (0.6, 1.0)
High: 0.8(0.5, 1.1)
Very high: 1.5 (0.9, 2.5)
Results: Unexposed population used as the reference group
Confounding: Age, race, sex, diagnosis year, insurance provider, and smoking status
Ducatman et al.
(2015, 3859843)
Medium
United States
2005-2006
Cross-
sectional
Men from C8 Health
Study,
Ages 20-49,
Ages 50-69,
9,169;
3,819
Serum Prostate- Regression
Mean(SD): specific coefficient ((3) per
40.22 (3.50) antigen ln-unit increase in
(PSA) PFOA
level GM ratio (GMR)
(PSA < 4.0 ng/mL
vs.
PSA > 4.0 ng/mL)
Confounding: Age, smoking status, average alcohol intake, and body mass index
Age 20-49
(3=1, p-value = 0.9
GMR = 1.15 (0.67, 1.98)
Age 50-69
(3=1, p-value = 0.72
GMR = 0.96 (0.77, 1.2)
Ghisari et al.
(2017, 3860243)
Medium
Denmark
1996-2002
Nested
case-
control
Adult women, 283
Serum
Cases: 4.87
Controls: 4.90
Breast
cancer
Cohort RR= 1.17(0.63,2.17)
CYP19 CC RR = 7.24 (1.00, 52), p-
value < 0.05
No significant associations observed
for remaining genotypes
Relative risk ratio
(RR) per ln-unit
increase in PFOA.
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
D-262
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design Population, Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
Hurley et al.
(2018, 5080646)
Medium
California,
US
2011-2015
Nested
case-
control
Adult women, 1,760 Serum
Median (min-
max):
Cases: 2.350
(0.042-39.100)
Controls: 2.475
(0.096-20.200)
Breast OR per loglO-unit T2: 0.901 (0.705, 1.152)
cancer increase inPFOA, or T3: 0.925 (0.715, 1.197)
(invasive) bytertiles Per unit increase: 0.733 (0.496, 1.081),
p-value = 0.11
Results: Lowest tertile used as the reference group
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, United States Nested
5412451) 1959-1967 case-
Medium control
Adult daughters of
women in CHDS
cohort, 310 controls,
102 cases
Perinatal serum
Cases: 30.5
(14.1-55.8)
Controls: 0.4
(0.2-0.6)
Breast OR per log2-unit
cancer increase in PFOA
"found no associations;" No results
reported
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. France Nested
(2020,5381529) 1990-2013 case-
Medium control
Postmenopausal Serum
women, Ages 40-65 6.64 (1.29-
in 1990, 194 cases, 21.39)
194 controls
Breast ORs by quartiles, Overall:
cancer and by estrogen Q2: 1.69 (0.89, 3.21)
(ER) or progesterone Q3: 0.88 (0.43, 1.8)
receptor (PR) status Q4: 0.92 (0.43, 1.98)
p-trend = 0.43
ER negative:
ORs of 3-7
p-trend = 0.59
PR negative: ORs of 1-4
p-trend = 0.90
Results: Lowest quartile used as the reference group
D-263
-------
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: 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
United States Nested
1993-2014 case-
control
Adults, 55-74, 648
Ages 55-59, 190
Ages 60-65,224
Ages 65+, 234
Males 432
Females 216
Serum Renal cell ORs per log2-unit
5.5(4.0-7.3 carcinoma increase in PFOA or
ug/L) by quartiles (total
cohort only)
Q2: 1.47 (0.77,2.8)
Q3: 1.24 (0.64,2.41)
Q4: 2.63 (1.33, 5.2)
p-trend = 0.007
Per unit increase: 1.71 (1.23, 2.37)
55-59:2.1 (1.21,3.34)
60-65: 1.6(1,2.45)
65+: 1.79 (1.21,2.77)
p-heterogeneity = 0.66
Males: 1.7(1.31,2.35)
Females: 1.79 (1.1,2.95)
p-heterogeneity = 0.87
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
Fry and Power
US NHANES Cohort
Adults,
Serum
Cancer
Hazard ratio per SD
0.94 (0.8, 1.11), p-value = 0.45
(2017, 4181820)
2003-2006
Ages 60+, 1,032
Median (SE):
mortality
unit increase in
Medium
23.7 (0.7) ng/g
PFOA
lipid
Confounding: Age, gender, race/ethnicity, and smoking status
Steenland et al.
United States Retrospect
Adult workers, 3,713
Drinking
Cancers
IRR by quartiles
Bladder cancer no lag:
(2015,2851015)
2008-2011 ive
water/occupatio
with and
Q2: 0.32 (0.08, 1.33)
Low
occupation
nal, serum
without a
Q3: 0.95 (0.28,3.14)
al cohort
Median =113;
10-year
Q4: 0.23 (0.05, 0.93)
Cumulative
lag:
p-trend logPFOA cumulative
exposure, 25th-
bladder,
exposure = 0.04
75 th percentiles
colorectal,
p-trend via quartiles = 0.19
with or without
D-264
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design Population, Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
10-year lag:
0.8-7.04 or
3.03-
11.42 ng/mL-
year
melanoma
, prostate
Bladder cancer with lag:
Q2: 0.55 (0.12,2.61)
Q3: 0.47 (0.1,2.21)
Q4: 0.31 (0.06, 1.54)
p-trend logPFOA cumulative
exposure = 0.06
p-trend via quartiles = 0.03
Colorectal, melanoma and prostate
cancers report p-trends of 0.10 or
greater
Results: Lowest quartile used as the reference group
Confounding: Gender, race, education, BMI, smoking, alcohol consumption
Christensen et al.
(2016, 3858533)
Low
Wisconsin, Cross- Male anglers, Ages Serum
US, 2012- sectional 50+, 154 2.50 (1.80-
2013 3.30)
Confounding: Age, BMI, work status, alcohol consumption
Cancer OR per unit increase
(any) in PFOA
1.5 (1.08,2.17)
Girardi and
Italy
Occupatio Male workers, 154
Occupational,
Mortality:
Mortality risk ratio
Merler (2019,
1960-2018
nal
serum
Liver
(RR) by tertiles for
6315730)
Retrospect
GMby
cancer,
PFAS plant workers
Low
ive Cohort
tertiles = 1,700;
liver
vs. nearby metal
13,051; and
cancer or
factory workers
81,934 ng/mL-
cirrhosis,
years
lung
Standardized
cancer,
mortality ratio in
malignant
each cumulative
neoplasm,
PFOA tertile
malignant
neoplasms
of
lymphatic
and
hematopoi
Malignant neoplasms of lymphatic and
hematopoietic tissues
RRT1: 1.44 (0.18, 11.8)
RRT2: 1.8 (0.22, 14.6)
RRT3: 5.06 (1.61, 16)
p-trend < 0.001
Any malignant neoplasm p-
trend = 0.008
All other mortalities not significant
D-265
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference,
Confidence
Location,
Years
Design Population, Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
etic
tissues
Confounding: Age at risk, calendar period
Lin et al. (2020,
6835434)
Low
Tsai et al. (2020,
6833693)
Low
China
Case-
Children,
Serum
2014-2017 control Ages < 16, 84
Germ cell OR per unit increase 1.03 (0.99,1.08)
tumors in PFOA
13.89 (8.05-
21.37)
Confounding: Infectious disease, cosmetics usage, barbecued food consumption, filtered water use, indoor decorating, living near farmland
(maternal behaviors/factors during pregnancy)
Plasma Breast OR per ln-unit
Mean(GM): cancer increase in PFOA
2.15 (1.77)
Taiwan
2014-2016
Case-
control
Adult women, 239
Age 50 or younger,
120
Age over 50, 119
Total cohort: 1.14 (0.66, 1.96)
Age 50 or younger: 0.78 (0.4, 1.51)
Age over 50: 0.89 (0.59, 1.34)
Confounding: Pregnancy history, oral contraception use, abortion, BMI, menopause, and education level
Itoh et al. (2021,
9959632)
Low
Japan
2001-2005
Case-
control
Serum
5.57 (3.98-
7.62)
Breast
cancer
OR by quartiles
Q2: 0.37 (0.19, 0.73), p-value < 0.05
Q3: 0.39 (0.18, 0.84), p-value < 0.05
Q4: 0.20 (0.08. 0.51), p-value < 0.05
p-trend = 0.001
Adult women,
Ages 20-74,
802 (401 breast
cancer cases, 401
controls)
Results: Lowest quartile used as the reference group
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,
10176563)
Low
China Case-
2016-2017 control
Adult men, 96
Adult women, 223
Serum
Case: 7.7 (4.4-
12.8); Control:
10.9(7.9-16.1)
Thyroid
cancer
OR by quartiles
Total
Q2: 0.24 (0.12,0.50)
Q3: 0.24 (0.11,0.49)
Q4: 0.20 (0.09, 0.44)
p-trend < 0.001
Male:
Q2: 0.15 (0.03,0.76)
Q3: 0.18 (0.04, 0.85)
Q4: 0.32 (0.08, 1.34)
D-266
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference, Location,
Confidence
Years
Design Population, Ages, N
Exposure
Matrix,
Levels
(ng/mL)a
Outcome Comparison
Select Resultsb
P-trend = 0.313
Female:
Q2: 0.31 (0.14,0.71)
Q3: 0.28 (0.12,0.63)
Q4: 0.25 (0.10,0.59)
p-trend = 0.006
Results: Lowest quartile used as the reference group
Confounding: Age, sex, and diabetes status
Omoike et al. United States Cross-
(2021,7021502) 2005-2012 sectional
Low
Adults from
NHANES,
Ages > 20 years,
6,652
Serum
3.20 (2.00-
4.90)
Cancers: OR per unit increase Ovarian cancer:
ovarian, in PFOA, or by
breast, quartiles
uterine,
and
prostate
Q2: 0.07 (0.07, 0.072)
Q3: 0.69 (0.68,0.70)
Q4: 1.77(1.75, 1.79)
p-trend < 0.001
Per unit increase: 1.015 (1.013, 1.017)
Breast cancer:
Q2: 2.40 (2.38, 2.42)
Q3: 1.39 (1.38, 1.40)
Q4: 2.30 (2.28, 2.31)
p-trend < 0.001
Per unit increase:
1.089 (1.089, 1.090)
Uterine cancer:
Per unit increase:
0.912(0.910,0.914)
Prostate cancer:
Per unit increase:
0.944 (0.943, 0.944)
Results: Lowest quartile used as the reference group
Confounding: Age, sex, education, race/ethnicity, PIR, BMI, and serum cotinine
D-267
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Notes: CHDS = The Child Health and Development Studies GM = geometric mean; HR = hazard ratio; IRR = incidence rate ratio; NHANES = National Health and Nutrition
Examination Survey; OR = odds ratio; SD = standard deviation; SE = standard error; SMR = standardized mortality ratio.
a Exposure levels reported as median (25th-75th percentile) in ng/mL 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-268
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix E. Benchmark Dose Modeling
E.l Epidemiology Studies
E.l.l Modelling 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 PFOA
Exposure Measured at Five Years of Age
Budtz-j0rgensen and Grandjean (2018, 5083631) fit multivariate models of PFOA 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 PFOS (as
log2[PFOS]), and without PFOS. 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 PFOA 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 PFOA
exposure without adjustment for PFOS using a likelihood ratio test (p = 0.76; see Budtz-
J0rgensen and Grandjean (2018, 5083631) Table 3), or for the model that did adjust for PFOS
(log2[PFOS]) (p = 0.69).
Table E-l summarizes the results from Budtz-j0rgensen and Grandjean (2018, 5083631) for
PFOA 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).9 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)).
9 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-269
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-l. Results specific to the slope from the linear analyses of PFOA 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
™ . u PFOS
Model shape ,. , ,
adjusted
Slope (P)
per ng/mL
SE(P)
ng/mL
Slope (P) fit
Lower
bound slope
(Plb)
ng/mL
PFOA at Age 5
PFOA at Age 5
Linear No
Linear Yes
-0.197
-0.185
0.0630
0.0697
p = 0.002
p = 0.008
-0.301
-0.299
Notes: SE = standard error
Interpretation of results in Table E-l:
• PFOA is a significant predictor in the single-PFAS model (P = -0.197; p = 0.002).
• Effects of PFOA in the single-PFAS model are attenuated when log2[PFOS] is included in
the model (P = -0.185; p = 0.008).
• The point estimate results for PFOA (P) in the single-PFAS model are potentially
confounded by PFOS since there was a 5% reduction in the effect size for PFOA from
-0.197 to -0.185 when controlling for PFOS.
• One explanation is that PFOS was a confounder of the PFOA 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 point estimate is the best representation of any
effect of PFOA.
• 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 5% difference in the BMD and a negligible
difference in the BMDL when PFOS is included in the model.
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
E-270
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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, 10428559} reviewed by the SAB
PFAS Review Panel, EPA relied on the BMDL modeling approach published in Budtz-
J0rgensen and Grandjean (2018, 5083631), which used a 5% fixed change in the distribution of
antibody concentrations as the BMR to derive BMDs and BMDLs. 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) mention the same concentration. However, the 2018 WHO
update {WHO, 2018, 10406857} argues that:
"...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
E-271
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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
Statistically, the Technical Guidance additionally suggests that studies of developmental effects
can support lower BMRs. 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-Jorgensen 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,
E-272
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 PFOA. 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 was (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 width of
the IQR 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-273
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
0.25
si
CD
-Q
O
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)]
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 PFOA 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 PFOS
Estimated with control of PFOS
BMR BMD (ng/mL) BMDL (ng/mL) BMD (ng/mL) BMDL (ng/mL)
P = -0.197 per ng/mL 0lb = -0.301 per ng/mL p = -0.185 per ng/mL 0lb = -0.299 per ng/mL
!/> SD 5.30 3.47a 5.66 3.49
1 SD 10.6 6.94 11.3 6.98
Notes:
a Denotes the selected POD.
The lowest serum PFOA concentration measured at age five years was 0.8 ng/mL, the 5th
percentile was 2.4 ng/mL, and the 10th percentile was 2.8 ng/mL {Grandjean, 2021, 9959716}
so the estimated BMDL for a BMR of V2 SD (BMDL', sd = 3.47 ng/mL) in the single-PFAS
model is well within the observed range (Table E-2). No information was available to judge the
E-274
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 PFOA well (p = 0.002).
The BMD>, sd estimate from the multi-PFAS models is 5% higher than the BMD', sd estimate
from the models with just PFOA, and the BMDL , sd estimates are the same. 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 PFOA in light of potential confounding, the two BMDL, sd estimates
are the same (3.53 ng/mL). EPA advanced the derivation based on results that did not control for
PFOS because this model appeared to fit PFOA better (p = 0.002 vs. 0.006) and there was no
uncertainty due to potential confounding in the BMDL. Overall confidence in the BMDLs for
tetanus was judged to be high.
For immunotoxicity related to tetanus associated with PFOA exposure measured at age five
years, the POD is based on a BMR of Vi SD and a BMDLy2 sd of 3.47 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 PFOA
Exposure Measured Perinatally
Budtz-j0rgensen and Grandjean (2018, 5083631) fit multivariate models of PFOA 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 PFOS (as log2[PFOS]), and without PFOS.
Three model shapes of PFOA 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
PFOA exposure without adjustment for PFOS using a likelihood ratio test (p = 0.25; see Budtz-
J0rgensen and Grandjean (2018, 5083631) Table 3), or for the model that did adjust for PFOS
(log2[PFOS]) (p = 0.26).
Table E-3 summarizes the results from Budtz-j0rgensen and Grandjean (2018, 5083631) for
tetanus 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 tetanus antibody concentrations
in Table 2 of Budtz-j0rgensen and Grandjean (2018, 5083631).9
Table E-3. Results of the linear analyses of PFOA 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 PFOS adjusted
Slope (P)
per ng/mL
SE(P)
ng/mL
Slope (P) fit
Lower
bound slope
(Plb)
ng/mL
Perinatal PFOA
Perinatal PFOA
Linear No
Linear Yes
-0.135
-0.126
0.0601
0.0685
p = 0.03
p = 0.07
-0.234
-0.239
E-275
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Note: SE = standard error.
Interpretation of results in Table E-3:
• PFOA is a significant predictor in the single-PFAS model (P = -0.135; p = 0.03).
• Effects are attenuated when log2[PFOS] are included in the model (P = -0.126; p = 0.07).
• The point estimate results for PFOA are potentially confounded by PFOS since there was
a 7% reduction in the effect size for PFOA from -0.135 to -0.126 when controlling for
PFOS.
• One explanation is that PFOS was a confounder of the PFOA 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
PFOA.
• 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 7% difference in the BMD and a 3% 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, 10428559} 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 one 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
E-276
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 IQR of the tetanus antibody concentrations
in the earlier birth cohort at age five years in IU/mL was (0.1, 0.51). Log2-tranforming these
values provides the IQR in log2(IU/mL) as (-3.32, -0.97). Assuming that these log2-transformed
values are similar to the normal distribution, the width of 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). The IQR 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 IQR 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).10 To show the
impact of the BMR on these results, Table E-4 presents the BMDs and BMDLs at BMRs of '/2
SD and 1 SD.
Table E-4. BMDs and BMDLs for effect of PFOA measured perinatally and anti-tetanus
antibody concentrations at age five years {Budtz-Jorgensen, 2018, 5083631}
Estimated without control of PFOS
Estimated with control of PFOS
BMR
BMD (ng/mL)
BMDL (ng/mL)
BMD (ng/mL)
BMDL (ng/mL)
P = -0.135 per ng/mL
Plb = -0.234 per ng/mL
P = -0.126 per ng/mL
Plb = -0.239 per ng/mL
V2SD
5.76
3.31a
6.17
3.25
1 SD
11.5
6.62
12.3
6.49
Notes:
a Denotes the selected POD.
The lowest perinatal maternal serum PFOA concentration measured was 0.8 ng/mL, the 5th
percentile was 1.7 ng/mL, and the 10th percentile was 2.0 ng/mL {Grandjean, 2021, 9959716} so
the estimated BMDLs for a BMR of V2 SD (BMDLi/2 sd = 3.31 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 PFOA well.
The BMD>, sd estimate from the multi-PFAS models is 7% lower than the BMD', sd estimate
from the models with just PFOA, and the BMDL , sd estimates is 3% lower. 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 PFOA in light of potential confounding, the two BMDL , sd estimates
are comparable (3.35 ng/mL vs. 3.25 ng/mL). EPA advanced the derivation based on results that
did not controls for PFOS because this model appeared to fit PFOA data better (p = 0.02 vs.
10 Pooled variance for tetanus in five-year olds = [(502-1 )(1.'74)2+(298—1 )(1.17)2]/[502+298—2] =2.41. The pooled SD is the
square root of 2.41 which is 1.55 log2(IU/mL).
E-277
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
0.07) and there was little uncertainty due to potential confounding in the BMDL. Overall
confidence in the BMDLs for tetanus was judged to be high.
For immunotoxicity related to tetanus associated with PFOA exposure measured at age five
years, the POD is based on a BMR of Vi SD and a BMDLy2 sd of 3.31 ng/mL.
E.1.1.1.3 Timmerman et al. (2021, 9416315)
Timmerman et al. (2021, 9416315) analyzed data from Greenlandic children ages 7-12 and fit
multivariate models of PFOA against loglO-transformed anti-tetanus antibody concentrations
measured at the same time as PFOA, 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 backtransformed to express the percent difference in
antibody concentrations at each ng/mL increase in serum PFOA concentrations in children,
which was -8 (95% CI: -30, 21) (Table 4, Timmerman et al. (2021, 9416315)). Using the
equation provided below, EPA estimated the regression slope as -0.036 (95% CI: -0.155,
0.083).
Percent Difference = (10^ — 1) x 100
Following the approach described previously for Budtz-j0rgensen and Grandjean (2018,
5083631), EPA derived BMDs and BMDLs for both a one SD change in the distribution of
log 10 (tetanus antibody concentrations) as a standard reporting level, and V2 SD change in the
distribution of loglO (tetanus antibody concentrations). The SD of the loglO (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 (2021, 9416315)). Logio-
transforming these values provides the 25th and 75th percentiles in logio (IU/mL) as (-0.60,
0.34). Assuming that these logio-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, 2017). Thus, SD = IQR/1.35, and the SD of tetanus antibodies
in log 10 (IU/mL) is (0.34 - (-0.60))/1.35 = 0.70 logio (IU/mL).
Table E-5. BMDs and BMDLs for effect of serum PFOA in children on anti-tetanus
antibody concentrations using a BMR of Vz SD change in logio (tetanus antibodies
concentration) and a BMR of 1 SD change in logio (tetanus antibodies concentration)
Timmerman et al. (2021, 9416315)
BMR
BMD (ng/mL)
BMDL (ng/mL)
P = -0.036 per ng/mL
P = -0.155 per ng/mL
!/2SD
9.66
2.26
1 SD
19.3
4.52
Note: 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),
EPA calculated that 8.4% of those values would be below the cutoff value of 0.1 IU/mL. A BMR
of V2 SD resulted in 19% of the values being below that cutoff which is 10.6% extra risk. This
suggest that in this case a BMR of V2 SD may not be a reasonably good estimate of 5% extra risk.
E-278
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Note that this BMDL is based on a non-significant PFOA regression parameter (P) estimated as
-0.013 (95% CI: -0.036, 0.013) (Timmerman, 2021, 9416315), and thus this POD is identified
with lower confidence.
For immunotoxicity related to tetanus associated with PFOA exposure measured at ages
five to ten years old, the POD estimated for comparison purposes was based on a BMR of
Vz SD and a BMDLy2 sd of 2.26 ng/mL.
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 2.3 to 12.1, with the selected POD of 3.47. The comparison
POD of 2.26 ng/mL is considered lower confidence because it is based on a non-significant
PFOA regression parameter.
Table E-6. BMDLs for effect of PFOA on anti-tetanus antibody concentrations using a
BMR of1/! SD
Study
Effect
BMDLy, sd (ng/mL)
ViSD
Budtz-Jorgcnscn
and Grandjean
(2018, 5083631)
Budtz-Jorgcnscn
and Grandjean
(2018, 5083631)
Timmerman et al. PFOA and anti-tetanus antibody concentrations
(2021, 9416315) at ages 7-10 years
PFOA at age five years and anti-tetanus
antibody concentrations at age seven years
PFOA perinatally and anti-tetanus antibody
concentrations at age five years
3.47
3.31
2.26
1.05 log2(lU/mL)
0.78 log2(lU/mL)
0.35 logio(lU/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 PFOA
Exposure Measured at Five Years of Age
Budtz-j0rgensen and Grandjean (2018, 5083631) fit multivariate models of PFOA 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 PFOS (as
log2[PFOS]), and without PFOS. Three model shapes were evaluated by Budtz-j0rgensen and
Grandjean (2018, 5083631) using likelihood ratio tests: a linear model of PFOA, 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 PFOA exposure without
adjustment for PFOS using a likelihood ratio test (p = 0.86; see Budtz-j0rgensen and Grandjean
(2018, 5083631) Table 3), or for the model that did adjust for PFOS (log2[PFOS]) (p = 0.92).
E-279
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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).9
Table E-7. Results specific to the slope from the linear analyses of PFOA measured at age
five years and log2(diphtheria antibodies) 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 PFOS adjusted
Slope (P)
per ng/mL
SE(P)
ng/mL
Slope (P) fit
Lower
bound slope
(Plb)
ng/mL
PFOA at Age 5
PFOA at Age 5
Linear No
Linear Yes
-0.126
-0.0867
0.0588
0.0649
p = 0.03
p = 0.18
-0.223
-0.194
Note: SE = standard error.
Interpretation of results in Table E-7:
• PFOA is a significant predictor in the single-PFAS model (P = -0.126; p = 0.03).
• Effects are attenuated when log2[PFOS] are included in the model (P = -0.0867;
p = 0.18).
• The point estimate results for PFOA are potentially confounded by PFOS since there was
a 30% reduction in the effect size for PFOA from -0.126 to -0.0867 when controlling for
PFOS.
• One explanation is that PFOS was a confounder of the PFOA 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
PFOA.
• 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 30% difference in the BMD and 15% 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, 10428559} 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
E-280
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 one SD change in the distribution of log2(diphtheria antibody concentrations),
and V2 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 (1993,
10228565) mentions the same concentration, but Galazka et al. (1993, 10228565) 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 V2 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 et al.
(1993, 10228565) mentions a case fatality rate of 5-10%) and PFOA-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)
E-281
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 interquartile range (IQR) of the diphtheria antibody concentrations at age 7 years in IU/mL
was (0.4, 1.6). Log2-tranforming these values provides the IQR in log2(IU/mL) as (-1.32, 0.68).
Assuming that these log2-transformed values are similar to the normal distribution, the width of
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 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 PFOA 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 PFOS
Estimated with control of PFOS
BMR
BMD (ng/mL) BMDL (ng/mL)
P = -0.126 per ng/mL 0lb = -0.223 per ng/mL
BMD (ng/mL) BMDL (ng/mL)
P = -0.0867 per ng/mL 0lb = -0.194 per ng/mL
V2SD
1 SD
5.88 3.32a
11.8 6.64
8.53 3.82
17.1 7.64
Notes:
a Denotes the selected POD.
The lowest serum PFOA concentration measured at age five years was 0.8 ng/mL, the 5th
percentile was 2.4 ng/mL, and the 10th percentile was 2.8 ng/mL {Grandjean, 2021, 9959716}
so the estimated BMDL for a BMR of V2 SD (BMDLi/2 sd = 3.32 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 PFOA well (p = 0.03).
The BMD>, sd estimate from the multi-PFAS models is 44% higher than the BMD', sd estimate
from the model with just PFOA, and the BMDLusnis 15% 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 PFOA in
light of potential confounding, the two BMDL , sd estimates which serve as the PODs are
comparable (3.30 ng/mL vs. 3.80 ng/mL). EPA advanced the POD based on results that did not
controls for PFOS because this model appeared to fit PFOA data better (p = 0.04 vs. 0.18) and
there was low uncertainty due to potential confounding in the BMDL. However, confidence was
diminished by 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 confidence.
For immunotoxicity related to diphtheria, associated with PFOA measured at age five
years, the POD is based on a BMR of V2 SD and a BMDLy2 sd of 3.32 ng/mL.
E-282
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E. 1.1.2.2 Budtz-J0rgensen and Grandjean (2018, 5083631) Results for
Decreased Diphtheria Antibody Concentrations at Five Years of Age and PFOA
Exposure Measured Perinatally
Budtz-j0rgensen and Grandjean (2018, 5083631) fit multivariate models of PFOA 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 PFOS (as log2[PFOS]), and without PFOS. Three model shapes were evaluated by
Budtz-j0rgensen and Grandjean (2018, 5083631) using likelihood ratio tests: a linear model of
PFOA, 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}. There was evidence that the piecewise-linear model fit better than the linear
model for the PFOA exposure without adjustment for PFOS (p = 0.012; see in Budtz-j0rgensen
and Grandjean (2018, 5083631), Table 3), and for the model that adjusted for PFOS
(log2[PFOS]) (p = 0.05). 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).9
Table E-9. Results of the analyses of PFOA 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
PFOS
adjusted
Slope (P)
per ng/mL
SE(P)
Slope (P) fit
Lower
bound slope
(Plb)
Perinatal PFOA
Piecewise
No
-0.495
0.163
p = 0.003
-0.764
Perinatal PFOA
Piecewise
Yes
-0.347
0.180
p = 0.05
-0.644
Notes: SE = standard error.
Interpretation of results in Table E-9:
• PFOA is a significant predictor in the single-PFAS model (P = -0.495; p = 0.003).
• Effects of PFOA are attenuated when PFOS is in the model (P = -0.347; p = 0.05).
• Results for PFOA are potentially confounded by PFOS since there was a 30% change in
the effect size for PFOA from -0.495 to -0.347 when controlling for PFOS
• One explanation is that PFOS was a confounder of the PFOA 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
PFOA.
E-283
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• 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 30% 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, 10428559} 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 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 authors.
EPA then pooled the summary statistics to describe the common SD. The IQR of the diphtheria
antibody concentrations in the earlier birth cohort at age five years in IU/mL was (0.05, 0.4).
Log2-tranforming these values provides the IQR in log2(IU/mL) as (-4.32, -1.32). Assuming
that these log2-transformed values are similar to the normal distribution, the width of 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 IQR of the diphtheria 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 IQR 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
E-284
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
log2(IU/mL).11 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 PFOA measured perinatally and anti-
diphtheria antibody concentrations at age five years {Budtz-Jorgensen, 2018, 5083631}
Estimated without control of PFOS
Estimated with control of PFOS
BMR
BMD (ng/mL)
P = -0.495 per ng/mL
BMDL (ng/mL)
Plb = -0.764 per ng/mL
BMD (ng/mL) BMDL (ng/mL)
P = -0.347 per ng/mL Plb = -0.644 per ng/mL
V2SD
1 SD
1.92
3.84
1.24a
2.49
2.74 1.47
5.47 2.95
Notes:
a Denotes the selected POD.
