vvEPA
DRAFT FOR PUBLIC COMMENT
EPA Document No.
EP A-822-P-23 -002
Economic Analysis for the
Proposed Per- and Polyfluoroalkyl Substances
National Primary Drinking Water Regulation
Appendices
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Economic Analysis for the Proposed Per- and Polyfluoroalkyl Substances
National Primary Drinking Water Regulation Appendices
Prepared by:
U.S. Environmental Protection Agency
Office of Water
Office of Groundwater and Drinking Water
Standards and Risk Management Division
Washington, DC 20460
EPA Document Number: EPA-822-P-23-002
MARCH 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Disclaimer
This document has been reviewed in accordance with EPA policy and approved for publication.
Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.
Proposed PFAS Rule Economic Analysis
i
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Contents
Appendix A. Framework of Bayesian Hierarchical Markov Chain Monte Carlo
Occurrence Model A-l
A. 1 Data Selection A-l
A. 2 Conceptual Model Structure A-4
A.3 Model Implementation A-6
Appendix B. Affected Population B-l
Appendix C. Cost Analysis Results C-l
C.l PWS-Level Cost Details C-l
C.l.l Mean Annual Cost for all Community Water Systems C-l
C.l.2 Mean Annual Cost for all Non-Transient Non-Community Water Systems .... C-5
C.l.3 Mean Annual Cost for Community Water Systems that Treat or Change
Water Source C-9
C. 1.4 Mean Annual Cost for Non-Transient Non-Community Water Systems that
Treat or Change Water Source C-13
C.l.5 Distribution of Small Community Water System Costs C-17
C.l.6 Distribution of Small Non-Community Non-Transient Water System Costs. C-21
C.l.7 Distribution of Small Community Water System Costs that Treat or Change
Water Source C-25
C.l.8 Distribution of Small Non-Community Water Non-Transient System Costs
that Treat or Change Water Source C-29
C.2 Household-Level Cost Details C-33
C.2.1 Household Costs for all Community Water Systems C-33
C.2.2 Household Costs for Community Water Systems that Treat or Change Water
Source C-37
Appendix D. PFOA and PFOS Serum Concentration-Birth Weight Relationship D-l
D. 1 Weight of Evidence of Birth Weight Effects D-l
D.2 Review of Available Meta-Analyses D-l
D.3 Exposure-Response Functions Based on Epidemiological Studies D-9
Appendix E. Effects of Reduced Birth Weight on Infant Mortality E-l
E. 1 Birth Weight-Mortality Relationship E-l
E.2 Basis for Updated Birth Weight-Mortality Relationship E-3
E.3 Development of the Analytical Dataset E-5
E.3.1 Data Sources E-5
E.3.2 Dataset Development E-6
E.3.3 Identification of Infant Mortality Risk Factors E-6
E.4 Development of Variables E-7
E.5 Summary Statistics E-ll
Proposed PFAS Rule Economic Analysis ii March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
E.6 Estimation Methods E-13
E.7 Results and Discussion E-14
E.7.1 Mortality Regression Models E-14
E.7.2 Comparison to Prior Studies E-22
E.8 Limitations and Uncertainties E-22
Appendix F. Serum Cholesterol Dose Response Functions F-l
F.l Data Sources F-l
F.l.l Literature Review and Studies Identification for the Meta-Analysis F-l
F.l.2 Assessment of Study Applicability to the Meta-Analysis F-2
F.2 Meta-Analysis F-8
F.3 Extraction of Slope Values for TC and HDLC F-8
F.4 Methods and Key Assumptions F-9
F.4.1 Slope Estimation for PFOA F-10
F.4.2 Slope Estimation for PFOS F-16
F.4.3 Sensitivity Analyses F-22
F.4.4 Limitations and Uncertainties F-22
Appendix G. CVD Benefits Model Details and Input Data G-l
G. 1 Model Overview and Notation G-l
G.2 Hard CVD Event Incidence Estimation G-8
G.2.1 Probability of the First Hard CVD Event G-8
G.2.2 Prevalence of Past Hard CVD Events G-l 1
G.2.3 Distribution of Fatal and Non-Fatal First Hard CVD Events G-13
G.2.4 Post-Acute CVD Mortality G-l6
G.2.5 Survivors of the first hard CVD event at ages 40-65 G-18
G.2.6 Survivors of the first hard CVD event at ages 66+ G-20
G.3 Detailed CVD Model Calculations G-21
G.3.1 Baseline Recurrent Calculations Without Explicit Treatment of the CVD
Population G-22
G.3.2 Baseline Recurrent Calculations with Explicit Treatment of the CVD
Population G-22
G.3.3 Regulatory Alternative Recurrent Calculations with Explicit Treatment of
the CVD Population G-25
G.3.4 Recurrent Estimation of Post-Acute CVD Mortality G-27
G.3.5 Risk Reduction Calculations G-28
G.4 ASCVD Model Validation G-29
G.5 CVD Model Inputs G-30
Appendix H. Cancer Benefits Model Details and Input Data H-l
H. 1 Details on the Cancer Life Table Approach H-l
H. 1.1 Evolution of Model Population (B,A) under Baseline Pollutant Exposure H-3
H. 1.2 Evolution of Model Population (B,A) under the Regulatory Alternative
Pollutant Exposure H-6
Proposed PFAS Rule Economic Analysis
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
H. 1.3 Health Effects and Benefits Attributable to the Regulatory Alternatives H-6
H.2 Cancer Life Table Model Input Data H-7
H.3 Baseline Kidney Cancer Statistics H-9
H.4 Baseline Bladder Cancer Statistics H-37
H.5 RCC Valuation Data H-43
Appendix I. Trihalomethane Co-Removal Model Details and Analysis 1-1
I.1 Data Analysis 1-1
1.2 Discussion of Other Models 1-5
1.3 THM4 Reduction Results 1-6
1.4 Sampling Points from the Fourth Six Year Review Plants with Granular Activated
Carbon Treatment 1-12
Appendix J. Value of a Statistical Life Updating J-13
Appendix K. Benefits Sensitivity Analyses K-l
K. 1 Overview of the Hypothetical Exposure Reduction K-l
K.2 Estimation of Blood Serum PFOA, PFOS, and PFNA K-2
K.3 CVD Sensitivity Analyses K-3
K.4 Birth Weight Sensitivity Analyses K-8
K.5 RCC Sensitivity Analyses K-l 1
Appendix L. Uncertainty Characterization Details and Input Data L-l
L.l Cost Analysis Uncertainty Characterization L-l
L.l.l Total Organic Carbon Concentration Uncertainty L-l
L.l.2 Compliance Technology Unit Cost Curve Selection Uncertainty L-l
L.2 Benefits Analysis Uncertainty Characterization L-2
L.2.1 Exposure-Response Function Uncertainty L-4
L.2.2 Population Attributable Fraction Uncertainty L-l
Appendix M. Environmental Justice M-l
M.l Demographic Profile of Category 4 and 5 PWS Service Areas M-l
M.2 Exposure Analysis Results M-4
M.2.1 Baseline Scenario M-4
M.2.2 Hypothetical Regulatory Scenario #1: UCMR 5 MRLs M-7
M.2.3 Hypothetical Regulatory Scenario #2: 10.0 ppt M-8
Appendix N. Supplemental Cost Analyses N-l
N. 1 Cost Analysis for Very Large Systems N-l
N.2 Hazardous Waste Disposal Cost Impacts N-2
N.3 Incremental Treatment Cost of Other PFAS N-4
Appendix O. Appendix References 0-13
Proposed PFAS Rule Economic Analysis iv March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
List of Tables
Table A-l: System and Sample Counts for Contributions to the Supplemental State Dataset
by State A-3
Table B-l: Summary of Inputs and Data Sources Used to Estimate Affected Population B-2
Table C-l: Mean Annualized Cost per CWSs, Proposed Option (PFOA and PFOS MCLs of
4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021) C-l
Table C-2: Mean Annualized Cost per CWSs, Option la (PFOA and PFOS MCLs of 4 ppt)
(Commercial Cost of Capital, $2021) C-2
Table C-3: Mean Annualized Cost per CWSs, Option lb (PFOA and PFOS MCLs of 5.0
ppt) (Commercial Cost of Capital, $2021) C-3
Table C-4: Mean Annualized Cost per CWSs, Option lc (PFOA and PFOS MCLs of 10.0
ppt) (Commercial Cost of Capital, $2021) C-4
Table C-5: Mean Annualized Cost per NTNCWS, Proposed Option (PFOA and PFOS
MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021) C-5
Table C-6: Mean Annualized Cost per NTNCWS, Option la (PFOA and PFOS MCLs of 4
ppt) (Commercial Cost of Capital, $2021) C-6
Table C-7: Mean Annualized Cost per NTNCWS, Option lb (PFOA and PFOS MCLs of
5.0 ppt) (Commercial Cost of Capital, $2021) C-7
Table C-8: Mean Annualized Cost per NTNCWS, Option lc (PFOA and PFOS MCLs of
10.0 ppt) (Commercial Cost of Capital, $2021) C-8
Table C-9: Mean Annualized Cost per CWSs that Treat or Change Water Source, Proposed
Option (PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital,
$2021) C-9
Table C-10: Mean Annualized Cost per CWSs that Treat or Change Water Source, Option
la (PFOA and PFOS MCLs of 4 ppt) (Commercial Cost of Capital, $2021) C-10
Table C-l 1: Mean Annualized Cost per CWSs that Treat or Change Water Source, Option
lb (PFOA and PFOS MCLs of 5.0 ppt) (Commercial Cost of Capital, $2021) C-l 1
Table C-12: Mean Annualized Cost per CWSs that Treat or Change Water Source, Option
lc (PFOA and PFOS MCLs of 10.0 ppt) (Commercial Cost of Capital, $2021) C-12
Table C-13: Mean Annualized Cost per NTNCWSs that Treat or Change Water Source,
Proposed Option (PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of
Capital, $2021) C-13
Table C-14: Mean Annualized Cost per NTNCWSs that Treat or Change Water Source,
Option la (PFOA and PFOS MCLs of 4 ppt) (Commercial Cost of Capital, $2021) C-14
Table C-15: Mean Annualized Cost per NTNCWSs that Treat or Change Water Source,
Option lb (PFOA and PFOS MCLs of 5.0 ppt) (Commercial Cost of Capital, $2021) C-15
Proposed PFAS Rule Economic Analysis
v
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-16: Mean Annualized Cost per NTNCWSs that Treat or Change Water Source,
Option lc (PFOA and PFOS MCLs of 10.0 ppt) (Commercial Cost of Capital, $2021) C-16
Table C-17: Distribution of Annualized Cost for Small CWSs, Proposed Option (PFOA
and PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021) C-17
Table C-18: Distribution of Annualized Cost for Small CWSs, Option la (PFOA and PFOS
MCLs of 4 ppt) (Commercial Cost of Capital, $2021) C-18
Table C-19: Distribution of Annualized Cost for Small CWSs, Option lb (PFOA and PFOS
MCLs of 5.0 ppt) (Commercial Cost of Capital, $2021) C-19
Table C-20: Distribution of Annualized Cost for Small CWSs, Option lc (PFOA and PFOS
MCLs of 10.0 ppt) (Commercial Cost of Capital, $2021) C-20
Table C-21: Distribution of Annualized Cost for Small NTNCWSs, Proposed Option
(PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021) C-21
Table C-22: Distribution of Annualized Cost for Small NTNCWSs, Option la (PFOA and
PFOS MCLs of 4.0 ppt) (Commercial Cost of Capital, $2021) C-22
Table C-23: Distribution of Annualized Cost for Small NTNCWSs, Option lb (PFOA and
PFOS MCLs of 5.0 ppt) (Commercial Cost of Capital, $2021) C-23
Table C-24: Distribution of Annualized Cost for Small NTNCWSs, Option lc (PFOA and
PFOS MCLs of 10.0 ppt) (Commercial Cost of Capital, $2021) C-24
Table C-25: Distribution of Annualized Cost for Small CWSs that Treat or Change Water
Source, Proposed Option (PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial
Cost of Capital) C-25
Table C-26: Distribution of Annualized Cost for Small CWSs that Treat or Change Water
Source, Option la (PFOA and PFOS MCLs of 4.0 ppt) (Commercial Cost of Capital,
$2021) C-26
Table C-27: Distribution of Annualized Cost for Small CWSs that Treat or Change Water
Source, Option lb (PFOA and PFOS MCLs of 5.0 ppt) (Commercial Cost of Capital,
$2021) C-27
Table C-28: Distribution of Annualized Cost for Small CWSs that Treat or Change Water
Source, Option lc (PFOA and PFOS MCLs of 10.0 ppt) (Commercial Cost of Capital,
$2021) C-28
Table C-29: Distribution of Annualized Cost for Small NTNCWSs that Treat or Change
Water Source, Proposed Option (PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0)
(Commercial Cost of Capital, $2021) C-29
Table C-30: Distribution of Annualized Cost for Small NTNCWSs that Treat or Change
Water Source, Option la (PFOA and PFOS MCLs of 4.0 ppt) (Commercial Cost of Capital,
$2021) C-30
Proposed PFAS Rule Economic Analysis
vi
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-31: Distribution of Annualized Cost for Small NTNCWSs that Treat or Change
Water Source, Option lb (PFOA and PFOS MCLs of 5.0 ppt) (Commercial Cost of
Capital, $2021) C-31
Table C-32: Distribution of Annualized Cost for Small NTNCWSs that Treat or Change
Water Source, Option lc (PFOA and PFOS MCLs of 10.0 ppt) (Commercial Cost of
Capital, $2021) C-32
Table C-33: Mean Annualized Cost per Household, Proposed Option (PFOA and PFOS
MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021) C-33
Table C-34: Mean Annualized Cost per Household, Option la (PFOA and PFOS MCLs of
4.0 ppt) (Commercial Cost of Capital, $2021) C-34
Table C-35: Mean Annualized Cost per Household, Option lb (PFOA and PFOS MCLs of
5.0 ppt) (Commercial Cost of Capital, $2021) C-35
Table C-36: Mean Annualized Cost per Household, Option lc (PFOA and PFOS MCLs of
10.0 ppt) (Commercial Cost of Capital, $2021) C-36
Table C-37: Mean Annualized Cost per Household in CWSs that Treat or Change Water
Source, Proposed Option (PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial
Cost of Capital, $2021) C-37
Table C-38: Mean Annualized Cost per Household in CWSs that Treat or Change Water
Source, Option la (PFOA and PFOS MCLs of 4.0 ppt) (Commercial Cost of Capital,
$2021) C-38
Table C-39: Mean Annualized Cost per Household in CWSs that Treat or Change Water
Source, Option lb (PFOA and PFOS MCLs of 5.0 ppt) (Commercial Cost of Capital,
$2021) C-39
Table C-40: Mean Annualized Cost per Household in CWSs that Treat or Change Water
Source, Option lc (PFOA and PFOS MCLs of 10.0 ppt) (Commercial Cost of Capital,
$2021) C-40
Table D-l: Data Sources for PFOA/PFOS Meta-Analyses of Birth Weight Effects D-4
Table E-l: Comparison of Sample Means for Singletons between the 1989 Natality-
Mortality Detail File and the combined 2016-2018 Period/Cohort Linked Birth-Infant
Death Data Files E-4
Table E-2: Variables Used in Singleton Mortality Regression Analysis E-8
Table E-3: Maternal and Infant Characteristics of the Study Population E-12
Table E-4: Odds Ratios and Marginal Effects for the Non-Hispanic Black, Non-Hispanic
White, and Hispanic Mortality Regression Models E-l8
Table E-5: Comparison of Ma et al. (2010) and the EPA Analysis E-22
Table E-6: Limitations and Uncertainties in the Analysis of the Birth Weight-Mortality
Relationship E-23
Proposed PFAS Rule Economic Analysis vii March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table F-l: Studies Selected for Inclusion in the Meta-Analyses F-5
Table F-2: Results forPFOA Meta-Analyses F-l 1
Table F-3: Results for PFOS Meta-Analyses F-17
Table F-4: Limitations and Uncertainties in the Analysis of the Serum Cholesterol Dose
Response Functions F-23
Table G-l: CVD Life Table Model Elements and Notation Summary G-6
Table G-2: ASCVD Model Coefficients G-9
Table G-3: Estimated Past Hard CVD Event Prevalence per 100,000 G-12
Table G-4: Estimated First Hard CVD Event Incidence and Distribution by CVD Event
Type G-l 3
Table G-5: Probability of Hospital Death for a Hard CVD Event G-15
Table G-6: Estimated Distribution of Fatal and Non-Fatal First Hard CVD Events G-l 6
Table G-l. Post-Acute All-Cause Mortality After the First Myocardial Infarction G-18
Table G-8: Post-Acute Mortality After the First Myocardial Infarction G-l9
Table G-9: Post-Acute CVD Mortality Following the First Myocardial Infarction and First
Ischemic Stroke in the Population Aged 66 Years or Older G-21
Table G-10: A Mapping of CVD Model Calculations by Initial Cohort Age, Current Cohort
Age, and Estimation Type G-22
Table G-l 1: Summary of ASCVD Model Validation G-30
Table G-12: Summary of Inputs and Data Sources Used in the CVD Model G-30
Table H-l: Health Risk Model Variable Definitions H-2
Table H-2: Summary of Data Sources Used in Cancer Lifetime Risk Models H-7
Table H-3: Summary of Baseline Kidney Cancer Incidence Data Used in the Model H-10
Table H-4: Summary of Race/Ethnicity-Specific Baseline Kidney Cancer Incidence Data
Used in the Model H-l 1
Table H-5: Summary of Relative and Absolute Kidney Cancer Survival Used in the Model. H-14
Table H-6: Summary of Race/Ethni city-Specific Relative and Absolute Kidney Cancer
Survival Used in the Model H-l8
Table H-7: Summary of All-Cause and Kidney Cancer Mortality Data Used in the Model.... H-32
Table H-8: Summary of Race/Ethnicity-Specific All-Cause and Kidney Cancer Mortality
Data Used in the Model H-3 3
Table H-9: Summary of Baseline Bladder Cancer Incidence Data Used in the Model H-37
Proposed PFAS Rule Economic Analysis
viii
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-10: Summary of Relative and Absolute Bladder Cancer Survival Used in the
Model 11-39
Table H-l 1: Summary of All-Cause and Bladder Cancer Mortality Data Used in the Model. H-42
Table H-12: Studies Reviewed Related to Kidney Cancer Medical Treatment Costs H-44
Table 1-1: ICR TSD Predictions for ATHM4 Based on Disinfectant 1-7
Table 1-2: ICR TSD Predictions for ATHM4 for V2 Year GAC Replacement Based on
Disinfectant Type, EBCT, and Source Water Type 1-8
Table 1-3: ICR TSD Predictions for ATHM4 for One Year GAC Replacement Based on
Disinfectant Type, EBCT, and Source Water Type 1-9
Table 1-4: ICR TSD Predictions for ATHM4 for 1 V2 Year GAC Replacement Based on
Disinfectant Type, EBCT, and Source Water Type 1-10
Table 1-5: ICR TSD Predictions for ATHM4 for Two Year GAC Replacement Based on
Disinfectant Type, EBCT, and Source Water Type 1-11
Table 1-6: Sampling Point IDs for each PWSID were Extracted and Matched for the Years
that Represent Before/After GAC Treatment (Example: PWSID AL0000577) 1-12
Table J-l: Estimated VSL Series J-14
Table J-2: Summary of Inputs and Data Sources Used for Valuation J-16
Table K-l: Overview of Hypothetical Exposure Reductions K-2
Table K-2: Overview of CVD Exposure-Response Scenarios K-4
Table K-3: Exposure-Response Information for CVD Biomarkers K-5
Table K-4: Summary of CVD Sensitivity Analysis for Hypothetical Exposure Reduction 1
(PFOA+PFOS) K-6
Table K-5: Overview of Birth Weight Exposure-Response Scenarios K-8
Table K-6: Exposure-Response Information for Birth Weight K-9
Table K-7: Summary of Birth Weight Sensitivity Analysis K-10
Table K-8: Overview of RCC Exposure-Response Scenarios K-l 1
Table K-9: Exposure-Response Information for RCC K-l 1
Table K-10: Summary of RCC Sensitivity Analysis K-12
Table L-l: Quantified Sources of Uncertainty in Benefits Estimates L-3
Table M-l: Number of Category 4 PWSs and Population Served by Size and State M-2
Table M-2: Number of Category 5 PWSs and Population Served by Size and State M-2
Table M-3: Population Served by Category 4 and 5 PWSs Compared to Percent of U.S.
Population by Demographic Group M-3
Proposed PFAS Rule Economic Analysis
ix
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table M-4: Baseline Scenario: Population Served by Category 4 and 5 PWS Service Areas
Above Baseline Thresholds and as a Percent of Total Population Served M-5
Table M-5: Average PFAS Concentrations (ppt) by Demographic Group in the Baseline,
Category 4 and 5 PWS Service Areas M-6
Table M-6: Hypothetical Regulatory Scenario #1: Demographic Breakdown of Population
Served by Category 4 and 5 PWS Service Areas Above UCMR 5 MRL and as a Percent of
Total Population Served M-9
Table M-7: Reductions in Average PFAS Concentrations (ppt) by Demographic Group in a
Hypothetical Regulatory Scenario with Maximum Contaminant Level at the UCMR 5
MRLs, Category 4 and 5 PWS Service Areas M-10
Table M-8: Hypothetical Regulatory Scenario #2: Demographic Breakdown of Population
Served by Category 4 and 5 PWS Service Areas Above 10.0 ppt and as a Percent of Total
Population Served M-l 1
Table M-9: Reductions in Average PFAS Concentrations (ppt) by Demographic Group in a
Hypothetical Regulatory Scenario with Maximum Contaminant Level at 10.0 ppt, Category
4 and 5 PWS Service Areas M-12
Table N-l: Characteristics of PWSs Serving a Retail Population Greater than One Million N-l
Table N-3: Model System Characteristics for Ground Water Systems N-4
Table N-4: Model System Characteristics for Surface Water Systems N-5
Table N-5: PFAS Occurrence Assumptions for Model System Analysis (ppt) N-6
Table N-6: Annualized Costs for Baseline Systems ($l,000's per year) N-7
Table N-7: Results for Type 1 Systems for High PFAS Occurrence N-9
Table N-8: Results for Type 1 Systems for Medium PFAS Occurrence N-10
Table N-9: Results for Type 2 Systems for High PFAS Occurrence N-l 1
Table N-10: Results for Type 2 Systems for Medium PFAS Occurrence N-l2
Proposed PFAS Rule Economic Analysis
x
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
List of Figures
Figure D-l: Results and Confidence Limits from PFOA, PFOS Meta-Analyses: Changes in
BW (grams) per Change in Serum PFAS Levels (ng/mL) D-8
Figure E-l: Comparison of Change in Incidence of Infant Death per 1 g Increase in Birth
Weight by Gestational Age Category and Race/Ethnicity (Deaths per 1,000 Births) E-15
Figure F-l: Diagram of Literature Retained for Use in the Meta-Analysis and Data Sources.... F-3
Figure F-2: Forest Plots Showing the Beta Coefficients Relating PFOA Concentrations to
TC and HDLC in Each Study Reporting Linear Associations, and Pooled Estimates After
Random-Effects Meta-Analysis F-12
Figure F-3: Filled-in Funnel Plots to Evaluate Publication Bias of the PFOA and TC (Left)
or HDLC (Right) Association in Studies Reporting Linear Associations F-13
Figure F-4: Forest Plots Showing the Beta Coefficients Relating TC and HDLC to PFOA
Concentrations in Each Study, and Pooled Estimates After Random-Effects Meta-Analysis. .F-14
Figure F-5: Filled-in Funnel Plots to Evaluate Publication Bias of the PF OA and TC (Left)
or HDLC (Right) Association F-15
Figure F-6: Forest Plots Showing the Beta Coefficients Relating TC and HDLC to PFOS
Concentrations in Each Study Reporting Linear Associations, and Pooled Estimates After
Random-Effects Meta-Analysis F-18
Figure F-7: Filled-in Funnel Plots to Evaluate Publication Bias of the PFOS and TC (Left)
or HDLC (Right) Association in Studies Reporting Linear Associations F-19
Figure F-8: Forest Plots Showing the Beta Coefficients Relating PFOS Concentrations to
TC and HDLC in Each Study, and Pooled Estimates After Random-Effects Meta-Analysis... F-20
Figure F-9: Filled-in Funnel Plots to Evaluate Publication Bias of the PFOS and TC (Left)
or HDLC (Right) Association F-21
Figure G-l: Overview of Life Table Calculations in the CVD Model G-3
Figure G-2: CVD Model Calculations Tracking CVD and Non-CVD Subpopulations for a
Specific Current Age of Cohort G-5
Figure 1-1: Example Breakthrough Curve for THM4 from the ICR Dataset with Logistic
Fit Functions Shown 1-2
Figure 1-2: Example Percent Removal Results vs. Time based on Logistic Plots Shown in
Figure 1-1 1-3
Figure 1-3: Mean Percentage Removal (Shaded Area ± 1 Standard Deviation) 1-4
Figure 1-4: Probability Density Function of Concentration Difference at 2 years of Carbon
Life (Subdivided by TOC level) 1-5
Proposed PFAS Rule Economic Analysis
xi
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Acronyms and Abbreviations
AE
Adverse Events
AFFF
Aqueous Film Forming Foam
AHRQ
Agency for Healthcare Research and Quality
ANGIDX
Angina, or Angina Pectoris, As Defined in the Medical Exposure Panel Survey
APFO
Ammonium Perfluorooctanoate Production
ASCVD
Atherosclerotic Cardiovascular Disease
AT SDR
Agency for Toxic Substances and Disease Registry
BEA
Bureau of Economic Analysis
BIRTH
Birth Characteristics
BLS
Bureau of Labor Statistics
BP
Blood Pressure
BW
Birth Weight
CAGR
Compound Annual Growth Rate
CDC
Centers for Disease Control and Prevention
CHD
Coronary Heart Disease
CHDDX
Coronary Heart Disease, as Defined in the Medical Exposure Panel Survey
CHMS
Canadian Health Measures Survey
CI
Confidence Interval
COI
Cost Of Illness
CPI
Consumer Price Index
CVD
Cardiovascular Disease
DBP
Disinfection Byproduct
DL
Detection Level
DS
Distribution System
EA
Economic Analysis
Proposed PFAS Rule Economic Analysis
xii
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
EBCT Empty Bed Contact Time
EIA Energy Information Administration
EJ Environmental Justice
EPA/OST U.S. Environmental Protection Agency Office of Science and Technology
EP Entry Point
FIPS Federal Information Processing Standards
GAC Granular Activated Carbon
GDP Gross Domestic Product
GFR Glomerular Filtration Rate
GW Ground Water
HCUP Healthcare Cost and Utilization Project
HDLC High-Density Lipoprotein Cholesterol
HESD Health Effects Support Document
HMO Health Maintenance Organization
ICER Incremental Cost-Effectiveness Ratio
ICR Information Collection Request
IR Incidence Ratio
IS Ischemic Stroke
KC Kidney Cancer
LBW Low Birth Weight
LCB Lower Confidence Bound
MCL Maximum Contaminant Level
MDEM Maternal Demographic and Socioeconomic Characteristics
MEPS Medical Expenditure Panel Survey
tv/tttw Heart Attack, or Myocardial Infarction, as Defined in the Medical Exposure Panel
IVI11J X
Survey
MR Point of Maximum Residence
Proposed PFAS Rule Economic Analysis
XIII
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
mRCC Metastatic Renal Cell Carcinoma
MRF Maternal Risk and Risk Mitigation Factors
MRL Minimum Reporting Level
NCCN National Comprehensive Cancer Network
NCHS National Center for Health Statistics
NHANES National Health and Nutrition Examination Survey
NPDWR National Primary Drinking Water Regulation
NVSS National Vital Statistics System
OGWDW Office Of Groundwater and Drinking Water
Other Kind of Heart Disease or Condition, As Defined in the Medical Exposure Panel
(JhLKiDA 0
Survey
OLS Ordinary Least Squares
OSHA Occupational Safety and Health Administration
OW Offi ce of Water
PAF Population Attributable Fraction
PBPK Pharmacologically Based Pharmacokinetic
PDV Present Discounted Value
PDYPP Personal Disposable Income Per Capita
PFAS Poly- and Perfluoroalkyl Substances
PFBS Perfluorobutane Sulfonic Acid
PFDA Perfluorodecanoic Acid
PFNA Perfluorononanoic Acid
PFOA Perfluorooctanoic Acid
PFOS Perfluorooctanesulfonic Acid
PK Pharmacokinetic
PPPM Per Patient Per Month
PWS Public Water Systems
Proposed PFAS Rule Economic Analysis
XIV
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PWSID
Public Water System ID
QALY
Quality-Adjusted Life-Years
RCC
Renal Cell Carcinoma
ROB
Risk of Bias
RSSCT
Rapid Small-Scale Column Tests
SAB
Science Advisory Board
SD
Standard Deviation
SDWIS
Safe Drinking Water Information System
SE
Standard Error
SEER
Surveillance, Epidemiology, and End Results
STRKDX
Stroke Diagnosis, As Defined in the Medical Exposure Panel Survey
SW
Surface Water
TC
Total Cholesterol
TOC
Total Organic Carbon
THM4
Four Regulated Trihalomethanes
TSD
Treatment Study Database
UCB
Upper Confidence Bound
UCMR
Unregulated Contaminant Monitoring Rule
VSL
Value of a Statistical Life
WS
Water System Facility Point
WTP
Willingness to Pay
Proposed PFAS Rule Economic Analysis
xv
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix A. Framework of Bayesian Hierarchical
Markov Chain Monte Carlo Occurrence Model
This appendix is adapted from Cadwallader et al. (2022) and details the Bayesian hierarchical
Markov chain Monte Carlo model developed by EPA to estimate national occurrence of poly-
and perfluoroalkyl substances (PFAS) at public water systems (PWSs) prior to the
implementation of drinking water treatment technologies and under theoretical regulatory
scenarios (Cadwallader et al. (2022). EPA used the occurrence model to define the universe of
PWSs that could be required to treat their drinking water to reduce PFAS levels under the
regulatory alternatives. EPA has used similar hierarchical model structures to inform analyses in
previous regulatory actions (U.S. EPA, 2000a; U.S. EPA, 2005).
A.l Data Selection
Data collected for the third Unregulated Contaminant Monitoring Rule (UCMR 3) served as the
primary dataset for this model due to its nationally representative design. While large PWSs
included in UCMR 3 represent a census, not all small PWSs were required to monitor. Rather, a
statistically representative national sample of 800 small PWSs were selected using a population-
weighted stratified random sampling design to select small PWSs with broad geographic
distribution representative of all source water types and size categories (U.S. EPA, 2012).
Because UCMR 3 included only a sample of small systems, there is greater uncertainty in the
occurrence estimates for small systems compared to large systems.
Because there was a relatively small fraction of UCMR 3 samples with PFAS concentrations
reported above minimum reporting levels (MRLs), EPA incorporated state PFAS monitoring
datasets to supplement UCMR 3 data in the occurrence model. These datasets, which have
generally been collected more recently than UCMR 3, generally have lower reporting limits
because the analytical methods have matured rapidly over the last 10 years, allowing laboratories
to reliably measure PFAS at concentrations approximately 3 and 30 times lower than for UCMR
3. While the model can incorporate results below reporting limits in the fitting process via
cumulative distribution functions, such results are less informative than reported values. Thus,
state datasets using lower reporting limits than those used in UCMR 3 helped to inform the
model through higher fractions of reported values. The introduction of additional state datasets
consisting of samples that were collected more recently than UCMR 3 broadened the temporal
range of data used to fit the model. EPA anticipates that, if temporal trends are significant, the
addition of more recent state data will only bias the results towards present day.
EPA collected state occurrence data using broad internet searches.1 and downloaded publicly
available monitoring data from state government websites as of August 2021. While
comprehensive information about methods used and reporting was not fully available for all of
the state monitoring programs, nearly all (at least 97.1%) of the state data incorporated in the
occurrence model were analyzed using EPA-approved PFAS drinking water analysis methods,
including EPA Methods 533, 537, and 537.1. Of these methods, the most commonly used
method was EPA Method 537.1.
1 Search terms included "PFAS", "drinking water", "occurrence", "monitoring", and "state", or a specific state name.
Proposed PFAS Rule Economic Analysis
A-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Additionally, if the state data met certain specifications, EPA assumed that they were statistically
comparable with the UCMR 3 data and could be used to inform the national occurrence model.
In making these determinations, EPA performed quality assurance on the state data as they were
reported and described online. The implemented quality assurance procedures included verifying
that the data utilized to inform the national model were inclusive of finished drinking water
samples only, reporting or detection limits were available for any samples reported as below a
reporting limit, PFOA, PFOS, PFHpA, and PFHxS were reported as individual chemical
analytes, and reported state data were for distinct state monitoring efforts {i.e., they were not also
a part of UCMR 3 monitoring). If any of this information could not be verified based on the
descriptions that states provided on their public websites or within the downloadable data, those
state data were not incorporated within the national occurrence model.
Further, the supplemental state data were limited to samples collected from systems that were
also included in UCMR 3. The purpose of this was to prevent biasing the dataset towards states
for which the data from additional PWSs were available and to maintain the nationally
representative set of systems selected for UCMR 3. Using these criteria, 17 states were identified
as having some state monitoring data to be included in fitting the national occurrence model.
These states included: Arizona, California, Colorado, Delaware, Georgia, Illinois, Kentucky,
Maine, Massachusetts, Michigan, New Hampshire, New Jersey, North Dakota, Ohio,
Pennsylvania, South Carolina, and Vermont (Arizona Department of Environmental Quality,
2021; California Division of Drinking Water, 2020; Colorado Department of Public Health and
Environment, 2020; Delaware Office of Drinking Water, 2021; Georgia Environmnetal
Protection Division, 2020; Illinois Environmental Protection Agency, 2021; Kentucky
Department for Environmental Protection, 2019; Maine Department of Environmental
Protection, 2020; Massachusetts Department of Energy and Environmental Affairs, 2021;
Michigan Environment, 2021; New Hampshire Department of Environmental Services, 2021;
New Jersey Department of Environmental Protection, 2021; North Dakota Department of
Environmental Quality, 2020; Ohio Department of Health, 2020; Pennsylvania Department of
Environmental Protection, 2020; South Carolina Department of Health and Environmental
Control, 2020; Vermont Department of Environmental Conservation, 2021). According to state
websites, these state data represent samples collected between March 2016 through May 2021.
The dataset used to fit the model included all data available in the final UCMR 3 dataset for
PFOS, PFOA, PFHpA, and PFHxS2 (U.S. EPA, 2017). This amounted to 36,972 samples each
for PFOS, PFOA, and PFHpA, and 36,971 UCMR 3 samples for PFHxS. Of these four PFAS,
1,114 samples had results reported at or above the UCMR 3 MRL3. The additional state datasets
included to supplement the UCMR 3 data included 6,645 PFOS samples, 6,656 PFOA samples,
4,715 PFHpA samples, and 5,114 PFHxS samples collected at systems that were included in
UCMR 3. Of these samples, 2,200 (33%) were reported values for PFOS, 2,694 (40%) were
reported values for PFOA, 932 (20%) were reported values for PFHpA, and 1,269 (25%) were
reported values for PFHxS. The remainder were listed as being below their respective reporting
limits.