The lowest serum PFOA concentration measured perinatally was 0.8 ng/mL, the 5th percentile
was 1.7 ng/mL, and the 10th percentile was 2.0 ng/mL {Grandjean, 2021, 9959716} so the
estimated BMD for a BMR of V2 SD (BMDL', sd = 1.24 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 PFOA well.
The BMD>, sd estimate from the multi-PFAS model is 43% higher than the BMD', sd estimated
from the model with just PFOA, and the BMDLusnis 19% 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 (1.24 ng/mL vs. 1.47 ng/mL) and EPA advanced the
derivation based on results that did not control for PFOS because this model appeared to fit
PFOA well (p = 0.003 vs. 0.05) and there was moderate uncertainty due to potential confounding
in the BMD and low uncertainty in the BMDL. Medium confidence in the BMDLs from PFOA
linear model (1.24 ng/mL) without control of PFOS since the model fit well and these BMDLs
show low uncertainty about confounding.
For immunotoxicity related to diphtheria, associated with PFOA measured at age five years, the
POD is based on a BMR of y2 SD and a BMDL^ SD of 1.24 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 PFOA against loglO-transformed anti-diphtheria antibody concentrations
measured at the same time as PFOA, 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 PFOA concentrations in children,
which was -22 (95% CI: -48, 16) (Table 4, Timmerman et al. (2021, 9416315)). Using the
equation provided below, EPA estimated the regression slope as -0.11 (95% CI: -0.28, 0.06).
11 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-285
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Percent Difference = (10^ — 1) x 100
Following the approach 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 median (25th, 75th percentiles) of 0.07 (0.02 and 0.28)
diphtheria antibody concentrations in IU/mL (Table 1 in Timmerman (2021, 9416315)). Logio -
transforming these values provides the 25th and 75th percentiles in logio (IU/mL) as (-1.70,
-0.55). Assuming that these logio -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. Thus, SD = IQR/1.35, and the SD of tetanus antibodies in logio
(IU/mL) is (-0.55 - (—1.70))/1.35 = 0.85 logio (IU/mL).
Table E-ll. BMDs and BMDLs for effect of PFOA 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 (diphtheria antibodies
concentration)
BMR
BMD (ng/mL)
BMDL (ng/mL)
P = -0.11 per ng/mL
P = -0.28 per ng/mL
!/2SD
3.93
1.49
1 SD
7.87
2.99
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 V2 SD resulted in 75% of the values being below that cutoff which is 18%
extra risk. This suggests that in this case the BMR of V2 SD may not be a reasonably good
estimate of 5% extra risk. This POD is considered lower confidence because it is based on a non-
significant PFOA regression parameter.
For immunotoxicity related to tetanus associated with PFOA 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 BMDL>/2sd of 1.49 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
diphtheria antibody concentrations. The selected and comparison PODs were based on a BMR of
V2 SD, resulting in BMDLs ranging from 1.24 to 14.5 ng/mL, with the selected POD of 3.32. The
comparison POD of 1.49 is considered lower confidence because it is based on a non-significant
PFOA regression parameter.
E-286
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-12. BMDLs for effect of PFOA on anti-diphtheria antibody concentrations using a
BMRof'/i SD
Study name Effect BMDL (ng/mL) Vi SD
Budtz-Jorgcnscn PFOA at age five years on anti-diphtheria 3.32 0.74 log2(IU/mL)
and Grandjean antibody concentrations at age seven years
(2018, 5083631)
Budtz-Jorgensen PFOA perinatally on anti-diphtheria antibody 1.24 0.95 log2(IU/mL)
and Grandjean concentrations at age five years
(2018, 5083631)
Timmerman et al. PFOA and anti-diphtheria antibody 1.49 0.48 logio(IU/mL)
(2021, 9416315) concentrations at ages 7-10 years
E.1.1.3 Modeling Results for Decreased HiB Antibody Concentrations
Abraham et al. (2020, 6506041) is a low confidence study; however, it is the only available
epidemiological study providing data on HiB antibody concentrations in children. In accordance
with the IRIS Handbook, data from a low confidence study may be considered for dose response
if "only low confidence studies had adequate data for... derivation" {U.S. EBA, 2022, 10476098}.
Though Abraham et al. (2020, 6506041) did not move forward for toxicity value derivation since
medium confidence studies exist for this endpoint (see above), the EBA is providing the results of
the decreased HiB antibody concentrations modeling as another line of support for the RfD based
on decreased antibody production in children. Abraham et al. (2020, 6506041) provides evidence
of decreased antibody production for a different vaccine type and is in another population outside
of the Faroe Islands.
NOAEC/LOAEC method. Abraham et al. (2020, 6506041) examined associations between
diminished vaccine response and serum levels of BFOA in 1 year old children in Germany. The
study evaluated dose-response associations using BFOA serum concentrations divided in
quintiles and deciles, and ANOVA was used to evaluated differences in mean antibody response
between the groups. If differences between the groups were present, the study authors derived a
NOAEC as the highest dose quantile below the first one showing a significant difference to the
lowest-dose quantile {Abraham, 2020, 6506041}. The results from this study for HiB antibody
levels are shown in Table E-13 with the NOAEC bolded. The p-values in this table were
calculated by the study authors using t-tests. As shown, the NOAECs was 19.4 ng/mL, the dose
in the 4th quintile. The authors also presented data by BFOA deciles, which involved smaller
sample sizes and less statistical power, but showed a similar NOAEC in the 8th decile of
20.5 ng/mL (Table S2, Abraham et al. (2020, 6506041)).
BMP method. EBA conducted dose response modeling of decreased HiB antibody levels in the
Abraham et al. (2020, 6506041) study using BMDS v3.3rcl0. BMRs of a change in the mean
equal to V2 and 1 SD from the control mean were used. Continuous models were used to fit the
dose-response data, summarized in Table E-13. The BMD modeling results are summarized in
Table E-14.
E-287
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-13. Dose-Response Modeling Data for Decreased HiB Antibody Levels in Abraham
et al. (2020, 6506041)
Dose
(ng/mL)
N
Mean Response (mg/dL)a
p-valueb
3.4
20
1.84 ±0.68
8.5
20
1.84 ±0.71
0.98
14.5
20
1.84 ±0.84
0.98
19.4
20
1.50 ±0.55
0.09
25.7
20
1.19 ±0.60
0.003
Notes:
a Data are presented as mean ± standard deviation.
bp-value for differences in anybody levels compared to lowest dose.
For immunotoxicity related to HiB antibody levels associated with PFOA exposure
measured at age one years old, the POD was based on a BMR of Vz SD and a BMDLy2 sd of
10.04 ng/mL.
E-288
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-14. Summary of Benchmark Dose Modeling Results for Decreased HiB Antibody Levels in Abraham et al. (2020,
6506041)
Model"
Goodness of Fit
p-value AIC
Scaled Residual
Dose Grout) BMDisd
Dose Group p Control Dose (ng/mL)
near BMDisd „ "fifr Group
blVIUo.ssi)
BMDLisd
(ng/mL)
BMDo.ssd
(ng/mL)
BMDLo.ssd
(ng/mL)
Exponential 3
Exponential 5
Hill
Polynomial Degree 3
Polynomial Degree 2
Power
Linear
0.72 207.49
0.12
0.69 207.58
0.73 206.11
0.69 207.56
0.36 208.06
0.15
-0.21
0.11
-0.79
-0.49
0.00
1.18
-0.47
-0.18
-0.52
1.37
-0.17
-0.01
1.18
-0.25
-0.51
-0.18
-0.80
25.36
20.63
25.16
25.07
25.39
23.42
19.18
18.33
19.98
15.46
20.21
19.36
0.00
19.91
17.72
20.32
11.71
11.68
13.63
0.00
10.04
9.21
10.24
7.73
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-289
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E. 1.1.4 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 PFOA
(median =1.1 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 the
regression coefficient and 95% confidence interval from the multivariate regression analyses of
maternal PFOA and anti-Rubella antibody levels (-0.40, 95% CI: -0.64, -0.17). Granum et al.
(2013, 1937228) also reported the summary statistics of rubella antibodies levels at the age of 3
(median = 1.9, 25th, 75th percentiles: 1.5, 2.1 OD). Upon investigation of the extra risk using the
distributional data and a cutoff value of 0.1 IU/mL EPA determined that this data was not
suitable for application of a BMR of 1 SD and V2 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. 1.1.5 Modeling Results for Decreased Influenza Antibody
Concentrations
EPA also considered Looker et al. (2014, 2850913) for modelling. Looker et al. (2014, 2850913)
was included in the 2016 Health Assessment of PFOA and observed an association with PFOA
in an adult population where PFOA exposure predominated (the C8 Health Study population).
Elevated PFOA serum concentrations (median = 31.6 ng/mL) were associated with reduced
antibody titer rise following vaccination, particularly to A/H3N2 influenza virus, and an
increased risk of not attaining the antibody threshold considered to offer long-term protection.
However, most estimated associations were statistically nonsignificant, and results were
inconsistent by vaccine type and by outcome classification (antibody titer rise following
vaccination, antibody titer ratio (postvaccine: prevaccine), seroconversion, and seroprotection).
Confidence intervals were relatively wide, especially when separated into quartiles, but some
results, most notably the reduction in seroprotection, were significantly lower in higher serum
PFOA concentration quartiles. Given the lack of dose-response trend and the lack of adequate
effect estimates (reflecting change in vaccine antibody per unit change in PFOA), EPA could not
model the results from Looker et al. (2014, 2850913)
E-290
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.1.2 Modeling Results for Decreased Birthweight
Six high confidence studies {Chu, 2020, 6315711; Govarts, 2016, 3230364; Sagiv, 2018,
4238410; Starling, 2017, 3858473; Wikstrom, 2020, 6311677; Yao, 2021, 9960202} reported
decreased birth weight in infants whose mothers were exposed to PFOA. These candidate studies
offer a variety of PFOA 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 (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.
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 U.S.
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. EPA's America's Children and the Environment (ACE) Biomonitoring on
Perfluorochemicals report (https://www.epa.gov/americaschildrenenvironment/data-tables-
biomonitorine-perfluorochemicals-pfcs) provides in Table B6b the median blood serum levels of
PFOA of 1.1 ng/mL in 2015-2016 in woman ages 16 to 49, using National Health and Nutrition
Examination Survey (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 five epidemiological studies.
E.1.2.1.1 Chu et al. (2020, 6315711)
Chu et al. (2020, 6315711) reported a P coefficient of-73.6 g (95% CI: -126.4, 20.9) per
ln(ng/mL) increase for the association between birth weight and maternal PFOA 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 (1.5 ng/mL) and the 25th and 75th percentiles (1.0 and
2.6 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:
pi = ln(q50) = Zn( 1.5) = 0.43 (1)
E-291
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
o- = ln(q75/q25)/1.349 = Zn(2.6/1.0)/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-45.2 g (95% CI: -77.6, -12.8) 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
dichotomous data, and therefore is an advantageous approach12. 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 2,500 g. In this case, the mean birth weight
would be 3169.2 grams. Given the median exposure of 1.1 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:
12 While the explicit application of the hybrid approach is not commonly used in IRIS dose/concentration/exposure-response
analyses, the more commonly used SD-defrnition 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-292
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
b = y - nix = 3261.6 g - (-45.2 sCf^)"1) 1.1^ = 3311.4(3)
The BMD was calculated by rearranging the equation y = mx + b and solving for x, using
3311.4 g for the b term and -45.2 for the m term. Doing so results in a value of 3.1 ng/mL:
x = (y-b)/m = (3169.2 g - 3311.4 g)/(-45.2 gO^r)'1) = 3.1 ng/mL
mL
To calculate the BMDL, the method is essentially the same except that the lower limit (LL) on
the P coefficient (-77.6) 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:
Upper Limit — Lower Limit 12.8 1 ( 77.6,g*) ng
SE ~ 192 ~ 3S2 ~ 16-5 9^mL>
Then the corresponding 95% one-sided lower bound on the P coefficient can be calculated as:
95% one - sided LL = B - 1.645(SE(B)) = -45.2 gO^r)'1 - 1.645 (16.5 gO^r)'1)
mL v mL J
n9 1
=-i2A^r
Using this value for the m term results in a BMDL value of 2.0 ng/mL maternal serum
concentration.
E 1.2.1.2 Govarts et al. (2016, 3230364)
Govarts et al. (2016, 3230364) reported a P coefficient of-34.5 g (95% CI: -129.0, 60.0) per
IQR increase in Z-score of ln(ng/mL) PFOA exposures, corresponding to a P coefficient of-53.4
g (95%) CI: -199.5, 92.8) per ln(ng/mL) increase, for the association between birth weight and
PFOA concentrations in umbilical cords plasma samples in a United States cohort. Given the
reported study-specific median (1.5 ng/mL) and the 25th and 75th percentiles (1.1 and
2.1 ng/mL) of the exposure, EPA estimated the mean (0.42) and standard deviation (0.48) of the
log normally distributed exposure. The re-expressed P coefficient is -34.3 g (95% CI: -128.2,
59.7) per ng/mL, and the intercept b is 3,299.4 g. A BMD of 3.8 ng/mL is calculated from
Govarts et al. (2016, 3230364) using the same approach as above with the same values for the
mean birth weight in the US.
To calculate the BMDL, the same procedure as above is used to calculate the corresponding 95%
one-sided lower limit for the re-expressed P coefficient from the re-expressed lower limit on the
95% two-sided confidence interval of-128.2 g per ng/mL. Using the lower limit value (-113.1 g
per ng/mL), a BMDL of 1.2 ng/mL is calculated.
E-293
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.1.2.1.3 Sogiv et al. (2018, 4238410)
Sagiv et al. (2018, 4238410) reported a P coefficient of-18.5 g (95% CI: -45.4, 8.3) per IQR
increase in PFOA (ng/mL), corresponding to a P coefficient of-4.9 g (95% CI: -11.9, 2.2)
per ng/mL increase, for the association between birth weight and maternal PFOA 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 3267.0 g based on the P coefficient of-4.9 g
per ng/mL and the corresponding 95% one-sided lower limits for the P coefficient is -10.8 g
per ng/mL. A BMD of 20.1 ng/mL and a BMDL of 9.1 ng/mL are calculated using the same
approach as above.
E.1.2.1.4 Starling et al. (2017, 3858473)
Starling et al. (2017, 3858473) reported a P coefficient of-51.4 g (95% CI: -97.2, -5.7) per
ln(ng/mL) for the association between birth weight and maternal PFOA 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 (1.1 ng/mL) and the 25th and 75th percentiles
(0.7 and 1.6 ng/mL) of the exposure, EPA estimated the mean (0.10) and standard deviation
(0.61) of the log normally distributed exposure. The re-expressed P coefficient is -45.0 g (95%
CI: -85.1, -5.0) per ng/mL and the intercept b is 3311.1 g. The 95% one-sided lower limit for
the re-expressed P coefficient is -78.6 g per ng/mL. The values of the BMD and BMDL are
3.2 ng/mL and 1.8 ng/mL, respectively.
E.1.2.1.5 Wikstrom et al. (2020, 6311677)
Wikstrom et al. (2020, 6311677) reported a P coefficient of-68.0 g (95% CI: -112.0, -24.0) per
ln(ng/mL) for the association between birth weight and maternal PFOA 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 (1.6 ng/mL) and the 25th and 75th percentiles
(1.1 and 2.3 ng/mL) of the exposure, EPA estimated the mean (0.48) and standard deviation
(0.54) of the log normally distributed exposure. The re-expressed P coefficient is -41.0 g (95%
CI: -67.5, -14.5) per ng/mL and the intercept b is 3306.7 g. The 95% one-sided lower limit for
the re-expressed P coefficient is -63.3 g per ng/mL. The values of the BMD and BMDL are
3.4 ng/mL and 2.2 ng/mL, respectively.
E.1.2.1.6 Yao et al. (2021, 9960202)
Yao et al. (2021, 9960202) reported a P coefficient of-25.2 g (95% CI: -75.3, 24.9) per
ln(ng/mL) for the association between birth weight and maternal PFOA serum concentrations
(collected within 3 days of delivery) in a China cohort. Given the cohort-specific median
(42.8 ng/mL) and the 25th and 75th percentiles (25.1 and 73.1 ng/mL) of the exposure reported
in Han et al. (2018, 5080230), EPA estimated the mean (3.76) and standard deviation (0.79) of
the log normally distributed exposure. The re-expressed P coefficient is -0.6 g (95% CI: -1.6,
0.5) per ng/mL and the intercept b is 3262.2 g. The 95% one-sided lower limit for the re-
expressed P coefficient is -1.5 g per ng/mL. The values of the BMD and BMDL are
168.5 ng/mL and 63.2 ng/mL, respectively.
E. 1.2.2 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
E-294
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
extreme ("adverse") response at zero dose (P(0)). However, this exact percentage of 8.27% was
calculated without accounting for the existence of background PFOA exposure in the US
population (i.e., 8.27% is not the tail probability of response at zero dose). Thus, EPA considers
an alternative control-group response distribution (N(jj.c,
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-15. BMDs and BMDLs for effect of PFOA on decreased birth weight, by using percentage (8.27%) of live births
falling below the public health definition of low birth weight, or alternative study-specific tail probability
Exact Percentage
Exposure
Exposure
Reported p ^e"
Intercept
95% one-
(P(0) =
8.27%)
Alternative Tail Probability3
Study
Median
Distribution
expressed B
SE ofp
sided LL ¦
(IQR)
(H, <*)
(95% CI) (95% CI)
b
ofp
BMD
BMDL
P(0)
BMD
BMDL
(ng/mL)
(ng/mL)
(ng/mL)
(ng/mL)
Chu et al.
1.5 (1.0-2.6)
(0.43, 0.75)
-73.6 (-126.4, -45.2 (-77.6,
(2020,
-20.9) g/ln(ng/mL) -12.8) g/ng/mL
3311.4
16.5
-72.4
3.1
2.0
8.48%
3.3
2.0
6315711)
Govarts et al.
1.5 (1.1-2.1)
(0.42, 0.48)
-34.5 (-129.0, -34.3 (-128.2,
(2016,
60.0) g/ln(ng/mL) 59.7) g/ng/mL
3299.4
47.9
-113.1
3.8
1.2
8.80%
4.2
1.3
3230364)
Sagiv et al.
5.8(4.1-7.9)
(1.76, 0.49)
-18.5 (-45.4, -4.9 (-11.9,
(2018,
8.3) g/IQR 2.2) g/ng/mL
3267.0
3.6
-10.8
20.1
9.1
9.71%
27.7
12.5
4238410)
(ng/mL)
Starling et al.
1.1 (0.7-1.6)
(0.1,0.61)
-51.4 (-97.2, -45.0 (-85.1,
(2017,
-5.7) g/ln(ng/mL) -5.0) g/ng/mL
3311.1
20.4
-78.6
3.2
1.8
8.48%
3.3
1.9
3858473)
Wikstrom et
1.6(1.1-2.3)
(0.48, 0.54)
-68.0 (-112.0, -41.0 (-67.5,
al. (2020,
-24.0) g/ln(ng/mL) -14.5) g/ng/mL
3306.7
13.5
-63.3
3.4
2.2
8.60%
3.6
2.3
6311677)
Yao etal.
42.8(25.1,
(3.76, 0.79)
-25.2 (-75.3, -0.6 (-1.6,
(2021,
73.1)
24.9) g/ln(ng/mL) 0.5) g/ng/mL
3262.2
0.6
-1.5
168.5
63.2
9.84%
241.6
90.6
9960202)
Notes:
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 U.S. population.
E-296
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
ACE Biomonitoring on Perfluorochemicals also provides the median blood serum levels of PFOA
among women ages 16 to 49 in 1999-2000 (4.6 ng/mL), in 2009-2010 (2.2 ng/mL), and in 2013-
2014 (1.4 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-16, demonstrate the robustness of EPA's approaches with alternative assumptions on
background exposures.
Table E-16. BMDs and BMDLs for effect of PFOA 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 Alternative Tail Probability1"
, Background Intercept (P(0)-8.27%)
Stud.v r 17
aiuuy
Exposure"
b
BMD
(ng/mL)
BMDL
(ng/mL)
P(0)
BMD
(ng/mL)
BMDL
(ng/mL)
Wikstrom
1.1
3306.7
3.4
2.2
8.60%
3.6
2.3
et al. (2020,
1.4
3319.0
3.7
2.4
8.28%
3.7
2.4
6311677)
2.2
3351.8
4.5
2.9
7.46%
3.9
2.5
4.6
3450.2
6.9
4.4
5.38%
4.8
3.1
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 PFOA exposure, the POD selected from the
available epidemiologic literature is 2.2 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 PFOA 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 PFOA 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 PFOA exposure measures across various populations. Dong et
al., (2019, 5080195) investigated the 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 PFOA and TC using a hybrid model {Crump, 1995, 2258}. The cut-off point for
adverse response (i.e., elevated TC) was set at the upper 5th percentile of TC values in the lowest
E-297
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOA 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 BMDioand BMDLioof 10.5 and 5.6 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 PFOA 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 3.95 ng/mL for the period 1999-2018, excluding adults taking cholesterol
medications, up to 9.11 ng/mL for the period 2017-2018, for all adults. The BMDL values for a
BMR of 5% ranged from 2.29 ng/mL for the period 1999-2018, excluding adults taking
cholesterol medications, up to 5.28 ng/mL for the period 2017-2018, for all adults. The BMD
values for a BMR of 10% ranged from 8.79 ng/mL for the period 1999-2018, excluding adults
taking cholesterol medications, up to 13.85 ng/mL for the period 2017-2018, for all adults. The
BMDL values for a BMR of 10% ranged from 5.10 ng/mL for the period 1999-2018, excluding
adults taking cholesterol medications, up to 8.03 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
1.48 mg/dL TC per ng/mL PFOA (95% CI: 0.2, 2.8). After correspondence with the study author,
EPA obtained an updated estimated coefficient of 1.44 mg/dL TC per ng/mL PFOA (95% CI: 0.2,
2.69), 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 the 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 PFOA
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 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.
SE =
Upper Limit — Lower Limit 2.6"
3.92
E-298
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
EPA estimated the distribution of TC assuming a normal distribution and the estimated mean
PFOA levels for each of the analysis periods (Table E-17).
Table E-17. NHANES mean and standard deviation of TC (mg/dL) and mean PFOA
(ng/mL)
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?
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 PFOA (x)
3.43
3.43
3.90
3.90
3.37
3.37
1.80
1.80
For the BMD analyses, the response of interest is having elevated serum cholesterol, defined as
greater than or equal to 240 mg/dL, which is the cutoff that the American Heart Association
recommends (www.heart.ore). The baseline probability P(0) of such a response is 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
PFOA 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 PFOA 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 US population, the mean TC
is y (tabulated above) and the mean PFOA 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-299
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
/240 — y\
P(d) = P(TC > 240) = 1 -
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
period available, a BMR of 5% and a BMDLs of 2.29 ng/mL. A comparison BMDL of
4.39 ng/mL based on the most period available is also considered.
E.1.3.2 Steenland et al. (2009,1291109)
Mean serum TC
The same hybrid approach described previously was also applied to Steenland et al. (2009,
1291109) using natural log-transformed values. Steenland et al. (2009, 1291109) reported in
Table 4 linear regression coefficient for change in ln-transformed TC per ln(PFOA): 0.0112 with
a standard deviation of 0.00076. The NHANES data used in this approached is summarized in
Table E-19 and BMD/BMDL values are presented in Table E-20.
Table E-19. NHANES mean and standard deviation of ln(TC) (ln(mg/dL)) and mean
ln(PFOA) (ln(ng/mL))
Time Period
1999-
1999-
2003-
2003-
2003-
2003-
2017- 2017-
2018
2018
2014
2014
2018
2018
2018 2018
Taking prescribed cholesterol
medication?
No
No
No
No
Mean ln(TC) (y)
5.26
5.27
5.26
5.27
5.25
5.26
5.22 5.24
Standard Deviation ln(TC) (S)
0.21
0.21
0.21
0.21
0.21
0.21
0.22 0.21
Mean ln(PFOA) (x)
0.94
0.94
1.11
1.11
0.92
0.92
0.37 0.37
Table E-20. BMD and BMDL for effect of PFOA on increased cholesterol in Steenland et al.
(2009,1291109)
Time Period
1999-
1999-
2003-
2003-
2003-
2003-
2017-
2017-
2018
2018
2014
2014
2018
2018
2018
2018
Taking prescribed
cholesterol medication?
No
No
No
No
BMR=5%
BMD (ng/mL)
8.08
4.99
9.17
5.44
13.60
7.25
99.46
44.76
BMDL (ng/mL)
6.54
4.25
7.33
4.58
10.45
5.93
62.48
30.48
BMR=10%
BMD (ng/mL)
199.25
115.42
224.12
126.37
340.22
168.85
2590.14 1012.48
BMDL (ng/mL)
116.69
71.43
129.70
77.49
188.76
100.55
1170.53 503.12
EPA also conducted dose-response modeling using mean serum TC reported across PFOA deciles
from Table 3 in Steenland et al. (2009, 1291109). 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 in ln-transformed serum TC (Table 3 in Steenland et al. (2009,
1291109) and summarized in Table E-21. 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-22.
E-301
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-21. Regression Results for Serum Total Cholesterol by Deciles of serum PFOA
from Steenland et al. (2009,1291109)
Decile
Dose
(ng/mL)
N
Regression coefficient3
(SD)
1
5.8
4629
0.00 (0.192)
2
9.7
4629
0.01 (0.192)
3
13.6
4629
0.02 (0.192)
4
17.9
4629
0.03 (0.192)
5
24.0
4629
0.04 (0.192)
Notes:
a Regression coefficient, change in the natural log of total cholesterol.
E-302
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-22. Summary of Benchmark Dose Modeling Results for Increased Mean Serum Total Cholesterol in Steenland et al.
(2009,1291109)
Model"
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group
near BMDisd
Dose Group
near BMDo.ssd
Control Dose
Group
BMDisd
(ng/mL)
BMDLisd
(ng/mL)
BMDo.ssd
(ng/mL)
BMDLo.ssd
(ng/mL)
Exponential 3
0.00
-10692.03
-0.93
-0.93
-2.86
40.73
39.46
33.41
32.20
Exponential 5
-
-
-
-
-
-
-
-
-
Hill
-
-
-
-
-
-
-
-
-
Polynomial
0.44
-10703.20
-0.45
-0.45
-0.76
77.42
51.38
43.06
35.34
Degree 3
Polynomial
0.30
-10702.46
-0.46
-0.46
-0.98
72.03
56.71
42.14
36.23
Degree 2
Power
0.74
-10705.63
-0.62
-0.62
-0.48
86.48
70.77
43.24
37.70
Linear
0.74
-10705.63
-0.62
-0.62
-0.48
86.48
75.27
43.24
37.64
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.
E-303
-------
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 from Steenland et al. (2009, 1291109) for having an elevated TC level
are shown in Table E-23. Sample sizes, mean PFOA 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-24. Note that this approach did not generate any viable models.
Table E-23. Odds ratios for elevated serum TC by quartiles of serum PFOA from
Steenland et al. (2009,1291109)
Quartile
Dose
(ng/mL)
N
Incidence
OR
95% CI
1
6.55
11575
1431
1
Ref
2
19.85
11434
1687
1.21
1.12, 1.31
3
46.75
11478
1866
1.33
1.23, 1.43
4
441
11477
2082
1.38
1.28, 1.50
Notes: OR = odds ratio; Ref = reference value.
E-304
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-24. Summary of Benchmark Dose Modeling Results for Elevated Total Cholesterol in Steenland et al. (2009,1291109)
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Dose Group Control
near BMDio near BMDs Dose Group
BMDio
(ng/mL)
BMDLio
(ng/mL)
BMDs
(ng/mL)
BMDLs
(ng/mL)
Dichotomous Hill
_b
-
-
0.15
-
-
-
15.63
4.74
Gamma
<0.0001
39352.35
-0.52
-0.52
-5.27
925.23
788.57
450.43
383.91
Log-Logistic
<0.0001
39352.05
-0.55
-0.55
-5.25
947.03
803.50
448.60
381.25
Multistage Degree 3
<0.0001
39352.35
-0.52
-0.52
-5.27
925.23
702.09
450.43
383.86
Multistage Degree 2
<0.0001
39352.35
-0.52
-0.52
-5.27
925.23
785.94
450.43
383.76
Multistage Degree 1
<0.0001
39352.35
-0.52
-0.52
-5.27
925.23
788.56
450.43
383.72
Weibull
<0.0001
39352.35
-0.52
-0.52
-5.27
925.23
788.57
450.44
383.91
Logistic
<0.0001
39353.79
-0.42
-0.42
-5.37
839.18
728.71
457.78
398.00
Log-Probit
0.00
39305.25
-2.00
-2.00
-2.00
0.38
0.18
0.00
0.00
Probit
<0.0001
39353.58
-0.43
-0.43
-5.35
851.39
737.27
456.88
396.03
Quantal Linear
<0.0001
39352.35
-0.52
-0.52
-5.27
925.23
788.57
450.43
383.91
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.
aNo viable models. No model was selected.
b BMD Computation failed
E-305
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Given the potential impact of taking cholesterol medication on the true association between
PFOA and increased TC, the results based on the data excluding such possibility is considered
higher confidence. As illustrated in Table E-23 there was a slight decline over time in PFOA
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 PFOA 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 4.25 ng/mL.