2 PFBS and PFNA were not included in this model because 19 reported values across the country from the primary dataset
(UCMR 3) were insufficient for fitting the national model (Cadwallader et al., 2022).
3 MRLs under UCMR 3 were as follows: PFOS 40 ppt; PFOA 20 ppt; PFNA 20 ppt; PFHxS 30 ppt; PFHpA 10 ppt; and PFBS 90
ppt.
Proposed PFAS Rule Economic Analysis
A-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table A-l provides information on the number of systems and samples included in each
supplemental state dataset. Reporting limits in state datasets varied both across and within
datasets but were primarily in the lower single digits in parts per trillion (ppt) for all four PFAS
included in the model, though for some samples the limits reported were as high as the UCMR 3
limits or as low as sub-1 ppt. The particularly low limits associated with some samples may be
associated with method detection limits rather than more conservative reporting limits.
Table A-l: System and Sample Counts for Contributions to the Supplemental
State Dataset by State
State
Systems
PFOS Samples
PFOA Samples
PFHpA
PFHxS
Included
Samples
Samples
AZ
2
190
189
0
0
CA
65
1,913
1,913
1,721
1,723
CO
52
95
95
95
95
DE
1
34
34
0
0
GA
1
2
2
2
2
IL
97
321
321
319
319
KY
23
25
25
25
25
MA
65
434
434
436
436
ME
1
3
3
3
3
MI
58
160
160
151
160
ND
1
1
1
1
1
NH
20
334
336
176
331
NJ
148
2,676
2,686
1,566
1,566
OH
145
234
234
0
234
PA
51
91
91
91
91
SC
31
104
104
101
100
VT
10
28
28
28
28
Abbreviations: PFOA - perfluorooctanoic acid; PFOS - perfluorooctanesulfonic acid; PFHpA - perfluoroheptanoic acid;
PFHxS - Perfluorohexane sulfonate.
Further, there were several instances where approximate values were provided in state data when
the sample results were above a method detection limit but below the quantitation limit. In these
cases, EPA used the reported values assuming that the uncertainty introduced by using these
values would be small in comparison to within-system variability. While certain systems may
have adapted treatment since the time that data were collected, the data included in the
occurrence model represent a best estimate of the current state of occurrence. Note that both
samples with results reported as specific measured concentrations and samples with
concentrations reported as lower than a reporting limit were used to fit the model. While the
latter help to provide information to the model, samples providing a measured result are much
more informative.
Proposed PFAS Rule Economic Analysis
A-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A.2 Conceptual Model Structure
The Bayesian hierarchical model presented here uses log transformed data. Unless otherwise
noted, all of the following discussions, equations, distributions are based upon the use of PFOA,
PFOS, PFHpA, and PFHxS data that have been log transformed with the natural log.
EPA tested several model variants. These variants all featured a hierarchical structure with a
multivariate normal distribution of system-level means and system-level normal distributions,
which were assumed to have been the parent distributions for the individual sample results. Thus,
for each variant, EPA assumed lognormality for system-level medians as well as within-system
occurrence. Lognormality is a common assumption for environmental contaminant
concentrations and constitutes a core assumption made here (Lockwood et al., 2001; Ott, 1995).
The exploration of alternative distributions is inhibited by the large fraction of samples found
below their respective reporting limits. Similar Bayesian hierarchical model approaches have
been used in past drinking water occurrence assessments conducted by EPA and others,
including for arsenic and Cryptosporidium parvum (Crainiceanu et al., 2003; Lockwood et al.,
2001; Ott, 1995). The exploration of alternative distributions is inhibited by the large fraction of
samples found below their respective reporting limits.
Model variants differed by inclusion of parameters specific to system size (small versus large)
and source water type (ground water versus surface water). These parameters included:
independent correlation matrices, between-system standard deviations, within-system standard
deviations, and fixed factor shifts of system-level means. EPA included fixed factor shifts in
model variants to allow the model to explore whether systems of certain categories (e.g., large or
small, ground water or surface water), might generally appear to have higher or lower
concentrations of each chemical. EPA compared these model variants using 5-fold cross
validation. EPA selected the model that performed best in the 5-fold cross validation exercise
(described below).
EPA assumed that system-level means were distributed multivariate normally. This was done to
allow the model to fit and utilize a covariance matrix among system-level means for the four
PFAS included. Before adjustment for system-specific factors, the system-level means for PFOS,
PFOA, PFHpA, and PFHxS were assumed to be distributed as:
Equation A-l:
mUraw.i ~ MVNorm(MU, 2)
Where i is the system index and equal to 1, ..., nsys, nsys is the number of PWSs informing the
model, muraw i is a vector of length 4, with the four values indicating unadjusted system-level
means for PFOS, PFOA, PFHpA, and PFHxS. MU is a vector of length 4 providing the grand
national means for large PWSs, 1 is the covariance matrix for system-level means. 1 is related to
the correlation matrix and between-system standard deviation as shown in Equation A-2.
Equation A-2:
S = diag(aB) * fl * diag(aB)
Proposed PFAS Rule Economic Analysis
A-4
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Where ob is a vector of between-system standard deviations and Q is the correlation matrix of
system-level means for PFOS, PFOA, PFHpA, and PFHxS. For small systems, a fixed factor
shift was then applied to muraw i. This is shown in Equation A-3.
Equation A-3:
mill = muraw j + (bSM * SM[)
Here bSM is a vector of length 4 indicating an adjustment to be added to the unadjusted system
level mean (muraw i) if a system is small. SMt is a binary indicating whether system i is small
(1) or large (0). muL is a vector of length 4, with the four values indicating adjusted system-level
means for PFOS, PFOA, PFHpA, and PFHxS. Samples are then assumed to be normally
distributed according to Equation A-4: if the sample is either from a large system (serving more
than 10,000) or is a PFHpA or PFHxS sample.
Equation A-4:
jijk ~ Norm(muijk,aWjk)
Where y represents sample results and j is a sample index and equal to 1 ? ... ^samp-, where Wsamp is
the total number of samples . Here i is the indicator for the system at which the sample y,-yfc was
collected and k is an indicator for the contaminant that Jijkis a sample of (i.e., PFOS, PFOA,
PFHpA, or PFHxS). Thus, yijk represents the /h sample of contaminant k collected from system
i. muik represents the kth element of muL shown in Equation A-3, % is a vector of length 4
providing the within-system standard deviation for each chemical included in the model. Thus
ow k represents the kth element of ow.
Within-system standard deviations specific to small systems were fit for PFOS and PFOA. awsm
replaces ow in Equation A-4 when the sample is either PFOS or PFOA collected at a small (sm)
system. Model variants that included within-system standard deviations specific to small systems
for all 4 chemicals as well as no within-system standard deviations specific to small systems
were both included in the cross-validation model comparison, but both were outperformed by the
model presented here. The limited reported values of PFHxS and PFHpA at small systems
relative to PFOS and PFOA made the fitting of within-system standard deviations specific to
small systems highly uncertain for these chemicals and adversely affected the model's predictive
performance. Because of this, EPA used within-system standard deviations pooled across both
system size categories for PFHxS and PFHpA.
Proposed PFAS Rule Economic Analysis
A-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
A3 Model Implementation
EPA conducted the data import, model setup, and assessment of model output using the R
programming language and the RStudio IDE (R Core Team, 2021; RStudio Team, 2020). The
Agency used Rstan to access the Stan probabilistic programming language and execute the
model (Stan Development Team, 2020; Stan Development Team, 2021). The R packages
reshape2 and dplyr were used for data handling (Wickham, 2007; Wickham et al., 2020). The R
packages bayesplot, ggplot, and ggpubr were used for data visualization (Gabry et al., 2020;
Kassambara, 2020; Wickham, 2016).
Stan uses Hamiltonian Monte Carlo No-U-Turn-Sampling for Markov chain Monte Carlo. EPA
ran models with 4 chains of 5,000 iterations, 2,000 of which were warmup, thinned by 3.
Thinning was used to balance memory limitations with desired effective sample size. Additional
sampler parameters included: adapt delta = 0.95, max treedepth = 12, and seed = 1337. EPA
used Shinystan (Gabry et al., 2018) to confirm that the effective sample size exceeded 1,000 for
all parameters that were not predefined values, such as the diagonal of a correlation matrix,
which is 1 by definition. EPA also used Shinystan to confirm chain mixing. No divergent
samples were observed.
For samples that were reported values (i.e., observed), the log probability was incremented using
the log of the normal density for the reported value given the system-level mean and within-
system deviation. For samples reporting the result as below the reporting limit rather than an
observed value, the log probability was incremented as the log of the cumulative normal
distribution at the reporting limit given the system-level mean and within-system standard
deviation.
EPA optimized the model via non-centered parameterization and Cholesky factorization of the
multivariate normal distribution. Additional information on handling of samples below a
reporting limit and model reparameterization are available in the Stan User's Guide sections on
"Censored data" and "Reparameterization", respectively (Stan Development Team, 2021). EPA
used weakly informative prior distributions. Prior distributions serve as a way to reflect
probabilistic beliefs for model parameters prior to seeing data. The decision to use weakly
informative priors allowed for the improvement of computational efficiency by providing loose
guidance towards sensical values for model parameters without influencing posterior
distributions in any substantive matter.
Proposed PFAS Rule Economic Analysis
A-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix B. Affected Population
This appendix describes the data sources used to evaluate the population potentially affected by
human health risk reductions due to reductions in drinking water exposure to per- and
polyfluoroalkyl substances (PFAS). Table B-l describes the data elements used to assess the
affected population in EPA's analysis of the benefits of reducing PFAS levels in drinking water.
These elements include the Safe Drinking Water Information System (SDWIS) 2021 quarter 4
(Q4) dataset (U.S. EPA, 2021b), and U.S. Census Bureau (2020).
The EPA SDWIS dataset provides information reported by states on drinking water systems, as
required by the Safe Drinking Water Act. The dataset generally includes information on system
name, identification number (public water system [PWS] ID), the cities or counties served, the
number of people served, the type of system (community, transient, or non-transient), whether
the system operates year-round or seasonally, and characteristics of the system's source water.
The U.S. Census provides detailed county-level population data by 5-year age-range, sex, race,
and ethnicity from 2010 to 2019. EPA first calculated, for each county, the average population
for each age-range/sex/race/ethnicity cohort over this 10-year period to determine a "typical-
year" demographic distribution for each county. EPA then calculated the proportion of each
county's population in each age-range/sex/race/ethnicity cohort in each of the 10-years. Finally,
EPA estimated the proportion of each county's population in each age/sex/race/ethnicity cohorts
by equally distributing the population in each 5-year age-range equally over the five years.
To determine the population proportions for each PWS, EPA took the following steps:
1. For PWSs for which EPA had information on the boundary of the PWS service area (see
Chapter 9):
a. Calculate the population-weighted proportion of the PWS's service area in each
county.
b. Use the values from (a) as weights, along with the county-level age-specific
sex/race/ethnicity population cohort data, to estimate the PWS's population served in
each age/sex/race/ethnicity cohort.
2. For PWSs for which EPA did not have information on the boundary of the PWS service area:
a. Developed a crosswalk between the primary SDWIS county name and the county
Federal Information Processing Standards (FIPS) codes used by the US. Census.
b. Used the PWS primary county age/sex/race/ethnicity population cohort data to
determine the PWS's population served in each age/sex/race/ethnicity cohort.
3. For PWSs for which EPA did not have information on the boundary or the primary county:
a. Used national age/sex/race/ethnicity population cohort data to determine the
PWS's population served in each age/sex/race/ethnicity cohort.
Proposed PFAS Rule Economic Analysis
B-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table B-l: Summary of Inputs and Data Sources Used to Estimate Affected Population
Data Element Modeled Variability Data Source
Notes
Initial Total
Population
Percentage of
Population in a
Demographic
Population
Subgroup
Location: PWS
Age: integer ages 0-84,
85+
Sex: males, females
Race/Ethnicity: non-
Hispanic White, non-
Hispanic Black,
Hispanic, other
Location: U.S. counties
SDWIS 2021
(U.S. EPA,
2021b)
U.S. Census
Bureau (2020):
Annual County
Resident
Population
Estimates by
Age,
Sex, Race, and
Hispanic Origin:
April 1, 2010 to
July 1,2019.
Public water system inventory from EPA's
SDWIS Q4 in 2021. EPA uses the SDWIS 2021
population data as the initial total population per
PWS.
The original data source contains total
population by race/ethnicity, sex, and 5-year age
groups.
Abbreviations: PWS - public water system; SDWIS - Safe Drinking Water Information System.
Proposed PFAS Rule Economic Analysis
B-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix C. Cost Analysis Results
This appendix provides additional cost output details. Section C.l provides PWS-level costs by
system type, primary source water, ownership, and system size category. Costs are provided for
all systems as well as for only those systems that must treat or change water source to comply
with the regulatory option. Section C.2 provides estimates of household costs.
C.l PWS-Level Cost Details
Section C.l provides PWS-level costs by system type, primary source water, ownership, and
system size category. Costs are provided for all systems as well as for only those systems that
must treat or change water source to comply with the regulatory option.
C.l.l Mean Annual Cost for all Community Water Systems
Table C-l: Mean Annualized Cost per CWSs, Proposed Option (PFOA and PFOS
MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Private
Ground
Less than 100
$904
$1,263
$1,680
Private
Ground
100 to 500
$1,357
$1,941
$2,598
Private
Ground
500 to 1,000
$1,838
$2,813
$3,928
Private
Ground
1,000 to 3,300
$2,975
$4,484
$6,261
Private
Ground
3,300 to 10,000
$5,981
$10,374
$15,395
Private
Ground
10,000 to 50,000
$67,442
$82,756
$98,919
Private
Ground
50,000 to 100,000
$114,170
$198,210
$297,240
Private
Ground
100,000 to 1,000,000
$252,900
$405,310
$626,760
Private
Surface
Less than 100
$1,040
$1,704
$2,533
Private
Surface
100 to 500
$1,552
$2,431
$3,473
Private
Surface
500 to 1,000
$1,762
$3,380
$5,380
Private
Surface
1,000 to 3,300
$2,511
$4,683
$7,573
Private
Surface
3,300 to 10,000
$5,265
$10,990
$18,306
Private
Surface
10,000 to 50,000
$65,305
$82,341
$100,160
Private
Surface
50,000 to 100,000
$106,190
$153,630
$203,000
Private
Surface
100,000 to 1,000,000
$1,442,200
$1,684,800
$1,921,000
Public
Ground
Less than 100
$892
$1,291
$1,752
Public
Ground
100 to 500
$1,412
$2,055
$2,787
Public
Ground
500 to 1,000
$1,993
$2,853
$3,922
Public
Ground
1,000 to 3,300
$3,408
$4,995
$6,879
Public
Ground
3,300 to 10,000
$8,334
$12,306
$16,911
Public
Ground
10,000 to 50,000
$81,937
$90,066
$98,714
Public
Ground
50,000 to 100,000
$154,560
$194,310
$237,500
Public
Ground
100,000 to 1,000,000
$628,370
$781,090
$960,360
Public
Surface
Less than 100
$1,103
$1,838
$2,810
Public
Surface
100 to 500
$1,757
$2,630
$3,663
Public
Surface
500 to 1,000
$2,247
$3,533
$5,140
Proposed PFAS Rule Economic Analysis
C-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-l: Mean Annualized Cost per CWSs, Proposed Option (PFOA and PFOS
MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Public
Surface
1,000 to 3,300
$3,524
$5,566
$7,901
Public
Surface
3,300 to 10,000
$8,909
$13,241
$18,217
Public
Surface
10,000 to 50,000
$78,401
$85,772
$93,705
Public
Surface
50,000 to 100,000
$122,530
$143,390
$165,420
Public
Surface
100,000 to 1,000,000
$533,340
$617,920
$707,580
Abbreviations: CWS - Community Water System.
Table C-2: Mean Annualized Cost per CWSs, Option la (PFOA and PFOS MCLs of 4
ppt) (Commercial Cost of Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$888
$1,261
$1,680
Private
Ground
100 to 500
$1,334
$1,934
$2,585
Private
Ground
500 to 1,000
$1,854
$2,796
$3,868
Private
Ground
1,000 to 3,300
$2,953
$4,450
$6,234
Private
Ground
3,300 to 10,000
$5,979
$10,261
$15,410
Private
Ground
10,000 to 50,000
$65,310
$79,729
$95,342
Private
Ground
50,000 to 100,000
$91,548
$172,320
$268,120
Private
Ground
100,000 to 1,000,000
$246,080
$389,920
$603,790
Private
Surface
Less than 100
$1,047
$1,701
$2,534
Private
Surface
100 to 500
$1,547
$2,425
$3,447
Private
Surface
500 to 1,000
$1,712
$3,365
$5,385
Private
Surface
1,000 to 3,300
$2,425
$4,660
$7,553
Private
Surface
3,300 to 10,000
$5,238
$10,883
$17,388
Private
Surface
10,000 to 50,000
$64,728
$80,458
$98,766
Private
Surface
50,000 to 100,000
$104,600
$150,890
$202,940
Private
Surface
100,000 to 1,000,000
$1,390,600
$1,623,400
$1,861,700
Public
Ground
Less than 100
$871
$1,289
$1,808
Public
Ground
100 to 500
$1,411
$2,049
$2,781
Public
Ground
500 to 1,000
$1,921
$2,840
$3,933
Public
Ground
1,000 to 3,300
$3,361
$4,959
$7,023
Public
Ground
3,300 to 10,000
$8,276
$12,164
$16,853
Public
Ground
10,000 to 50,000
$80,197
$87,864
$96,239
Public
Ground
50,000 to 100,000
$151,640
$189,130
$231,820
Public
Ground
100,000 to 1,000,000
$580,840
$731,850
$904,440
Public
Surface
Less than 100
$1,087
$1,834
$2,802
Public
Surface
100 to 500
$1,754
$2,622
$3,663
Public
Surface
500 to 1,000
$2,243
$3,518
$5,136
Public
Surface
1,000 to 3,300
$3,509
$5,541
$7,863
Proposed PFAS Rule Economic Analysis
C-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-2: Mean Annualized Cost per CWSs, Option la (PFOA and PFOS MCLs of 4
ppt) (Commercial Cost of Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water Category
Public Surface 3,300 to 10,000 $8,879 $13,154 $18,087
Public Surface 10,000 to 50,000 $76,668 $84,314 $92,037
Public Surface 50,000 to 100,000 $120,050 $140,270 $161,820
Public Surface 100,000 to 1,000,000 $513,240 $598,820 $688,670
Abbreviations: CWS - Community Water System.
Table C-3: Mean Annualized Cost per CWSs, Option lb (PFOA and PFOS MCLs of
5.0 ppt) (Commercial Cost of Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$730
$1,016
$1,383
Private
Ground
100 to 500
$1,080
$1,544
$2,133
Private
Ground
500 to 1,000
$1,476
$2,205
$3,136
Private
Ground
1,000 to 3,300
$2,256
$3,471
$5,044
Private
Ground
3,300 to 10,000
$4,489
$7,926
$12,195
Private
Ground
10,000 to 50,000
$51,098
$62,760
$76,022
Private
Ground
50,000 to 100,000
$68,449
$131,150
$212,960
Private
Ground
100,000 to 1,000,000
$172,010
$285,530
$450,550
Private
Surface
Less than 100
$823
$1,375
$2,120
Private
Surface
100 to 500
$1,229
$1,941
$2,812
Private
Surface
500 to 1,000
$1,309
$2,658
$4,433
Private
Surface
1,000 to 3,300
$1,693
$3,605
$6,021
Private
Surface
3,300 to 10,000
$3,613
$8,317
$13,926
Private
Surface
10,000 to 50,000
$48,955
$63,164
$78,915
Private
Surface
50,000 to 100,000
$82,435
$120,590
$163,410
Private
Surface
100,000 to 1,000,000
$1,176,200
$1,387,000
$1,610,600
Public
Ground
Less than 100
$712
$1,038
$1,413
Public
Ground
100 to 500
$1,127
$1,627
$2,254
Public
Ground
500 to 1,000
$1,507
$2,232
$3,098
Public
Ground
1,000 to 3,300
$2,532
$3,838
$5,403
Public
Ground
3,300 to 10,000
$6,179
$9,366
$13,160
Public
Ground
10,000 to 50,000
$64,417
$71,078
$77,915
Public
Ground
50,000 to 100,000
$119,530
$150,760
$185,190
Public
Ground
100,000 to 1,000,000
$471,660
$599,690
$742,700
Public
Surface
Less than 100
$898
$1,490
$2,290
Public
Surface
100 to 500
$1,364
$2,075
$2,962
Public
Surface
500 to 1,000
$1,676
$2,747
$3,995
Public
Surface
1,000 to 3,300
$2,680
$4,252
$6,096
Proposed PFAS Rule Economic Analysis
C-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-3: Mean Annualized Cost per CWSs, Option lb (PFOA and PFOS MCLs of
5.0 ppt) (Commercial Cost of Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water Category
Public Surface 3,300 to 10,000 $6,721 $10,102 $14,011
Public Surface 10,000 to 50,000 $60,686 $66,147 $72,448
Public Surface 50,000 to 100,000 $88,688 $106,470 $125,280
Public Surface 100,000 to 1,000,000 $405,400 $471,320 $541,870
Abbreviations: CWS - Community Water System.
Table C-4: Mean Annualized Cost per CWSs, Option lc (PFOA and PFOS MCLs of
10.0 ppt) (Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Private
Ground
Less than 100
$393
$528
$702
Private
Ground
100 to 500
$558
$776
$1,051
Private
Ground
500 to 1,000
$691
$1,059
$1,519
Private
Ground
1,000 to 3,300
$977
$1,581
$2,389
Private
Ground
3,300 to 10,000
$1,722
$3,538
$5,974
Private
Ground
10,000 to 50,000
$19,811
$26,317
$33,672
Private
Ground
50,000 to 100,000
$19,453
$45,545
$87,450
Private
Ground
100,000 to 1,000,000
$37,601
$82,560
$146,030
Private
Surface
Less than 100
$460
$728
$1,095
Private
Surface
100 to 500
$637
$987
$1,450
Private
Surface
500 to 1,000
$639
$1,287
$2,300
Private
Surface
1,000 to 3,300
$697
$1,618
$2,943
Private
Surface
3,300 to 10,000
$1,171
$3,541
$6,806
Private
Surface
10,000 to 50,000
$19,238
$26,682
$35,690
Private
Surface
50,000 to 100,000
$35,527
$57,657
$82,458
Private
Surface
100,000 to 1,000,000
$532,810
$692,740
$855,930
Public
Ground
Less than 100
$371
$541
$754
Public
Ground
100 to 500
$565
$799
$1,103
Public
Ground
500 to 1,000
$707
$1,042
$1,449
Public
Ground
1,000 to 3,300
$1,098
$1,683
$2,418
Public
Ground
3,300 to 10,000
$2,633
$4,038
$5,760
Public
Ground
10,000 to 50,000
$29,654
$33,106
$36,905
Public
Ground
50,000 to 100,000
$52,779
$70,554
$90,600
Public
Ground
100,000 to 1,000,000
$200,950
$270,410
$350,070
Public
Surface
Less than 100
$478
$756
$1,224
Public
Surface
100 to 500
$696
$1,012
$1,435
Public
Surface
500 to 1,000
$767
$1,261
$1,948
Public
Surface
1,000 to 3,300
$1,086
$1,764
$2,620
Proposed PFAS Rule Economic Analysis
C-4
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-4: Mean Annualized Cost per CWSs, Option lc (PFOA and PFOS MCLs of
10.0 ppt) (Commercial Cost of Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water Category
Public Surface 3,300 to 10,000 $2,616 $4,135 $5,953
Public Surface 10,000 to 50,000 $25,209 $28,007 $30,854
Public Surface 50,000 to 100,000 $30,672 $38,542 $47,165
Public Surface 100,000 to 1,000,000 $163,970 $199,200 $238,560
Abbreviations: CWS - Community Water System.
C.1.2 Mean Annual Cost for all Non-Transient Non-Community
Water Systems
Table C-5: Mean Annualized Cost per NTNCWS, Proposed Option (PFOA and PFOS
MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Private
Ground
Less than 100
$1,023
$1,421
$1,918
Private
Ground
100 to 500
$1,334
$1,943
$2,642
Private
Ground
500 to 1,000
$1,543
$2,487
$3,614
Private
Ground
1,000 to 3,300
$2,251
$3,888
$5,962
Private
Ground
3,300 to 10,000
$1,919
$8,285
$16,229
Private
Ground
10,000 to 50,000
$458
$39,962
$226,000
Private
Surface
Less than 100
$954
$1,812
$2,946
Private
Surface
100 to 500
$1,378
$2,665
$4,499
Private
Surface
500 to 1,000
$1,236
$3,827
$7,903
Private
Surface
1,000 to 3,300
$1,769
$5,369
$10,507
Private
Surface
3,300 to 10,000
$3,029
$15,334
$33,118
Private
Surface
10,000 to 50,000
$8,885
$70,753
$154,530
Private
Surface
100,000 to 1,000,000
$902
$240,660
$2,044,000
Public
Ground
Less than 100
$975
$1,411
$1,920
Public
Ground
100 to 500
$1,320
$1,959
$2,676
Public
Ground
500 to 1,000
$1,489
$2,383
$3,422
Public
Ground
1,000 to 3,300
$2,272
$3,973
$6,106
Public
Ground
3,300 to 10,000
$916
$9,171
$22,933
Public
Ground
10,000 to 50,000
$39,264
$122,780
$233,030
Public
Surface
Less than 100
$746
$1,853
$3,638
Public
Surface
100 to 500
$1,004
$2,674
$5,136
Public
Surface
500 to 1,000
$647
$3,685
$9,165
Public
Surface
1,000 to 3,300
$1,261
$6,725
$14,525
Proposed PFAS Rule Economic Analysis
C-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-5: Mean Annualized Cost per NTNCWS, Proposed Option (PFOA and PFOS
MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Public
Surface
3,300 to 10,000
$1,247
$13,484
$35,246
Public
Surface
10,000 to 50,000
$1,140
$65,055
$194,550
Public
Surface
50,000 to 100,000
$591
$83,260
$506,000
Abbreviations: NTNCWS - Non-Transient, Non-Community Water Systems.
Table C-6: Mean Annualized Cost per NTNCWS, Option la (PFOA and PFOS MCLs
of 4 ppt) (Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Private
Ground
Less than 100
$1,018
$1,417
$1,917
Private
Ground
100 to 500
$1,347
$1,936
$2,639
Private
Ground
500 to 1,000
$1,594
$2,476
$3,614
Private
Ground
1,000 to 3,300
$2,251
$3,864
$5,962
Private
Ground
3,300 to 10,000
$1,919
$8,228
$16,227
Private
Ground
10,000 to 50,000
$458
$39,525
$226,580
Private
Surface
Less than 100
$953
$1,808
$2,946
Private
Surface
100 to 500
$1,286
$2,656
$4,499
Private
Surface
500 to 1,000
$1,236
$3,804
$7,361
Private
Surface
1,000 to 3,300
$1,597
$5,325
$10,471
Private
Surface
3,300 to 10,000
$2,699
$15,070
$31,326
Private
Surface
10,000 to 50,000
$8,885
$68,997
$148,280
Private
Surface
100,000 to 1,000,000
$902
$237,510
$2,044,000
Public
Ground
Less than 100
$972
$1,407
$1,920
Public
Ground
100 to 500
$1,328
$1,953
$2,676
Public
Ground
500 to 1,000
$1,473
$2,374
$3,381
Public
Ground
1,000 to 3,300
$2,090
$3,950
$6,105
Public
Ground
3,300 to 10,000
$916
$9,118
$22,619
Public
Ground
10,000 to 50,000
$36,094
$119,980
$223,080
Public
Surface
Less than 100
$746
$1,849
$3,732
Public
Surface
100 to 500
$965
$2,665
$5,136
Public
Surface
500 to 1,000
$647
$3,670
$9,709
Public
Surface
1,000 to 3,300
$1,257
$6,676
$14,525
Public
Surface
3,300 to 10,000
$1,183
$13,313
$33,960
Public
Surface
10,000 to 50,000
$1,114
$64,144
$193,970
Public
Surface
50,000 to 100,000
$591
$81,895
$506,000
Abbreviations: NTNCWS - Non-Transient, Non-Community Water Systems.
Proposed PFAS Rule Economic Analysis
C-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-7: Mean Annualized Cost per NTNCWS, Option lb (PFOA and PFOS MCLs
of 5.0 ppt) (Commercial Cost of Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$812
$1,143
$1,549
Private
Ground
100 to 500
$1,067
$1,542
$2,097
Private
Ground
500 to 1,000
$1,195
$1,951
$2,811
Private
Ground
1,000 to 3,300
$1,681
$3,008
$4,701
Private
Ground
3,300 to 10,000
$1,129
$6,329
$13,742
Private
Ground
10,000 to 50,000
$458
$30,335
$225,300
Private
Surface
Less than 100
$771
$1,457
$2,477
Private
Surface
100 to 500
$987
$2,118
$3,665
Private
Surface
500 to 1,000
$879
$3,001
$6,298
Private
Surface
1,000 to 3,300
$1,125
$4,173
$8,294
Private
Surface
3,300 to 10,000
$2,055
$11,733
$28,079
Private
Surface
10,000 to 50,000
$6,347
$55,661
$135,260
Private
Surface
100,000 to 1,000,000
$902
$164,110
$1,051,400
Public
Ground
Less than 100
$758
$1,134
$1,598
Public
Ground
100 to 500
$1,049
$1,550
$2,136
Public
Ground
500 to 1,000
$1,154
$1,861
$2,672
Public
Ground
1,000 to 3,300
$1,570
$3,060
$4,809
Public
Ground
3,300 to 10,000
$761
$6,925
$18,498
Public
Ground
10,000 to 50,000
$32,771
$99,358
$203,830
Public
Surface
Less than 100
$588
$1,491
$3,104
Public
Surface
100 to 500
$775
$2,108
$4,245
Public
Surface
500 to 1,000
$556
$2,867
$7,962
Public
Surface
1,000 to 3,300
$918
$5,149
$11,711
Public
Surface
3,300 to 10,000
$1,048
$10,131
$28,423
Public
Surface
10,000 to 50,000
$879
$47,923
$150,450
Public
Surface
50,000 to 100,000
$591
$58,768
$489,030
Abbreviations: NTNCWS - Non-Transient, Non-Community Water Systems.
Proposed PFAS Rule Economic Analysis
C-7
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-8: Mean Annualized Cost per NTNCWS, Option lc (PFOA and PFOS MCLs
of 10.0 ppt) (Commercial Cost of Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$435
$595
$793
Private
Ground
100 to 500
$535
$768
$1,035
Private
Ground
500 to 1,000
$560
$911
$1,396
Private
Ground
1,000 to 3,300
$686
$1,350
$2,264
Private
Ground
3,300 to 10,000
$751
$2,775
$7,244
Private
Ground
10,000 to 50,000
$458
$12,056
$82,018
Private
Surface
Less than 100
$452
$778
$1,396
Private
Surface
100 to 500
$569
$1,080
$2,093
Private
Surface
500 to 1,000
$565
$1,436
$3,479
Private
Surface
1,000 to 3,300
$647
$1,863
$4,444
Private
Surface
3,300 to 10,000
$1,453
$5,342
$14,663
Private
Surface
10,000 to 50,000
$3,279
$26,457
$78,374
Private
Surface
100,000 to 1,000,000
$902
$36,179
$6,586
Public
Ground
Less than 100
$410
$592
$821
Public
Ground
100 to 500
$515
$761
$1,087
Public
Ground
500 to 1,000
$525
$860
$1,321
Public
Ground
1,000 to 3,300
$639
$1,309
$2,218
Public
Ground
3,300 to 10,000
$612
$2,824
$9,931
Public
Ground
10,000 to 50,000
$2,556
$46,452
$110,060
Public
Surface
Less than 100
$386
$763
$1,935
Public
Surface
100 to 500
$491
$1,057
$2,415
Public
Surface
500 to 1,000
$449
$1,349
$3,768
Public
Surface
1,000 to 3,300
$588
$2,228
$6,340
Public
Surface
3,300 to 10,000
$840
$4,285
$17,050
Public
Surface
10,000 to 50,000
$828
$17,287
$86,559
Public
Surface
50,000 to 100,000
$591
$11,283
$6,000
Abbreviations: NTNCWS - Non-Transient, Non-Community Water Systems.