E.133 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 US. Median PFOA levels at
baseline were comparable to those from NHANES 1999-2000, 4.9 (25th, 75th percentiles: 3.5,
6.7 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. Both analyses showed evidence of an association between PFOA and
increased TC.
EPA conducted dose-response modeling using mean serum TC reported across PFOA quartiles
using data 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 PFAS
concentrations (ng/mL), summarized in Table E-25. BMRs of a change in the mean equal to '/2
and 1 SDs from the control mean were chosen. The BMD modeling results are summarized in
Table E-26.
Table E-25. Adjusted Mean Differences in Serum Total Cholesterol by Quartiles of Serum
PFOA (ng/mL) from Lin et al. (2019,1291109)
Dose
(ng/mL)
N
Mean TCa b
2.6
221
0.00 ±35.85
4.2
222
2.00 ±36.68
5.6
227
10.13 ±35.47*
8.4
228
13.36 ±36.40*
Notes:
a Data are presented as mean ± standard deviation.
b Adjusted mean difference in lipid levels (mg/dL) per quartile of baseline plasma PFOA concentration
(ng/mL); *p < 0.05.
E-306
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-26. Summary of Benchmark Dose Modeling Results for Increase Mean Serum Total Cholesterol from Lin et al. (2019,
5187597)
Goodness of Fit Scaled Residual
Model"
p-value
AIC
Dose Group
near BMDisd
Dose Group
near
BMDo.ssd
Control Dose
Group
BMDisd
(ng/mL)
BMDLisd
(ng/mL)
BMDo.ssd
(ng/mL)
BMDLo.ssd
(ng/mL)
Exponential 3
0.09
8983.84
-0.31
-0.31
-1.04
11.63
9.85
9.41
8.52
Exponential 5
_b
-
-
-
-
-
-
-
-
Hill
-
-
-
-
-
-
-
-
-
Polynomial
0.10
8983.81
-0.39
-0.39
-0.30
13.63
10.28
8.64
5.14
Degree 3
Polynomial
0.33
8981.29
-0.38
-0.38
0.03
14.57
10.53
7.29
5.27
Degree 2
Power
0.34
8981.24
-0.38
0.01
0.01
14.53
10.63
-7.27
0.00
Linear
0.34
8981.24
-0.38
-0.38
0.01
14.53
10.56
7.27
5.28
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.
a Selected model in bold.
b BMD Computation failed
E-307
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E. 1.3.4 Summary of Modeling Results for Increased Cholesterol
Table E-27 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 2.29 to 5.28 ng/mL with the selected POD of 2.29 ng/mL.
Table E-27. BMDLs for effect of PFOA on serum total cholesterol using a BMR of 5%
Study name
Effect
BMD (ng/mL)
BMDL (ng/mL)
Dong et al. (2019, 5080195)
Exclude those prescribed
cholesterol medication,
1999-2018
3.95
2.29
Steenland et al. (2009, 1291109)
Exclude those prescribed
cholesterol medication
4.99
4.25
Lin et al. (2019, 5187597)
7.27
5.28
E.1.4 Modeling Results for Liver Toxicity
This updated review indicated that PFOA is associated with increases in the liver enzyme ALT
(See Main PFOA Document). Four medium confidence studies were selected as candidates for
POD derivation. The two largest studies of PFOA and ALT in adults are Gallo et al. (2012,
1276142) and Darrow et al. (2016, 3749173), both conducted in over 30,000 adults from the C8
Study Project (for detailed descriptions of the study and findings, see Main PFOA Document and
Table D-6). The main differences between the two studies are reflected in exposure assessment:
Gallo et al. (2012, 1276142) includes measured PFOA serum concentrations, while Darrow et al.
(2016, 3749173) based PFOA exposure on modeled PFOA serum levels. Two additional studies
{Lin, 2010, 1291111; Nian, 2019, 5080307} were considered by EPA for POD derivation
because they reported significant associations in general populations in the U.S and a high
exposed population in China, respectively. In an NHANES adult population, Lin et al. (2010,
1291111) observed elevated ALT levels per log-unit increase in PFOA. The association between
PFOA and liver enzymes was more evident in obese subjects, as well as subjects with insulin
resistance and/or metabolic syndromes. When dividing the serum PFOA into quartiles in the
fully adjusted models in subjects with a body mass index > 30 kg/m2, the ALT level trend across
the serum PFOA quartiles was significant. 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 PFOA 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 a large population of 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 PFOA, as
well significant increases in odds ratios of elevated ALT. Median serum PFOA concentrations in
this study were 6.2 (ng/mL).
Both Gallo et al. (2012, 1276142) and Darrow et al. (2016, 3749173) studies evaluated the
relationship between PFOA and ALT using two general types of analyses. In the first, subjects
were divided into quantiles of PFOA exposure (quintiles in Darrow et al. (2016, 3749173) and
E-308
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
deciles in Gallo et al. (2012, 1276142)), and linear regression models were used to compare
mean ALT levels by each non-reference quantile vs. mean ALT level in the lowest quantile. In
the second type of analysis, a logistic regression evaluated ORs for having an ALT level above a
certain cutoff for each non-reference quantile compared to the lowest (reference) quantile. The
cutoff values used to define elevated ALT levels in both studies 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 these studies.
E.l.4.1.1 Gallo et al. (2012, 1276142)
Elevated ALT
NOAEC/LOAEC method. The results of the logistic regression analysis of elevated ALT across
deciles of PFOA are presented in Table E-28. The mean, median and ranges of PFOA
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-28. The NOAEC
is bolded and is the mean PFOA serum concentration in the highest decile of PFOA 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 PFOA). The NOAEC based on the
elevated ALT data from Gallo et al. (2012, 1276142) is 9.78 ng/mL.
BMP method. EPA used 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 the number of participants with and without
elevated ALT for each decile of PFOA were obtained (Table E-28).
Table E-28. Odds Ratios for Elevated ALT by Decile of PFOA serum concentrations
(ng/mL) from Gallo et al. (2012,1276142)
Decile
Minimum
Maximum
Median
Mean
OR
95% CI
Participants
without
Participants
with Elevated
ALT
Total
(ng/mL)
(ng/mL)
(ng/mL) (ng/mL)
Elevated
(N)
ALT
0
0.25
7.9
5.8
5.46
1
ref
4,201
408
4,609
1
8.0
11.5
9.7
9.76
1.09
0.94,1.26
4,123
450
4,573
2
11.6
15.5
13.5
13.5
1.19
1.03, 1.37
4,184
504
4,688
3
15.6
20.7
17.9
18.0
1.26
1.09, 1.45
4,137
541
4,678
4
20.8
27.9
24.0
24.1
1.4
1.22, 1.62
4,069
570
4,639
5
28.0
39.3
33.0
33.2
1.39
1.21, 1.60
4,126
555
4,681
6
39.4
57.0
47.2
47.5
1.31
1.14, 1.52
4,125
518
4,643
7
57.1
89.0
70.8
71.7
1.42
1.23, 1.64
4,100
542
4,642
8
89.1
189.3
118.1
124.9
1.4
1.21, 1.62
4,119
531
4,650
9
189.4
22412
355.8
522.0
1.54
1.33, 1.78
4,074
575
4,649
Notes:
The NOAEC is bolded.
ALT = alanine transaminase; OR = odds ratio.
E-309
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Applying BMDS v3.3rcl0 using a BMR of 10% and 5% to the data for all ten deciles did not
result in any viable models. Applying BMDS v3.3rcl0 to the data for the 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 PFOA levels are
summarized in Table E-29, Table E-30, Table E-31, and Table E-32. The approaches provide
similar BMDLs, with slightly higher values for unadjusted and adjusted models, using mean and
median concentration, ranging from 36.0 to 39.2 for a 10% BMR, and from 18.3 to 19.2 for a 5%
BMR. The gist approach resulted in BMD(BMDL) values of 10.4 (9.0) ng/mL and 8.3(7.5)
ng/mL for BMRs of 10% and 5%, respectively.
E-310
-------
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 Mean PFOA Serum Concentration
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group
near BMDio
Dose Group
near BMDs
Control Dose
Group
BMDio
(ng/mL)
BMDLio
(ng/mL)
BMDs
(ng/mL)
BMDLs
(ng/mL)
Dichotomous Hill
_b
-
-
0.01
0.00
-
-
37.57
1.59
Gamma
0.88
15710.72
-0.47
-0.47
-0.36
49.86
39.02
24.27
19.00
Log-Logistic
0.89
15710.66
-0.45
-0.45
-0.33
50.88
39.15
24.10
18.73
Multistage Degree 3
0.21
15715.51
-0.73
-0.73
-1.08
40.01
32.72
27.39
17.39
Multistage Degree 2
0.88
15710.72
-0.47
-0.47
-0.36
49.86
38.03
24.27
19.00
Multistage Degree 1
0.88
15710.72
-0.47
-0.47
-0.36
49.86
39.01
24.27
19.00
Weibull
0.88
15710.72
-0.47
-0.47
-0.36
49.86
39.02
24.27
19.00
Logistic
0.75
15711.31
-0.55
-0.55
-0.57
44.00
36.13
25.45
21.06
Log-Probit
0.95
15712.09
-0.11
0.09
0.07
53.24
31.56
12.42
0.24
Probit
0.78
15711.21
-0.54
-0.54
-0.54
44.85
36.58
25.30
20.79
Quantal Linear
0.88
15710.72
-0.47
-0.47
-0.36
49.86
39.02
24.27
19.00
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-311
-------
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 Mean PFOA Serum Concentration
Goodness of Fit Scaled Residual
BMDio BMDLio BMDs BMDLs
p-value AIC D()SC5''/I()II1'|) D()SC Con';r()l (ng/mL) (ng/mL) (ng/mL) (ng/mL)
near BMDio near BMDs Dose Group
Dichotomous Hill
_b
-
-
0.00
0.00
-
-
44.13
19.91
Gamma
0.88
15710.72
-0.47
-0.47
-0.36
49.86
36.33
24.27
19.00
Log-Logistic
0.89
15710.66
-0.45
-0.45
-0.33
51.49
36.52
24.39
19.17
Multistage Degree 3
0.88
15710.72
-0.47
-0.47
-0.36
49.86
30.23
24.27
18.99
Multistage Degree 2
0.88
15710.72
-0.47
-0.47
-0.36
49.86
33.44
24.27
19.00
Multistage Degree 1
0.88
15710.72
-0.47
-0.47
-0.36
49.86
39.02
24.27
19.00
Weibull
0.88
15710.72
-0.47
-0.47
-0.36
49.86
35.91
24.27
19.00
Logistic
0.75
15711.31
-0.55
-0.55
-0.57
41.20
33.25
23.56
19.10
Log-Probit
0.94
15712.10
-0.15
-0.15
0.04
80.42
42.50
26.88
19.41
Probit
0.78
15711.21
-0.54
-0.54
-0.54
42.36
34.02
23.66
19.08
Quantal Linear
0.88
15710.72
-0.47
-0.47
-0.36
49.86
39.02
24.27
19.00
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-312
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-31. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et al. (2012,1276142) Using the
Unadjusted Median PFOA 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
Dichotomous Hill
_b
-
-
0.00
0.00
-
-
30.38
1.63
Gamma
0.86
15710.82
-0.49
-0.49
-0.42
48.82
38.18
23.77
18.59
Log-Logistic
0.87
15710.76
-0.48
-0.48
-0.39
49.77
38.59
23.57
18.29
Multistage Degree 3
0.41
15713.97
-0.69
-0.69
-0.70
41.72
34.14
25.48
17.84
Multistage Degree 2
0.86
15710.82
-0.49
-0.49
-0.42
48.82
37.41
23.77
18.59
Multistage Degree 1
0.86
15710.82
-0.49
-0.49
-0.42
48.82
38.18
23.77
18.59
Weibull
0.86
15710.82
-0.49
-0.49
-0.42
48.82
38.18
23.77
18.59
Logistic
0.72
15711.45
-0.56
-0.56
-0.63
43.35
35.62
25.08
20.78
Log-Probit
0.96
15712.05
-0.10
0.06
0.06
38.13
28.57
6.49
0.28
Probit
0.75
15711.34
-0.55
-0.55
-0.60
44.15
36.03
24.92
20.49
Quantal Linear
0.86
15710.82
-0.49
-0.49
-0.42
48.82
38.18
23.77
18.59
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-313
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-32. Summary of Benchmark Dose Modeling Results for Elevated ALT in Gallo et al. (2012,1276142) Using the
Adjusted, No Intercept Median PFOA Serum Concentration
Goodness of Fit Scaled Residual
Model3
p-value
AIC
Dose Group
near BMD10
Dose Group
near BMDs
Control Dose
Group
BMD10
(ng/mL)
BMDLio
(ng/mL)
BMDs
(ng/mL)
BMDLs
(ng/mL)
Dichotomous Hill
_b
-
-
-0.01
0.00
-
-
39.27
19.50
Gamma
0.86
15710.82
-0.49
-0.49
-0.42
48.82
35.81
23.77
18.59
Log-Logistic
0.87
15710.76
-0.48
-0.48
-0.39
50.41
36.03
23.88
18.77
Multistage Degree 3
0.86
15710.82
-0.49
-0.49
-0.42
48.82
29.64
23.77
18.59
Multistage Degree 2
0.86
15710.82
-0.49
-0.49
-0.42
48.82
32.80
23.77
18.59
Multistage Degree 1
0.86
15710.82
-0.49
-0.49
-0.42
48.82
38.17
23.77
18.58
Weibull
0.86
15710.82
-0.49
-0.49
-0.42
48.82
35.40
23.77
18.59
Logistic
0.72
15711.45
-0.56
-0.56
-0.63
40.38
32.55
23.08
18.70
Log-Probit
0.95
15712.07
-0.14
-0.14
0.04
82.86
42.57
26.64
18.96
Probit
0.75
15711.34
-0.55
-0.55
-0.60
41.50
33.30
23.18
18.68
Quantal Linear
0.86
15710.82
-0.49
-0.49
-0.42
48.82
38.18
23.77
18.59
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-314
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Hybrid method. The hybrid method uses the regression slope from the linear regression model of
ln-transformed ALT and ln-PFOA 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 0.022 (95% CI: 0.018, 0.025) In ALT (IU/L) per In ng/mL
PFOA (Table 2, Gallo et al. (2012, 1276142), model 3).
Using a normal approximation, the standard error of the regression coefficient is estimated as:
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 most recent sex-specific upper reference limits reported in Valenti et al.
(2021, 10369689). These are slightly lower and more health protective than the cutoff values
used in the original study (45 IU/L for men and 34 IU/L for women).
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 reported regression 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 PFOA data for each NHANES cycle 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 PFOA (Table E-33). The unrounded values were
used in the calculations:
Table E-33. NHANES mean and standard deviation of ln(ALT) (In IU/L) and mean PFOA
(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 IU/L) (5)
0.46
0.41
0.46
0.41
0.48
0.42
Mean In PFOA (In ng/mL)
(x)
1.10
0.80
1.08
0.78
0.50
0.25
Notes: ALT = alanine transaminase.
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
SE =
Upper Limit — Lower Limit 0.025 — 0.018
= 0.0018
3.92
3.92
E-315
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
weights. Second, the lognormal estimate of P(0), "P(0) Lognormal," was calculated assuming
that ALT is lognormally distributed using the equation:
fln(C) — mean(ln ALT))
P( 0) Lognormal = 1-
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(In C — y\
P(d) = P{ALT > C) = P(ln ALT >lnC)= 1 -
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.1.4.1.2 Nian et al. (2019, 5080307)
NOAEC/LOAEC method. Categorical data, which can be used to develop NOAECs, were not
available 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 PFOA was 7.4 (95% CI: 3.9, 11.0) (Table 3, Nian et al. (2019,
5080307)). The reported regression coefficient P, which is also referred to as m, was calculated
from the percent change expressed as (e|3-l)*100, resulting in a slope of 0.071 (95% CI: 0.038,
0.104) In ALT (IU/L) per In ng/mL PFOA. The estimated BMDs and BMDLs are presented in
Table E-36.
Table E-36. BMD and BMDL for effect of PFOA (ng/mL) on increased ALT in Nian et al.
(2019, 5080307)
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
5.90
4.61
6.27
4.50
3.31
2.42
BMDL
4.88
3.76
5.08
3.68
2.72
2.03
BMR=5%, P(0) Lognormal
BMD
8.44
6.16
8.35
5.95
4.55
3.48
BMDL
6.31
4.63
6.24
4.50
3.43
2.63
BMR=10%, P(0) Empirical
BMD
15.57
11.65
16.38
11.23
8.55
6.13
BMDL
9.81
7.33
10.14
7.10
5.40
3.96
BMR=10%, P(0)
Lognormal
BMD
21.06
14.81
20.87
14.16
11.22
8.29
BMDL
12.20
8.71
12.07
8.39
6.57
4.91
E.1.4.1.3 Dor row et al. (2016, 3749173)
NOAEC/LOAEC method. The results of the linear regression analysis of elevated ALT across
quintiles of PFOA are presented in Table E-37. The PFOA dose levels in each quintile of
exposure were calculated as the midpoint of the reported quintile ranges (Table 2 in Darrow et al.
(2016, 3749173)). The NOAEC is bolded and is the mean PFOA serum concentration in the
highest quintile of PFOA that did not show a statistically significant change in ALT, which in
this case is the 2nd quintile, compared to the reference category (the lowest quintile of PFOA).
The NOAEC based on the elevated ALT data from Darrow et al. (2016, 3749173) is 8.6 ng/mL.
Table E-37. Linear Regression results for In (ALT) by quintiles of serum PFOA
Concentration in Darrow et al. (2016, 3749173)
Quintile
Dose
(ng/mL)
N
Regression
coefficient3
95% CI
1
2
4.20
8.60
6145
6145
Ref
0.001
Ref
-0.016,0.018
E-318
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Quintile
Dose
(ng/mL)
N
Regression
coefficient3
95% CI
3
19.05
6145
0.023
0.007, 0.040
4
54.10
6145
0.036
0.019,0.053
5"
1,8120.2
6145
0.048
0.031,0.066
Notes:
a Linear regression coefficient for ln-transformed ALT.
Hybrid method. The previously described hybrid method was implemented using data from
Darrow et al. (2016, 3749173). The regression model adjusted for age, sex, BMI, alcohol
consumption, regular exercise, smoking status, education, insulin resistance, fasting status,
history of working at DuPont place, and race. The reported regression coefficient P, which is also
referred to as m, 0.012 In ALT (IU/L) per In ng/mL PFOA (95% CI: 0.009, 0.016). The values of
the BMD and BMDL are presented in Table E-38.
Table E-38. BMD and BMDL for effect of PFOA (ng/mL) on increased ALT in Darrow et
al. (2016, 3749173)
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
170.00
170.39
261.56
162.90
102.05
57.26
BMDL
70.30
65.99
98.12
63.45
41.44
24.95
BMR=5%, P(0) Lognormal
BMD
1425.99
953.63
1444.21
857.24
676.42
487.93
BMDL
370.41
253.45
372.90
232.26
181.68
133.07
BMR=10%, P(0) Empirical
BMD
54468.17
42371.98
79307.85
37331.45
29002.22
14336.82
BMDL
6380.83
4913.47
8530.54
4432.63
3425.39
1867.46
BMR=10%, P(0) Lognormal
BMD
328872.91
176813.58
335682.39
148268.33
146339.24
85701.65
BMDL
26005.03
15003.59
26339.57
13022.55
12133.22
7551.51
BMDS method. Darrow et al. (2016, 3749173) the increased mean ALT concentration (Table
E-39) was modeled using BMDS v3.3rcl0. BMRs of a change in the mean equal to V2 and 1 SDs
from the control mean were chosen (Table E-40). No viable models were identified.
Table E-39. Dose-Response Modeling Data for Increased Mean ALT Concentration in
Darrow et al. (2016, 3749173)
Dose
Mean Responsea b
(ng/mL)
N
4.20
6145
0.000 ± 0.48
8.60
6145
0.001 ±0.48
19.05
6145
0.023 ± 0.47
54.10
6145
0.036 ±0.48
Notes:
E-319
-------
DRAFT FOR PUBLIC COMMENT
a Data are presented as mean ± standard deviation.
b Linear regression coefficient for ln-transformed ALT
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-40. Summary of Benchmark Dose Modeling Results for Increased Mean ALT concentrations in Darrow et al. (2016,
3749173)
Goodness of Fit Scaled Residual
Model3
p-value
AIC
Dose Group near
BMDisd
Dose Group
near BMDossd
Control Dose
Group
BMDisd
BMDLisd
BMDo.ssd
BMDLo.ssd
Exponential 3
<0.0001
35008.88
-0.04
-0.04
-0.02
689.70
0.00
631.70
0.00
Exponential 5
_b
-
-
-
-
-
-
-
-
Hill
-
-
-
-
-
-
-
-
-
Polynomial
Degree 3
<0.0001
34974.38
-0.37
-0.37
-2.43
5840.29
1369.18
2920.14
1060.82
Polynomial
Degree 2
<0.0001
34974.38
-0.38
-0.38
-2.42
5836.79
2087.90
2918.39
1424.88
Power
<0.0001
34974.38
-0.37
-0.37
-2.42
5836.99
5037.10
2918.50
2219.70
Linear
<0.0001
34974.38
-0.37
-0.37
-2.42
5836.99
4554.37
2918.50
2277.24
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-321
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E. 1.4.1.4 Summary of Modeling Results for Liver Toxicity
Table E-41 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 3.76 to
65.99 ng/mL.
Table E-41. BMDLs for effect of PFOA on serum ALT using a BMR of 5%
Study name
BMDL (ng/mL)
Galloetal. (2012, 1276142)
17.93
Darrow et al. (2016, 3749173)
65.99
Nian et al. (2019, 5080307)
3.76
E.1.5 Modeling Results for Cancer
This updated review indicated that there is an increase in risk for kidney or renal cell carcinoma
(RCC) and testicular cancers and PFOA exposure {Shearer, 2021, 7161466; Chang, 2014,
2850282; Bartell, 2021, 7643457}. Although newer studies generally show no association, there
is some evidence that PFOA may be related to breast cancer risk especially in participants with
specific polymorphisms or specific types of tumors {Ghisari, 2017, 3860243; Mancini, 2019,
5381529}. Two occupational studies {Steenland, 2015, 2851015; Girardi, 2019, 6315730}
support an increase in risk for liver cancer, malignant neoplasm of the lymphatic and
hematopoietic tissue, as well as an increasing trend in prostate cancer that did not reach statistical
significance. No associations were found for colorectal cancer in either the general population or
occupational studies, or for lung cancer in occupational studies.
Results are most consistent for kidney cancer in adults based on a new nested case-control study
{Shearer, 2021, 7161466}, two C8 Health Project studies {Barry, 2013, 2850946; Vieira, 2013,
2919154} and an occupational mortality study {Steenland, 2012, 2919168} from the 2016 HA.
For dose-response modelling, Shearer (2021, 7161466) was selected as the key study. For
sensitivity analyses, EPA also considered the C8 Health Project study {Vieira, 2013, 2919154}.
Considerations included study population (general population vs. occupational or high-exposed
populations), statistical power and study quality.
The high exposure occupational study by Steenland and Woskie (2012, 2919168) evaluated
kidney cancer mortality in workers from West Virginia and observed significant elevated risk of
kidney cancer death in the highest exposure quartile (> 2,384 ppm-years). This study was limited
by the small number of observed cancer cases (six kidney cancer deaths). This study was not
used for dose-response analysis because information on a range of exposures more relevant to
the general population were available from the Shearer et al. (2021, 7161466) and Vieira et al.
(2013, 2919154). The study by Barry et al. (2013, 2850946) was not used for dose-response
analysis because it was performed in the same study area as the Vieira et al. (2013, 2919154)
study and these two studies likely involved a number of the same participants. In addition, Barry
et al. (2013, 2850946) could not be used in the sensitivity analysis because it lacked the
necessary exposure measurements for CSF calculation. In this study, estimated PFOA
concentrations are provided in Table 2 for community level and worker level. However,
E-322
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
combined exposure levels of each quartile of the overall study population were not reported.
Without overall exposure data in each quartile, CSF calculations are not feasible.
The study by Raleigh et al. (2014, 2850270) was not selected because of the concerns of
exposure assessment methods and study quality. This study used modeled estimates of PFOA air
concentrations in the workplace rather than biomonitoring measurements. This is a concern
because the assessment lack of information about the degree to which inhaled PFOA is absorbed
in humans and factors that may affect the absorption, as well as PFOA exposure data in non-
production workers was not based on actual measurements. In addition, this study did not
observe an association between PFOA and kidney cancer. The possible reasons of this study
could have missed to identify the association between PFOA, and kidney cancer include
relatively small numbers of cases, lack of information adjustment on risk factors of kidney
cancer such as smoking status and BMI, and the methods for exposure assessment.
Shearer et al. (2021, 7161466} is a multi-center case-control study nested within the National
Cancer Institute's (NCI) Prostate, Lung, Colorectal, and Ovarian Screening Trial (PLCO). The
PLCO is a randomized clinical trial of the use of serum biomarkers for cancer screening. The
cases in this study {Shearer 2021, 7161466} included all the participants of the screening arm of
the PLCO trial who were newly diagnosed with RCC during the follow-up period (N = 326). All
cases were histopathologically confirmed. Controls were selected from among participants of the
PLCO trial screening arm who had never had RCC. Controls were individually matched to the
RCC cases by age at enrollment, sex, race/ethnicity, study center, and year of blood draw. PFOA
concentrations were measured in the baseline serum samples collected between 1993 and 2002.
Median PFOA levels in controls was 5.0 ng/mL, comparable with 4.8 ng/mL in adults 60 and
over from NHHANES 1999-2000. The analyses accounted for numerous confounders including
BMI, smoking, history of hypertension, estimated glomerular filtration rate, previous freeze-thaw
cycle, calendar and study year of blood draw, sex, race and ethnicity, study center. Socio-
economic status was not explicitly considered in the analyses.
There was a statistically significant increase in odds of RRC per doubling of PFOA (OR = 1.71,
95% CI: 1.23, 2.37) and in the highest vs. lowest quartile (OR = = 2.63, 95% CI: 1.33, 5.2).
Although non-significant elevated risks were observed in the second and third quartiles, there
was a statistically significant increasing trend with increasing PFOA exposure across quartiles
(p-trend = 0.007). Statistically significant increased odds of RCC were observed in participants
ages 55-59 years, and in men and in women, separately.
For sensitivity analyses, EPA also considered the C8 Health Project study {Vieira, 2013,
2919154}. The Shearer et al. (2021, 7161466) and Viera et al. (2013, 2919154) have
considerable differences with respect to several study design aspects. These include the types of
outcomes considered (RCC vs. any kidney cancer), the type of exposure assessment (serum
biomarker vs. modeled exposure), source population (multi-center us. Ohio and WV regions),
study size (324 cases and 324 matched controls vs. 59 cases and 7,585 registry-based controls).
Additionally, the dramatically different regression slopes resulting from the two studies (0.0981,
95% CI: 0.0025, 0.1937 vs. 0.0122, 95% CI: 0.006, 0.0238 per ng/mL PFOA, from Shearer et al.
(2021, 7161466) and Vieira et al. (2013, 2919154), respectively), are an indication that the
studies have considerable differences.
E-323
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.1.5.1.1 Cancer Slope Factor (CSF) Calculations
E.1.5.1.2 Shearer et al. (2021, 7161466)
The methods used to calculate CSFs based on data from Shearer et al. (2021, 7161466) are based
on those used by US EPA for its CSF calculation for TCE {U.S. EPA, 2011, 9642147} and by
OEHHA for its PHG for arsenic {OEHHA, 2004, 10369748}.
The underlying model involves a linear regression between PFOA exposure and cancer relative
risk used to estimate the dose-response between PFOA and RCC risk, of the form:
RR = 1 + bx
This was calculated using a weighted linear regression of the quartile specific RRs. The variable
b is then the slope of the excess risk (RR-1) and PFOA dose (x,) in each non-reference exposure
group, and can be estimated as follows {Rothman, 2008, 1260377}{U.S. EPA, 2011, 9642147}:
HwixiRRi~ HWjXj
b = V 2
Z WiXi2
where (w,) are the weights defined as the inverse of the variance of each RR,. Since the incidence
of kidney cancer is relatively low and because the cases and controls were matched on age, the
ORs represent a good approximation of the underlying RRs. Thus, the variance of the quartile
specific ORs can be used to estimate the weights (wt) as follows {U.S. EPA, 2011, 9642147}:
1 1
" Var{ORi) " 2 (InUCLj - lnLCLt\2
URi X V 2 x 1.96 )
where UCLi and LCL, are the upper and lower 95% CIs of the quartile specific ORs (Table E-
46). The PFOA dose levels (pci) in each quartile of exposure were calculated as the midpoint of
the reported range (Table E-42). Since the intercept of the regression is set at 1 for a dose of 0,
the midpoint of the lowest quartile was subtracted from each of the midpoint of the upper
quartiles.