Proposed PFAS Rule Economic Analysis
C-8
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.1.3 Mean Annual Cost for Community Water Systems that
Treat or Change Water Source
Table C-9: Mean Annualized Cost per CWSs that Treat or Change Water Source,
Proposed Option (PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost
of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Private
Ground
Less than 100
$13,793
$15,375
$17,165
Private
Ground
100 to 500
$22,643
$25,014
$28,091
Private
Ground
500 to 1,000
$30,327
$35,471
$41,384
Private
Ground
1,000 to 3,300
$47,843
$56,147
$64,536
Private
Ground
3,300 to 10,000
$98,144
$122,540
$149,250
Private
Ground
10,000 to 50,000
$242,260
$283,750
$327,210
Private
Ground
50,000 to 100,000
$453,640
$642,690
$873,600
Private
Ground
100,000 to 1,000,000
$618,460
$903,530
$1,317,400
Private
Surface
Less than 100
$13,825
$22,440
$33,672
Private
Surface
100 to 500
$25,990
$34,414
$44,293
Private
Surface
500 to 1,000
$31,117
$49,160
$70,681
Private
Surface
1,000 to 3,300
$51,733
$71,512
$94,466
Private
Surface
3,300 to 10,000
$99,840
$143,310
$189,180
Private
Surface
10,000 to 50,000
$320,010
$376,680
$441,640
Private
Surface
50,000 to 100,000
$455,230
$580,310
$722,250
Private
Surface
100,000 to 1,000,000
$3,070,400
$3,677,400
$4,323,600
Public
Ground
Less than 100
$12,509
$15,804
$19,872
Public
Ground
100 to 500
$24,659
$27,579
$30,985
Public
Ground
500 to 1,000
$33,895
$37,684
$42,053
Public
Ground
1,000 to 3,300
$60,164
$65,292
$71,266
Public
Ground
3,300 to 10,000
$124,420
$138,050
$153,510
Public
Ground
10,000 to 50,000
$310,720
$332,940
$359,380
Public
Ground
50,000 to 100,000
$572,190
$666,560
$767,330
Public
Ground
100,000 to 1,000,000
$1,839,100
$2,243,500
$2,696,400
Public
Surface
Less than 100
$14,608
$23,242
$34,898
Public
Surface
100 to 500
$31,258
$38,394
$46,282
Public
Surface
500 to 1,000
$44,745
$54,166
$64,710
Public
Surface
1,000 to 3,300
$80,587
$89,394
$100,290
Public
Surface
3,300 to 10,000
$174,280
$192,360
$211,670
Public
Surface
10,000 to 50,000
$390,180
$413,650
$440,960
Public
Surface
50,000 to 100,000
$558,040
$617,230
$680,550
Public
Surface
100,000 to 1,000,000
$1,955,300
$2,180,000
$2,409,300
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-9
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-10: Mean Annualized Cost per CWSs that Treat or Change Water Source,
Option la (PFOA and PFOS MCLs of 4 ppt) (Commercial Cost of Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$13,797
$15,360
$17,165
Private
Ground
100 to 500
$22,620
$24,958
$27,881
Private
Ground
500 to 1,000
$30,091
$35,340
$41,144
Private
Ground
1,000 to 3,300
$47,736
$55,900
$64,492
Private
Ground
3,300 to 10,000
$95,732
$121,890
$150,170
Private
Ground
10,000 to 50,000
$234,880
$273,690
$318,220
Private
Ground
50,000 to 100,000
$380,760
$554,250
$763,320
Private
Ground
100,000 to 1,000,000
$602,820
$873,550
$1,260,600
Private
Surface
Less than 100
$13,825
$22,417
$32,893
Private
Surface
100 to 500
$25,929
$34,364
$44,729
Private
Surface
500 to 1,000
$31,760
$49,023
$68,984
Private
Surface
1,000 to 3,300
$51,733
$71,315
$94,466
Private
Surface
3,300 to 10,000
$100,280
$142,600
$187,850
Private
Surface
10,000 to 50,000
$312,310
$368,670
$432,680
Private
Surface
50,000 to 100,000
$446,710
$571,210
$718,700
Private
Surface
100,000 to 1,000,000
$2,945,100
$3,548,000
$4,185,100
Public
Ground
Less than 100
$12,508
$15,786
$19,872
Public
Ground
100 to 500
$24,647
$27,529
$30,749
Public
Ground
500 to 1,000
$33,905
$37,582
$41,824
Public
Ground
1,000 to 3,300
$59,904
$65,018
$71,258
Public
Ground
3,300 to 10,000
$123,730
$137,210
$152,030
Public
Ground
10,000 to 50,000
$303,220
$325,520
$351,740
Public
Ground
50,000 to 100,000
$560,950
$650,410
$750,540
Public
Ground
100,000 to 1,000,000
$1,709,000
$2,109,700
$2,556,100
Public
Surface
Less than 100
$14,707
$23,217
$34,389
Public
Surface
100 to 500
$31,284
$38,328
$46,405
Public
Surface
500 to 1,000
$45,319
$54,047
$64,642
Public
Surface
1,000 to 3,300
$80,343
$89,200
$99,333
Public
Surface
3,300 to 10,000
$174,620
$191,760
$211,100
Public
Surface
10,000 to 50,000
$384,370
$407,370
$432,800
Public
Surface
50,000 to 100,000
$549,840
$605,030
$661,580
Public
Surface
100,000 to 1,000,000
$1,894,700
$2,117,600
$2,346,500
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-10
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-ll: Mean Annualized Cost per CWSs that Treat or Change Water Source,
Option lb (PFOA and PFOS MCLs of 5.0 ppt) (Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Private
Ground
Less than 100
$13,628
$15,348
$17,285
Private
Ground
100 to 500
$22,264
$24,902
$28,287
Private
Ground
500 to 1,000
$29,179
$34,987
$41,682
Private
Ground
1,000 to 3,300
$46,419
$55,271
$65,374
Private
Ground
3,300 to 10,000
$92,947
$119,610
$151,040
Private
Ground
10,000 to 50,000
$212,950
$253,590
$298,520
Private
Ground
50,000 to 100,000
$322,410
$483,190
$682,510
Private
Ground
100,000 to 1,000,000
$481,610
$720,550
$1,054,200
Private
Surface
Less than 100
$13,299
$22,383
$35,730
Private
Surface
100 to 500
$24,251
$34,417
$46,479
Private
Surface
500 to 1,000
$29,347
$48,679
$73,251
Private
Surface
1,000 to 3,300
$48,169
$70,702
$96,783
Private
Surface
3,300 to 10,000
$92,104
$140,100
$195,870
Private
Surface
10,000 to 50,000
$296,000
$354,490
$422,720
Private
Surface
50,000 to 100,000
$422,310
$569,950
$744,160
Private
Surface
100,000 to 1,000,000
$2,692,900
$3,294,000
$3,963,200
Public
Ground
Less than 100
$11,802
$15,778
$20,539
Public
Ground
100 to 500
$24,239
$27,445
$30,997
Public
Ground
500 to 1,000
$33,307
$37,342
$42,096
Public
Ground
1,000 to 3,300
$58,301
$64,247
$70,591
Public
Ground
3,300 to 10,000
$119,090
$133,890
$149,780
Public
Ground
10,000 to 50,000
$291,670
$313,770
$341,640
Public
Ground
50,000 to 100,000
$528,450
$626,110
$737,620
Public
Ground
100,000 to 1,000,000
$1,578,200
$1,986,000
$2,443,400
Public
Surface
Less than 100
$13,946
$23,116
$35,821
Public
Surface
100 to 500
$30,195
$38,103
$47,319
Public
Surface
500 to 1,000
$43,413
$53,838
$65,474
Public
Surface
1,000 to 3,300
$77,609
$88,477
$100,280
Public
Surface
3,300 to 10,000
$171,080
$190,810
$211,650
Public
Surface
10,000 to 50,000
$371,540
$395,540
$422,410
Public
Surface
50,000 to 100,000
$513,520
$569,380
$632,290
Public
Surface
100,000 to 1,000,000
$1,758,500
$1,984,400
$2,231,100
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-ll
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-12: Mean Annualized Cost per CWSs that Treat or Change Water Source,
Option lc (PFOA and PFOS MCLs of 10.0 ppt) (Commercial Cost of Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$12,623
$15,266
$18,265
Private
Ground
100 to 500
$20,928
$24,867
$29,633
Private
Ground
500 to 1,000
$25,195
$34,328
$45,727
Private
Ground
1,000 to 3,300
$39,923
$53,253
$69,516
Private
Ground
3,300 to 10,000
$67,601
$111,670
$160,040
Private
Ground
10,000 to 50,000
$149,440
$190,110
$236,170
Private
Ground
50,000 to 100,000
$138,770
$250,590
$413,130
Private
Ground
100,000 to 1,000,000
$181,810
$335,330
$556,430
Private
Surface
Less than 100
$11,312
$21,631
$48,766
Private
Surface
100 to 500
$19,433
$34,132
$54,990
Private
Surface
500 to 1,000
$0
$45,387
$87,847
Private
Surface
1,000 to 3,300
$31,944
$67,316
$119,200
Private
Surface
3,300 to 10,000
$0
$126,060
$232,900
Private
Surface
10,000 to 50,000
$219,550
$286,960
$371,720
Private
Surface
50,000 to 100,000
$357,660
$546,630
$798,650
Private
Surface
100,000 to 1,000,000
$1,811,000
$2,375,100
$3,006,700
Public
Ground
Less than 100
$9,959
$15,815
$24,391
Public
Ground
100 to 500
$22,245
$27,144
$32,751
Public
Ground
500 to 1,000
$29,548
$36,349
$43,505
Public
Ground
1,000 to 3,300
$54,295
$61,836
$70,426
Public
Ground
3,300 to 10,000
$104,700
$125,700
$148,080
Public
Ground
10,000 to 50,000
$248,310
$274,750
$301,260
Public
Ground
50,000 to 100,000
$428,340
$557,010
$696,310
Public
Ground
100,000 to 1,000,000
$1,272,400
$1,737,100
$2,343,700
Public
Surface
Less than 100
$11,050
$21,773
$47,885
Public
Surface
100 to 500
$25,066
$37,813
$55,612
Public
Surface
500 to 1,000
$34,959
$53,196
$74,908
Public
Surface
1,000 to 3,300
$69,223
$86,744
$106,090
Public
Surface
3,300 to 10,000
$157,500
$186,670
$220,730
Public
Surface
10,000 to 50,000
$338,060
$367,130
$399,760
Public
Surface
50,000 to 100,000
$391,500
$461,120
$536,130
Public
Surface
100,000 to 1,000,000
$1,406,600
$1,610,100
$1,842,900
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-12
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.1.4 Mean Annual Cost for Non-Transient Non-Community
Water Systems that Treat or Change Water Source
Table C-13: Mean Annualized Cost per NTNCWSs that Treat or Change Water
Source, Proposed Option (PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0)
(Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Private
Ground
Less than 100
$14,944
$16,802
$19,106
Private
Ground
100 to 500
$22,289
$25,430
$28,973
Private
Ground
500 to 1,000
$26,994
$33,645
$41,449
Private
Ground
1,000 to 3,300
$41,545
$53,745
$68,213
Private
Ground
3,300 to 10,000
$40,954
$116,000
$182,440
Private
Ground
10,000 to 50,000
$0
$73,652
$450,020
Private
Surface
Less than 100
$12,304
$22,749
$38,986
Private
Surface
100 to 500
$19,960
$35,121
$58,014
Private
Surface
500 to 1,000
$23,099
$48,140
$91,951
Private
Surface
1,000 to 3,300
$35,604
$72,650
$126,920
Private
Surface
3,300 to 10,000
$0
$146,250
$287,490
Private
Surface
10,000 to 50,000
$48,432
$251,710
$501,260
Private
Surface
100,000 to 1,000,000
$0
$238,530
$2,044,000
Public
Ground
Less than 100
$13,147
$16,806
$20,754
Public
Ground
100 to 500
$23,273
$26,988
$31,412
Public
Ground
500 to 1,000
$29,074
$34,958
$41,922
Public
Ground
1,000 to 3,300
$46,306
$59,460
$74,131
Public
Ground
3,300 to 10,000
$0
$115,160
$236,120
Public
Ground
10,000 to 50,000
$201,020
$386,810
$638,040
Public
Surface
Less than 100
$11,006
$22,625
$54,345
Public
Surface
100 to 500
$19,067
$36,564
$70,084
Public
Surface
500 to 1,000
$0
$40,721
$123,550
Public
Surface
1,000 to 3,300
$0
$87,328
$162,990
Public
Surface
3,300 to 10,000
$0
$137,840
$297,070
Public
Surface
10,000 to 50,000
$0
$325,460
$827,950
Public
Surface
50,000 to 100,000
$0
$81,787
$506,000
Abbreviations: NTNCWS - Non-Transient, Non-Community Water Systems.
Proposed PFAS Rule Economic Analysis
C-13
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-14: Mean Annualized Cost per NTNCWSs that Treat or Change Water
Source, Option la (PFOA and PFOS MCLs of 4 ppt) (Commercial Cost of Capital,
$2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$14,755
$16,765
$19,090
Private
Ground
100 to 500
$22,386
$25,372
$29,041
Private
Ground
500 to 1,000
$26,642
$33,558
$41,519
Private
Ground
1,000 to 3,300
$41,261
$53,550
$68,193
Private
Ground
3,300 to 10,000
$38,732
$115,590
$183,890
Private
Ground
10,000 to 50,000
$0
$72,962
$450,020
Private
Surface
Less than 100
$12,212
$22,713
$38,986
Private
Surface
100 to 500
$20,047
$35,037
$58,014
Private
Surface
500 to 1,000
$22,725
$47,930
$91,951
Private
Surface
1,000 to 3,300
$35,604
$72,214
$120,710
Private
Surface
3,300 to 10,000
$0
$145,010
$283,220
Private
Surface
10,000 to 50,000
$46,976
$245,720
$497,570
Private
Surface
100,000 to 1,000,000
$0
$235,400
$2,044,000
Public
Ground
Less than 100
$13,147
$16,775
$20,754
Public
Ground
100 to 500
$23,291
$26,946
$31,412
Public
Ground
500 to 1,000
$28,669
$34,876
$42,260
Public
Ground
1,000 to 3,300
$45,327
$59,252
$74,131
Public
Ground
3,300 to 10,000
$0
$114,760
$236,120
Public
Ground
10,000 to 50,000
$202,700
$378,980
$638,040
Public
Surface
Less than 100
$11,006
$22,602
$54,345
Public
Surface
100 to 500
$18,746
$36,491
$70,084
Public
Surface
500 to 1,000
$0
$40,562
$118,960
Public
Surface
1,000 to 3,300
$0
$86,982
$163,570
Public
Surface
3,300 to 10,000
$0
$136,790
$294,960
Public
Surface
10,000 to 50,000
$0
$322,080
$869,420
Public
Surface
50,000 to 100,000
$0
$80,434
$506,000
Abbreviations: NTNCWS - Non-Transient, Non-Community Water Systems.
Proposed PFAS Rule Economic Analysis
C-14
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-15: Mean Annualized Cost per NTNCWSs that Treat or Change Water
Source, Option lb (PFOA and PFOS MCLs of 5.0 ppt) (Commercial Cost of Capital,
$2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$14,574
$16,753
$19,228
Private
Ground
100 to 500
$21,694
$25,274
$29,239
Private
Ground
500 to 1,000
$25,841
$33,159
$42,008
Private
Ground
1,000 to 3,300
$38,717
$53,029
$70,227
Private
Ground
3,300 to 10,000
$0
$107,710
$184,870
Private
Ground
10,000 to 50,000
$0
$56,633
$450,020
Private
Surface
Less than 100
$11,736
$22,512
$44,040
Private
Surface
100 to 500
$18,147
$34,672
$61,094
Private
Surface
500 to 1,000
$0
$45,520
$101,190
Private
Surface
1,000 to 3,300
$0
$69,878
$137,120
Private
Surface
3,300 to 10,000
$0
$134,110
$288,170
Private
Surface
10,000 to 50,000
$0
$224,800
$480,670
Private
Surface
100,000 to 1,000,000
$0
$162,070
$1,051,400
Public
Ground
Less than 100
$12,677
$16,754
$21,459
Public
Ground
100 to 500
$22,399
$26,878
$31,692
Public
Ground
500 to 1,000
$28,044
$34,737
$43,256
Public
Ground
1,000 to 3,300
$44,274
$58,753
$75,683
Public
Ground
3,300 to 10,000
$0
$100,990
$234,660
Public
Ground
10,000 to 50,000
$170,360
$360,500
$599,940
Public
Surface
Less than 100
$0
$21,337
$56,764
Public
Surface
100 to 500
$0
$34,996
$72,257
Public
Surface
500 to 1,000
$0
$34,438
$106,300
Public
Surface
1,000 to 3,300
$0
$79,410
$163,840
Public
Surface
3,300 to 10,000
$0
$117,080
$281,100
Public
Surface
10,000 to 50,000
$0
$264,010
$804,430
Public
Surface
50,000 to 100,000
$0
$57,368
$485,750
Abbreviations: NTNCWS - Non-Transient, Non-Community Water Systems.
Proposed PFAS Rule Economic Analysis
C-15
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-16: Mean Annualized Cost per NTNCWSs that Treat or Change Water
Source, Option lc (PFOA and PFOS MCLs of 10.0 ppt) (Commercial Cost of Capital,
$2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$13,443
$16,618
$20,349
Private
Ground
100 to 500
$19,735
$24,953
$31,471
Private
Ground
500 to 1,000
$22,088
$32,706
$48,603
Private
Ground
1,000 to 3,300
$29,695
$50,688
$80,990
Private
Ground
3,300 to 10,000
$0
$66,598
$178,520
Private
Ground
10,000 to 50,000
$0
$22,172
$163,500
Private
Surface
Less than 100
$0
$19,528
$55,631
Private
Surface
100 to 500
$0
$29,873
$81,076
Private
Surface
500 to 1,000
$0
$28,629
$106,660
Private
Surface
1,000 to 3,300
$0
$46,594
$150,160
Private
Surface
3,300 to 10,000
$0
$84,427
$298,160
Private
Surface
10,000 to 50,000
$0
$150,420
$439,000
Private
Surface
100,000 to 1,000,000
$0
$34,439
$0
Public
Ground
Less than 100
$10,234
$16,663
$25,514
Public
Ground
100 to 500
$19,918
$26,542
$35,205
Public
Ground
500 to 1,000
$24,337
$34,172
$48,183
Public
Ground
1,000 to 3,300
$34,816
$56,918
$88,928
Public
Ground
3,300 to 10,000
$0
$50,638
$192,610
Public
Ground
10,000 to 50,000
$0
$268,350
$591,950
Public
Surface
Less than 100
$0
$12,436
$39,864
Public
Surface
100 to 500
$0
$23,895
$79,534
Public
Surface
500 to 1,000
$0
$15,610
$83,872
Public
Surface
1,000 to 3,300
$0
$45,908
$156,730
Public
Surface
3,300 to 10,000
$0
$57,878
$252,450
Public
Surface
10,000 to 50,000
$0
$112,850
$543,790
Public
Surface
50,000 to 100,000
$0
$10,060
$0
Abbreviations: NTNCWS - Non-Transient, Non-Community Water Systems.
Proposed PFAS Rule Economic Analysis
C-16
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
CI.5 Distribution of Small Community Water System Costs
Table C-17: Distribution of Annualized Cost for Small CWSs, Proposed Option (PFOA and
PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021)
Annualized Cost Per CWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$159
$159
$159
$330
$2,993
Private
Ground
100 to 500
$193
$193
$211
$553
$3,542
Private
Ground
500 to 1,000
$185
$185
$316
$812
$4,452
Private
Ground
1,000 to 3,300
$193
$193
$361
$1,224
$6,684
Private
Ground
3,300 to 10,000
$320
$368
$852
$2,436
$12,812
Private
Surface
Less than 100
$257
$257
$258
$610
$3,585
Private
Surface
100 to 500
$297
$297
$304
$972
$4,468
Private
Surface
500 to 1,000
$280
$280
$318
$1,171
$5,468
Private
Surface
1,000 to 3,300
$267
$267
$393
$1,339
$6,547
Private
Surface
3,300 to 10,000
$332
$334
$644
$2,246
$13,966
Public
Ground
Less than 100
$159
$159
$159
$343
$2,996
Public
Ground
100 to 500
$193
$193
$212
$565
$3,603
Public
Ground
500 to 1,000
$185
$185
$313
$731
$4,386
Public
Ground
1,000 to 3,300
$193
$193
$359
$1,107
$6,288
Public
Ground
3,300 to 10,000
$320
$544
$876
$2,541
$13,311
Public
Surface
Less than 100
$257
$257
$259
$650
$3,683
Public
Surface
100 to 500
$297
$297
$299
$803
$4,499
Public
Surface
500 to 1,000
$280
$280
$308
$1,175
$5,493
Public
Surface
1,000 to 3,300
$267
$267
$365
$1,206
$6,942
Public
Surface
3,300 to 10,000
$332
$332
$594
$1,833
$13,241
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-17
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-18: Distribution of Annualized Cost for Small CWSs, Option la (PFOA and PFOS
MCLs of 4 ppt) (Commercial Cost of Capital, $2021)
Annualized Cost Per CWS
Ownership Source Population 10th 25th 50th 75th 90th
Water
Served Size
Category
Percentile
Percentile
Percentile
Percentile
Percentile
Private
Ground
Less than 100
$159
$159
$159
$307
$2,946
Private
Ground
100 to 500
$193
$193
$210
$462
$3,212
Private
Ground
500 to 1,000
$185
$185
$316
$685
$3,920
Private
Ground
1,000 to 3,300
$193
$193
$360
$995
$5,527
Private
Ground
3,300 to 10,000
$320
$368
$847
$2,288
$10,697
Private
Surface
Less than 100
$257
$257
$258
$493
$3,131
Private
Surface
100 to 500
$297
$297
$303
$629
$3,342
Private
Surface
500 to 1,000
$280
$280
$316
$796
$3,808
Private
Surface
1,000 to 3,300
$267
$267
$392
$842
$4,199
Private
Surface
3,300 to 10,000
$332
$334
$622
$1,898
$9,849
Public
Ground
Less than 100
$159
$159
$159
$310
$2,938
Public
Ground
100 to 500
$193
$193
$212
$450
$3,209
Public
Ground
500 to 1,000
$185
$185
$313
$607
$3,759
Public
Ground
1,000 to 3,300
$193
$193
$359
$931
$5,375
Public
Ground
3,300 to 10,000
$320
$544
$871
$2,436
$11,874
Public
Surface
Less than 100
$257
$257
$259
$521
$3,190
Public
Surface
100 to 500
$297
$297
$299
$607
$3,277
Public
Surface
500 to 1,000
$280
$280
$307
$697
$3,442
Public
Surface
1,000 to 3,300
$267
$267
$364
$773
$3,698
Public
Surface
3,300 to 10,000
$332
$332
$594
$1,456
$5,735
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-18
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-19: Distribution of Annualized Cost for Small CWSs, Option lb (PFOA and PFOS
MCLs of 5.0 ppt) (Commercial Cost of Capital, $2021)
Annualized Cost Per CWS
Ownership Source Population 10th 25th 50th 75th 90th
Water
Served Size
Category
Percentile
Percentile
Percentile
Percentile
Percentile
Private
Ground
Less than 100
$159
$159
$159
$278
$2,705
Private
Ground
100 to 500
$193
$193
$211
$382
$2,898
Private
Ground
500 to 1,000
$185
$185
$311
$540
$3,119
Private
Ground
1,000 to 3,300
$193
$193
$359
$709
$3,802
Private
Ground
3,300 to 10,000
$320
$367
$818
$1,829
$5,966
Private
Surface
Less than 100
$257
$257
$258
$444
$2,726
Private
Surface
100 to 500
$297
$297
$303
$551
$2,999
Private
Surface
500 to 1,000
$280
$280
$315
$605
$3,121
Private
Surface
1,000 to 3,300
$267
$267
$391
$659
$3,185
Private
Surface
3,300 to 10,000
$332
$334
$605
$1,418
$5,278
Public
Ground
Less than 100
$159
$159
$159
$279
$2,661
Public
Ground
100 to 500
$193
$193
$212
$375
$2,920
Public
Ground
500 to 1,000
$185
$185
$310
$480
$3,078
Public
Ground
1,000 to 3,300
$193
$193
$358
$666
$3,734
Public
Ground
3,300 to 10,000
$320
$531
$857
$1,879
$6,569
Public
Surface
Less than 100
$257
$257
$259
$450
$2,800
Public
Surface
100 to 500
$297
$297
$299
$531
$3,002
Public
Surface
500 to 1,000
$280
$280
$306
$564
$3,071
Public
Surface
1,000 to 3,300
$267
$267
$362
$622
$3,152
Public
Surface
3,300 to 10,000
$332
$332
$593
$1,103
$3,749
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-19
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-20: Distribution of Annualized Cost for Small CWSs, Option lc (PFOA and PFOS
MCLs of 10.0 ppt) (Commercial Cost of Capital, $2021)
Annualized Cost Per CWS
Ownership Source Population 10th 25th 50th 75th 90th
Water
Served Size
Category
Percentile
Percentile
Percentile
Percentile
Percentile
Private
Ground
Less than 100
$159
$159
$159
$276
$389
Private
Ground
100 to 500
$193
$193
$211
$358
$660
Private
Ground
500 to 1,000
$185
$185
$310
$460
$928
Private
Ground
1,000 to 3,300
$193
$193
$358
$576
$1,311
Private
Ground
3,300 to 10,000
$320
$366
$747
$1,394
$2,877
Private
Surface
Less than 100
$257
$257
$258
$431
$610
Private
Surface
100 to 500
$297
$297
$302
$496
$967
Private
Surface
500 to 1,000
$280
$280
$313
$516
$1,276
Private
Surface
1,000 to 3,300
$267
$267
$389
$515
$1,194
Private
Surface
3,300 to 10,000
$332
$334
$594
$1,071
$2,608
Public
Ground
Less than 100
$159
$159
$159
$276
$406
Public
Ground
100 to 500
$193
$193
$211
$358
$671
Public
Ground
500 to 1,000
$185
$185
$309
$387
$806
Public
Ground
1,000 to 3,300
$193
$193
$358
$554
$1,254
Public
Ground
3,300 to 10,000
$320
$528
$848
$1,434
$3,041
Public
Surface
Less than 100
$257
$257
$258
$432
$664
Public
Surface
100 to 500
$297
$297
$299
$494
$942
Public
Surface
500 to 1,000
$280
$280
$303
$515
$1,038
Public
Surface
1,000 to 3,300
$267
$267
$358
$499
$1,126
Public
Surface
3,300 to 10,000
$332
$332
$547
$894
$2,097
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-20
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.1.6 Distribution of Small Non-Community Non-Transient
Water System Costs
Table C-21: Distribution of Annualized Cost for Small NTNCWSs, Proposed Option (PFOA
and PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021)
Annualized Cost Per NTNCWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$155
$159
$162
$374
$3,011
Private
Ground
100 to 500
$160
$193
$201
$471
$3,142
Private
Ground
500 to 1,000
$156
$185
$205
$529
$3,342
Private
Ground
1,000 to 3,300
$169
$193
$293
$831
$4,322
Private
Ground
3,300 to 10,000
$320
$320
$489
$1,667
$8,428
Private
Surface
Less than 100
$257
$257
$269
$575
$3,283
Private
Surface
100 to 500
$297
$297
$375
$784
$3,809
Private
Surface
500 to 1,000
$280
$280
$471
$1,220
$5,953
Private
Surface
1,000 to 3,300
$267
$267
$389
$1,395
$5,981
Private
Surface
3,300 to 10,000
$339
$575
$1,167
$3,890
$23,117
Public
Ground
Less than 100
$159
$159
$161
$367
$3,012
Public
Ground
100 to 500
$193
$193
$196
$424
$3,094
Public
Ground
500 to 1,000
$185
$185
$188
$440
$3,112
Public
Ground
1,000 to 3,300
$193
$193
$251
$661
$3,777
Public
Ground
3,300 to 10,000
$320
$344
$600
$1,340
$14,323
Public
Surface
Less than 100
$257
$257
$271
$665
$3,202
Public
Surface
100 to 500
$297
$297
$324
$799
$3,784
Public
Surface
500 to 1,000
$280
$280
$309
$1,023
$4,630
Public
Surface
1,000 to 3,300
$267
$268
$477
$1,380
$7,205
Public
Surface
3,300 to 10,000
$332
$366
$753
$2,133
$29,501
Abbreviations: NTNCWS - Non-Transient Non-Community Water System.
Proposed PFAS Rule Economic Analysis
C-21
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-22: Distribution of Annualized Cost for Small NTNCWSs, Option la (PFOA and
PFOS MCLs of 4.0 ppt) (Commercial Cost of Capital, $2021)
Annualized Cost Per NTNCWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$155
$159
$162
$373
$3,008
Private
Ground
100 to 500
$160
$193
$201
$469
$3,133
Private
Ground
500 to 1,000
$156
$185
$205
$525
$3,327
Private
Ground
1,000 to 3,300
$169
$193
$292
$826
$4,270
Private
Ground
3,300 to 10,000
$320
$320
$488
$1,663
$8,336
Private
Surface
Less than 100
$257
$257
$268
$574
$3,277
Private
Surface
100 to 500
$297
$297
$375
$782
$3,798
Private
Surface
500 to 1,000
$280
$280
$470
$1,213
$5,897
Private
Surface
1,000 to 3,300
$267
$267
$389
$1,390
$5,912
Private
Surface
3,300 to 10,000
$339
$574
$1,163
$3,833
$22,435
Public
Ground
Less than 100
$159
$159
$161
$367
$3,006
Public
Ground
100 to 500
$193
$193
$196
$424
$3,089
Public
Ground
500 to 1,000
$185
$185
$188
$439
$3,103
Public
Ground
1,000 to 3,300
$193
$193
$251
$658
$3,747
Public
Ground
3,300 to 10,000
$320
$344
$600
$1,335
$14,199
Public
Surface
Less than 100
$257
$257
$271
$664
$3,197
Public
Surface
100 to 500
$297
$297
$324
$796
$3,769
Public
Surface
500 to 1,000
$280
$280
$309
$1,021
$4,615
Public
Surface
1,000 to 3,300
$267
$268
$477
$1,371
$7,120
Public
Surface
3,300 to 10,000
$332
$365
$751
$2,119
$28,949
Abbreviations: NTNCWS - Non-Transient Non-Community Water System.
Proposed PFAS Rule Economic Analysis
C-22
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-23: Distribution of Annualized Cost for Small NTNCWSs, Option lb (PFOA and
PFOS MCLs of 5.0 ppt) (Commercial Cost of Capital, $2021)
Annualized Cost Per NTNCWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$155
$159
$162
$304
$2,781
Private
Ground
100 to 500
$160
$193
$200
$378
$2,842
Private
Ground
500 to 1,000
$156
$185
$204
$391
$2,917
Private
Ground
1,000 to 3,300
$169
$193
$288
$605
$3,177
Private
Ground
3,300 to 10,000
$320
$320
$480
$1,350
$4,849
Private
Surface
Less than 100
$257
$257
$268
$484
$2,745
Private
Surface
100 to 500
$297
$297
$373
$616
$3,046
Private
Surface
500 to 1,000
$280
$280
$460
$890
$3,957
Private
Surface
1,000 to 3,300
$267
$267
$384
$1,081
$3,814
Private
Surface
3,300 to 10,000
$339
$568
$1,078
$3,191
$13,604
Public
Ground
Less than 100
$159
$159
$161
$306
$2,716
Public
Ground
100 to 500
$193
$193
$196
$359
$2,792
Public
Ground
500 to 1,000
$185
$185
$188
$354
$2,824
Public
Ground
1,000 to 3,300
$193
$193
$249
$482
$2,968
Public
Ground
3,300 to 10,000
$320
$344
$593
$1,059
$7,723
Public
Surface
Less than 100
$257
$257
$271
$503
$2,442
Public
Surface
100 to 500
$297
$297
$323
$615
$2,836
Public
Surface
500 to 1,000
$280
$280
$307
$765
$3,223
Public
Surface
1,000 to 3,300
$267
$268
$465
$1,062
$4,291
Public
Surface
3,300 to 10,000
$332
$364
$707
$1,782
$17,685
Abbreviations: NTNCWS - Non-Transient Non-Community Water System.
Proposed PFAS Rule Economic Analysis
C-23
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-24: Distribution of Annualized Cost for Small NTNCWSs, Option lc (PFOA and
PFOS MCLs of 10.0 ppt) (Commercial Cost of Capital, $2021)
Annualized Cost Per NTNCWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$155
$159
$162
$276
$587
Private
Ground
100 to 500
$160
$193
$200
$331
$706
Private
Ground
500 to 1,000
$156
$185
$203
$342
$804
Private
Ground
1,000 to 3,300
$169
$193
$281
$478
$1,146
Private
Ground
3,300 to 10,000
$320
$320
$477
$974
$2,114
Private
Surface
Less than 100
$257
$257
$268
$438
$774
Private
Surface
100 to 500
$297
$297
$370
$547
$1,110
Private
Surface
500 to 1,000
$280
$280
$450
$645
$1,583
Private
Surface
1,000 to 3,300
$267
$267
$380
$753
$1,804
Private
Surface
3,300 to 10,000
$339
$551
$963
$2,336
$4,662
Public
Ground
Less than 100
$159
$159
$161
$276
$559
Public
Ground
100 to 500
$193
$193
$196
$325
$632
Public
Ground
500 to 1,000
$185
$185
$188
$313
$649
Public
Ground
1,000 to 3,300
$193
$193
$245
$371
$946
Public
Ground
3,300 to 10,000
$320
$344
$586
$850
$1,797
Public
Surface
Less than 100
$257
$257
$271
$435
$787
Public
Surface
100 to 500
$297
$297
$323
$514
$1,056
Public
Surface
500 to 1,000
$280
$280
$305
$504
$1,361
Public
Surface
1,000 to 3,300
$267
$268
$445
$748
$1,755
Public
Surface
3,300 to 10,000
$332
$364
$641
$1,319
$3,910
Abbreviations: NTNCWS - Non-Transient Non-Community Water System.