The standard error and 95% CIs for the regression slope b can then be calculated as follows:
SEb = -r^— and 95% CIb = b ± SEb
Table E-42. ORs for the association between PFOA serum concentrations and RCC in
Shearer et al. (2021, 7161466) and data used for CSF calculations
PFOA
Range
(ng/mL)
Xi
ORi
LCIi
UCIi
Var(ORi)
Wi
WiXi
WiXi2
WiXiORi
cases
controls
<4
0 (reference)
1
-
-
47
81
4.0-5.5
2.75
1.47
0.77
2.80
0.234
4.267
11.734
32.267
17.248
83
79
5.5-7.3
4.4
1.24
0.64
2.41
0.176
5.685
25.012
110.053
31.015
69
83
7.3-27.2
15.25
2.63
1.33
5.20
0.837
1.195
18.224
277.909
47.928
125
81
E-324
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The CSF is then calculated as the excess cancer risk associated with each ng/mL increase in
serum PFOA (CSFsemm). The CSFsemm was calculated by first converting the linear regression
model discussed above from the RR scale to the absolute risk scale. This was done assuming a
baseline risk (Ro) of RCC or kidney cancer in an unexposed or lower exposure reference group.
Since this is not available in a case-control study, the lifetime risk of RCC in US males is used.
The lifetime RCC risk was estimated by multiplying the lifetime risk of kidney cancer in US
males {American Cancer Society, 2020, 9642148} by the percentage of all kidney cancers that
are the RCC subtype (90%). This gives an Ro of 0.0202 x 90% = 0.0182. The CSFsemm was then
calculated as the product of the upper 95% CL of the dose-response slope (b) and Ro. The
estimated CSF serum IS 0. 00352 (ng/mL)"1 (Table E-43). The estimated The CSF serum based on the
estimated slope b is 0.00178 (ng/mL)"1
Table E-43. Internal CSF calculations for Shearer et al. (2021, 7161466) and Vieira et al.
(2013, 2919154) studies
Calculations
Shearer et al. (2021,
7161466)
Vieira et al. (2013,
2919154)3
Vieira et al. (2013,
2919154)b
E(WjXjORj)
96.19
1005.87
2592.49
E(WjXj)
54.97
655.39
1448.70
E(WjXj2)
420.23
28669.32
332109.95
SEb
0.0488
0.0059
0.0017
B
0.0981
0.0122
0.0034
CIb LCL
0.0025
0.0006
0.0000
CIb UCL
0.1937
0.0238
0.0068
Ro
0.01818
0.0202
0.0202
CSF serum- central
0.00178
0.00025
0.00007
CSF serum- UCL
0.00352
0.000481
0.000138
Notes:
a Highest exposure level excluded.
b Highest exposure level included.
One potential limitation of the weighted linear regression for estimating the dose-response
relationship between PFOA and relative risk of RCC, is that it ignores the covariance between
the estimated OR for each exposure quartile compared to the lowest quartile, since they come
from the same study and share a reference group. To evaluate the potential impact of the lack of
independence, 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. This approach is
compared to the regression coefficients obtained using the variance-weighted least squares
(vwls) approach which does not adjust for covariance between estimates. These analyses were
performed in STATA vl7.0 (StataCorp. 2021. Stata Statistical Software: Release 17. College
Station, TX: StataCorp LLC).
E-325
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
While these estimates obtained under the assumption of a linear relationship between the
exposure and the logarithm of the OR cannot be directly compared to the weighted linear
regression with fixed intercept used for deriving the CSF, the findings suggest that the lack of
independence in the study-specific ORs as minor impact on the CSF calculations (Figure E-3).
. gist logor dose if studyname=="ShearerJ 2021"., se(se) cov(n cases) cc
Generalized least-squares regression Number of obs = 3
Goodness-of-fit chi2(2) = 0.84 Model chi2(l) = 8,39
Prob > chi2 = 0.6570 Prob > chi2 = 0.0038
logor
Coefficient Std. err.
N
"U
V
N
[95% conf. interval]
dose
.0582322 .0201097
2.90 0.004
.0188178 .0976465
. gist logor dose if studynaine=="ShearerJ 2021"., se(se) cov(n cases) cc vwls
Variance-weighted least-squares regression Number of obs = 3
Goodness-of-fit chi2(2) = 0.44 Model chi2(l) = 9.06
Prob > chi2 = 0.8018 Prob > chi2 = 0.0026
logor
Coefficient Std. err.
z P > | z |
[95% conf. interval]
dose
.064746 .0215109
3.01 0.003
.0225854 .1069066
Figure E-3. Regression coefficients and 95% CIs between the log of the RCC ORs and
serum PFOA concentrations using data from Shearer et al. (2021, 7161466): adjusted (gist)
and unadjusted (vwls) for OR dependence
EPA considered evaluating the dose-response using the Benchmark Dose Software (BMDS).
However, categorical data from case-control studies cannot be used in the US EPA BMDS since
these models are based on cancer risk, and the data needed to calculate risks (i.e., the
denominators) are not available.
E.1.5.1.3 Sensitivity Analyses
Vieira et al. (2013, 2919154)
The Vieira et al. (2013, 2919154) study was a cancer registry-based case-control conducted in 13
counties in Ohio and West Virginia that surround the DuPont Washington Works PFOA facility
(C8 study area). The cancers of interest included kidney, pancreatic, testicular, and liver cancers.
The researchers selected these because they had been linked to PFOA in previous animal and
human studies. The controls were all other cancer types. Initially, all incident cancer cases
diagnosed from 1996 through 2005 were obtained from the Ohio Cancer Incidence Surveillance
System (OCISS) and the West Virginia Cancer Registry (WVCR), respectively. However, only
the OCISS provided the participants addresses, which could be used to develop individual
estimates of PFOA exposure at the time of diagnosis and 10 years before diagnosis. Those living
in one of the included counties, but outside of an exposed water district, were assigned to the
E-326
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
"unexposed" reference category. For participants residing in one of the exposed water districts,
PFOA exposure was categorized into groups of "low," "medium," and "high" based on the tertile
cutoff points in these participants. Cumulative exposure was assessed by summing the yearly
serum PFOA exposure estimates for the ten years prior to cancer diagnosis. Analyses were
adjusted for age, sex, diagnosis year, smoking status (current, past, unknown, or never), and
insurance provider (government-insured Medicaid, uninsured, unknown, or privately insured).
There was a statistically significant increase in the odds of kidney cancer when comparing both
the high (OR = 2.0; 95% CI: 1.3, 3.2) and the very high (OR = 2.0; 95% CI: 1.0, 3.9) exposure
categories to the unexposed reference population. The corresponding ORs were similar in the
high and very high categories of cumulative exposure (2.0 and 2.1, respectively) but were
slightly lower (1.8 and 1.7, respectively) in analyses without the 10-year lag. P-values for trends
or analyses using continuous estimates of PFOA exposure were not provided.
Using the data from Table E-44, the model fit was better when the highest exposure level was
excluded (Table E-43). With a lifetime kidney cancer of Ro of 0.0202, the CSFsemm was then
calculated as the product of the upper 95% CL of the dose-response slope (b) and Ro. When the
highest exposure group was excluded, the estimated CSF serum is 0.00048 (ng/mL)-1 (and 0.00025
(ng/mL)when based on the slope b) (Table E-43). When the highest exposure group was
included, the estimated CSFsemm is 0.00014 (ng/mL)"1 (and 0.00007 (ng/mL)"1 when based on the
slope b) (Table E-43).
Table E-44. ORs for the association between PFOA serum concentrations and RCC in
Vieira et al. (2013, 2919154) and data used for CSF calculations
PFOA
Range
(ng/mL)
Xi
ORi
LCIi
UCIi
Var(ORi)
Wi
WiXi
WiXi2
WiXiORi
cases
controls
0
0 (reference)
1.0
-
-
187
5957
3.7-12.8
8.25
0.8
0.4
1.5
0.073
13.743
113.382
935.400
90.705
11
446
12.9-30.7
21.8
1.2
0.7
2.0
0.103
9.682
211.074
4601.413
253.289
17
455
30.8-109
69.9
2.0
1.3
3.2
0.211
4.734
330.937
23132.508
661.874
22
339
110-655
382.5
2.0
1.0
3.9
0.482
2.074
793.309
303440.633
1586.618
9
142
Integrating two study CSFs
The Shearer et al. (2021, 7161466) and Vieira et al. (2013, 2919154) have considerable
differences with respect to several aspects including outcomes considered (RCC vs. any kidney
cancer), exposure assessment (serum biomarker vs. modeled exposure), source population
(multi-center nationally vs. Ohio and WV), study size (324 cases and 324 matched controls us.
59 cases and 7585 registry-based controls). Additionally, the dramatically different slopes
resulting from the two studies (0.0981, 95% CI: 0.0025, 0.1937 vs. 0.0122, 95% CI: 0.0006,
0.0238 from Shearer et al. (2021, 7161466) and Vieira et al. (2013, 2919154), respectively), are
an indication that the studies have considerable differences.
EPA performed a sensitivity analysis to derive a CSFsemm based on the pooled data from the two
studies. EPA pooled the study-specific slopes estimated as previously described using a random
effects REML approach. A pooled lifetime kidney cancer Ro was calculated as a weighted
E-327
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
average of the outcome specific Ro, weighted by the inverse of the sample size, applied to the
upper 95% CL of the pooled dose-response slope. When the highest exposure group from was
excluded {Vieira, 2013, 2919154}, the estimated CSF serum IS 0. 00242 (ng/mL)-1 (Table E-45).
When the highest exposure group from Vieira et al. (2013, 2919154) was included, the estimated
CSF serum IS 0. 00257 (ng/mL)"1 (Table E-45).
Table E-45.CSF calculations pooling dose-response for Shearer et al. (2021, 7161466) and
Vieira et al. (2013, 2919154) studies
Calculations
Shearer et al. (2021,7161466),
Vieira et al. (2013,2919154)3
Shearer et al. (2021, 7161466),
Vieira et al. (2013,2919154)b
Pooled b
0.041
0.038
CIb LCL
-0.038
-0.051
CIb UCL
0.121
0.128
Ro
0.02004
0.02004
CSF serum- central
0.00082
0.00076
CSF serum- UCL
0.00242
0.00257
Notes:
a Highest exposure level excluded
b Highest exposure level excluded
Another approach for deriving a combined CSF is to take the geometric mean of the study-
specific CSF serum, resulting in a combined CSF serum of 0.00130 (ng/mL)"1 and of 0.00070 (ng/mL)"
1 when the highest exposure group from Vieira et al. (2013, 2919154) was excluded or included,
respectively and the upper limits of the dose-response slopes were used.
However, in this particular situation, given that the two studies have considerable differences
listed above, EPA believes that these studies should not be combined in this manner.
E.2 Toxicology Studies
E.2.1 Butenhoff, 2012, 2919192
EPA conducted dose response modeling of the Butenhoff et al. (2012, 2919192) study using the
BMDS 3.2 program. This study addresses Ley dig cell adenomas in the testes in male Sprague-
Dawley Crl:COBS@CD(SD)BR rats.
E.2.1.1 Ley dig Cell Adenomas in the Testes
Increased incidence of Ley dig cell adenomas in the testes was observed in male Sprague-Dawley
Crl:COBS@CD(SD)BR rats. Dichotomous models were used to fit dose-response data. BMR of
4% and 10% change in the response were chosen. The 4% change was chosen because it is the
low end of the observed response within the study and the 10% change was chosen because it is
the recommended standard reporting level for comparison across chemicals 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 AUC for duration of the study (AUC) was
E-328
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
selected for this model because the AUC accounts for the accumulation of effects expected to
precede the increased incidence of Ley dig cell adenomas.
Table E-46. Dose-Response Modeling Data for Leydig Cell Adenomas in the Testes in Male
Sprague-Dawley Crl:COBS@CD(SD)BR Rats Following Exposure to PFOA {Butenhoff,
2012, 2919192}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L/day)
Number per Group
Incidence
0
0
50
0
1.3
43,263.7
50
2
14.2
167,102.5
50
7
BMD modeling results for Leydig cell adenomas in the testes are summarized in Table E-47 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 Multistage Degree 1 model had the AIC. The lower
bound on the dose level corresponding to the 95% lower confidence limit for a 4% change in the
response (BMDL4) from the selected Multistage Degree 1 model is 27,089.3 mg/L/day.
E-329
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-47. Summary of Benchmark Dose Modeling Results for Leydig Cell Adenomas in the Testes in Male Sprague-Dawley
Crl:COBS@CD(SD)BR Rats Following Exposure to PFOA {Butenhoff, 2012, 2919192}
Model"
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Dose Group Control
near BMD4 near BMD10 Dose Group
BMD4 BMDL4 BMD10 BMDL10 Basis for Model
(mg/L/day) (mg/L/day) (mg/L/day) (mg/L/day) Selection
Multistage
Degree 2
0.956 61.3
0.05
-0.03
44,791.1
27,088.1
115,604.6
69,904.0
Multistage
Degree 1
0.956 61.3
0.05
-0.03
-8.9 x e"
44,791.1
27,089.3
115,604.6
69,901.5
EPA selected the
Multistage
Degree 1 model.
All 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.
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL =benchmark dose lower limit; BMD4 =dose level corresponding to a 4% change in the response;
BMDL4 = lower bound on the dose level corresponding to the 95% lower confidence limit for a 4% change in the response; BMD10 = dose level corresponding to a 10% change in
the response; BMDL10 = lower bound on the dose level corresponding to the 95% lower confidence limit for a 10% change in the response.
a Selected model in bold.
E-330
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
^—Estimated Probability
Response at BMD
— — Linear Extrapolation
O Data
BMD
BMDL
Figure E-4. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 1 Model for Leydig Cell Adenomas in the Testes in Male Sprague-Dawley
Crl:COBS@CD(SD)BR Rats Following Exposure to PFOA with BMR 4% Extra Risk
{Butenhoff, 2012, 2919192}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.2 Dewitt, 2008,1290826
EPA conducted dose response modeling of the Dewitt et al. (2008, 1290826) study using the
BMDS 3.2 program. This study addresses serum sheep red blood cells (SRBC)-specific IgM
antibody titers in female C57BL/6N mice (Study I) and SRBC-specific IgM antibody titers in
female C57BL/6N mice (Study II).
E.2.2.1 Serum Sheep Red Blood Cells-specific IgM antibody titers in
Female C57BL/6N Mice (Study I)
Decreased mean response of SRBC-specific IgM antibody titers was observed in female
C57BL/6N mice (Study I). Continuous models were used to fit dose-response data. A benchmark
response (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-48. The 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 effects leading to decreased response of
SRBC-specific IgM antibody titers.
E-331
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-48. Dose-Response Modeling Data for Serum Sheep Red Blood Cells-specific IgM
Antibody Titers in Female C57BL/6N Mice (Study I) Following Exposure to PFOA
{Dewitt, 2008,1290826}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Mean Response (log2 to
reach 0.5 OD)a
0
0
8
8.0 ± 0.3b
3.75
73.0
8
7.1 ±0.6
7.5
90.8
8
6.8 ±0.3
15
103.7
8
6.1 ± 0.8
30
118.3
8
5.6 ±0.8
Notes:
a Data are presented as mean ± standard deviation.
b Standard deviations were calculated from standard errors.
BMD modeling results for serum SRBC-specific IgM antibody titers are summarized in Table
E-49 and Figure E-5. The best fitting model was the Polynomial Degree 4 model based on
adequate p-values (greater than 0.1), and the Polynomial Degree 4 model had the lowest AIC.
The BMDLisd from the selected Polynomial Degree 4 model is 18.2 mg/L.
Table E-49. Summary of Benchmark Dose Modeling Results for Serum Sheep Red Blood
Cells-specific IgM Antibody Titers in Female C57BL/6N Mice (Study I) Following
Exposure to PFOA (nonconstant variance) {Dewitt, 2008,1290826}
Goodness of Fit Scaled Residual
BMDisd BMDLisd Basis for Model
1VIOUCI
p-value
AIC
Dose Group
Control Dose
(mg/L)
(mg/L)
Selection
near BMD
Group
Exponential 2
0.0183
77.4
-0.29
-0.29
15.5
10.9
EPA selected the
Exponential 3
0.2736
72.0
-0.31
-0.02
47.5
28.8
Polynomial Degree
4 model. All
Exponential 4
0.0183
77.4
-0.30
-0.30
15.6
10.9
models, except
Exponential 5
0.2736
72.0
-0.31
-0.02
47.4
28.8
Exponential 2,
Hill
0.1148
73.9
-0.27
-0.03
45.8
27.7
Exponential 4, and
Polynomial
Degree 4
0.5269
69.6
-0.05
-0.05
31.9
18.2
Linear, had
adequate fit (p-
Polynomial
Degree 3
0.5127
69.7
-0.16
-0.05
38.4
20.2
values greater than
0.1), and the
Polynomial Degree
Polynomial
0.4705
69.9
-0.13
-0.10
43.7
23.0
4 model had the
Degree 2
lowest AIC.
Power
0.2888
71.9
-0.27
-0.03
45.7
27.0
Linear
0.0323
76.2
-0.36
-0.36
17.2
12.1
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-332
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
9
8 fFt
7
C.
1 ^^^_T
o
(D
£ S
-L Estimated Probability
c
o
Q.
Response at BMD
QJ ^
Ctl
3
O Data
2
BMD
1
BMDL
1
n
u
0
20 40 60 80 100
Dose
Figure E-5. Plot of Mean Response by Dose with Fitted Curve for the Polynomial Degree 4
Model for Serum Sheep Red Blood Cells-specific IgM Antibody Titers in Female
C57BL/6N Mice (Study I) Following Exposure to PFOA {Dewitt, 2008, 1290826}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.2.2 Serum Sheep Red Blood Cells-specific IgM antibody titers in
Female C57BL/6N Mice (Study II)
Decreased mean response of serum SRBC-specific IgM antibody titers was observed in female
C57BL/6N mice (Study II). 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-50. The Ciastzavg was selected for this model.
Table E-50. Dose-Response Modeling Data for Serum Sheep Red Blood Cells-specific IgM
Antibody Titers in Female C57BL/6N Mice (Study II) Following Exposure to PFOA
{Dewitt, 2008,1290826}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Mean Response (log2 to
reach 0.5 OD)a
0
0
8
7.9 ± 0.3b
0.94
24.2
8
8.0 ±0.3
1.88
45.3
8
7.8 ±0.3
3.75
73.1
8
7.4 ±0.3
7.50
91.5
8
7.3 ±0.3
Notes:
a Data are presented as mean ± standard deviation.
b Standard deviations were calculated from standard errors.
E-333
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The benchmark dose (BMD) modeling results for serum SRBC-specific IgM antibody titers are
summarized in Table E-51. No models provided an adequate fit, therefore a NOAEL
approach was taken for this endpoint.
Table E-52. Summary of Benchmark Dose Modeling Results for Serum Sheep Red Blood
Cells-specific IgM Antibody Titers in Female C57BL/6N Mice (Study II) Following
Exposure to PFOA (nonconstant variance) {Dewitt, 2008,1290826}
Goodness of Fit
Scaled Residual
BMDisd
BMDLisd
Basis for Model
Model
p-value
AIC
Dose Group
Control Dose
(mg/L)
(mg/L)
Selection
near BMD
Group
Exponential 2
<0.0001
25.6
1.0
-1.4
39.8
27.6
No models had
Exponential 3
<0.0001
24.3
0.2
-0.7
56.5
36.6
adequate fit (p-
Exponential 4
<0.0001
25.6
1.0
-1.4
39.7
27.6
values greater than
0.1).
Exponential 5
<0.0001
24.3
0.2
-0.7
56.6
36.6
Hill
<0.0001
24.4
0.0
-0.5
53.7
40.7
Polynomial
Degree 4
<0.0001
22.4
0.2
-0.6
58.2
36.8
Polynomial
Degree 3
<0.0001
22.4
0.2
-0.6
58.2
36.8
Polynomial
Degree 2
<0.0001
22.4
0.2
-0.6
58.2
36.8
Power
<0.0001
24.4
0.2
-0.7
56.5
36.5
Linear
<0.0001
25.4
1.0
-1.4
40.3
28.3
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.
E.2.3 Lou, 2006,, 1276159
EPA conducted dose response modeling of the Lau et al. (2006, 1276159) study using the BMDS
3.2 program. This study addresses prenatal loss (% live per litter) and maternal body weight
change in Po female CD-I mice, and fetal body weight and time to eye opening in Fi male and
female CD-I mice.
E.2.3.1 Prenatal Loss (% live per litter)
Increased mean response of prenatal loss 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 0.5 standard
deviations from the control mean was chosen. The doses and response data used for the modeling
are listed in Table E-53. The Cavg,dam,gest and Cmax,dam were both considered and shown below
because prenatal loss could be a result of exposure during a sensitive window of development
where a Cmax metric is expected to better correlate with the effect or an accumulation of exposure
where an average concentration metric is expected to better correlate with the effect.
E-334
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-53. Dose-Response Modeling Data for Prenatal Loss in Po Female CD-I Mice
Following Exposure to PFOA {Lau, 2006,1276159}
Administered Dose
(mg/kg/day)
Internal Dose
Cavg,dam,gest
(mg/L)
Cmax,dam
(mg/L)
Number per group
Mean Response (%
live per litter)3
0
0
0
42
4.1 ± 9.1b
1
33.9
62.0
15
1.0 ±2.7
3
74.9
114.9
16
7.4 ± 10
5
91.6
135.9
20
2.4 ±3.6
10
112.6
177.4
14
7.7 ± 12.3
20
139.7
252.9
5
25.9 ±26.2
Notes:
a Data are presented as mean ± standard deviation.
b Standard deviations were calculated from standard errors.
The benchmark dose (BMD) modeling results for prenatal loss using Cavg,dam,gest and Cmax,dam are
summarized in Table E-54 and Table E-55, respectively. No models provided an adequate fit,
therefore a NOAEL approach was taken for this endpoint.
Table E-54. Summary of Benchmark Dose Modeling Results for Prenatal Loss using
Cavg,dam,gest in Po Female CD-I Mice Following Exposure to PFOA (nonconstant variance)
{Lau, 2006,1276159}
Model
Goodness of Fit Scaled Residual
BMDo.ssd
p-value AIC D°Se S°"P ControlDose (mg/L,
1 near BMD Group
BMDLo.ssd Basis for Model
(mg/L) Selection
Exponential
< 0.000
833.0
0.58
1.49
79.5
64.1
No models had
2
1
adequate fit (p-
Exponential
3
< 0.000
1
812.9
0.53
0.21
119.5
105.5
values greater than
0.1).
Exponential
4
< 0.000
1
837.7
3.09
0.01
224.6
218.2
Exponential
5
< 0.000
1
810.5
0.06
0.24
112.9
107.2
Hill
< 0.000
1
810.5
0.08
0.24
113.1
106.7
Polynomial
< 0.000
813.3
-0.09
0.32
110.1
95.8
Degree 5
1
Polynomial
< 0.000
815.9
-0.03
0.33
110.4
90.8
Degree 4
1
Polynomial
< 0.000
819.6
2.17
0.02
188.4
149.7
Degree 3
1
Polynomial
< 0.000
827.4
2.57
-3.6xe~3
280.9
147.8
Degree 2
1
Power
< 0.000
1
812.1
0.46
0.21
117.4
106.9
E-335
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Model
Goodness of Fit Scaled Residual
BMDo.ssd BMDLo.ssd
n-value AIC D«se Group Control Dose (mg/L) (mg/L)
1 near BMD Group
Basis for Model
Selection
Linear
<0.000 837.7
1
3.11
-0.04
279.1
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDOo.5sd = 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 Lower limit includes zero; BMDL not estimated.
Table E-55. Summary of Benchmark Dose Modeling Results for Prenatal Loss using
Cmax,dam in Po Female CD-I Mice Following Exposure to PFOA (nonconstant variance)
{Lau, 2006,1276159}
Model
Goodness of Fit
p-value AIC
Scaled Residual
BMDo.ssd BMDLo.ssd
Dose Group Control Dose (mg/L) (mg/L)
near BMD Group
Basis for Model
Selection
Exponential < 0.000 852.0
2 1
Exponential <0.000 816.9
3 1
Exponential < 0.000 837.2
4 1
Exponential <0.000 814.4
1
5
Hill
<0.000 819.4
1
Polynomial <0.000 813.0
Degree 5 1
Polynomial <0.000 813.4
Degree 4 1
Polynomial <0.000 815.1
Degree 3 1
Polynomial < 0.000 822.4
Degree 2 1
<0.000 817.1
Power
<0.000 834.4
Linear ^
-9999
1.42
2.87
0.48
-9999
1.40
1.52
1.75
2.18
1.44
2.94
-0.71
-0.07
-0.80
0.25
-0.08
-0.24
-0.25
-0.02
-0.09
-0.09
-0.27
-9999
339.8
1467.4
190.0
65535.0
371.4
476.2
574.9
952.4
508.4
6308.9
145.0
956.9
161.8
a
253.7
140.5
287.8
359.8
352.2
765.5
No models had
adequate fit (p-
values greater than
0.1).
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDOo.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 Lower limit includes zero; BMDL not estimated.
E.2.3.2 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. A BMR of a change in the mean equal to
E-336
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
0.5 standard deviations and a BMR of a 5% decrease in pup weight were chosen. The doses and
response data used for the modeling are listed in Table E-56. The Cavg,pup,gest was selected for this
model 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 fetal body weight.
Table E-56. Dose-Response Modeling Data for Fetal Body Weight in Fi Male and Female
CD-I Mice Following Exposure to PFOA {Lau, 2006,1276159}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Mean Response (g)a
0
0
42
1.1 ±0.1b
1
8.5
15
1.0±0.1
3
18.7
16
1.0 ±0.2
5
22.9
20
1.0 ±0.2
10
28.1
14
1.0 ±0.2
20
34.9
5
0.9 ±0.2
Notes:
a Data are presented as mean ± standard deviation.
b Standard deviations were calculated from standard errors.
The BMD modeling results for fetal body weight are summarized in Table E-57. No models
provided an adequate fit, therefore a NOAEL approach was taken for this endpoint.
E-337
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-57. Summary of Benchmark Dose Modeling Results for Fetal Body Weight Change in Fi Male and Female CD-I Mice
Following Exposure to PFOA (constant variance) {Lau, 2006,1276159}
Goodness of Fit Scaled Residual
Model
p-value
AIC
Dose Group
near
BMDo.ssd
Dose Group
near BMDs
Control
Dose Group
BMDo.ssd
(mg/L)
BMDLo.ssd
(mg/L)
BMDs
(mg/L)
BMDLs
(mg/L)
Basis for Model
Selection
Exponential 2
0.561
-95.1
-0.2
-1.3
0.5
17.8
11.6
12.4
8.3
No models had
Exponential 3
0.561
-95.1
-0.2
-1.3
0.5
17.8
11.6
12.4
8.3
adequate fit for
the constant or
Exponential 4
0.395
-93.1
-0.2
-1.3
0.5
17.8
1.2
12.4
a
non-constant
variance (p-
Exponential 5
0.586
-94.1
-0.1
2.6 x e~3
2.6 x e"3
4.7
1.3
2.7
a
values were less
than 0.05).
Hill
0.611
-94.2
-0.2
4.3 x e~3
4.3 x e~3
5.0
a
2.5
a
Polynomial
Degree 5
0.555
-95.0
-0.3
-1.3
0.6
19.0
12.2
13.5
9.2
Polynomial
Degree 4
0.551
-95.0
-0.2
-1.3
0.5
18.3
12.2
12.9
8.9
Polynomial
Degree 3
0.551
-95.0
-0.2
-1.3
0.5
18.3
12.2
12.9
8.9
Polynomial
Degree 2
0.551
-95.0
-0.2
-1.3
0.5
18.3
12.2
12.9
8.9
Power
0.551
-95.0
-0.2
-1.3
0.5
18.3
12.2
12.9
8.9
Linear
0.551
-95.0
-0.2
-1.3
0.5
18.3
12.2
12.9
8.9
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; BMD5 = dose level corresponding to a 5% change; 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; BMDL5 = lower bound on the dose level corresponding to the 95% lower
confidence limit for a 5% change.
a Lower limit includes zero; BMDL not estimated.
E-338
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.2.33 Time to Eye Opening
Decreased mean response of time to eye opening 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 1 standard deviations from the control mean was selected, and a BMR of a change in the
mean equal to 0.5 standard deviations from the control mean is provided for comparison
purposes. The doses and response data used for the modeling are listed in Table E-58. The
average concentration normalized per day during gestation, Cavg,Pup,gest, average concentration
normalized per day during lactation (Cavg,puP,iact), maximum fetal concentration during gestation
(Cmax,puP,gest), and maximum pup concentration during lactation (Cmax,pup,iact) were all considered
and shown below because time to eye opening could be a result of exposure during a sensitive
window of development where a Cmax metric is expected to better correlate with the effect or an
accumulation of exposure where an average concentration metric is expected to better correlate
with the effect and time to eye opening could be due to exposure during the gestation or lactation
lifestages. The Cavg,PuP,gest was selected for this model.