Proposed PFAS Rule Economic Analysis
C-24
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
CI.7 Distribution of Small Community Water System Costs that
Treat or Change Water Source
Table C-25: Distribution of Annualized Cost for Small CWSs that Treat or Change Water
Source, Proposed Option (PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost
of Capital)
Annualized Cost Per CWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$8,242
$8,683
$9,301
$11,346
$27,138
Private
Ground
100 to 500
$12,388
$13,573
$15,824
$22,255
$41,253
Private
Ground
500 to 1,000
$18,503
$20,456
$24,163
$35,680
$62,893
Private
Ground
1,000 to 3,300
$25,828
$31,113
$41,075
$62,626
$101,440
Private
Ground
3,300 to 10,000
$42,808
$61,678
$104,750
$148,750
$200,140
Private
Surface
Less than 100
$11,017
$11,659
$12,692
$17,258
$38,601
Private
Surface
100 to 500
$15,480
$17,234
$20,195
$28,916
$62,180
Private
Surface
500 to 1,000
$21,658
$24,401
$28,921
$44,857
$80,445
Private
Surface
1,000 to 3,300
$28,556
$35,544
$47,013
$71,191
$116,060
Private
Surface
3,300 to 10,000
$47,538
$65,132
$105,250
$161,590
$222,040
Public
Ground
Less than 100
$8,513
$9,037
$9,724
$12,272
$28,057
Public
Ground
100 to 500
$13,841
$15,669
$18,696
$24,900
$45,429
Public
Ground
500 to 1,000
$20,264
$22,802
$27,529
$37,874
$63,959
Public
Ground
1,000 to 3,300
$29,277
$36,300
$49,322
$79,479
$116,780
Public
Ground
3,300 to 10,000
$47,826
$69,046
$120,390
$175,150
$242,740
Public
Surface
Less than 100
$11,478
$12,340
$13,568
$18,255
$38,665
Public
Surface
100 to 500
$17,641
$20,038
$24,264
$33,456
$69,933
Public
Surface
500 to 1,000
$25,435
$29,535
$35,977
$49,550
$98,337
Public
Surface
1,000 to 3,300
$36,552
$45,699
$60,955
$108,180
$158,060
Public
Surface
3,300 to 10,000
$61,341
$109,550
$157,880
$223,440
$315,330
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-25
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-26: Distribution of Annualized Cost for Small CWSs that Treat or Change Water
Source, Option la (PFOA and PFOS MCLs of 4.0 ppt) (Commercial Cost of Capital, $2021)
Annualized Cost Per CWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$8,241
$8,683
$9,300
$11,327
$27,122
Private
Ground
100 to 500
$12,388
$13,571
$15,813
$22,150
$41,135
Private
Ground
500 to 1,000
$18,495
$20,435
$24,093
$35,472
$62,521
Private
Ground
1,000 to 3,300
$25,792
$31,035
$40,863
$62,207
$100,980
Private
Ground
3,300 to 10,000
$42,645
$61,280
$104,060
$147,920
$198,890
Private
Surface
Less than 100
$11,017
$11,659
$12,690
$17,222
$38,422
Private
Surface
100 to 500
$15,480
$17,229
$20,182
$28,840
$62,034
Private
Surface
500 to 1,000
$21,656
$24,391
$28,876
$44,625
$80,068
Private
Surface
1,000 to 3,300
$28,551
$35,499
$46,932
$70,801
$115,570
Private
Surface
3,300 to 10,000
$47,559
$64,780
$104,660
$160,580
$220,770
Public
Ground
Less than 100
$8,512
$9,037
$9,723
$12,242
$28,005
Public
Ground
100 to 500
$13,840
$15,663
$18,681
$24,805
$45,299
Public
Ground
500 to 1,000
$20,250
$22,778
$27,458
$37,693
$63,747
Public
Ground
1,000 to 3,300
$29,229
$36,204
$49,100
$78,996
$116,420
Public
Ground
3,300 to 10,000
$47,688
$68,318
$119,840
$173,990
$241,120
Public
Surface
Less than 100
$11,473
$12,336
$13,563
$18,180
$38,580
Public
Surface
100 to 500
$17,638
$20,030
$24,244
$33,353
$69,691
Public
Surface
500 to 1,000
$25,416
$29,510
$35,935
$49,329
$97,973
Public
Surface
1,000 to 3,300
$36,514
$45,642
$60,830
$107,860
$157,780
Public
Surface
3,300 to 10,000
$61,146
$109,170
$157,500
$222,660
$314,150
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-26
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-27: Distribution of Annualized Cost for Small CWSs that Treat or Change Water
Source, Option lb (PFOA and PFOS MCLs of 5.0 ppt) (Commercial Cost of Capital, $2021)
Annualized Cost Per CWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$8,243
$8,684
$9,299
$11,315
$27,068
Private
Ground
100 to 500
$12,404
$13,592
$15,814
$22,000
$41,107
Private
Ground
500 to 1,000
$18,457
$20,361
$23,896
$34,884
$61,286
Private
Ground
1,000 to 3,300
$25,648
$30,705
$40,201
$61,181
$98,083
Private
Ground
3,300 to 10,000
$41,805
$59,107
$99,511
$143,270
$192,070
Private
Surface
Less than 100
$11,024
$11,617
$12,659
$17,347
$36,341
Private
Surface
100 to 500
$15,445
$17,190
$20,104
$28,731
$60,478
Private
Surface
500 to 1,000
$22,051
$24,239
$28,622
$43,537
$73,417
Private
Surface
1,000 to 3,300
$29,009
$35,094
$45,915
$69,367
$108,070
Private
Surface
3,300 to 10,000
$49,697
$62,806
$99,489
$153,560
$208,260
Public
Ground
Less than 100
$8,513
$9,035
$9,721
$12,311
$27,622
Public
Ground
100 to 500
$13,869
$15,683
$18,655
$24,690
$45,364
Public
Ground
500 to 1,000
$20,235
$22,712
$27,257
$37,317
$63,297
Public
Ground
1,000 to 3,300
$29,104
$35,900
$48,432
$77,739
$115,360
Public
Ground
3,300 to 10,000
$46,991
$65,063
$117,300
$169,370
$235,640
Public
Surface
Less than 100
$11,510
$12,312
$13,534
$18,227
$36,911
Public
Surface
100 to 500
$17,579
$19,934
$24,072
$32,931
$67,725
Public
Surface
500 to 1,000
$25,309
$29,420
$35,748
$49,163
$94,752
Public
Surface
1,000 to 3,300
$36,310
$45,398
$60,422
$106,750
$156,470
Public
Surface
3,300 to 10,000
$60,592
$107,090
$156,240
$221,180
$314,330
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-27
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-28: Distribution of Annualized Cost for Small CWSs that Treat or Change Water
Source, Option lc (PFOA and PFOS MCLs of 10.0 ppt) (Commercial Cost of Capital, $2021)
Annualized Cost Per CWS
Ownership
Source
Population
10th
25th
50th
75th
90th
Water
Served Size
Percentile
Percentile
Percentile
Percentile
Percentile
Category
Private
Ground
Less than 100
$8,246
$8,685
$9,296
$11,309
$26,576
Private
Ground
100 to 500
$12,443
$13,644
$15,814
$21,811
$42,334
Private
Ground
500 to 1,000
$18,167
$20,056
$23,343
$33,375
$55,838
Private
Ground
1,000 to 3,300
$25,043
$29,561
$38,032
$57,040
$86,431
Private
Ground
3,300 to 10,000
$46,598
$54,349
$80,681
$121,300
$160,810
Private
Surface
Less than 100
$11,874
$11,996
$12,704
$16,406
$29,507
Private
Surface
100 to 500
$15,695
$16,927
$19,498
$28,151
$49,859
Private
Surface
500 to 1,000
$25,697
$25,938
$27,971
$36,375
$59,476
Private
Surface
1,000 to 3,300
$36,662
$37,534
$43,464
$59,249
$88,349
Private
Surface
3,300 to 10,000
$67,604
$68,920
$83,604
$117,100
$171,040
Public
Ground
Less than 100
$8,484
$8,978
$9,675
$12,569
$25,391
Public
Ground
100 to 500
$13,874
$15,675
$18,507
$24,307
$44,767
Public
Ground
500 to 1,000
$19,996
$22,340
$26,368
$35,849
$59,638
Public
Ground
1,000 to 3,300
$28,623
$34,810
$46,314
$73,135
$110,780
Public
Ground
3,300 to 10,000
$45,527
$59,221
$107,720
$157,310
$219,500
Public
Surface
Less than 100
$12,720
$12,814
$13,541
$16,792
$29,851
Public
Surface
100 to 500
$17,401
$19,634
$23,454
$32,686
$58,677
Public
Surface
500 to 1,000
$24,877
$28,671
$34,581
$48,272
$79,172
Public
Surface
1,000 to 3,300
$34,856
$44,117
$58,833
$100,210
$147,770
Public
Surface
3,300 to 10,000
$59,278
$98,619
$150,840
$215,500
$308,880
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-28
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.1.8 Distribution of Small Non-Community Water Non-
Transient System Costs that Treat or Change Water Source
Table C-29: Distribution of Annualized Cost for Small NTNCWSs that Treat or Change
Water Source, Proposed Option (PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0)
(Commercial Cost of Capital, $2021)
Annualized Cost Per NTNCWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$8,151
$8,652
$9,402
$13,927
$29,307
Private
Ground
100 to 500
$12,366
$13,514
$15,869
$22,370
$43,434
Private
Ground
500 to 1,000
$18,336
$20,363
$23,726
$32,058
$56,838
Private
Ground
1,000 to 3,300
$25,223
$30,622
$39,135
$56,911
$88,201
Private
Ground
3,300 to 10,000
$68,153
$70,086
$84,488
$109,320
$145,570
Private
Surface
Less than 100
$11,004
$11,389
$12,714
$18,294
$34,841
Private
Surface
100 to 500
$16,242
$17,512
$20,804
$31,341
$53,902
Private
Surface
500 to 1,000
$27,455
$27,751
$30,531
$40,657
$67,812
Private
Surface
1,000 to 3,300
$36,257
$37,665
$45,202
$65,710
$101,980
Private
Surface
3,300 to 10,000
$78,055
$79,016
$94,687
$136,070
$213,890
Public
Ground
Less than 100
$8,391
$8,962
$9,798
$14,498
$31,389
Public
Ground
100 to 500
$13,771
$15,456
$18,566
$24,145
$45,771
Public
Ground
500 to 1,000
$20,454
$22,936
$26,097
$33,347
$57,315
Public
Ground
1,000 to 3,300
$29,309
$35,505
$44,022
$64,724
$100,390
Public
Ground
3,300 to 10,000
$82,369
$82,494
$87,770
$102,890
$144,070
Public
Surface
Less than 100
$13,233
$13,292
$14,048
$18,239
$31,632
Public
Surface
100 to 500
$19,511
$19,896
$22,515
$31,137
$51,418
Public
Surface
500 to 1,000
$34,025
$34,025
$34,464
$37,435
$55,439
Public
Surface
1,000 to 3,300
$50,739
$51,412
$58,817
$76,914
$113,290
Public
Surface
3,300 to 10,000
$94,800
$94,963
$101,130
$119,070
$165,810
Abbreviations: NTNCWS - Non-Transient Non-Community Water System.
Proposed PFAS Rule Economic Analysis
C-29
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-30: Distribution of Annualized Cost for Small NTNCWSs that Treat or Change
Water Source, Option la (PFOA and PFOS MCLs of 4.0 ppt) (Commercial Cost of Capital,
$2021)
Annualized Cost Per NTNCWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$8,151
$8,651
$9,400
$13,862
$29,201
Private
Ground
100 to 500
$12,365
$13,512
$15,856
$22,258
$43,245
Private
Ground
500 to 1,000
$18,332
$20,349
$23,689
$31,923
$56,483
Private
Ground
1,000 to 3,300
$25,193
$30,542
$39,029
$56,643
$87,732
Private
Ground
3,300 to 10,000
$68,040
$69,972
$84,199
$108,800
$145,070
Private
Surface
Less than 100
$10,999
$11,381
$12,704
$18,261
$34,774
Private
Surface
100 to 500
$16,240
$17,498
$20,769
$31,252
$53,635
Private
Surface
500 to 1,000
$27,406
$27,699
$30,447
$40,557
$67,438
Private
Surface
1,000 to 3,300
$36,244
$37,639
$45,074
$65,228
$101,290
Private
Surface
3,300 to 10,000
$77,908
$78,795
$94,100
$134,480
$211,810
Public
Ground
Less than 100
$8,390
$8,961
$9,796
$14,433
$31,194
Public
Ground
100 to 500
$13,769
$15,452
$18,554
$24,100
$45,656
Public
Ground
500 to 1,000
$20,448
$22,921
$26,072
$33,251
$57,021
Public
Ground
1,000 to 3,300
$29,264
$35,443
$43,904
$64,387
$99,974
Public
Ground
3,300 to 10,000
$82,239
$82,352
$87,613
$102,430
$143,460
Public
Surface
Less than 100
$13,238
$13,296
$14,061
$18,231
$31,586
Public
Surface
100 to 500
$19,511
$19,893
$22,504
$31,063
$51,281
Public
Surface
500 to 1,000
$33,894
$33,894
$34,338
$37,251
$55,138
Public
Surface
1,000 to 3,300
$50,602
$51,259
$58,474
$76,432
$112,550
Public
Surface
3,300 to 10,000
$94,451
$94,612
$100,620
$118,000
$163,920
Abbreviations: NTNCWS - Non-Transient Non-Community Water System.
Proposed PFAS Rule Economic Analysis
C-30
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-31: Distribution of Annualized Cost for Small NTNCWSs that Treat or Change
Water Source, Option lb (PFOA and PFOS MCLs of 5.0 ppt) (Commercial Cost of Capital,
$2021)
Annualized Cost Per NTNCWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$8,154
$8,652
$9,399
$13,780
$29,332
Private
Ground
100 to 500
$12,370
$13,514
$15,827
$22,103
$43,086
Private
Ground
500 to 1,000
$18,126
$20,127
$23,360
$31,478
$54,595
Private
Ground
1,000 to 3,300
$25,002
$30,024
$38,197
$55,500
$83,953
Private
Ground
3,300 to 10,000
$70,785
$71,520
$79,719
$97,765
$131,930
Private
Surface
Less than 100
$11,217
$11,461
$12,609
$18,004
$32,108
Private
Surface
100 to 500
$16,937
$17,750
$20,640
$30,245
$50,656
Private
Surface
500 to 1,000
$28,890
$28,976
$30,699
$37,838
$62,611
Private
Surface
1,000 to 3,300
$39,132
$39,774
$44,641
$60,545
$95,018
Private
Surface
3,300 to 10,000
$82,605
$82,822
$91,849
$119,600
$188,480
Public
Ground
Less than 100
$8,385
$8,952
$9,800
$14,386
$30,867
Public
Ground
100 to 500
$13,759
$15,442
$18,494
$23,999
$45,449
Public
Ground
500 to 1,000
$20,357
$22,801
$25,939
$33,147
$55,606
Public
Ground
1,000 to 3,300
$28,855
$34,887
$43,242
$62,864
$95,930
Public
Ground
3,300 to 10,000
$78,626
$78,695
$80,908
$89,118
$120,970
Public
Surface
Less than 100
$13,635
$13,653
$14,102
$16,892
$28,898
Public
Surface
100 to 500
$20,649
$20,804
$22,536
$29,155
$48,972
Public
Surface
500 to 1,000
$30,515
$30,515
$30,653
$32,028
$43,871
Public
Surface
1,000 to 3,300
$51,409
$51,669
$55,832
$67,481
$98,713
Public
Surface
3,300 to 10,000
$87,944
$88,024
$91,161
$101,240
$134,750
Abbreviations: NTNCWS - Non-Transient Non-Community Water System.
Proposed PFAS Rule Economic Analysis
C-31
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-32: Distribution of Annualized Cost for Small NTNCWSs that Treat or Change
Water Source, Option lc (PFOA and PFOS MCLs of 10.0 ppt) (Commercial Cost of Capital,
$2021)
Annualized Cost Per NTNCWS
Ownership
Source
Water
Population
Served Size
Category
10th
Percentile
25th
Percentile
50th
Percentile
75th
Percentile
90th
Percentile
Private
Ground
Less than 100
$8,158
$8,642
$9,394
$13,498
$29,755
Private
Ground
100 to 500
$12,351
$13,479
$15,684
$21,714
$42,285
Private
Ground
500 to 1,000
$18,233
$19,774
$22,724
$30,351
$45,974
Private
Ground
1,000 to 3,300
$28,443
$29,839
$35,451
$48,171
$68,049
Private
Ground
3,300 to 10,000
$54,781
$54,819
$55,576
$58,802
$73,180
Private
Surface
Less than 100
$12,831
$12,842
$13,162
$15,351
$26,237
Private
Surface
100 to 500
$19,558
$19,604
$20,335
$24,123
$39,400
Private
Surface
500 to 1,000
$24,041
$24,041
$24,165
$24,977
$33,683
Private
Surface
1,000 to 3,300
$36,481
$36,509
$37,094
$39,732
$54,838
Private
Surface
3,300 to 10,000
$69,885
$69,885
$70,472
$74,324
$100,370
Public
Ground
Less than 100
$8,375
$8,883
$9,779
$13,988
$27,831
Public
Ground
100 to 500
$13,670
$15,313
$18,162
$23,729
$42,742
Public
Ground
500 to 1,000
$20,347
$22,160
$25,161
$32,310
$47,559
Public
Ground
1,000 to 3,300
$32,900
$34,495
$40,468
$55,548
$78,082
Public
Ground
3,300 to 10,000
$46,464
$46,464
$46,535
$47,197
$54,139
Public
Surface
Less than 100
$10,362
$10,362
$10,385
$10,678
$14,112
Public
Surface
100 to 500
$18,747
$18,747
$18,963
$20,109
$28,531
Public
Surface
500 to 1,000
$15,568
$15,568
$15,570
$15,646
$17,328
Public
Surface
1,000 to 3,300
$37,082
$37,082
$37,527
$39,003
$48,527
Public
Surface
3,300 to 10,000
$49,931
$49,931
$50,141
$51,539
$59,786
Abbreviations: NTNCWS - Non-Transient Non-Community Water System.
Proposed PFAS Rule Economic Analysis
C-32
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.2 Household-Level Cost Details
Section C.2 provides estimates of household costs by primary source water, ownership, and
system size category. Costs are provided for all CWSs as well as for only CWSs that must treat
or change water source to comply with the regulatory option.