Table E-58. Dose-Response Modeling Data for Time to Eye Opening in Fi Male and Female
CD-I Mice Following Exposure to PFOA {Lau, 2006,1276159}
Administered
Internal Dose
Number
per group
Mean
Response
(days)3
Dose
(mg/kg/day)
Cavg,pup,gest
(mg/L)
Cavg,pup,lact
(mg/L)
Cmax,pup,gest
(mg/L)
Cmax, pup, la ct
(mg/L)
0
0
0
0
0
22
14.8 ± 0.5b
1
8.8
21.7
16.0
28.7
8
15.2 ±0.6
3
19.1
33.0
28.8
44.0
8
15.5 ±0.3
5
23.2
35.1
34.0
46.9
17
16.0 ±0.8
10
28.3
37.7
44.4
50.5
13
17.2 ± 1.1
20
35.1
40.9
63.3
58.4
3
17.9 ± 1.4
Notes:
a Data are presented as mean ± standard deviation.
b Standard deviations were calculated from standard errors.
For Cavg,pup,gest, the benchmark dose (BMD) modeling results for time to eye opening are
summarized in Table E-59 and Figure E-6. The best fitting model was the Polynomial Degree 2
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 Polynomial Degree 2 model
had the lowest AIC. The BMDLisd from the selected Polynomial Degree 2 model is 10.1 mg/L.
E-339
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-59. Summary of Benchmark Dose Modeling Results for Time to Eye Opening using Cavg,PuP,gest in Fi Male and Female
CD-I mice Following Exposure to PFOA (nonconstant variance) {Lau, 2006,1276159}
Goodness of Fit Scaled Residual
Dose
Dose
Control
Dose
Group
BMDo.ssd
BMDLo.ssd
BMDisd
BMDLisd
Basis for
Model"
p-value
AIC
Group
near
BMDo.ssd
Group
near
BMDisd
(mg/L)
(mg/L)
(mg/L)
(mg/L)
Model
Selection
Exponential
0.006
162.8
0.5
-0.5
0.5
3.7
2.9
7.4
S &
EPA selected
2
D. O
the Polynomial
Exponential
0.152
155.4
1.3
-0.5
-0.5
12.0
7.4
16.0
11.3
Degree 2
3
model. All
Exponential
0.001
165.6
0.5
-0.6
0.5
3.6
2.8
7.1
5.6
models, except
4
Exponential 2,
Exponential
0.072
157.4
1.4
-0.4
-0.6
12.9
7.9
16.6
11.9
4, and 5, and
5
Linear, had
Hill
0.123
156.3
0.6
0.6
-1.1
17.1
8.9
19.5
18.2
adequate fit (p-
values greater
Polynomial
0.104
156.7
0.8
-0.6
-0.2
9.2
4.4
14.6
8.8
than 0.1), the
Degree 5
BMDLs were
Polynomial
0.104
156.7
0.8
-0.6
-0.2
9.1
4.6
14.6
9.1
sufficiently
Degree 4
close (less than
Polynomial
0.202
154.8
1.0
-0.6
-0.3
10.2
5.4
14.9
10.0
threefold
Degree 3
difference),
Polynomial
0.180
154.4
1.0
1.0
-0.2
9.8
5.7
13.8
10.1
and the
Degree 2
Polynomial
Power
0.155
155.4
1.3
-0.5
-0.5
12.5
7.8
16.3
11.7
Degree 2
model had the
Linear
0.004
163.6
0.5
-0.6
0.5
3.6
2.8
7.1
5.6
lowest AIC.
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.5sd = dose level corresponding to a change in the mean equal to
0.5 standard deviation(s) 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(s) from the control mean; 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-340
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
25
20
8 15 <3=
£
O
Q.
LO
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-60. Summary of Benchmark Dose Modeling Results for Time to Eye Opening using Cavg,PuP,iact in Fi Male and Female
CD-I mice Following Exposure to PFOA (nonconstant variance) {Lau, 2006,1276159}
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Dose Group Control
Group near near Dose
BMDo.ssd BMDisd Group
BMDo.ssd BMDLo.ssd BMDisd BMDLisd
(mg/L) (mg/L) (mg/L) (mg/L)
Basis for Model
Selection
Exponential 2
Exponential 3
<0.0001
0.060
177.8
157.5
0.4
-0.6
-1.7
-0.6
0.4
-0.7
5.7
28.5
4.4
27.9
11.4
31.3
8.8
30.7
EPA selected the
Hill model. The
Hill model had
Exponential 4
<0.0001
180.4
0.4
-1.8
0.4
5.6
4.3
11.1
8.6
adequate fit (p-
Exponential 5
Hill
Polynomial
Degree 5
Polynomial
Degree 4
Polynomial
Degree 3
0.027
0.235
0.048
0.040
0.005
159.3
154.4
157.8
157.8
162.8
-0.5
0.5
1.2
1.0
0.3
-0.5
0.5
-1.4
1.0
0.3
-0.8
-1.1
-0.2
-0.1
0.2
28.9
31.7
24.1
22.5
19.0
27.9
28.7
13.4
14.4
13.4
31.6
33.1
28.2
26.8
23.9
30.6
32.3
22.0
21.5
19.9
values greater
than 0.1), the
BMDLs were
sufficiently close
(less than
threefold
difference), and
the Hill model
had the lowest
Polynomial
Degree 2
0.0005
168.3
-0.6
-0.6
0.4
13.7
10.1
19.4
16.1
AIC.
Power
0.066
157.3
-0.5
-0.5
-0.8
28.9
27.6
31.6
28.2
Linear
<0.0001
178.4
0.4
-1.8
0.4
5.6
4.3
11.1
8.5
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; 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-342
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
25
Dose
Estimated Probability
Response at BMD
O Data
BMD
BMDL
Figure E-7. Plot of Mean Response by Dose with Fitted Curve for the Selected Hill Model
for Time to Eye Opening using Cavg,PuP,iact in Fi Male and Female CD-I Mice Following
Exposure to PFOA {Lau, 2006,1276159}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
For Cmax.pup.gest, the benchmark dose (BMD) modeling results for time to eye opening are
summarized in Table E-61 and Figure E-8. The best fitting model was the Power 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 Power model had the lowest AIC. The
BMDLisd from the selected Power model is 14.6 mg/L.
E-343
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-61. Summary of Benchmark Dose Modeling Results for Time to Eye Opening using Cmax,pup,gest in Fi Male and Female
CD-I mice Following Exposure to PFOA (nonconstant variance) {Lau, 2006,1276159}
Goodness of Fit Scaled Residual
Model"
p-value
AIC
Dose
Group
near
BMDo.ssd
Dose
Group
near
BMDisd
Control
Dose
Group
BMDo.ssd
(mg/L)
BMDLo.ssd
(mg/L)
BMDisd
(mg/L)
BMDLisd
(mg/L)
Basis for
Model
Selection
Exponential
2
0.013
160.8
0.6
-0.89
0.64
5.8
4.6
11.5
9.1
EPA selected
the Power
Exponential
3
0.111
156.1
0.5
0.45
-0.02
13.8
8.4
20.2
14.2
model. All
models, except
Exponential
4
0.009
161.7
0.6
-0.97
0.61
5.6
5.5
11.1
10.9
Exponential 2,
Exponential 4,
Exponential
5
0.056
157.9
0.6
0.57
-0.08
14.7
9.4
20.9
15.1
Exponential 5,
and Linear,
Hill
0.149
155.9
0.4
0.45
-1.04
25.7
11.5
29.0
17.5
had adequate
fit (p-values
Polynomial
Degree 5
0.123
155.9
0.5
0.55
-0.07
14.5
7.0
20.8
13.5
greater than
0.1), the
Polynomial
Degree 4
0.122
155.9
0.6
0.55
-0.06
14.6
7.0
20.9
13.5
BMDLs were
sufficiently
Polynomial
Degree 3
0.123
155.9
0.6
0.55
-0.07
14.5
7.0
20.8
13.5
close (less than
threefold
Polynomial
Degree 2
0.123
155.9
0.5
0.55
-0.08
14.5
7.7
20.8
13.9
difference),
and the Power
Power
0.124
155.9
0.5
0.54
-0.05
14.4
8.8
20.7
14.6
model had the
lowest AIC.
Linear
0.009
161.7
0.6
-0.97
0.61
5.6
4.4
11.1
8.7
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.5sd = dose level corresponding to a change in the mean equal to
0.5 standard deviation(s) 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(s) from the control mean; 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-344
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
25
20
15
£
O
Q.
LO
CD
10
Estimated Probability
^^Response at BMD
O Data
BMD
BMDL
10
20
30
Dose
40
50
60
Figure E-8. Plot of Mean Response by Dose with Fitted Curve for the Selected Power
Model for Time to Eye Opening using Cmax,pup,gest in Fi Male and Female CD-I Mice
Following Exposure to PFOA {Lau, 2006,1276159}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
For Cmax.pup.iact, the benchmark dose (BMD) modeling results for time to eye opening are
summarized in Table E-62 and Figure E-9. 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 AIC. The
BMDLisd from the selected Hill model is 43.0 mg/L.
E-345
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-62. Summary of Benchmark Dose Modeling Results for Time to Eye Opening using Cmax,pup,iact in Fi Male and Female
CD-I mice Following Exposure to PFOA (nonconstant variance) {Lau, 2006,1276159}
Goodness of Fit Scaled Residual
Model"
p-value
AIC
Dose
Group
near
BMDo.ssd
Dose
Group
near
BMDisd
Control
Dose
Group
BMDo.ssd
(mg/L)
BMDLo.ssd
(mg/L)
BMDisd
(mg/L)
BMDLisd
(mg/L)
Basis for
Model
Selection
Exponential 2
<0.0001
176.4
0.4
-1.7
0.4
7.7
6.0
15.3
11.8
EPA selected
Exponential 3
0.013
160.9
1.3
-1.2
-0.4
33.5
32.5
38.6
33.7
the Hill
11.4
model. The
Exponential 4
<0.0001
179.0
0.4
-1.7
0.4
7.5
5.7
14.9
Hill had
adequate fit
Exponential 5
0.005
162.7
1.4
-1.1
-0.5
34.5
33.0
39.3
.3 / .o
(p-values
Hill
0.228
154.5
0.5
0.5
-1.1
42.3
39.5
44.1
43.0
greater than
0.1), the
Polynomial
0.037
158.4
0.9
-1.3
-0.1
30.0
15.3
36.7
27.0
BMDLs were
Degree 5
sufficiently
Polynomial
0.023
159.5
0.9
0.9
-0.2
29.9
16.7
35.8
27.3
close (less
Degree 4
than threefold
Polynomial
0.010
161.5
0.4
0.4
0.2
25.6
17.7
32.3
26.6
difference),
Degree 3
and the Hill
Polynomial
0.001
166.6
-0.6
-0.6
0.4
18.5
13.7
26.1
21.7
model had the
Degree 2
37.5
lowest AIC.
Power
0.014
160.7
1.4
-1.1
-0.5
34.5
30.4
39.3
Linear
<0.0001
177.0
0.4
-1.7
0.4
7.4
5.7
14.9
11.4
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.5sd = dose level corresponding to a change in the mean equal to
0.5 standard deviation(s) 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(s) from the control mean; 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-346
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
25
20
d) 15
LO
£
O
Q.
LO
CD
10
Estimated Probability
^^Response at BMD
O Data
BMD
BMDL
10
20
30
Dose
40
50
Figure E-9. Plot of Mean Response by Dose with Fitted Curve for the Selected Hill Model
for Time to Eye Opening using Cmax,pup,iact in Fi Male and Female CD-I Mice Following
Exposure to PFOA {Lau, 2006,1276159}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.4 Li, 2018, 5084746
EPA conducted dose response modeling of the Li et al. (2018, 5084746) study using the BMDS
3.2 program. This study addresses fetal body weight in Fi male and female Kunming mice and
maternal body weight in Po female Kunming mice.
E.2.4.1 Fetal Body Weight
Decreased mean response of fetal body weight was observed in Fi male and female Kunming
mice. 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-63. The Cavg.Pup.gest was selected for this model 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 fetal body weight.
Table E-63. Dose-Response Modeling Data for Fetal Body Weight in Fi Male and Female
Kunming Mice Following Exposure to PFOA {Li, 2018, 5084746}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Mean Response (g)a
0
0
10
1.5 ±0.01
1
8.5
10
1.5 ±0.01
5
22.9
10
1.3 ±0.01
10
28.1
10
1.0 ±0.10
E-347
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Administered Dose
Internal Dose
(mg/kg/day)
(mg/L)
Number per Group
Mean Response (g)a
20
34.9
10
0.9 ±0.05
Notes:
a Data are presented as mean ± standard deviation.
The benchmark dose (BMD) modeling results for fetal body weight are summarized in Table
E-64. No models provided an adequate fit, therefore a NOAEL approach was taken for this
endpoint.
E-348
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-64. Summary of Benchmark Dose Modeling Results for Fetal Body Weight in Fi Male and Female Kunming Mice
Following Exposure to PFOA (constant variance) {Li, 2018, 5084746}
Goodness of Fit Scaled Residual
Model
p-value
AIC
Dose Group
near
BMDo.ssd
Dose Group
near
BMDs
Control
Dose Group
BMDo.ssd
(mg/L)
BMDLo.ssd
(mg/L)
BMDs
(mg/L)
BMDLs
(mg/L)
Basis for Model
Selection
Exponential 2
<0.0001
-73.7
-2.5
-2.5
-2.5
2.5
2.1
3.7
3.3
No models had
Exponential 3
<0.0001
-116.1
0.5
0.5
-0.8
9.3
7.5
13.0
11.1
adequate fit (p-
values were less
Exponential 4
<0.0001
-73.7
-2.5
-2.5
-2.5
2.5
2.1
3.7
3.3
than 0.1 or Test 2
p-values were
Exponential 5
0.992
-151.8
1.4xe-3
1.4xe-3
-7.2 xe~3
16.6
15.4
19.6
18.7
less than 0.05).
Hill
1.000
-151.8
7.5 xe~8
7.5 xe~8
-2.3 xe~4
18.2
17.0
20.4
20.0
Polynomial
Degree 4
<0.0001
-114.5
0.8
0.8
-0.8
8.2
5.7
11.9
9.6
Polynomial
Degree 3
<0.0001
-114.5
0.8
0.8
-0.8
8.2
5.7
11.9
9.6
Polynomial
Degree 2
<0.0001
-114.5
0.8
0.8
-0.8
8.2
5.7
11.9
9.6
Power
<0.0001
-112.6
0.8
0.8
-0.9
8.1
6.2
11.7
9.7
Linear
<0.0001
-83.8
-2.7
2.4
-2.7
2.7
2.3
4.3
4.0
Notes'. AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; 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 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-349
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.2.5 Loveless, 2008, 988599
EPA conducted dose response modeling of the Loveless et al. (2008, 988599) study using the
BMDS 3.2 program. This study addresses focal necrosis in male Crl:CD(SD)IGS BR rats and
focal necrosis, individual cell necrosis, and IgM serum titer in male Crl:CD-l(ICR)BR mice.
E. 2.5.1 Focal Necrosis in Male Crl:CD-l (ICR)BR Mice
Increased incidence of focal necrosis was observed in male Crl:CD-l(ICR)BR mice.
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-65. The 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 effect resulting in focal necrosis.
Table E-65. Dose-Response Modeling Data for Focal Necrosis in Male Crl:CD-l(ICR)BR
Mice Following Exposure to PFOA {Loveless, 2008, 988599}
Administered Dose
(mg/kg/day)
Dose
(mg/L)
Number per Group
Incidence
0
0
19
0
0.3
27.7
20
1
1
70.5
20
3
10
119.2
20
4
30
158.9
19
7
The BMD modeling results for focal necrosis are summarized in Table E-66 and Figure E-10.
The best fitting model was the Dichotomous Hill model based on adequate p-values (greater than
0.1), and the Dichotomous Hill model had the lowest BMDL. The BMDLio from the selected
Dichotomous Hill model is 10.0 mg/L.
Table E-66. Summary of Benchmark Dose Modeling Results for Focal Necrosis in Male
Crl:CD-l(ICR)BR Mice Following Exposure to PFOA {Loveless, 2008, 988599}
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control Dose
near BMD Group
BMDio
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Dichotomous
0.809
76.3
0.10
-0.001
52.3
10.0
EPA selected the
Hill
Dichotomous Hill
Gamma
0.824
76.3
0.12
-0.018
52.6
30.5
model as it had the
Log-Logistic
lowest RMDT, All
0.936
74.3
0.10
-0.001
52.3
27.0
models had
Multistage
0.972
74.1
0.28
-0.001
55.0
30.9
adequate fit (p-
Degree 4
values greater than
Multistage
0.870
76.2
0.26
-0.001
55.0
30.8
0.1).
Degree 3
Multistage
0.847
76.2
0.20
-0.001
54.2
30.7
Degree 2
E-350
-------
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)
BMDLio
(mg/L)
Multistage
Degree 1
0.906
74.4
-0.24
-0.001
45.0
30.2
Weibull
0.829
76.3
0.14
-0.001
53.0
30.6
Logistic
0.760
75.5
0.76
-0.735
83.8
65.2
Log-Probit
0.781
76.4
0.02
-0.001
50.4
12.9
Probit
0.798
75.3
0.69
-0.656
78.7
60.8
Basis for Model
Selection
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.
1
0.9
0.8
Estimated Probability
^^Response at BMD
O Data
BMD
BMDL
Figure E-10. Plot of Incidence Rate by Dose with Fitted Curve for the Selected
Dichotomous Hill Model for Focal Necrosis in Male Crl:CD-l(ICR)BR Mice Following
Exposure to PFOA {Loveless, 2008, 988599}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.5.2 Individual Cell Necrosis in Mole Crl:CD-l(ICR)BR Mice
Increased incidence of individual cell necrosis was observed in male Crl:CD-l(ICR)BR mice.
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-67. The 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 effect resulting in cell necrosis.
E-351
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-67. Dose-Response Modeling Data for Individual Cell Necrosis in Male Crl:CD-
1(ICR)BR Mice Following Exposure to PFOA {Loveless, 2008, 988599}
Administered Dose
(mg/kg/day)
Dose
(mg/L)
Number per Group
Incidence
0
0
19
0
0.3
27.7
20
0
1
70.5
20
11
10
119.2
20
20
30
158.9
19
19
The BMD modeling results for individual cell necrosis are summarized in Table E-68 and Figure
E-l 1. The best fitting model was the Probit 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 Probit model had the lowest Akaike
information criterion (AIC). The BMDLio from the selected Probit model is 36.0 mg/L.
Table E-68. Summary of Benchmark Dose Modeling Results for Individual Cell Necrosis in
Male Crl:CD-l(ICR)BR Mice Following Exposure to PFOA {Loveless, 2008, 988599}
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control Dose
near BMD Group
BMDio
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Dichotomous
1.000
29.5
-0.001
-0.001
61.7
42.2
EPA selected the
Hill
Probit model. All
Gamma
0.990
31.7
-0.085
-0.001
49.4
36.7
models, except
Log-Logistic
1.000
31.5
-0.001
-0.001
61.7
42.2
Multistage Degree
1, had adequate fit
Multistage
0.981
30.3
-0.616
-0.001
42.7
31.5
(p-values greater
Degree 4
than 0.1), the
Multistage
0.840
32.3
-1.020
-0.001
35.4
26.9
BMDLs were
Degree 3
sufficiently close
Multistage
0.283
38.8
-1.767
-0.001
23.6
18.2
(less than threefold
Degree 2
difference), and the
Probit model had
Multistage
0.001
60.1
-0.001
-0.001
7.0
5.4
the lowest AIC.
Degree 1
Weibull
1.000
29.6
0.041
-0.001
50.3
50.1
Logistic
1.000
29.5
<0.0001
-0.001
61.2
39.1
Log-Probit
1.000
31.5
<0.0001
-0.001
61.6
39.1
Probit
1.000
29.5
<0.0001
<0.0001
58.1
36.0
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-352
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-ll. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Probit
Model for Individual Cell Necrosis in Male Crl:CD-l(ICR)BR Mice Following Exposure to
PFOA {Loveless, 2008, 988599}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.5.3 IgM Serum Titer in Male Crl:CD-l(ICR)BR Mice
Decreased mean response of IgM serum titer was observed in male Crl:CD-l(ICR)BR 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-69. The 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 effect resulting in decreased mean response of IgM serum titer.
Table E-69. Dose-Response Modeling Data for IgM Serum Titer in Male Crl:CD-l(ICR)BR
Mice Following Exposure to PFOA {Loveless, 2008, 988599}
Administered Dose
Dose
(mg/kg/day)
(mg/L)
Number per Group
Mean Response (mg/dL)a
0
0
20
8.9 ±0.6
0.3
27.7
20
8.9 ±0.8
1
70.5
20
8.4 ±0.7
10
119.2
20
7.2 ±0.8
30
158.9
20
6.4 ±0.8
Note:
a Data are presented as mean ± standard deviation.
E-353
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The BMD modeling results for IgM serum titer are summarized in Table E-70 and Figure E-12.
The best fitting model was the Exponential 3 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 Exponential 3 model had the lowest AIC. The BMDLisd from the selected
Exponential 3 model is 57.6 mg/L.
Table E-70. Summary of Benchmark Dose Modeling Results for IgM Serum Titer in Male
Crl:CD-l(ICR)BR Mice Following Exposure to PFOA (constant variance) {Loveless, 2008,
988599}
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMDisd
(mg/L)
BMDLisd
(mg/L)
Basis for Model
Selection
Exponential 2
0.004
239.1
1.0
-1.9
42.4
35.4
EPA selected the
Exponential 3
Exponential 4
Exponential 5
0.527
0.004
0.261
228.9
239.1
230.9
0.5
1.0
0.5
-0.4
-1.9
-0.4
75.0
42.4
75.1
57.6
35.4
57.6
Exponential 3
model. All models,
except Exponential
2, Exponential 4
Hill
0.901
229.6
0.0
-0.1
80.3
62.2
and Linear, had
Polynomial
Degree 4
Polynomial
Degree 3
Polynomial
0.334
0.334
0.334
229.8
229.8
229.8
0.5
0.5
0.5
-0.5
-0.5
-0.5
75.3
75.3
75.3
54.4
54.4
54.4
adequate fit (p-
values greater than
0.1), the BMDLs
were sufficiently
close (less than
threefold
Degree 2
difference), and
Power
0.422
229.3
0.6
-0.5
74.0
55.9
the Exponential 3
Linear
0.018
235.7
0.9
-1.8
45.8
38.9
model had the
lowest AIC.
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-354
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
H
o cn
8
7
/
Estimated Probability
Response at BMD
O Data
QJ U
to
c
° s
Q.
tn
(D
C£ A
t
o
BMD
3
T
BMDL
Z
1
1
n
0 20 40
60 80 100 120 140
Dose
Figure E-12. Plot of Mean Response by Dose with Fitted Curve for the Selected
Exponential 3 Model for IgM Serum Titer in Male Crl:CD-l(ICR)BR Mice Following
Exposure to PFOA {Loveless, 2008, 988599}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.6 NTP, 2020, 7330145
EPA conducted dose response modeling of the NTP (2020, 7330145) study using the BMDS 3.2
program. This study addresses hepatocyte single cell death, necrosis in the liver, relative kidney
weight (right), hepatocellular adenomas, hepatocellular adenoma or carcinoma, and pancreatic
acinar cell adenoma in Fi male Sprague-Dawley rats and uterine adenocarcinoma in Fi female
Sprague-Dawley rats.
E. 2.6.1 Hepatocyte Single Cell Death
Increased incidence of hepatocyte single cell death was observed in Fi male Sprague-Dawley
rats. Dichotomous models were used to fit dose-response data. A benchmark response (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-71. The
Cavg.puP.total 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 effects
leading to an increased incidence of hepatocyte single cell death.
Table E-71. Dose-Response Modeling Data for Hepatocyte Single Cell Death in Fi Male
Sprague-Dawley Rats Following Exposure to PFOA {NTP, 2020, 7330145}
Administered Dose Internal Dose
Number per Group Incidence
(ppm)a (mg/L)
0/ 0 0 50 1
300 / 0 0.4 50 1
E-355
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Administered Dose
(ppm)a
Internal Dose
(mg/L)
Number per Group
Incidence
0/ 20
72.6
50
1
300 / 20
73.6
50
3
0/ 40
113.5
50
11
300 / 40
115.2
50
5
0/ 80
161.7
50
24
300 / 80
161.8
50
29
Note:
a Doses are presented as perinatal exposure/postnatal exposure.
Hepatocyte single cell death was assessed (1) following postweaning exposure, (2) following
perinatal and postweaning exposure, and (3) using a pooled method. The pooled method used the
dose response data associated with both postweaning exposure (1) and perinatal and postweaning
exposure (2).
The BMD modeling results for hepatocyte single cell death following postweaning exposure to
PFOA are summarized in Table E-72 and Figure E-13. The best fitting model was the Multistage
Degree 3 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
3 model had the lowest 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 3 model
is 77.1 mg/L.
Table E-72. Summary of Benchmark Dose Modeling Results for Hepatocyte Single Cell
Death in Fi Male Sprague-Dawley Rats Following Postweaning Exposure to PFOA {NTP,
2020, 7330145}
Goodness of Fit Scaled Residual
Model"
BMD io
BMDLio
Basis for Model
p-value
AIC
Dose Group
Control Dose
(mg/L)
(mg/L)
Selection
near BMD
Group
Dichotomous
_b
149.5
9.4 x e"4
0.03
104.5
85.9
EPA selected the
Hill
Multistage Degree
Gamma
0.308
148.6
0.64
0.29
98.8
82.2
3 model. All
models, except
Log-Logistic
0.262
148.9
0.67
0.32
98.5
81.5
Dichotomous Hill,
Multistage
0.354
148.2
-1.31
0.46
89.9
77.1
Multistage Degree
Degree 3
2, Multistage
Multistage
0.064
152.8
-2.00
0.52
73.8
61.9
Degree 1, and
Probit, had
adequate fit (p-
Degree 2
Multistage
0.001
162.8
-2.81
0.42
44.6
33.8
values greater than
Degree 1
0.1), the BMDLs
Weibull
0.200
149.3
0.80
0.32
98.4
80.1
were sufficiently
Logistic
close (less than
threefold
0.222
148.7
-1.24
1.02
92.3
77.8
Log-Probit
0.389
148.3
0.52
0.27
98.5
82.8
difference), and the
Probit
0.090
149.7
-1.37
1.67
86.8
72.1
Multistage Degree
"3 hctH thp
lowest AIC.
E-356
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
Figure E-13. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 3 Model for Hepatocyte Single Cell Death in Fi Male Sprague-Dawley Rats
Following Postweaning Exposure to PFOA {NTP, 2020, 7330145}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
The benchmark dose (BMD) modeling results for hepatocyte single cell death following
perinatal and postweaning exposure to PFOA are summarized in Table E-73 and Figure E-14.
The best fitting model was the Gamma 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 lower bound on the dose level corresponding to
the 95% lower confidence limit for a 10% response level BMDLio from the selected Gamma
model is 100.1 mg/L.
Table E-73. Summary of Benchmark Dose Modeling Results for Hepatocyte Single Cell
Death in Fi Male Sprague-Dawley Rats Following Perinatal and Postweaning Exposure to
PFOA {NTP, 2020, 7330145}
Scaled Residual
Model3
Dichotomous
Hill
Gamma
Log-Logistic
Goodness of Fit
p-value AIC
-b 142.1
0.427 138.7
0.320 140.1
Dose Group
near BMD
-0.07
-0.69
-0.07
Control Dose
Group
-0.68
-0.57
-0.68
BMDio
(mg/L)
121.2
114.8
121.1
BMDLio
(mg/L)
101.4
100.1
101.4
Basis for Model
Selection
EPA selected the
Gamma model.
The Gamma, Log-
Logistic, Weibull,
and Log-Probit had
adequate fit (p-
E-357
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control Dose
near BMD Group
BMDio
(mg/L)
BMDLio
(mg/L)
Multistage
Degree 3
Multistage
Degree 2
Multistage
Degree 1
Weibull
Logistic
0.043 144.1
0.002 151.1
163.2
< 0.000
1
0.330 140.0
0.044 141.9
Log-Probit 0.308 140.1
Probit
0.004 145.0
-0.31
-1.20
-2.14
-0.12
-1.46
-0.01
-0.04
0.40
0.52
0.43
-0.64
1.86
-0.72
2.53
88.7
72.5
43.0
121.2
97.1
121.1
89.4
77.5
61.5
32.7
98.3
82.4
105.2
74.7
Basis for Model
Selection
values greater than
0.1), the BMDLs
were sufficiently
close (less than
threefold
difference), and the
Gamma model had
the lowest AIC.
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 Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
Figure E-14. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Gamma
Model for Hepatocyte Single Cell Death in Fi Male Sprague-Dawley Rats Following
Perinatal and Postweaning Exposure to PFOA {NTP, 2020, 7330145}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
The benchmark dose (BMD) modeling results for hepatocyte single cell death using a pooled
method are summarized in Table E-74 and Figure E-15. The best fitting model was the
E-358
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 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 is 90.9 mg/L.
Table E-74. Summary of Benchmark Dose Modeling Results for Hepatocyte Single Cell
Death in Fi Male Sprague-Dawley Rats Following Exposure to PFOA (Pooled) {NTP, 2020,
7330145}
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control Dose
near BMD Group
BMDio
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Dichotomou
0.273
287.8
1.16
-0.07
105.3
92.5
s Hill
Gamma
0.380
286.0
1.07
-0.14
105.2
92.1
Log-Logistic
0.399
285.8
1.16
-0.07
105.3
92.5
Multistage
0.170
289.7
1.25
0.02
104.9
91.1
Degree 7
Multistage
0.170
289.7
1.25
0.02
104.9
91.2
Degree 6
Multistage
0.285
287.7
1.25
0.01
104.9
91.5
Degree 5
Multistage
0.536
284.1
0.90
0.17
100.5
90.9
Degree 4
Multistage
0.209
288.3
-0.27
0.43
89.3
82.6
Degree 3
Multistage
0.005
299.9
-1.17
0.52
73.1
66.1
Degree 2
Multistage
< 0.000
322.0
-2.84
0.47
43.8
35.9
Degree 1
1
Weibull
0.413
285.7
1.23
0.01
104.9
91.8
Logistic
0.160
287.0
0.78
1.40
94.6
84.3
Log-Probit
0.350
286.1
1.07
-0.23
106.0
92.4
Probit
0.015
290.9
-0.18
2.08
88.0
77.6
EPA selected the
Multistage Degree
4 model. All
models, except
Multistage Degree
1 and 2, and
Probit, 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 = 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-359
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1
0.9
0.8
0.7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Necrosis in the liver was assessed (1) following postweaning exposure, (2) following perinatal
and postweaning exposure, and (3) using a pooled method. The pooled method used the dose
response data associated with both postweaning exposure (1) and perinatal and postweaning
exposure (2).
The BMD modeling results for necrosis in the liver following postweaning exposure to PFOA
are summarized in Table E-76 and Figure E-16. 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 Log-Logistic model had the
lowest AIC. The BMDLio from the selected Log-Logistic model is 15.3 mg/L.
Table E-76. Summary of Benchmark Dose Modeling Results for Necrosis in the Liver in Fi
Male Sprague-Dawley Rats Following Postweaning Exposure to PFOA {NTP, 2020,
7330145}
Model"
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMD io
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Dichotomous
_b
225.6
-6.5 x e"4
-1.8 x e"4
62.5
C
Hill
Gamma
0.160
224.8
-0.2
-0.2
26.4
20.7
Log-Logistic
0.307
223.6
-0.1
-0.1
20.9
15.3
Multistage
Degree 3
0.160
224.8
-0.2
-0.2
26.4
20.7
Multistage
Degree 2
0.160
224.8
-0.2
-0.2
26.4
20.7
Multistage
Degree 1
0.160
224.8
-0.2
-0.2
26.4
20.7
Weibull
0.160
224.8
-0.2
-0.2
26.4
20.7
Logistic
0.008
231.4
1.4
-1.7
52.4
43.9
Log-Probit
0.307
224.2
9.2 x e"4
9.2 x e"4
1.7
C
Probit
0.011
230.6
1.4
-1.5
49.3
41.6
EPA selected the
Log-Logistic
model. All models,
except the
Dichotomous Hill,
Logistic and
Probit, had
adequate fit (p-
values greater than
0.1), the BMDLs
were sufficiently
" close (less than
threefold
difference), and the
Log-Logistic
model had the
lowest AIC.
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 Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
c Lower limit includes zero; BMDL not estimated.
E-361
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-16. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Log-Logistic
Model for Necrosis in the Liver in Fi Male Sprague-Dawley Rats Following Postweaning
Exposure to PFOA {NTP, 2020, 7330145}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
The BMD modeling results for necrosis in the liver following perinatal and postweaning
exposure to PFOA are summarized in Table E-77 and Figure E-17. The Dichotomous Hill model
was saturated and while the Log-Probit model had adequate fit, the BMD/BMDL ratio was larger
than three. Of the remaining models, the selected model was the Multistage Degree 1 model
based on adequate p-values (greater than 0.1) and lowest AIC. The BMDLio from the selected
Multistage Degree 1 model is 26.9 mg/L.
Table E-77. Summary of Benchmark Dose Modeling Results for Necrosis in the Liver in Fi
Male Sprague-Dawley Rats Following Perinatal and Postweaning Exposure to PFOA
{NTP, 2020, 7330145}
Goodness of Fit Scaled Residual
Model3
p-value
AIC
Dose Group
near BMD
Control
Dose Group
BMD io
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Dichotomous
_b
198.1
0.225
-0.009
40.7
10.0
EPA selected the
Hill
Multistage Degree
Gamma
0.611
196.1
0.212
-0.007
38.6
27.0
1 model. All
Log-Logistic
0.585
196.1
0.225
-0.008
40.7
22.4
models, except
Dichotomous Hill,
Multistage
0.645
196.0
0.267
-0.018
37.8
27.0
had adequate fit (p-
Degree 3
values greater than
Multistage
0.627
196.1
0.246
-0.014
38.1
27.0
0.1). The Log-
Degree 2
Probit model had a
Multistage
0.869
194.1
0.013
0.013
34.8
26.9
BMD/BMDL ratio
Degree 1
greater than three.
Weibull
0.614
196.1
0.220
-0.007
38.6
27.0
The Multistage
E-362
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Model3
Goodness of Fit
Scaled Residual
BMDio
(mg/L)
BMDLio
(mg/L)
p-value
AIC
Dose Group
near BMD
Control
Dose Group
Logistic
0.267
196.7
1.149
-1.063
68.0
57.3
Log-Probit
0.567
196.1
0.222
-0.007
43.0
3.7
Probit
0.348
196.0
1.095
-0.863
63.8
53.7
Basis for Model
Selection
Degree 1 model
had the lowest AIC
of the remaining
models.
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 Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
0.9
0.8
0.7
Estimated Probability
0 20 40 60 80 100 120 140 160
Dose
Figure E-17. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 1 Model for Necrosis in the Liver in Fi Male Sprague-Dawley Rats Following
Perinatal and Postweaning Exposure to PFOA {NTP, 2020, 7330145}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
The BMD modeling results for necrosis using pooled methods are summarized in Table E-78 and
Figure E-18. All models except the Logistic and Probit model had adequate fit with p-values
(greater than 0.1). While the Dichotomous Hill and Log-Probit model had adequate fit, the
BMD/BMDL ratio was larger than three. Of the remaining models, the selected model was the
Log-Logistic model based on adequate p-values (greater than 0.1) and lowest AIC. The BMDLio
from the selected Log- Logistic model is 20.1 mg/L.
E-363
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-78. Summary of Benchmark Dose Modeling Results for Necrosis in the Liver in Fi
Male Sprague-Dawley Rats Following Exposure to PFOA (Pooled) {NTP, 2020, 7330145}
Goodness of Fit Scaled Residual
BMDio BMDLio Basis for Model
° C n-value AIC Dose Group Control (m«/L) (mg/L) Selection
p near BMD Dose Group
Dichotomous
0.213
0.284
0.377
420.9
418.3
417.4
-0.4
-0.5
-0.5
-t m -t
© © ©
35.5
30.1
25.1
4.9
25.2
20.1
EPA selected the
Hill
Gamma
Log-Logistic
Log-Logistic
model. All models,
except Logisitic
and Probit, had
Multistage
Degree 7
Multistage
Degree 6
Multistage
Degree 5
0.284
0.284
0.284
418.3
418.3
418.3
-0.5
-0.5
-0.5
0.3
0.3
0.3
30.1
30.1
30.1
25.2
25.2
25.2
adequate fit (p-
values greater than
0.1). The Log-
Logistic model was
selected based on
the lowest AIC
value for models
Multistage
Degree 4
0.284
418.3
-0.5
0.3
30.1
25.2
with sufficiently
close BMD and
Multistage
Degree 3
Multistage
Degree 2
0.284
0.284
418.3
418.3
-0.5
-0.5
© ©
30.1
30.1
25.2
25.2
BMDL values (less
than threefold
difference).
Multistage
Degree 1
0.284
418.3
-0.5
0.3
30.1
25.2
Weibull
0.284
418.3
-0.5
0.3
30.1
25.2
Logistic
0.011
428.1
2.3
-1.1
59.1
52.3
Log-Probit
0.308
419.0
-0.4
0.4
19.2
3.4
Probit
0.017
426.6
2.3
-0.9
55.5
49.2
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-364
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-18. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Log-Logistic
Model for Necrosis in the Liver in Fi Male Sprague-Dawley Rats Following Exposure to
PFOA (Pooled) {NTP, 2020, 7330145}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.6.3 Hepatocellular Adenomas
Increased incidence of hepatocellular adenomas was observed in Fi 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-79. The Cavg.Pup.totai was selected for this
model because this metric accounts for the accumulation of effects expected to precede the
increased incidence of hepatocellular adenomas.
Table E-79. Dose-Response Modeling Data for Hepatocellular Adenomas in Fi Male
Sprague-Dawley Rats Following Exposure to PFOA {NTP, 2020, 7330145}
Administered Dose
(ppm)a
Internal Dose
(mg/L)
Number per Group
Incidence
0/ 0
0
50
0
300/ 0
0.4
50
0
0/ 20
72.6
50
0
300 / 20
73.6
50
1
0/ 40
113.5
50
7
300 / 40
115.2
50
5
0/ 80
161.7
50
11
300 / 80
161.8
50
10
Notes:
a Doses are presented as perinatal exposure/postnatal exposure.
E-365
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Hepatocellular adenomas were assessed (1) following postweaning exposure, (2) following
perinatal and postweaning exposure, and (3) using a pooled method. The pooled method used the
dose response data associated with both postweaning exposure (1) and perinatal and postweaning
exposure (2).
The BMD modeling results for hepatocellular adenomas following postweaning exposure are
summarized in Table E-80 and Figure E-19. The best fitting model was the Multistage Degree 3
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 3 model
had the lowest AIC. The BMDLio from the selected Multistage Degree 3 model is 95.3 mg/L.
Table E-80. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenomas
in Fi Male Sprague-Dawley Rats Following Postweaning Exposure to PFOA {NTP, 2020,
7330145}
Goodness of Fit Scaled Residual
BMDio BMDLio Basis for Model
Moflpl
p-value AIC D°Se^™"l> nC°"'ro1
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1
0.9
0.8
0.7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Model3
Multistage
Degree 1
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMDio
(mg/L)
0.378
98.0
-0.1
-0.146
107.9
BMDLio
(mg/L)
73.4
Basis for Model
Selection
threefold
difference), and the
Multistage Degree
2 model had the
lowest AIC.
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.
1
0.9
0.8
0.7
cu 0.6
tn
8 0.5
cu
0.4
Estimated Probability
Response at BMD
— — Linear Extrapolation
O Data
BMD
BMDL
Figure E-20. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 2 Model for Hepatocellular Adenomas in Fi Male Sprague-Dawley Rats Following
Perinatal and Postweaning Exposure to PFOA {NTP, 2020, 7330145}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
The BMD modeling results for hepatocellular adenomas using pooled methods are summarized
in Table E-82 and Figure E-21. The best fitting model was the Multistage Degree 6 model based
on adequate p-values (greater than 0.1), and the BMDLs were sufficiently close (less than
threefold difference) among adequately fitted models. Two models (Multistage Degree 6 and 7)
had the same lowest AIC value. The BMDLio from the selected Multistage Degree 6 model is
104.2 mg/L.
E-368
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-82. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenomas
in Fi Male Sprague-Dawley Rats Following Exposure to PFOA (Pooled) {NTP, 2020,
7330145}
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMDio
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Multistage
Degree 7
Multistage
Degree 6
Multistage
Degree 5
Multistage
Degree 4
Multistage
Degree 3
Multistage
Degree 2
Multistage
Degree 1
0.843 191.9
0.834 191.9
0.754 193.9
0.754 193.9
0.843 191.9
0.684 195.7
0.179 202.7
0.3
0.3
0.3
0.3
0.3
0.9
0.6
-0.001
-0.001
-0.001
-0.001
-0.001
-0.001
-0.001
119.9
119.9
119.9
119.9
119.9
112.6
100.6
104.2
104.2
104.2
104.2
104.2
98.4
76.8
EPA selected the
Multistage Degree
6 model. All
models had
adequate fit (p-
values greater than
0.1), and the
BMDLs were
sufficiently close
(less than threefold
difference). The
Multistage Degree
6 had the lowest
AIC value.
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
E.2.6.4 Hepatocellular Adenoma or Carcinoma
Increased incidence of hepatocellular adenoma or carcinoma was observed in Fi 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-83. The Cavg,pup,total was
selected for this model because this metric accounts for the accumulation of effects expected to
precede the increased incidence of hepatocellular adenomas or carcinomas.
Table E-83. Dose-Response Modeling Data for Hepatocellular Adenoma or Carcinoma in
Fi Male Sprague-Dawley Rats Following Exposure to PFOA {NTP, 2020, 7330145}
Administered Dose
(ppm)a
Internal Dose
(mg/L)
Number per Group
Incidence
0/ 0
0
50
0
300/ 0
0.3
50
0
0/ 20
72.6
50
0
300 / 20
73.5
50
1
0/ 40
113.5
50
7
300 / 40
115.1
50
5
0/ 80
161.7
50
11
300 / 80
161.7
50
12
Notes:
a Doses are presented as perinatal exposure/postnatal exposure.
Hepatocellular adenoma or carcinoma was assessed (1) following postweaning exposure, (2)
following perinatal and postweaning exposure, and (3) using a pooled method. The pooled
method used the dose response data associated with both postweaning exposure (1) and perinatal
and postweaning exposure (2). The dose response data (1) following postweaning exposure was
the same between hepatocellular adenoma and hepatocellular adenoma or carcinoma therefore
this modeling information can be found in Table E-80 and Figure E-19.
The BMD modeling results for hepatocellular adenoma or carcinoma following postweaning
exposure to PFOA are summarized in Table E-84 and Figure E-22. The best fitting model was
the Multistage Degree 3 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 3 model had the lowest AIC. The BMDLio from the selected Multistage
Degree 3 model is 95.3 mg/L.
E-370
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-84. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenoma
or Carcinoma in Fi Male Sprague-Dawley Rats Following Postweaning Exposure to PFOA
{NTP, 2020, 7330145}
Goodness of Fit
Scaled Residual
BMDio
BMDLio
Basis for Model
Model3
p-value AIC
Dose Group
Control
(mg/L)
(mg/L)
Selection
near BMD
Dose Group
EPA selected the
Multistage
Degree 3
0.420 99.1
1.2
-0.001
117.1
95.3
Multistage Degree
3 model.
Multistage Degree
2 and 3 had
adequate fit (p-
Multistage
0.397 100.4
0.7
-0.001
108.8
88.7
values greater than
Degree 2
0.1), the BMDLs
were sufficiently
close (less than
threefold
Multistage
Degree 1
difference), and the
0.064 106.5
0.4
-0.001
94.1
65.3
Multistage Degree
3 model had the
lowest AIC.
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.
1
0.9
0.8
0.7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The BMD modeling results for hepatocellular adenoma or carcinoma following perinatal and
postweaning exposure to PFOA are summarized in Table E-85 and Figure E-23. The best fitting
model was the Multistage Degree 2 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 2 model had the lowest AIC. The BMDLio from the selected
Multistage Degree 2 model is 88.7 mg/L.
Table E-85. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenoma
or Carcinoma in Fi Male Sprague-Dawley Rats Following Perinatal and Postweaning
Exposure to PFOA {NTP, 2020, 7330145}
Goodness of Fit Scaled Residual
BMDio BMDLio Basis for Model
Moflpl
D-value AIC Dose Group Contro1 (mg/L) (mg/L) Selection
p near BMD Dose Group
EPA selected the
Multistage Degree
2 model. All
models had
adequate fit (p-
values greater than
0.1), the BMDLs
were sufficiently
close (less than
threefold
difference), and the
Multistage Degree
2 model had the
lowest AIC.
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.
Multistage 0 %1 m 5 Q 1 -0.001 117.5 95.8
Degree 3
Multistage q.752 10o.8 -0.2 -0.009 109.4 88.7
Degree 2
Multistage 0.199 104.8 -0.4 -0.155 94.9 65.9
Degree 1
E-372
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1
0.9
0.8
0.7
Estimated Probability
Response at BMD
— — Linear Extrapolation
O Data
BMD
BMDL
Figure E-23. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 2 Model for Hepatocellular Adenoma or Carcinoma in Fi Male Sprague-Dawley
Rats Following Perinatal and Postweaning Exposure to PFOA {NTP, 2020, 7330145}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
The BMD modeling results for hepatocellular adenoma or carcinoma using pooled methods are
summarized in Table E-86 and Figure E-24. The best fitting model was the Multistage Degree 2
model based on adequate p-values (greater than 0.1), and the BMDLs were sufficiently close
(less than threefold difference) among adequately fitted models. The BMDLio from the selected
Multistage Degree 3 model is 103.7 mg/L.
Table E-86. Summary of Benchmark Dose Modeling Results for Hepatocellular Adenoma
or Carcinoma in Fi Male Sprague-Dawley Rats Following Exposure to PFOA (Pooled)
{NTP, 2020, 7330145}
Model"
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMD io
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Multistage
Degree 7
Multistage
Degree 6
Multistage
Degree 5
Multistage
Degree 4
Multistage
Degree 3
Multistage
Degree 2
0.713
200.6
0.1
-0.001
117.3
103.7
0.713
200.6
0.1
-0.001
117.3
103.7
0.713
200.6
0.1
-0.001
117.3
103.8
0.819
198.6
0.1
-0.001
117.3
103.7
0.893
196.6
0.1
-0.001
117.3
103.7
0.759
199.2
0.7
-0.001
109.3
95.7
EPA selected the
Multistage Degree
3 model. All
models had
adequate fit (p-
values greater than
0.1), and the
BMDLs were
sufficiently close
(less than threefold
difference). The
Multistage Degree
E-373
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMDio
(mg/L)
BMDLio
(mg/L)
Multistage
Degree 1
0.179 207.3
0.5
-0.001
94.5
72.7
Basis for Model
Selection
3 had the lowest
AIC.
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.
1
0.9
0.8
0.7
Estimated Probability
Response at BMD
— — Linear Extrapolation
O Data
BMD
BMDL
Figure E-24. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 2 Model for Hepatocellular Adenoma or Carcinoma in Fi Male Sprague-Dawley
Rats Following Exposure to PFOA (Pooled) {NTP, 2020, 7330145}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.6.5 Pancreatic Acinar Cell Adenoma
Increased incidence of pancreatic acinar cell adenoma was observed in Fi 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-87. The Cavg.pup.totai was selected for
this model because this metric accounts for the accumulation of effects expected to precede the
increased of pancreatic acinar cell adenomas.
Table E-87. Dose-Response Modeling Data for Pancreatic Acinar Cell Adenoma in Fi Male
Sprague-Dawley Rats Following Exposure to PFOA {NTP, 2020, 7330145}
Administered Dose
(ppm)a
0/ 0
Internal Dose
(mg/L)
0
Number per Group
50
Incidence
E-374
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Administered Dose
(ppm)a
Internal Dose
(mg/L)
Number per Group
Incidence
300/ 0
0.4
50
7
0/ 20
72.6
50
28
300 / 20
73.6
50
18
0/ 40
113.5
50
26
300 / 40
115.2
50
30
0/ 80
161.7
50
32
300 / 80
161.8
50
30
Note:
a Doses are presented as perinatal exposure/postnatal exposure.
Pancreatic acinar cell adenoma was assessed (1) following postweaning exposure, (2) following
perinatal and postweaning exposure, and (3) using a pooled method. The pooled method used the
dose response data associated with both postweaning exposure (1) and perinatal and postweaning
exposure (2).
The BMD modeling results for pancreatic acinar cell adenoma following postweaning exposure
to PFOA are summarized in Table E-88 and Figure E-25. The best fitting model was the
Multistage Degree 2 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 2 model had the lowest AIC. The BMDLio from the selected Multistage
Degree 2 model is 33.9 mg/L.
Table E-88. Summary of Benchmark Dose Modeling Results for Pancreatic Acinar Cell
Adenoma in Fi Male Sprague-Dawley Rats Following Postweaning Exposure to PFOA
{NTP, 2020, 7330145}
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMD io
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
EPA selected the
Multistage
0.043
207.9
-1.2
0.2
56.5
31.7
Multistage Degree
Degree 3
2 model.
Multistage Degree
2 model had
Multistage
0.127
206.1
-1.5
0.3
51.8
33.9
adequate fit (p-
Degree 2
values greater than
0.1), the BMDLs
were sufficiently
close (less than
Multistage
Degree 1
threefold
0.005
213.5
-0.4
0.4
22.0
17.6
difference), and the
Multistage Degree
2 model had the
lowest AIC.
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-375
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Estimated Probability
Response at BMD
— — Linear Extrapolation
O Data
BMD
BMDL
Figure E-25. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 2 Model for Pancreatic Acinar Cell Adenoma in Fi Male Sprague-Dawley Rats
Following Postweaning Exposure to PFOA {NTP, 2020, 7330145}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
The BMD modeling results for pancreatic acinar cell adenoma following perinatal and
postweaning exposure are summarized in Table E-89 and Figure E-26. 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 is 15.7 mg/L.
Table E-89. Summary of Benchmark Dose Modeling Results for Pancreatic Acinar Cell
Adenoma in Fi Male Sprague-Dawley Rats Following Perinatal and Postweaning Exposure
to PFOA {NTP, 2020, 7330145}
Goodness of Fit Scaled Residual
BMDio BMDLio Basis for Model
Mode a
p-value AIC Dose Group Contro1 (mg/L) (mg/L) Selection
' near BMD Dose Group
EPA selected the
Multistage Degree
1 model. All
models had
adequate fit (p-
values greater than
0.1), the BMDLs
were sufficiently
close (less than
E-376
Multistage 0.178 248.3 0.1 0.1 20.6 15.7
Degree 3
Multistage 0.178 248.3 0.1 0.1 20.7 15.7
Degree 2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Model3
Multistage
Degree 1
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMDio
(mg/L)
0.404 246.3
0.1
0.1
20.2
BMDLio
(mg/L)
15.7
Basis for Model
Selection
threefold
difference), and the
Multistage Degree
1 model had the
lowest AIC.
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.
Figure E-26. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Multistage
Degree 1 Model Pancreatic Acinar Cell Adenoma in Fi Male Sprague-Dawley Rats
Following Perinatal and Postweaning Exposure to PFOA {NTP, 2020, 7330145}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
The BMD modeling results for pancreatic acinar cell adenoma using the pooled method are
summarized in Table E-90. No models provided an adequate fit, therefore a LOAEL approach
was taken for this endpoint.
E-377
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-90. Summary of Benchmark Dose Modeling Results for Pancreatic Acinar Cell
Adenoma in Fi Male Sprague-Dawley Rats Following Exposure to PFOA (Pooled) {NTP,
2020, 7330145}
Model
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMDio
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Multistage
Degree 7
Multistage
Degree 6
0.021
0.021
458.3
458.3
-2.4
-2.4
1 1
© ©
41.9
41.9
21.9
21.9
No models had
adequate fit (p-
values were less
than 0.1)
Multistage
Degree 5
0.021
458.3
-2.4
-0.9
41.9
21.9
Multistage
Degree 4
0.021
458.3
-2.4
-0.9
41.9
21.9
Multistage
Degree 3
0.021
458.3
-2.4
-0.9
41.9
21.9
Multistage
Degree 2
0.021
458.3
-2.4
-0.9
41.9
22.5
Multistage
Degree 1
0.012
460.1
1.3
-0.7
20.9
17.6
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.2.7 Song, 2018, 5079725
EPA conducted dose response modeling of the Song et al. (2018, 5079725) study using BMDS
3.2 program. This study addresses the offspring survival in Fi male and female Kunming mice.
E. 2.7.1 Offspring Survival
Decreased mean response of number of offspring survival was observed in Fi male and female
Kunming mice. Continuous models were used to fit dose-response data. BMR of a change in the
mean equal to 0.1 and 0.5 standard deviations from the control mean were chosen. The doses and
response data used for the modeling are listed in Table E-91. The Cavg,puP,gest, Cavg,puP,iact,
Cavg,pup,gest,iact, Cmax,pup,gest, and Cmax,pup,iact were considered and shown below because prenatal loss
could be a result of exposure during a sensitive window of development where a Cmax metric is
expected to better correlate with the effect or an accumulation of exposure and an average
concentration metric is expected to better correlate with the effect and this could occur during the
gestation or lactation lifestages. The Cavg,pup,gest,iact was selected for this model.
Table E-91. Dose-Response Modeling Data for Offspring Survival in Fi Male and Female
Kunming Mice Following Exposure to PFOA {Song, 2018, 5079725}
Administer
ed Dose
(mg/kg/day
)
Internal Dose
Cavg,pup,gest
(mg/L)
Cavg,pup,lact
(mg/L)
Cavg,pup,gest,la
(mg/L)
Cmax, pup, gest
(mg/L)
C max,pup,lact
(mg/L)
Number
per group
Mean
Response
(incidences
)a
E-378
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
0
0
0
0
0
0
10
15.1 ± 7.6b
1
8.5
21.1
15.4
15.5
27.9
10
13.0 ± 14.5
2.5
17.0
31.9
25.3
27.0
42.5
10
12.0 ± 10.1
5
22.9
35.1
29.6
34.0
46.8
10
6.4 ± 17.1
Notes:
a Data are presented as mean ± standard deviation.
b Standard deviations were calculated from standard errors.
For Cavg,pup,gest, the BMD modeling results for offspring survival are summarized in Table E-92
and Figure E-27. The best fitting model was the Polynomial Degree 2 model based on adequate
p-values (greater than 0.1), and the BMDLs were sufficiently close (less than threefold
difference) among adequately fitted models. The BMDLo.ssd from the selected Polynomial
Degree 2 model is 8.8 mg/L.
E-379
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-92. Summary of Benchmark Dose Modeling Results for Offspring Survival using Cavg,PuP,gest in Fi Male and Female
Kunming Mice Following Exposure to PFOA (constant variance) {Song, 2018, 5079725}
Model3 Goodness of Fit Scaled Residual
Dose Group
Dose Group
near BMDossd
Control
BMDo.isd
BMDLo.isd
BMDo.ssd
BMDLo.ssd
Basis for Model
P-
value
AIC
near
BMDo.isd
Dose
Group
(mg/L)
(mg/L)
(mg/L)
(mg/L)
Selection
Exponential
2
0.736
320.3
-0.15
0.53
-0.15
3.0
1.1
18.5
6.1
EPA selected the
Polynomial
Exponential
3
0.709
321.8
0.03
-0.01
0.25
15.0
1.1
21.5
6.6
Degree 2 model.
All models had
Exponential
4
0.736
320.3
-0.15
0.53
-0.15
3.0
1.1
18.5
6.1
adequate fit (p-
values greater
Exponential
5
0.709
321.8
0.03
-0.01
0.25
15.0
1.1
21.5
6.6
than 0.1), and the
BMDLs were
Hill
0.701
321.8
-1.2xe-4
-1.2xe-4
0.27
16.4
9.3
19.4
_b
sufficiently close
(less than
Polynomial
Degree 3
0.711
321.8
-0.25
-0.09
0.10
10.7
1.8
20.7
8.9
threefold
difference).
Polynomial
Degree 2
0.898
319.9
-0.23
-0.17
0.03
9.0
1.8
20.1
8.8
Power
0.718
321.8
0.06
-0.01
0.23
14.2
1.8
21.5
8.9
Linear
0.791
320.2
-0.16
0.53
-0.16
3.6
1.7
18.2
8.6
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.isd = dose level corresponding to a change in the mean equal to
0.1 standard deviations from the control mean; BMDLo.isd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.1
standard deviations from the control mean; BMDo.5sd = dose level corresponding to a change in the mean equal to 0.5 standard deviation from the control mean.
a Selected model in bold.
b Lower limit includes zero; BMDL not estimated.
E-380
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
£
O
Q.
LO
CD
DC
Estimated Probability
^^Response at BMD
O Data
BMD
BMDL
-10
Dose
Figure E-27. Plot of Mean Response by Dose with Fitted Curve for the Selected
Exponential 2 Model for Offspring Survival using Cavg,PuP,gest in Fi Male and Female CD-I
Mice Following Exposure to PFOA {Song, 2018, 5079725}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
For Cavg.pup.iact, the BMD modeling results for offspring survival are summarized in Table E-93
and Figure E-28. The best fitting model was the Polynomial Degree 3 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 Polynomial Degree 3 model had the lowest AIC. The
BMDLo 5sd from the selected Polynomial Degree 3 model is 15.2 mg/L.
E-381
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-93. Summary of Benchmark Dose Modeling Results for Offspring Survival using Cavg,PuP,iact in Fi Male and Female
Kunming Mice Following Exposure to PFOA (constant variance) {Song, 2018, 5079725}
Goodness of Fit
Scaled Residual
Model"
p-value
AIC
Dose Group
near
BMDo.isd
Dose Group
near BMDossd
Control
Dose
Group
BMDo.isd
(mg/L)
BMDLo.isd
(mg/L)
BMDo.ssd
(mg/L)
BMDLo.ss
(mg/L)
Exponential
2
0.586
320.8
-0.139
-0.782
-0.139
5.7
1.9
35.1
11.1
Exponential
3
0.701
321.8
0.001
-4.583 x e~4
0.271
30.9
2.3
34.4
13.1
Exponential
4
0.586
320.8
-0.139
-0.782
-0.139
5.7
1.9
35.1
11.1
Exponential
5
0.701
321.8
0.001
-4.615 x e~4
0.271
30.9
2.3
34.4
13.1
Hill
_b
323.8
0.004
-0.001
0.269
30.7
C
34.5
C
Polynomial
Degree 3
Polynomial
Degree 2
0.768
320.2
-0.158
0.581
-0.020
19.5
3.0
33.3
15.2
0.721
320.3
0.011
0.612
-0.112
14.6
3.0
32.6
15.0
Power
0.702
321.8
0.003
-4.564 x e~4
0.270
30.7
3.2
34.5
16.0
Basis for
Model
Selection
Linear
0.617
320.7
-0.174
0.574
-0.174
6.6
2.9
32.8
14.5
EPA selected
the
Polynomial
Degree 3
model. All
models,
except
Exponential 5
and Hill, had
adequate fit
(p-values
greater than
0.1), the
BMDLs were
sufficiently
close (less
than threefold
difference),
and the
Polynomial
Degree 3
model had the
lowest AIC.
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.isd = dose level corresponding to a change in the mean equal to
0.1 standard deviations from the control mean; BMDLo.isd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.1
standard deviations from the control mean; BMDo.5sd = dose level corresponding to a change in the mean equal to 0.5 standard deviation from the control mean.
a Selected model in bold.
b Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
c Lower limit includes zero; BMDL not estimated.
E-382
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-28. Plot of Mean Response by Dose with Fitted Curve for the Selected Polynomial
Degree 3 Model for Offspring Survival using Cavg,PuP,iact in Fi Male and Female CD-I Mice
Following Exposure to PFOA {Song, 2018, 5079725}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
For Cavg.pup.gest.iact, the BMD modeling results for offspring survival are summarized in Table
E-94 and Figure E-29. The best fitting model was the Polynomial Degree 3 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 Polynomial Degree 3 model had the lowest
AIC. The BMDLo 5sd from the selected Polynomial Degree 3 model is 12.3 mg/L.
E-383
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-94. Summary of Benchmark Dose Modeling Results for Offspring Survival using Cavg,pup,gest,iact in Fi Male and Female
Kunming Mice Following Exposure to PFOA (constant variance) {Song, 2018, 5079725}
Model"
Goodness of Fit
p-value AIC
Scaled Residual
Dose
Group
near
BMDo.isd
Dose Group
near
BMDo.ssd
Control
Dose
Group
BMDo.isd BMDLo.isd BMDo.ssd BMDLo.ssd
(mg/L) (mg/L) (mg/L) (mg/L)
Basis for
Model
Selection
Exponential
2
Exponential
3
Exponential
4
Exponential
5
Hill
Polynomial
Degree 3
Polynomial
Degree 2
Power
Linear
0.637
320.6
-0.16
0.54
-0.16
4.5
0.703
321.8
0.01
-2.96 x e~3
0.27
23.8
0.637
320.6
-0.16
0.54
-0.16
4.5
0.703
321.8
0.01
-2.96 x e~3
0.27
23.8
0.701
321.8
5.88 x e~5
2.09 x e~6
0.27
24.4
0.852
320.0
-0.23
-0.25
0.03
16.1
0.801
320.1
-0.09
0.54
-0.08
11.9
0.706
321.8
0.02
-4.97 x e~3
0.26
23.4
0.679
320.5
-0.19
0.57
-0.19
5.1
1.5
1.7
1.5
1.7
_b
2.5
2.4
2.5
2.3
27.2
28.7
27.2
28.7
28.1
27.5
26.7
28.8
25.7
10.1
8.8
10.1
_b
12.3
12.2
12.6
11.7
EPA selected
the Polynomial
Degree 3 model.
All models had
adequate fit (p-
values greater
than 0.1), the
BMDLs were
sufficiently
close (less than
threefold
difference), and
the Polynomial
Degree 3 model
had the lowest
AIC.
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.isd = dose level corresponding to a change in the mean equal to
0.1 standard deviations from the control mean; BMDLo.isd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.1
standard deviations from the control mean; BMDo.ssd = dose level corresponding to a change in the mean equal to 0.5 standard deviation from the control mean.
a Selected model in bold.
b Lower limit includes zero; BMDL not estimated.
E-384
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-29. Plot of Mean Response by Dose with Fitted Curve for the Selected Polynomial
Degree 3 Model for Offspring Survival using Cavg,pup,gest,iact in Fi Male and Female CD-I
Mice Following Exposure to PFOA {Song, 2018, 5079725}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
For Cmax.pup.gest, the benchmark dose (BMD) modeling results for offspring survival are
summarized in Table E-95 and Figure E-30. The best fitting model was the Polynomial Degree 2
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 Polynomial Degree 2 model
had the lowest AIC. The BMDLo 5sd from the selected Polynomial Degree 2 model is 13.4 mg/L.
E-385
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-95. Summary of Benchmark Dose Modeling Results for Offspring Survival using Cmax,pup,gest, in Fi Male and Female
Kunming Mice Following Exposure to PFOA (constant variance) {Song, 2018, 5079725}
Goodness of Fit
Scaled Residual
Model"
P" AIC
value
Dose
Group near
BMDo.isd
Dose
Group
near
BMDo.ssd
Control
Dose
Group
BMDo.isd
(mg/L)
BMDLo.isd
(mg/L)
BMDo.ssd
(mg/L)
BMDLo.ssd
(mg/L)
Basis for
Model
Selection
Exponential
2
0.686
320.4
-0.167
0.529
-0.167
4.8
1.7
29.1
9.5
EPA selected
the
Exponential
3
0.708
321.8
0.031
-0.009
0.252
24.4
1.8
32.3
10.6
Polynomial
Degree 2
model. All
models, except
Exponential
4
0.686
320.4
-0.167
0.529
-0.167
4.8
1.7
29.1
9.5
Exponential
5
0.701
323.8
0.029
-0.008
0.253
24.5
1.8
32.3
10.6
for the Hill
model, had
Hill
_b
323.8
8.798 x e~5
2.097 x e~4
0.272
26.0
16.5
30.6
C
adequate fit
(p-values
Polynomial
Degree 3
Polynomial
Degree 2
Power
0.667
0.867
0.717
321.9
320.0
321.8
-0.253
-0.157
0.058
-0.132
0.445
-0.013
0.073
-0.040
0.229
17.8
13.4
23.5
2.7
2.7
2.7
31.1
29.9
32.4
13.6
13.4
13.7
greater than
0.1), the
BMDLs were
sufficiently
close (less
than threefold
Linear
0.738 320.3
-0.193
0.543
-0.193
5.6
2.6
27.8
13.0
difference),
and the
Polynomial
Degree 2
model had the
lowest AIC.
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.isd = dose level corresponding to a change in the mean equal to
0.1 standard deviations from the control mean; BMDLo.isd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.1
standard deviations from the control mean; BMDo.ssd = dose level corresponding to a change in the mean equal to 0.5 standard deviation from the control mean.
a Selected model in bold.
b Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
c Lower limit includes zero; BMDL not estimated.
E-386
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure E-30. Plot of Mean Response by Dose with Fitted Curve for the Selected Polynomial
Degree 2 Model for Offspring Survival using Cmax,pup,gest in Fi Male and Female CD-I Mice
Following Exposure to PFOA {Song, 2018, 5079725}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
For Cmax.pup.iact, the benchmark dose (BMD) modeling results for offspring survival are
summarized in Table E-96 and Figure E-31. The best fitting model was the Polynomial Degree 3
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 Polynomial Degree 3 model
had the lowest AIC. The BMDLo 5sd from the selected Polynomial Degree 3 model is 20.3 mg/L.
E-387
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-96. Summary of Benchmark Dose Modeling Results for Offspring Survival using Cmax,pup,iact, in Fi Male and Female
Kunming Mice Following Exposure to PFOA (constant variance) {Song, 2018, 5079725}
Goodness of Fit
Scaled Residual
Model"
p-value
AIC
Dose
Group
near
BMDo.isd
Dose Group
near BMDo.ssd
Control
Dose
Group
BMDo.isd
(mg/L)
BMDLo.isd
(mg/L)
BMDo.ssd
(mg/L)
BMDLo.ssd
(mg/L)
Basis for
Model
Selection
Exponential
2
0.589
320.8
-0.140
-0.778
-0.140
7.6
2.6
46.6
14.8
EPA selected
the Polynomial
Exponential
3
0.701
321.8
0.002
-6.406 x e
0.270
41.1
3.0
45.9
17.5
Degree 3
model. All
Exponential
4
0.589
320.8
-0.140
-0.778
-0.140
7.6
2.6
46.6
14.8
models, except
the Hill model,
Exponential
5
_b
323.8
0.002
-4.655 x c ¦'
0.271
41.1
C
45.9
2.6
had adequate
fit (p-values
Hill
_b
323.8
0.005
-0.001
0.269
40.9
C
46.1
C
greater than
0.1), the
Polynomial
Degree 3
Polynomial
Degree 2
Power
0.772
0.725
0.702
320.2
320.3
321.8
-0.163
0.005
0.004
0.575
0.610
-8.674 x c ¦'
-0.017
-0.111
0.269
25.9
19.4
40.9
4.1
4.0
4.3
44.4
43.4
46.1
20.3
20.0
21.3
BMDLs were
sufficiently
close (less than
threefold
difference),
and the
Polynomial
Linear
0.620
320.6
-0.175
0.574
-0.175
8.7
3.9
43.5
19.4
Degree 3
model had the
lowest AIC.
Notes: AIC = Akaike information criterion; BMD = benchmark dose; BMDL = benchmark dose lower limit; BMDo.isd = dose level corresponding to a change in the mean equal to
0.1 standard deviations from the control mean; BMDLo.isd = lower bound on the dose level corresponding to the 95% lower confidence limit for a change in the mean equal to 0.1
standard deviations from the control mean; BMDo.ssd = dose level corresponding to a change in the mean equal to 0.5 standard deviation from the control mean.
a Selected model in bold.
b Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
c Lower limit includes zero; BMDL not estimated.
E-388
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
25
20
15 (*
K 10
£
O
Q.
LO
o 5
DC 3
-5
10
20
30
C)
40
Estimated Probability
^^Response at BMD
O Data
BMD
BMDL
-10
Dose
Figure E-31. Plot of Mean Response by Dose with Fitted Curve for the Selected Polynomial
Degree 3 Model for Offspring Survival using Cmax,pup,iact in Fi Male and Female CD-I Mice
Following Exposure to PFOA {Song, 2018, 5079725}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E.2.8 Wolf, 2007, 1332672
EPA conducted dose response modeling of the Wolf et al. (2007, 1332672) study using the
BMDS 3.2 program. This study addresses pup body weight change in Fi male and female CD-I
mice (in atero exposure), time to eye opening in Fi male and female CD-I mice (in atero and
lactational exposure), and dams with whole litter loss (%) in Po female CD-I mice.
E.2.8.1 Pup Body Weight Change
Decreased mean response of pup body weight change was observed in Fi male and female CD-I
mice (in atero exposure). Continuous models were used to fit dose-response data. A benchmark
response (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-97. The
Cavg.pup.gest dose metric was selected for this model 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 change.
Table E-97. Dose-Response Modeling Data for Pup Body Weight Change in Fi Male and
Female CD-I Mice Following Exposure to PFOA {Wolf, 2007,1332672}
Administered Dose
(mg/kg/day)
Internal Dose
(mg/L)
Number per Group
Mean Response (g)a
0
0
14
12.4 ± 1.2b
3
18.7
11
11.4 ± 1.3
5
22.9
13
9.6 ± 1.3
E-389
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Notes:
a Data are presented as mean ± standard deviation.
b Standard deviations were calculated from standard errors.
The BMD modeling results for pup body weight are summarized in Table E-98. No models
provided an adequate fit, therefore a LOAEL approach was taken for this endpoint.
Table E-98. Summary of Benchmark Dose Modeling Results for Pup Body Weight Change
in Fi Male and Female CD-I Mice Following Exposure to PFOA (constant variance) {Wolf,
2007,1332672}
Goodness of Fit Scaled Residual
BMDo.s sd BMDLo.s sd Basis for Model
1VIOUCI
p-value
AIC
Dose Group
near BMD
Control Dose
Group
(mg/L)
(mg/L)
Selection
Exponential 2
0.006
136.0
-0.3
-0.
.3
6.2
4.3
No models had
Exponential 3
a
130.5
1.9 x e~4
-8.5 :
x e~5
17.0
12.5
adequate fit (p-
Exponential 4
0.006
136.0
-0.3
-0.
.3
6.2
4.3
values greater
than 0.01).
Exponential 5
a
132.5
-8.9 x e~5
4.7 x
e~3
17.1
12.5
Hill
a
130.5
4.7 x e~5
-2.0 :
x e~4
18.0
12.2
Polynomial
0.055
132.2
1.5
-0.
.5
11.3
6.8
Degree 2
Power
a
130.5
-2.1 x e~7
1.3 x
e-7
17.0
12.2
Linear
0.008
135.6
-0.3
-0.
.3
6.5
4.7
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 Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
E.2.8.2 Time to Eye Opening
Increased mean response of time to eye opening was observed in Fi male and female CD-I mice
(in utero and lactational exposure). Continuous models were used to fit dose-response data.
BMR of a change in the mean equal to 1 standard deviations from the control mean was chosen.
The doses and response data used for the modeling are listed in Table E-99. The dose metrics,
Cavg,pup,gest, Cavg,pup,iact, Cavg,pup,gest,iact, Cmax,pup,gest, and Cmax,pup,iact were all considered and shown
below because time to eye opening could be a result of exposure during a sensitive window of
development where a Cmax metric is expected to better correlate with the effect or an
accumulation of exposure where an average concentration metric is expected to better correlate
with the effect and time to eye opening could be due to exposure during the gestation or lactation
lifestages.
E-390
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-99. Dose-Response Modeling Data for Time to Eye Opening Fi Male and Female
CD-I Mice Following Exposure to PFOA {Wolf, 2007,1332672}
Internal Dose
Administered Number Mean
Dose Cavg, Cavg, Cavg, per Response
l--max,pup,gest x^max,pup,lact
(mg/kg/day) pup,gest pup,iact pup,gest,iact Group (days)3
(mg/L) (mg/L) (mg/L)
0
0
0
0
0
0
14
14.8 ± 0.3b
3
18.7
32.9
26.6
28.7
43.9
12
15.8 ±0.7
5
22.6
35.1
29.6
34.0
46.8
12
15.9 ± 1.4
Notes:
a Data are presented as mean ± standard deviation.
b Standard deviations were calculated from standard errors.
The BMD modeling results for time to eye opening are summarized in Table E-100 for
Cavg,pup,gest, Table E-101 for Cavg,pup,lact, Table E-102 for Cavg,pup,gest,iact, Table E-103 for
Cmax,puP,gest, and Table E-104 for Cmax,puP,iact. No models provided an adequate fit. A LOAEL
approach was taken for this endpoint.
Table E-100. Summary of Benchmark Dose Modeling Results for Time to Eye Opening for
Cavg,pup,gest in Fi Male and Female CD-I Mice Following Exposure to PFOA (constant
variance) {Wolf, 2007, 1332672}
Goodness of Fit Scaled Residual
BMDisd BMDLisd Basis for Model
° C n-value AIC Dose Group Control (m«/L) (mg/L) Selection
p near BMD Dose Group
Exponential
0.740
102.5
0.26
-0.05
17.4
11.9
No models had
2
adequate fit for
Exponential
3
0.740
102.5
0.26
-0.05
17.4
11.9
constant and
nonconstant
Exponential
A
a
104.5
0.22
-0.03
17.2
1.0
variance (p-values
were less than
Exponential
5
a
104.5
0.22
-0.04
17.2
1.0
0.05).
Hill
_b
106.4
1.3 x e~3
-1.4 x e~4
16.4
C
Polynomial
0.755
102.5
0.24
-0.04
17.3
11.6
Degree 2
Power
0.755
102.5
0.24
-0.04
17.3
11.6
Linear
0.755
102.5
0.24
-0.04
17.3
11.6
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 Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
b Degrees of freedom are negative, (Goodness of fit test cannot be calculated).
c Lower limit includes zero; BMDL not estimated.
E-391
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-101. Summary of Benchmark Dose Modeling Results for Time to Eye Opening for
Cavg,pup,iact in Fi Male and Female CD-I Mice Following Exposure to PFOA (constant
variance) {Wolf, 2007, 1332672}
Goodness of Fit Scaled Residual
BMDisd BMDLisd Basis for Model
° C n-value AIC Dose Group Control (mg/L) (mg/L) Selection
^ near BMD Dose Group
Exponential
2
0.929
102.4
-0.1
2.8 >
< e~3
28.0
19.1
No models had
adequate fit for
Exponential
3
a
104.4
-0.1
3.9 >
< e~3
28.0
19.1
constant and
nonconstant
Exponential
A
a
104.4
-0.1
3.9 >
< e~3
27.8
18.7
variance (p-values
were less than
4
Exponential
5
_b
106.4
-0.1
4.5 >
< e~3
27.8
18.7
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 Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
b Degrees of freedom are negative, (Goodness of fit test cannot be calculated).
c Lower limit includes zero; BMDL not estimated.
Table E-102. Summary of Benchmark Dose Modeling Results for Time to Eye Opening for
Cavg,pup,gest,iact in Fi Male and Female CD-I Mice Following Exposure to PFOA (constant
variance) {Wolf, 2007, 1332672}
Goodness of Fit Scaled Residual
BMDisd BMDLisd Basis for Model
° C n-value AIC Dose Group Control (mg/L) (mg/L) Selection
^ near BMD Dose Group
Exponential
2
0.959
102.4
0.04
-4.3
x e~3
23.2
15.8
No models had
adequate fit for
Exponential
3
0.959
102.4
0.04
-4.1
x e~3
23.2
15.8
constant and
nonconstant
Exponential
A
a
104.4
0.02
-2.2 :
x e~3
22.9
1.3
variance (p-values
were less than
4
Exponential
5
a
104.4
0.02
-1.9
x e~3
22.9
1.3
0.05).
Hill
_b
106.4
-5.3 xe~3
5.2 x
e~4
25.2
C
Polynomial
Degree 2
0.968
102.4
0.08
-3.0 :
< e~3
23.1
15.5
Power
0.968
102.4
0.08
-3.2 :
< e~3
23.1
15.5
Linear
0.968
102.4
0.08
-3.2 :
x e~3
23.1
15.5
E-392
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
b Degrees of freedom are negative, (Goodness of fit test cannot be calculated)..
c Lower limit includes zero; BMDL not estimated.
Table E-103. Summary of Benchmark Dose Modeling Results for Time to Eye Opening for
Cmax,pup,gest in Fi Male and Female CD-I Mice Following Exposure to PFOA (constant
variance) {Wolf, 2007, 1332672}
Model
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMDisd
(mg/L)
BMDLisd
(mg/L)
Basis for Model
Selection
Exponential
2
0.818
102.5
0.17
-0.03
26.1
17.8
No models had
adequate fit for
Exponential
3
0.818
102.5
0.17
-0.03
26.1
17.8
constant and
nonconstant
Exponential
A
a
104.4
7.5 xe-5
-2.2 x e-5
22.7
1.5
variance (p-values
were less than
4
Exponential
5
a
104.4
-5.9xe~6
-1.5 x e-5
22.7
1.5
0.05).
Hill
_b
106.4
6.1xe-4
-4.3 x e~5
26.0
C
Polynomial
Degree 2
0.831
102.4
0.16
-0.02
25.9
17.4
Power
0.831
102.4
0.16
-0.02
25.9
17.4
Linear
0.831
102.4
0.16
-0.02
25.9
17.4
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 Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
b Degrees of freedom are negative, (Goodness of fit test cannot be calculated).
c Lower limit includes zero; BMDL not estimated.
Table E-104. Summary of Benchmark Dose Modeling Results for Time to Eye Opening
(Cmax,pup, lact) in Fi Male and Female CD-I Mice Following Exposure to PFOA (constant
variance) {Wolf, 2007, 1332672}
Model3
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMDisd
(mg/L)
BMDLisd
(mg/L)
Basis for Model
Selection
Exponential 0.934 102.4
2
Exponential -a 104.4
3
Exponential 0.929 102.4
4
Exponential
5
283.1
-0.06
-0.06
-0.07
0.17
3.8 x e~3
4.0 x e~3
3.9 x e~3
6.1
37.3
37.3
37.1
25.1
25.4
25.4
_b
20.6
No models had
adequate fit for
constant and
nonconstant
variance (p-values
were less than
0.05).
E-393
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Model3
Hill
Polynomial
Degree 2
Power
Linear
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group Control
near BMD Dose Group
BMDisd
(mg/L)
BMDLisd
(mg/L)
Basis for Model
Selection
0.929
106.4
104.4
104.4
102.4
7.3 >
1.3 >
-9.1
-0.07
-4.2
1.0 >
7.5
3.9
41.9
39.5
39.3
37.1
24.9
24.9
24.9
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 Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
b Lower limit includes zero; BMDL not estimated.
c Degrees of freedom are negative, (Goodness of fit test cannot be calculated).
E.2.8.3 Dams with Whole Litter Loss
Increased incidence of dams with whole litter loss was observed in Po female CD-I Mice.
Dichotomous models were used to fit dose-response data. BMR of 5% and 10% extra risk were
chosen. The doses and response data used for the modeling are listed in Table E-105. The
Cavg,dam,gest dose metric was selected for this model to consider an accumulation of exposure
where an average concentration metric is expected to better correlate with the effect.
Table E-105. Dose-Response Modeling Data for Dams with Whole Litter Loss in Po Female
CD-I Mice Following Exposure to PFOA {Wolf, 2007,1332672}
Administered Dose Internal Dose
Number per Group Incidence
(mg/kg/day) (mg/L)
0 0 39 1
3 74.9 25 1
5 91.6 30 5
The BMD modeling results for dams with whole litter loss are summarized in Table E-106 and
Figure E-32. The best fitting model was the Gamma 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 lower bound on the dose level
corresponding to the 95% lower confidence limit for a 5% response level (BMDLs) from the
selected Gamma model is 29.2 mg/L.
E-394
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-106. Summary of Benchmark Dose Modeling Results for Dams with Whole Litter Loss in Po Female CD-I Mice
Following Exposure to PFOA {Wolf, 2007,1332672}
Model"
Goodness of Fit
p-value AIC
Scaled Residual
Dose Group
near BMDs
Dose Group Control
near BMDio Dose Group
BMDs
(mg/L)
BMDLs
(mg/L)
BMDio
(mg/L)
BMDLio
(mg/L)
Basis for Model
Selection
Dichotomou
s Hill
Gamma
Log-
Logistic
Multistage
Degree 2
Multistage
Degree 1
Weibull
Logistic
Log-Probit
Probit
65535
52.7
1.6 x e~6
-1.4 x e~5
2.2 x e~5
81.1
C
86.4
C
0.712
48.9
-0.3
0.1
0.1
79.1
29.2
87.2
59.9
_b
50.7
5.8 x e~4
1.3 x e~4
2.9 x e~4
83.1
29.1
88.5
61.4
0.256
0.189
0.258
b
50.2
-0.9
0.7
0.1
62.0
23.5
0.241 50.3
-0.9
0.7
0.2
57.2
38.7
87.3
48.3
50.1
-1.0
0.8
0.1
46.4
22.2
95.3
45.7
50.7
-4.5 x e-5
-4.5 x e-5
1.6 x e~4
83.3
30.2
88.6
62.0
50.2
-0.9
0.6
0.3
59.4
41.5
86.3
63.9
50.7
-8.3 x e~4
-2.4 x e~4
4.3 x e~4
82.4
C
88.0
C
61.3
EPA selected the
Gamma model.
All models,
except the
Dichotomous
Hill, Log-
Logistic,
Weibull, and
Log-Probit had
adequate fit (p-
values greater
than 0.1), the
BMDLs were
sufficiently close
(less than
threefold
difference), and
the Gamma
model had the
lowest AIC.
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.
b Degrees of freedom = 0, saturated model (Goodness of fit test cannot be calculated).
c Lower limit includes zero; BMDL not estimated.
E-395
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1
0.9
0.8
0.7
0.6
OJ
l/l
2 0-5
0.4
0.3
0.2
0.1
0
q:
Frequentist Gamma Model with BMR of 10% Extra Risk for the BMD and 0.95 Lower
Confidence Limit for the BMDL
-^Estimated Probability
Response at BMD
C Data
BMD
BMDL
10
20
30
40
50
60
70
80
90
Dose
Figure E-32. Plot of Incidence Rate by Dose with Fitted Curve for the Selected Gamma
Model for Dams with Whole Litter Loss in Po Female CD-I Mice Following Exposure to
PFOA (constant variance) {Wolf, 2007,1332672}
BMD = benchmark dose; BMDL = benchmark dose lower limit.
E-l
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix F. Pharmacokinetic Modeling
F.l Animal Pharmacokinetic Model
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 main text (see Main PFOA Document).
F.l.lComporison 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).
k
~ 1 mg/kg, oral
|
O 10 mg/kg, oral
1(P ¦
~ A
A 60 mg/kg, oral
1Q1 '
=hp
\ °
. 0 O o
Iff3 '
0 20 40 60 00
Time (days)
Figure F-l. Experimentally Observed Serum Concentrations {Lou, 2009, 2919359} and
Median Predictions for a Single Oral Dose of 1,10, or 60 mg/kg PFOA to Female CD1 Mice
1 mg/kg oral dose represented by the squares and solid line; 10 mg/kg oral dose represented by the circles and dashed line;
60 mg/kg oral dose represented by the upward triangle and dotted line.
F-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
102
HT
1 101
u
8 io3
E
ill
1/1
< 10"1
o
u_
Q_
10"2
Figure F-2. Experimentally Observed Serum Concentrations {Kemper, 2003, 6302380} and
Median Prediction for a Single IV Dose of 1 mg/kg or an Oral Dose of 0.1,1, 5, or 25 mg/kg
PFOA to Male Sprague-Dawley Rats
1 mg/kg intravenous (IV) dose represented by the downward triangles and dashed line; 0.1 mg/kg oral dose represented by the
squares and solid line; 1 mg/kg oral dose represented by the circle and dashed line; 5 mg/kg oral dose represented by the upward
triangles and dotted line; 25 mg/kg oral dose represented by the diamonds and dash-dot line.
<>
.8 g-SSjoo.
Q O,-©-" WTi
"~ u ~~ i_r
~
o
V
A
0.1 mg/kg, oral
1 mg/kg, oral
1 mg/kg, IV
5 mg/kg, oral
O 25 rng/kg, oral
10"
10-
10°
Time (days)
101
Time (days)
Figure F-3. Experimentally Observed Serum Concentrations {Kemper, 2003, 6302380}
and Median Prediction for a Single IV Dose of 1 mg/kg or a Single Oral Dose of 0.1,1, 5, or
15 mg/kg PFOA to Female Sprague-Dawley Rats3
1 mg/kg intravenous (IV) dose represented by the downward triangles and dashed line; 0.1 mg/kg oral dose represented by the
squares and solid line; 1 mg/kg oral dose represented by the circle and dashed line; 5 mg/kg oral dose represented by the upward
triangles and dotted line; 15 mg/kg oral dose represented by the diamonds and dash-dot line.
a Change in slope from 1-10 days represents a transition to a "beta-phase" elimination in female rats.
F-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Literature reported concentration (mg/l)
Figure F-4. Model prediction summary for PFOA training data
Model predictions on the training data result in a mean squared log error (MSLE) of 0.395. Dashed lines represent +/- one-half
logio.
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.
F-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
| KI2
S
C M2
?
Cavg_pup_eesl
¦¦
Cavg_pupjact
Cavg_jxip_ges(_lac<
Cavg_pup_diet
Cavg_pup_total
Cavg
r_f_m
-0.6 -0.4 -0.2 0.0 0t2
value
-1.5 -1.0 -05 00 05 10
value
Adult model Developmental model
Figure F-5. PFOA Sensitivity Coefficients of the Adult Model and Developmental Model
As demonstrated in Figure F-5, the renal resorption mechanism (Tmc and KT) and the volume of
distribution (Vcc) represent the most sensitive pathways for concentrations in the adult animal.
This is to be expected because the renal resorption parameters govern the effective half-life of
PFOA in the adult. 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), PFOA 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 PFOA in the diet during growth is much larger than
the steady state concentration during the 21 days of lactational exposure.
F.1.2 Visual Inspection of Test Data sets 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.
F-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
3—n .-«i n E n n V
u
n 6 mg/kg, IV
O 6 mg/kg, oral
& 12 mg/kg, oral
O 45 mg/kg, oral
^ ¦-^ X O
10"2 10-:1 10° 101 10*
Time (days)
Figure F-6. Experimentally Observed Serum Concentrations {Dzierlenga, 2020, 5916078}
and Median Predictions for a Single IV Dose of 6 mg/kg or a Single Oral Dose of 6,12, or
45 mg/kg PFOA to Male Sprague-Dawley Rats
6 mg/kg intravenous (IV) dose represented by the squares and solid line; 6 mg/kg oral dose represented by the circles and solid
line; 12 mg/kg oral dose represented by the upward triangles and dashed line; 45 mg/kg oral dose represented by the diamonds
and dash-dot line.
Time (days)
Figure F-7. Experimentally Observed Serum Concentrations {Dzierlenga, 2020, 5916078}
and Median Predictions for a Single IV Dose of 40 mg/kg or a Single Oral Dose of 40, 80, or
320 mg/kg PFOA to Female Sprague-Dawley Ratsa'b
40 mg/kg intravenous (IV) dose represented by the squares and solid line; 40 mg/kg oral dose represented by the circles and solid
line; 80 mg/kg oral dose represented by the upward triangles and dashed line; 320 mg/kg oral dose represented by the diamonds
and dash-dot line.
a Change in slope from 1-10 days represents a transition to a "beta-phase" elimination in female rats.
b The poor fit to 320 mg/kg reflects a dose that is outside the scope of the currently parametrized model.
F-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
ID2 l
E
J 101
c
o
u
E
3
V
1/1 10° -
<
Q :
10":1 -
10"2 10":1 10° 101 102
Time (days)
Figure F-8. Experimentally Observed Serum Concentrations and Median Predictions for a
Single IV Dose of 1 mg/kg or an Oral Gavage Dose of 1 mg/kg PFOA {Kim, 2016, 3749289}
or an IV Dose of 20 mg/kg PFOA {Kudo, 2002,2990271} to Male Sprague-Dawley Rats
1 mg/kg intravenous (IV) dose represented by the squares and solid line; 6 mg/kg oral dose represented by the circles and solid
line; 20 mg/kg IV dose represented by the downward triangles and dashed line.
10*
3 Id1
C7i
£
c Iff3
o
t-J
E
to io-1
i/i
<
o
fc Iff-2
Iff"3
10"2 10"1 10° Iff1 102
Time (days)
Figure F-9. Experimentally Observed Serum Concentrations and Median Predictions for a
Single IV Dose of 1 mg/kg or an Oral Gavage Dose of 1 mg/kg PFOA {Kim, 2016, 3749289}
or an IV Dose of 20 mg/kg PFOA {Kudo, 2002, 2990271} to Female Sprague-Dawley Rats3
1 mg/kg intravenous (IV) dose represented by the squares and solid line; 6 mg/kg oral dose represented by the circles and solid
line; 20 mg/kg IV dose represented by the downward triangles and dashed line.
a Change in slope from 1-10 days represents a transition to a "beta-phase" elimination in female rats.
F-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
| 60
Q-u
40 ¦
< 20
O
1
;
i
4
i
i
i
i
20
40
60 BO
Time [days]
100
120
140
160
Figure F-10. Observed and Predicted PFOA Plasma Concentration in Female Sprague-
Dawley Rats following Perinatal, Lactational, and Post-weaning Exposure during Study 1
of NTP (2020, 7330145)a b
a Vertical black dashed and dash-dot lines represent the end of gestation and weaning, respectively.
b Top panel represents dam concentrations (mg/L) from conception (t=0 days) to weaning (t=43 days) while bottom panel
represents fetal/pup concentrations from conception (t=0 days) to postnatal week 16 (PNW 16) during interim evaluation. Each
simulation represents a dam daily dietary exposure of 0,150, or 300 ppm coupled with either 300 ppm or 1,000 ppm daily
dietary exposure to the pup post-weaning. Using the "dam/pup ppm" nomenclature, four total dosing scenarios are modeled:
0/300 ppm (square, solid line), 0/1000 ppm (circle, dot-dash line), 150/300 ppm (triangle, dashed line), and 300/1000 ppm
(diamond, dotted line) with corresponding PNW 16 pup plasma concentrations represented as color-matched circles. Dam
concentrations only tracked through the end of weaning.
60 B0
Time [days]
Figure F-ll. Observed and Predicted PFOA Plasma Concentrations in Male Sprague-
Dawley Rats following Perinatal, Lactational, and Post-weaning Exposure during Study 2
of NTP (2020, 7330145)a b
a Vertical black dashed and dash-dot lines represent the end of gestation and weaning, respectively.
b Top panel represents dam concentrations (mg/L) from conception (t=0 days) to weaning (t=43 days) while bottom panel
represents fetal/pup concentrations from conception (t=0 days) to postnatal week 16 (PNW 16) during interim evaluation. Each
simulation represents a dam daily dietary exposure of 300 ppm with 20 (solid line) 40 (dashed) and 80 (dot-dash) ppm daily
dietary exposure to the pup post-weaning. Black circles represent fetal and pup concentrations at gestation day 18 and
postnatal day 4 while the open square (20 ppm), open circle (40 ppm), and open diamond (80 ppm) represent the reported PFOA
plasma concentrations in pup at PNW 16. Dam concentrations only tracked through the end of weaning.
F-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
10'
Literature reported concentration (mg/LJ
Literature reported concentration (mg/L}
Adult test data
Fetal/infant test data
Figure F-12. Model Prediction Summary for PFOA Test Data
Left: Model predictions on the adult, single-dose test data result in a mean squared log error (MSLE) of 1.44. Right: Model
predictions on fetal/infant pharmacokinetics during development broken out by lifestages (pre-natal - green, lactation - orange,
post-weaning - blue) and species (rat - circle, mouse - x) with an MSLE of 0.285. Dashed lines represent +/- one-half logio
F.1.3 Consideration of Hinderliter et a I. (2006, 3749132) in the
Animal Model
Based on SAB's recommendation, EPA examined Hinderliter et al. (2006, 3749132) and
compared the reported pharmacokinetic data at 2-hours post dosing and at 24-hours post dosing
for the 3-, 4-, and 5- week animals given a single oral gavage PFOA dose of 10 or 30 mg/kg to
determine how the model predicts single-dose pharmacokinetics at this young age (Figure F-13).
During the post-weaning phase, the modeling framework in the analysis of the Hindierliter et al.
(2006, 3749132) study uses the Wambaugh et al. (2013, 2850932) model with reported juvenile
body weights to simulate the post-weaning animals. Across all three age groups, this approach
works reasonably well for juvenile male rats (blue and orange symbols in Figure F-13). As a
result of investigating Hinderliter et al. (2006, 3749132), EPA found an age-dependent change in
model predictions for the female juvenile rat (red symbols), where the Wambaugh et al. (2013,
2850932) model dramatically underpredicts the 3-week-old female rats at 24 hours post-dosing
while slightly underpredicting the 5-week-old female rats at 24 hours post-dosing. This is due to
the rapid female rat specific PFOA clearance in the Wambaugh et al. (2013, 2850932) model
which was parameterized on adult female rat pharmacokinetic data. One possibility is that this
model underprediction for young animals could be due to a not yet modeled age-dependent
change in PFOA urinary excretion as female pups mature to adult rats and could be attributed to
changes in OAT1/OAT3 expression as the pup ages. However, as outlined in Figure F-12, the
one compartment model approach for breastfed pups successfully predicts the reported pup pre-
natal and lactation life-stages. Additionally, Figures F-10, F-l 1, and F-12 demonstrate that the
switch to the Wambaugh et al. (2013, 2850932) for post-weaning and pup maturation
successfully predicts steady-state PFOA concentrations in the post-weaning male and female rats
F-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
at postnatal week 19 when the endpoint of interest from NTP (2020, 7330145) is measured.
While it might be possible to use the reported PK data in post-weaning, juvenile, rats from
Hinderliter et al. (2006, 3749132) to estimate an age-dependent clearance for these young rats,
EPA's assessment of the study indicates that, due to the single-dose study design and age at
which the measurements were reported (i.e., 3-5 weeks of age), incorporation of the results
would not impact the current risk estimation of the endpoints used in the NTP study because
those measurements were taken at 19 weeks of age with continuous dosing between 15 and 19
weeks.
Literature reported concentration (mg/L)
Figure F-13. Model Prediction Summary for PFOA Data from Hinderliter et al. (2006,
3749132)
F-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Model predictions of juvenile rats dosed with PFOA from Hinderliter et al. (2006, 3749132) using the adult toxicokinetic
parameters determined in Wambaugh et al. (2013, 2850932). Symbol color reflects the sex of the rat at the given hours post-
dosing where blue and orange represent male rats at 2 and 24 hours post-dosing, respectively. Female rats are represented as
green and red at 2 and 24 hours post-dosing, respectively. Symbol types represent the rat age when dosing began and correspond
to 3 weeks (circle), 4 weeks (x) and 5 weeks (square) of age. Dashed lines represent +/- one-half loglO. Female rats measured at
24 hours post-dosing represent the predicted concentrations falling outside the +/- one-half log 10 bounds.
F.2 Human Model Validation
As mentioned in the main document (see Main PFOA 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-14). 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 PFOS as well as PFOA. 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.
Age (yr) Age (yr)
Figure F-14. 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-15; Figure F-16). 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
F-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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
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-15. Predicted Child Serum Levels Compared to Reported Values
These values were calculated using the updated parameters with constant Vdand exposure relative to body weight.
MOBA = Norwegian Mother, Father, and Child Cohort Study.
F-12
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Observed PFOA (ng/ml) Observed PFOS (ng/ml)
Figure F-16. Comparison of Predicted and Observed Child Serum Levels
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-17). 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 PFOA 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 PFOA (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 PFOA 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 increased. This reflects the increased importance of excretion relative to the distribution
of incoming PFOA during the time period following lactational exposure.
F-13
-------
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
d
Child at 5 yr
a
-2
i-
-2
Figure F-17. 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.
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-18). 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-14
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Fromme, 2010
MOBA Cohort
Mogensen, 2015
Fromme, 2010
MOBA Cohort
Mogensen, 2015
Figure F-18. 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
MOBA = Norwegian Mother, Father, and Child Cohort Study.
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-19). As discussed in the main document (see Main PFOA Document), this
approach was not used for dosimetric extrapolation due primarily to the poor fit to the PFOA
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.
Age (yr) Age (yr)
Figure F-19. Predicted Child Serum Levels Compared to Reported Values with Constant
Volume of Distribution and Variable Exposure Based on Drinking Water Intake
F-15
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
MOBA = Norwegian Mother, Father, and Child Cohort Study.
F-16
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix G. Relative Source Contribution
G.l Background
EPA applies an 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., PFOA) 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 PFOA, other potential sources of exposure
include diet, ambient and indoor air, incidental soil and dust ingestion, and consumer products.
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
populations, 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
G-l
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
approach from a public health perspective and would result in a higher MCLG for those
disproportionately affected populations.
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 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}. Literature
that reported exposure measures from household media paired with occupant PFAS
concentrations in blood serum was identified. Second, systematic evidence mapping was
conducted for literature reporting measured occurrence of PFAS 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
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 PFOA and seven other frequently measured PFAS (PFOS, perfluorobutanoic acid
(PFBA), 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 according to the PECO criteria identified 486 studies
that moved on to full-text screening; only 6 studies fully addressed the protocol requirements
{Wu, 2014, 2533322; Makey, 2017, 3860102; Byrne, 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 using DistillerSR software.
Exposure intake calculations were performed to estimate a percentage of participant 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 the IRIS
Handbook {U.S. EPA, 2022, 2022, 10476098}. This systematic review provided evidence for an
estimated range of indoor exposure media's contribution to serum PFAS concentrations and
G-2
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
highlighted the limited availability of concordant measurement data from indoor exposure media
and participant serum.
The Deluca and coworkers review (2021 in prep., 9419129) 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., PFOA was measured in sera or a non-
indoor environmental medium). Because the combination of PFOA 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 by reviewing literature reporting measured occurrence of PFAS in exposure media. The
review focused on eight PFAS (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 that were published between 2003-2020. ICF's litstream™
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.
Published 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 for occurrence in 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 PFOA 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 PFOA Sources
PFOA is a synthetic, fully fluorinated, organic acid that is used in many types of consumer
products and in the production of fluoropolymers {U.S. EPA, 2016, 3982042}. PFOA has been
used in flame repellents, cosmetics, paints, polishes, and processing aids used in the manufacture
of nonstick coatings on cookware. It is one of a large group of perfluoroalkyl substances that are
used in consumer and industrial products to improve their resistance to stains, grease, and water.
Under EPA's PFOA Stewardship Program, the eight major companies of the
G-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
perfluoropolymer/fluorotelomer industry agreed to voluntarily reduce facility emissions and
product content of PFOA and related chemicals on a global basis by 95% by no later than 2010
and eliminate these substances from products by 2015 {U.S. EPA, 2006, 3005012}. Despite the
United States phase out of production, EPA has found widespread PFOA contamination in water,
sediments, and soils. Exposure to PFOA can occur through food including fish and shellfish,
house dust, air, and contact with consumer products.
G.3.1 Dietary Sources
Ingestion of food is a potentially significant source of exposure to PFOA and is often claimed to
be the dominant source of exposure 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 is
typically estimated to be about 2-3 ng/kg/day {Gleason, 2017, 5024840}. 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 gastrointestinal uptake {Trudel,
2008, 214241}. However, the estimates are highly uncertain due to analytical methods with poor
sensitivity, relatively few food items with detectable levels, and levels that can vary greatly
depending on sources or location {Gleason, 2017, 5024840}.
There is currently no comprehensive, nationwide Total Diet Study (TDS) for PFOA that can be
used to draw conclusions about the occurrence and potential risk of PFOA 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, collected for the TDS. 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, but not including PFOA: fish sticks (PFOS and PFNA), canned tuna
(PFOS and PFDA), and protein powder (PFOS). PFOA was not detected in any of the food
samples analyzed in FDA TDS samples of produce, meats, dairy and grain products in 2019 or
2021 {FDA, 2021, 9419076}. In a 2018 focused study near a PFAS production plant in the
Fayetteville, North Carolina area, PFOA was detected in several produce samples (cabbage,
collard greens, kale, mustard greens, swiss chard, and lettuce) {FDA, 2018, 9419064}. In bottled
water, PFOA was below the lower limit of quantification (LOQ; 4 ng/L) in all (30) analyzed
samples 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 general levels of PFAS in the U.S. food supply
{FDA, 2021, 9419076}. In a 2010 study, PFOA was detected in food samples collected from
five grocery stores in Texas {Schecter, 2010, 729962}; based on the results from this study and
on dietary intakes from the 2007 U.S. Department of Agriculture food availability data set, the
estimated daily exposure to PFOA per capita was 60 ng/day {U.S. EPA, 2016, 3982042}.
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 PFASs 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
G-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 PFOA for
children and adults were estimated as 0.30 and 0.18 ng/kg body weight per day, respectively.
The most important contributor was "Fish and other seafood," followed by "Eggs and egg
products," "Meat and meat products," and "Fruit and fruit products." "Vegetables and vegetable
products" and "Drinking water" were also found to be important contributors to dietary PFOA
exposure. It is unclear whether or not 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 PFOA, dietary intake was by far the greatest
contributor to aggregate exposure (contributing 92% of total estimated PFOA intake), but intake
from ingestion of house dust represented the dominant pathway for some of the top 20% most
highly exposed individuals. On average, measured serum concentrations of PFOA were similar
to modeled concentrations based on intakes. It is notable that while the authors reported
significant positive correlations between PFOA concentrations in serum and estimated intakes
based on surface dust and vacuum cleaner bag dust samples, correlations with estimated dietary
intakes were not significant, which the authors attributed to temporal variations in dietary intakes
over several 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 PFOA approximately
three times higher from solid food than from liquids.
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}
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
G-5
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
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 fluorotelomer alcohol (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, California.
Under FDA rules, there are dozens of PFAS chemicals still approved for food contact materials.
In 2018, 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 {Schaider, 2017, 3981864; Yuan, 2016, 3859226; Zabaleta, 2020,
6505866}. The data from the cited 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 the Schreder &
Dickman (2018, 9419077) report, inorganic fluoride was the analyte for the initial analysis. Fifty
six percent 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. Twenty fast food packaging samples of the 407 total samples were
selected for more extensive PFAS specific analysis. PFOA, PFHxA, and PFBS were among the
PFAS with the highest detection rates; PFOA was detected in 6/20 samples.
An analysis of popcorn bags, snack bags, and sandwich bags purchased in 2018 from
international vendors and grocery stores in the United States found little evidence of PFOA, with
only two popcorn bags with content above the limit of quantitation of 5.11 ng per gram of paper
{Monge Brenes, 2019, 5080553}. The authors presented these results as evidence of a reduction
in PFOA 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
G-6
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
measurable concentrations of PFOA in the 2 bags from the United States, levels typically at
about 4 ng/g in those from several European countries, and levels around 50 ng/g in bags from
China.
Yuan et al. (2016, 3859226) analyzed 25 food contact materials purchased in Columbus, Ohio
for PFAS as compared with 69 products purchased in China. The primary PFAS substances
detected were consistently the C6 to C14 telomer alcohols. In food packaging materials from
China, of the 15 detected perfluorinated carboxylic acids, PFOA was the most frequently
detected (90%) and was detected with the highest median concentration (1.72 ng/g). In contrast
to the products from China, the primary analyte from U.S. paper food contact products other than
popcorn bags was the 6:2 telomer alcohol. The authors also report a migration efficiency of
PFOA from paper bowl packaging into food stimulants of 1.58%. This is a relatively low
efficiency compared to several of the FTOHs which the authors reported to migrate with greater
than 90% efficiency.
Zabaleta et al. (2020, 6505866) looked at PFAS in 25 paper- and paperboard packaging materials
primarily collected in Spain. Except in the single microwave popcorn bag collected from China,
none of the perfluorocarboxylic acids (C3, C6, C7, C8, C 9, C10), including PFOA, were 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 C3 to C10 perfluorinated carboxylates. Zabaleta et al.
(2020, 6505866) also monitored migration of the PFAS carboxylates (C6 to C10) from
packaging materials into cereal, rice, or milk. For each PFAS studied the percent migration to
milk exceeded that to rice with the lowest percent migration being that to cereal. Percent
migration to foods decreased as the carbon chain length increased (C6 to C10) after a 6-month
period. The migration percentage of PFOA into cereal, rice, and milk powder products over
6 months ranged from 1.4%-5.6%.
G.3.1.2 Fish and Shellfish
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. PFOA has been detected in
freshwater fish fillet samples collected during several national studies in rivers and the Great
Lakes; however, PFOA is reported at a lower frequency and at lower levels compared to other
PFAS, including PFOS ( Table G-l).
Table G-l. Summary of EPA national freshwater fish tissue monitoring results for PFOA
Reference
Most Commonly Sampled
Species
Site Description Results
U.S. EPA (2010,
10369692)
Smallmouth bass
Largemouth bass Channel
catfish
162 urban river sites across No PFOA detections
the United States reported.
U.S. EPA (2015,
10369694)
Largemouth bass
Smallmouth bass Black
crappie
White crappie
Walleye/sauger Yellow
perch
349 urban and nonurban PFOA detected in 4% of
river sites across the United samples.
States Maximum detected
concentration 0.27 ng/g.
G-7
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Reference
Most Commonly Sampled
Species
Site Description
Results
White bass
Northern pike
Lake trout
Brown trout
Rainbow trout
Brook trout
U.S. EPA (2011,
10369695)
Lake trout Smallmouth bass 157 nearshore sites along
Walleye the U.S. shoreline of the
Great Lakes
PFOA detected in 12% of
samples.
Maximum detected
concentration 0.97 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
Rainbow Trout
Chinook Salmon
Yellow Perch
Brown Trout
Lake Trout
Walleye
152 nearshore sites along
the U.S. shoreline of the
Great Lakes
PFOA detected in 14% of
samples.
Maximum detected
concentration 1.93 ng/g.
Notes: U.S. EPA = United States Environmental Protection Agency
In addition, there are several available studies that assess PFAS concentrations in fish, shellfish,
and other aquatic species. In 2015, Penland et al. (2020, 6512132) measured PFAS
concentrations in invertebrates and vertebrates along the Yadkin - Pee Dee River in North
Carolina and South Carolina. PFOA was detected in whole body tissues of unionid mussels
(7.41 ng/g wet weight) and aquatic insects (10.68 ng/g wet weight), but was not detected in
Asian clam, snails, or crayfish. PFOA was measured in muscle tissue of 2/11 sampled fish
species: the channel catfish (21.19 ng/g wet weight) and notchlip redhorse (45.66 ng/g wet
weight).
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. Samples were analyzed for 16 PFAS, including PFOA. Brown crab and
shrimps had the highest average concentrations of PFOA (0.78 and 0.43 ng/g ww, respectively).
PFOA was also detected in farmed fish including eel and trout, and marine fish species including
cod, haddock, and sole. However, the PFAS with generally the highest percent detection and
average concentration in all sample types was PFOS.
In seafood samples collected for the FDA 2021-22 seafood survey, Young et al. (2022,
10601281), analyzed concentrations of 20 PFAS, including PFOA, in 8 of the most highly
consumed seafood products in the U.S. PFOA was detected most frequently (100% of samples;
G-8
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
n=10) and at the highest average concentrations (8,334 ppt) in clams and was also detected in
100% of crab samples (n=l 1; 300.9 ppt average concentration). The study reported detections in
cod (20% of samples; n=10; 103.5 ppt average concentration in samples with detections). PFOA
was not detected above the method detection limit (90 or 68 ppt) in tuna, salmon, shrimp, tilapia,
or pollock.
In summary, PFOA has been detected in fish and shellfish samples from freshwater and marine
fish and shellfish, as well as in both farmed and wild-caught samples. While most of the data
were collected from freshwater fish samples, recent studies suggest ingestion of many types of
fish and shellfish can be a potential source of exposure to PFOA. However, in contrast to PFOS,
PFOA concentrations in biotic media tend to be low, or below detection levels, highlighting the
lower overall bioaccumulation potential for this chemical, based on its physical-chemical
properties, including a shorter perfluorinated chain length, and a carboxylate head group. In
addition, trophic biomagnification is rarely observed in aquatic food webs with PFOA.
G.3.2 Consumer Product Uses
A targeted analysis of 29 U.S. and Canadian cosmetic products with high fluorine content
{Whitehead, 2021, 9416542} found high concentrations of fluorotelomer alcohols (FTOH),
including 8:2 FTOH, commonly present in the formulations. A fraction of 8:2 FTOH is believed
to undergo metabolic transformation into PFOA. In addition to direct contact with personal care
products, products and articles (and the use of these) may be sources in the indoor environment
that manifest as measured occurrence in house dust and indoor air. An earlier investigation of
consumer exposure to PFOA by Trudel et al. (2008, 214241) used mechanistic modeling
together with information on product-use habits to estimate oral and dermal exposures clothes,
carpet, upholstery, and food contact materials. Noting that PFOA may be contained as a
contaminant in older and in new products, the authors estimated exposure via both mill-treated
and home-treated carpets. The authors concluded that contact with consumer products is not a
significant contributor to total exposure, but that since PFOA may be a contaminant in even new
products, consumer exposure may continue to occur, particularly via both mill-treated and home-
treated carpets. The authors also point out that carpet and other textiles are likely to be
continuing sources of PFOA in house dust. In contrast, in an analysis of 116 articles of
commerce from the United States, U.S. EPA (2009, 1290922) identified carpets and related
products as potentially the most significant source of PFCAs out of 13 total product categories
analyzed. PFOA was detected in all 13 product types. Other important indoor sources of PFCAs
include floor wax/sealant and home textiles, upholstery, and apparel. In a similar analysis of 52
European products collected between 2014-2016, Borg and Ivarrson (2017, 9416541) reported
that PFOA was the most commonly detected PFAS and was detected in all samples except those
that did not contain any detectable levels of PFAS. Notably, the authors specifically targeted
products that were known or suspected to contain PFAS in their analyses.
Liu et al. (2014, 2324799) investigated trends in PFAS content of household goods between
2007 and 2011. They reported that while PFOA concentrations displayed an overall downward
trend with significant reductions observed in nearly all product categories, PFOA was still
detected in many products. Kotthoff et al. (2015, 2850246) similarly reported broad detection of
PFOA in a 2010 sampling effort that collected 115 European consumer products, including
carpets, leather, outdoor materials, cooking materials, and others. PFOA was detected in all but
G-9
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
one sample type, often at the highest median concentration compared to other PFCAs. FTOHs
were frequently detected at the highest median concentration overall. The products with the
highest concentrations of total PFAS included ski wax (median concentration of 15.5 |ig/kg
PFOA), leather products (median concentration of 12.4 |ig/m2), and outdoor materials (median
concentration of 6 |ig/m2 PFOA). PFOA has also been detected in textile samples of outdoor
apparel from Europe and Asia {Gremmel, 2016, 3858525; van der Veen, 2020, 6316195}. PFOA
was detected in jackets ranging from concentrations of 0.02-4.59 (J,g/m2 {Gremmel, 2016,
3858525}. Interestingly, the level of almost all individual PFAS, including PFOA, and total
PFAS increased when the textiles were subjected to weathering (i.e., increased ultraviolet (UV)
radiation, temperature, and humidity for 300 hours to mimic the average lifespan of outdoor
apparel) {van der Veen, 2020, 6316195}.
G.33 Indoor Dust
Several studies suggest that PFOA and its precursors in indoor air and/or house dust may be an
important exposure source for some individuals {Shoeib, 2011, 1082300; Schlummer, 2013,
2552131; Gebbink, 2015, 2850068; Poothong, 2020, 6311690}. PFOA is generally a dominant
ionic PFAS constituent in indoor air and 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}.
PFOA was measured at the highest concentrations (geometric mean concentrations ranging from
41.4-45.0 ng/g) and frequencies (ranging from 89-91% detected) in dust sampled from
Californian households {Wu, 2014, 2533322}. Similarly, PFOA was found at the second highest
levels (mean concentration of 1.98 ng/g) of 15 PFAS measured in dust samples taken from
households in Seoul, South Korea {Kim, 2019, 5080673}. PFOA was detected in all dust
samples from that study. Makey et al. (2017, 3860102) measured PFOA and PFOA precursors in
dust and found weak correlations between concentrations in dust and serum PFOA
concentrations in pregnant Canadian participants. One study in Alaska Natives found no
correlation between dust and serum PFOA concentrations {Byrne, 2017, 4165183}.
G.3.4 Ambient Air
Perfluoroalkyl chemicals have been found in ambient air globally, with the highest
concentrations observed or expected in urban areas and nearest to industrial facilities, areas
where AFFF firefighting foams are used, wastewater treatment plants, waste incinerators, and
landfills {Ahrens, 2011, 2325317}. Perfluorinated acids were measured in Albany, New York air
samples (gas mean concentration of 3.16 pg/m3 and particulate phase mean concentration of 2.03
pg/m3) {Kim, 2007, 1289790}. In Minneapolis, Minnesota, PFOA in the particulate phase
ranged from 1.6-5.1 pg/m3 and from 1.7-16.1 pg/m3 in the gas phase {MPCA, 2008, 9419086}.
Even remote areas far from urban centers have previously reported PFOA concentrations in air
samples: PFOA has been detected in Resolute Bay, Nunavut, Canada {Stock, 2007, 1289794},
as well as other Arctic environments {Butt, 2010, 1291056}.
PFOA is not listed as a hazardous air pollutant under the Clean Air Act. However, two states
(New York and Michigan) have set enforceable air emissions limits. Ambient air is a possible
source of exposure to PFOA for the general population; however, the contribution of air to total
G-10
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
exposure is likely low. For example, De Silva et al. (2021, 7542691) estimated that < 1% of
PFOA exposure to humans in the United States is from inhalation.
G.3.5 Other Exposure Considerations
PFOA has 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 increase 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 is the major contributor
to total PFOA 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 PFOA 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 exposure
(Box 8A; Figure G-l). Therefore, an RSC of 20% (0.20) should be used (Box 8B; Figure G-l).
G-ll
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure G-l. Application of the Exposure Decision Tree {U.S. EPA, 2000,19428} for PFOA
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 PFOA, food and indoor dust are potentially significant exposure 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 PFOA.
G-12
------- |