C.2.1 Household Costs for oil Community Water Systems
Table C-33: Mean Annualized Cost per Household, Proposed Option (PFOA and
PFOS MCLs of 4.0 ppt and HI of 1.0) (Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Private
Ground
Less than 100
$41
$59
$82
Private
Ground
100 to 500
$23
$33
$49
Private
Ground
500 to 1,000
$7
$10
$15
Private
Ground
1,000 to 3,300
$4
$7
$10
Private
Ground
3,300 to 10,000
$3
$5
$7
Private
Ground
10,000 to 50,000
$8
$9
$11
Private
Ground
50,000 to 100,000
$5
$8
$11
Private
Ground
100,000 to 1,000,000
$4
$6
$8
Private
Surface
Less than 100
$38
$66
$105
Private
Surface
100 to 500
$18
$32
$61
Private
Surface
500 to 1,000
$6
$11
$23
Private
Surface
1,000 to 3,300
$3
$6
$12
Private
Surface
3,300 to 10,000
$2
$4
$8
Private
Surface
10,000 to 50,000
$6
$7
$9
Private
Surface
50,000 to 100,000
$4
$5
$7
Private
Surface
100,000 to 1,000,000
$10
$12
$14
Public
Ground
Less than 100
$49
$73
$106
Public
Ground
100 to 500
$19
$28
$43
Public
Ground
500 to 1,000
$5
$8
$11
Public
Ground
1,000 to 3,300
$3
$5
$7
Public
Ground
3,300 to 10,000
$7
$11
$15
Public
Ground
10,000 to 50,000
$8
$8
$9
Public
Ground
50,000 to 100,000
$5
$7
$8
Public
Ground
100,000 to 1,000,000
$7
$8
$10
Public
Surface
Less than 100
$54
$95
$155
Public
Surface
100 to 500
$19
$31
$55
Public
Surface
500 to 1,000
$5
$9
$18
Public
Surface
1,000 to 3,300
$3
$5
$10
Public
Surface
3,300 to 10,000
$7
$11
$19
Public
Surface
10,000 to 50,000
$7
$8
$9
Public
Surface
50,000 to 100,000
$5
$6
$6
Public
Surface
100,000 to 1,000,000
$8
$9
$10
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-33
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-34: Mean Annualized Cost per Household, Option la (PFOA and PFOS
MCLs of 4.0 ppt) (Commercial Cost of Capital, $2021)
Ownership
Source
Population Served Size
5th Percentile
Mean
95th
Water
Category
Percentile
Private
Ground
Less than 100
$41
$59
$82
Private
Ground
100 to 500
$23
$33
$49
Private
Ground
500 to 1,000
$7
$10
$15
Private
Ground
1,000 to 3,300
$4
$6
$9
Private
Ground
3,300 to 10,000
$3
$5
$7
Private
Ground
10,000 to 50,000
$7
$9
$11
Private
Ground
50,000 to 100,000
$4
$7
$10
Private
Ground
100,000 to 1,000,000
$3
$5
$8
Private
Surface
Less than 100
$38
$66
$103
Private
Surface
100 to 500
$18
$32
$61
Private
Surface
500 to 1,000
$5
$11
$22
Private
Surface
1,000 to 3,300
$3
$6
$12
Private
Surface
3,300 to 10,000
$2
$4
$8
Private
Surface
10,000 to 50,000
$5
$7
$9
Private
Surface
50,000 to 100,000
$4
$5
$7
Private
Surface
100,000 to 1,000,000
$9
$11
$13
Public
Ground
Less than 100
$49
$72
$106
Public
Ground
100 to 500
$19
$28
$43
Public
Ground
500 to 1,000
$5
$8
$11
Public
Ground
1,000 to 3,300
$3
$5
$7
Public
Ground
3,300 to 10,000
$7
$11
$15
Public
Ground
10,000 to 50,000
$7
$8
$9
Public
Ground
50,000 to 100,000
$5
$6
$8
Public
Ground
100,000 to 1,000,000
$6
$8
$10
Public
Surface
Less than 100
$53
$95
$155
Public
Surface
100 to 500
$19
$30
$54
Public
Surface
500 to 1,000
$5
$9
$18
Public
Surface
1,000 to 3,300
$3
$5
$10
Public
Surface
3,300 to 10,000
$7
$11
$19
Public
Surface
10,000 to 50,000
$7
$8
$9
Public
Surface
50,000 to 100,000
$5
$5
$6
Public
Surface
100,000 to 1,000,000
$7
$8
$9
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-34
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-35: Mean Annualized Cost per Household, Option lb (PFOA and PFOS
MCLs of 5.0 ppt) (Commercial Cost of Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$33
$47
$66
Private
Ground
100 to 500
$18
$26
$38
Private
Ground
500 to 1,000
$5
$8
$12
Private
Ground
1,000 to 3,300
$3
$5
$7
Private
Ground
3,300 to 10,000
$2
$4
$6
Private
Ground
10,000 to 50,000
$6
$7
$8
Private
Ground
50,000 to 100,000
$3
$5
$8
Private
Ground
100,000 to 1,000,000
$2
$4
$6
Private
Surface
Less than 100
$30
$53
$87
Private
Surface
100 to 500
$14
$25
$47
Private
Surface
500 to 1,000
$4
$9
$18
Private
Surface
1,000 to 3,300
$2
$4
$9
Private
Surface
3,300 to 10,000
$1
$3
$6
Private
Surface
10,000 to 50,000
$4
$6
$7
Private
Surface
50,000 to 100,000
$3
$4
$6
Private
Surface
100,000 to 1,000,000
$8
$9
$11
Public
Ground
Less than 100
$37
$58
$86
Public
Ground
100 to 500
$15
$22
$35
Public
Ground
500 to 1,000
$4
$6
$9
Public
Ground
1,000 to 3,300
$3
$4
$6
Public
Ground
3,300 to 10,000
$5
$8
$12
Public
Ground
10,000 to 50,000
$6
$7
$7
Public
Ground
50,000 to 100,000
$4
$5
$6
Public
Ground
100,000 to 1,000,000
$5
$7
$8
Public
Surface
Less than 100
$44
$77
$130
Public
Surface
100 to 500
$15
$24
$42
Public
Surface
500 to 1,000
$4
$7
$15
Public
Surface
1,000 to 3,300
$2
$4
$6
Public
Surface
3,300 to 10,000
$5
$9
$14
Public
Surface
10,000 to 50,000
$6
$6
$7
Public
Surface
50,000 to 100,000
$3
$4
$5
Public
Surface
100,000 to 1,000,000
$6
$7
$8
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-35
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-36: Mean Annualized Cost per Household, Option lc (PFOA and PFOS MCLs
of 10.0 ppt) (Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Private
Ground
Less than 100
$18
$24
$33
Private
Ground
100 to 500
$9
$13
$18
Private
Ground
500 to 1,000
$2
$4
$6
Private
Ground
1,000 to 3,300
$1
$2
$3
Private
Ground
3,300 to 10,000
$1
$2
$3
Private
Ground
10,000 to 50,000
$2
$3
$4
Private
Ground
50,000 to 100,000
$1
$2
$3
Private
Ground
100,000 to 1,000,000
$1
$1
$2
Private
Surface
Less than 100
$18
$28
$45
Private
Surface
100 to 500
$8
$13
$18
Private
Surface
500 to 1,000
$2
$4
$7
Private
Surface
1,000 to 3,300
$1
$2
$3
Private
Surface
3,300 to 10,000
$0
$1
$3
Private
Surface
10,000 to 50,000
$2
$2
$3
Private
Surface
50,000 to 100,000
$1
$2
$3
Private
Surface
100,000 to 1,000,000
$4
$5
$6
Public
Ground
Less than 100
$20
$30
$44
Public
Ground
100 to 500
$7
$11
$15
Public
Ground
500 to 1,000
$2
$3
$4
Public
Ground
1,000 to 3,300
$1
$2
$3
Public
Ground
3,300 to 10,000
$2
$4
$5
Public
Ground
10,000 to 50,000
$3
$3
$3
Public
Ground
50,000 to 100,000
$2
$2
$3
Public
Ground
100,000 to 1,000,000
$2
$3
$4
Public
Surface
Less than 100
$24
$39
$66
Public
Surface
100 to 500
$8
$12
$17
Public
Surface
500 to 1,000
$2
$3
$4
Public
Surface
1,000 to 3,300
$1
$2
$2
Public
Surface
3,300 to 10,000
$2
$3
$5
Public
Surface
10,000 to 50,000
$2
$3
$3
Public
Surface
50,000 to 100,000
$1
$2
$2
Public
Surface
100,000 to 1,000,000
$2
$3
$3
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-36
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
C.2.2 Household Costs for Community Water Systems that Treat
or Change Water Source
Table C-37: Mean Annualized Cost per Household in CWSs that Treat or Change
Water Source, Proposed Option (PFOA and PFOS MCLs of 4.0 ppt and HI of 1.0)
(Commercial Cost of Capital, $2021)
Ownership
Source
Water
Population Served Size
Category
5th Percentile
Mean
95th Percentile
Private
Ground
Less than 100
$677
$677
$677
Private
Ground
100 to 500
$392
$392
$392
Private
Ground
500 to 1,000
$123
$123
$123
Private
Ground
1,000 to 3,300
$76
$76
$76
Private
Ground
3,300 to 10,000
$52
$52
$52
Private
Ground
10,000 to 50,000
$32
$32
$32
Private
Ground
50,000 to 100,000
$25
$25
$25
Private
Ground
100,000 to 1,000,000
$12
$12
$12
Private
Surface
Less than 100
$798
$798
$798
Private
Surface
100 to 500
$389
$389
$389
Private
Surface
500 to 1,000
$141
$141
$141
Private
Surface
1,000 to 3,300
$76
$76
$76
Private
Surface
3,300 to 10,000
$53
$53
$53
Private
Surface
10,000 to 50,000
$33
$33
$33
Private
Surface
50,000 to 100,000
$21
$21
$21
Private
Surface
100,000 to 1,000,000
$25
$25
$25
Public
Ground
Less than 100
$835
$835
$835
Public
Ground
100 to 500
$341
$341
$341
Public
Ground
500 to 1,000
$94
$94
$94
Public
Ground
1,000 to 3,300
$63
$63
$63
Public
Ground
3,300 to 10,000
$118
$118
$118
Public
Ground
10,000 to 50,000
$31
$31
$31
Public
Ground
50,000 to 100,000
$23
$23
$23
Public
Ground
100,000 to 1,000,000
$24
$24
$24
Public
Surface
Less than 100
$1,131
$1,131
$1,131
Public
Surface
100 to 500
$392
$392
$392
Public
Surface
500 to 1,000
$117
$117
$117
Public
Surface
1,000 to 3,300
$73
$73
$73
Public
Surface
3,300 to 10,000
$149
$149
$149
Public
Surface
10,000 to 50,000
$39
$39
$39
Public
Surface
50,000 to 100,000
$24
$24
$24
Public
Surface
100,000 to 1,000,000
$30
$30
$30
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-37
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-38: Mean Annualized Cost per Household in CWSs that Treat or Change
Water Source, Option la (PFOA and PFOS MCLs of 4.0 ppt) (Commercial Cost of
Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$676
$676
$676
Private
Ground
100 to 500
$391
$391
$391
Private
Ground
500 to 1,000
$122
$122
$122
Private
Ground
1,000 to 3,300
$76
$76
$76
Private
Ground
3,300 to 10,000
$52
$52
$52
Private
Ground
10,000 to 50,000
$30
$30
$30
Private
Ground
50,000 to 100,000
$21
$21
$21
Private
Ground
100,000 to 1,000,000
$12
$12
$12
Private
Surface
Less than 100
$797
$797
$797
Private
Surface
100 to 500
$388
$388
$388
Private
Surface
500 to 1,000
$140
$140
$140
Private
Surface
1,000 to 3,300
$76
$76
$76
Private
Surface
3,300 to 10,000
$53
$53
$53
Private
Surface
10,000 to 50,000
$32
$32
$32
Private
Surface
50,000 to 100,000
$20
$20
$20
Private
Surface
100,000 to 1,000,000
$24
$24
$24
Public
Ground
Less than 100
$834
$834
$834
Public
Ground
100 to 500
$341
$341
$341
Public
Ground
500 to 1,000
$94
$94
$94
Public
Ground
1,000 to 3,300
$62
$62
$62
Public
Ground
3,300 to 10,000
$117
$117
$117
Public
Ground
10,000 to 50,000
$30
$30
$30
Public
Ground
50,000 to 100,000
$22
$22
$22
Public
Ground
100,000 to 1,000,000
$23
$23
$23
Public
Surface
Less than 100
$1,129
$1,129
$1,129
Public
Surface
100 to 500
$391
$391
$391
Public
Surface
500 to 1,000
$117
$117
$117
Public
Surface
1,000 to 3,300
$73
$73
$73
Public
Surface
3,300 to 10,000
$148
$148
$148
Public
Surface
10,000 to 50,000
$39
$39
$39
Public
Surface
50,000 to 100,000
$24
$24
$24
Public
Surface
100,000 to 1,000,000
$30
$30
$30
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-38
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-39: Mean Annualized Cost per Household in CWSs that Treat or Change
Water Source, Option lb (PFOA and PFOS MCLs of 5.0 ppt) (Commercial Cost of
Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$674
$674
$674
Private
Ground
100 to 500
$389
$389
$389
Private
Ground
500 to 1,000
$121
$121
$121
Private
Ground
1,000 to 3,300
$75
$75
$75
Private
Ground
3,300 to 10,000
$51
$51
$51
Private
Ground
10,000 to 50,000
$28
$28
$28
Private
Ground
50,000 to 100,000
$19
$19
$19
Private
Ground
100,000 to 1,000,000
$10
$10
$10
Private
Surface
Less than 100
$794
$794
$794
Private
Surface
100 to 500
$387
$387
$387
Private
Surface
500 to 1,000
$139
$139
$139
Private
Surface
1,000 to 3,300
$76
$76
$76
Private
Surface
3,300 to 10,000
$52
$52
$52
Private
Surface
10,000 to 50,000
$31
$31
$31
Private
Surface
50,000 to 100,000
$20
$20
$20
Private
Surface
100,000 to 1,000,000
$22
$22
$22
Public
Ground
Less than 100
$830
$830
$830
Public
Ground
100 to 500
$339
$339
$339
Public
Ground
500 to 1,000
$94
$94
$94
Public
Ground
1,000 to 3,300
$62
$62
$62
Public
Ground
3,300 to 10,000
$114
$114
$114
Public
Ground
10,000 to 50,000
$29
$29
$29
Public
Ground
50,000 to 100,000
$21
$21
$21
Public
Ground
100,000 to 1,000,000
$22
$22
$22
Public
Surface
Less than 100
$1,147
$1,147
$1,147
Public
Surface
100 to 500
$387
$387
$387
Public
Surface
500 to 1,000
$116
$116
$116
Public
Surface
1,000 to 3,300
$73
$73
$73
Public
Surface
3,300 to 10,000
$147
$147
$147
Public
Surface
10,000 to 50,000
$37
$37
$37
Public
Surface
50,000 to 100,000
$22
$22
$22
Public
Surface
100,000 to 1,000,000
$28
$28
$28
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-39
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table C-40: Mean Annualized Cost per Household in CWSs that Treat or Change
Water Source, Option lc (PFOA and PFOS MCLs of 10.0 ppt) (Commercial Cost of
Capital, $2021)
Ownership Source Population Served Size 5th Percentile Mean 95th Percentile
Water
Category
Private
Ground
Less than 100
$661
$661
$661
Private
Ground
100 to 500
$386
$386
$386
Private
Ground
500 to 1,000
$119
$119
$119
Private
Ground
1,000 to 3,300
$72
$72
$72
Private
Ground
3,300 to 10,000
$47
$47
$47
Private
Ground
10,000 to 50,000
$20
$20
$20
Private
Ground
50,000 to 100,000
$10
$10
$10
Private
Ground
100,000 to 1,000,000
$5
$5
$5
Private
Surface
Less than 100
$766
$766
$766
Private
Surface
100 to 500
$381
$381
$381
Private
Surface
500 to 1,000
$130
$130
$130
Private
Surface
1,000 to 3,300
$72
$72
$72
Private
Surface
3,300 to 10,000
$48
$48
$48
Private
Surface
10,000 to 50,000
$24
$24
$24
Private
Surface
50,000 to 100,000
$19
$19
$19
Private
Surface
100,000 to 1,000,000
$16
$16
$16
Public
Ground
Less than 100
$823
$823
$823
Public
Ground
100 to 500
$334
$334
$334
Public
Ground
500 to 1,000
$91
$91
$91
Public
Ground
1,000 to 3,300
$59
$59
$59
Public
Ground
3,300 to 10,000
$107
$107
$107
Public
Ground
10,000 to 50,000
$25
$25
$25
Public
Ground
50,000 to 100,000
$19
$19
$19
Public
Ground
100,000 to 1,000,000
$18
$18
$18
Public
Surface
Less than 100
$1,107
$1,107
$1,107
Public
Surface
100 to 500
$381
$381
$381
Public
Surface
500 to 1,000
$114
$114
$114
Public
Surface
1,000 to 3,300
$72
$72
$72
Public
Surface
3,300 to 10,000
$144
$144
$144
Public
Surface
10,000 to 50,000
$34
$34
$34
Public
Surface
50,000 to 100,000
$18
$18
$18
Public
Surface
100,000 to 1,000,000
$24
$24
$24
Abbreviations: CWS - Community Water System.
Proposed PFAS Rule Economic Analysis
C-40
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix D. PFOA and PFOS Serum
Concentration-Birth Weight Relationship
This appendix describes the methods used to estimate relationships between birth weight (BW)
and serum per- and polyfluoroalkyl substances (PFAS) based on available studies. EPA used
these relationships to estimate incremental changes in birth weight associated with reduced
exposure to PFAS, namely perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid
(PFOS).
D.l Weight of Evidence of Birth Weight Effects
In the Health Effects Support Document (HESD) for PFOA (U.S. EPA, 2016b), EPA
characterized the evidence for PFOA effects on birth weight as "plausible" based on human and
animal study data, and four of the five endpoints used for derivation of an RfD were lowered
fetal weights in rodents. The HESD for PFOS (U.S. EPA, 2016a) indicated that, despite
considerable uncertainty, the available human data "suggest an association of prenatal serum
PFOS with deficits in mean birth weight and with LBW [low birth weight]." The Agency for
Toxic Substances and Disease Registry (ATSDR, 2018) listed reduced birth weight as one of the
endpoints for which the available evidence "suggested" a relationship between human PFAS
exposure and effect. Negri et al. (2017), considering both toxicological and epidemiological
evidence, concluded that a causal relationship between PFOA and PFOS exposure and reduced
birth weight was "likely". The most recent syntheses of evidence, EPA's Toxicity Assessments
and Proposed Maximum Contaminant Level Goals for PFOA and PFOS in Drinking Water,
found clear evidence of an association between PFOA and PFOS and birth weight in both
toxicological and epidemiological studies (U.S. EPA, 2023a; U.S. EPA, 2023b). Based on these
findings, EPA's Office of Ground Water and Drinking Water (OGWDW) derived exposure-
response estimates for both compounds.
D.2 Review of Available Meta-Analyses
EPA's OGWDW reviewed literature identified in the EPA Office of Water, Office of Science
and Technology (OW/OST) literature reviews on the relationship between PFAS and birth
weight to identify previous estimates of serum PFAS-birth weight relationships. Many
epidemiological studies and several meta-analyses of existing studies have identified associations
between perfluorinated compound exposure and indices of fetal growth (primarily reduced birth
weight) (ATSDR, 2018; Johnson et al., 2014; Verner et al., 2015; Negri et al., 2017; Steenland et
al., 2018; Dzierlenga, Crawford, et al., 2020). Most studies of the relationship between maternal
serum PFOA and birth weight reported negative (i.e., inverse) relationships, while the evidence
for PFOS was more variable, as described below. Note that EPA's review was based primarily
on secondary sources; OGWDW did not conduct a systematic literature search or independent
risk of bias (ROB) analyses for any identified systematic reviews and meta-analyses. Rather,
EPA relied on previous authors who have analyzed the literature using different protocols related
to literature relevance, study quality, and ROB. However, OW/OST has evaluated
epidemiological literature for PFOA/PFOS as part of a systematic review to update the 2016
HESDs for PFOS and PFOA.
Proposed PFAS Rule Economic Analysis
D-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The five studies considered by EPA/OST for PFOA report the following slope estimates (in birth
weight g per ng/mL serum): -4.9 (Sagiv et al., 2018), -20.7 (Govarts et al., 2016 The five studies
considered by EPA/OST for PFOA report the following slope estimates (in birth weight g per
ng/mL serum): -4.9 (Sagiv et al., 2018), -20.7 (Govarts et al., 2016), -41.0 (Wikstrom et al.,
2019), -45.0 (Starling et al., 2017), and -45.2 (Chu et al., 2020). Compare these estimates to the
selected slope estimate from Negri et al. (2017) of -12.8 g per ng/mL. The four studies
considered by EPA/OST for PFOS report the following slope estimates (in birth weight g per
ng/mL serum): -1.1 (Sagiv et al., 2018), -5.5 (Starling et al., 2017), -8.4 (Wikstrom et al., 2019),
and -11.0 (Chu et al., 2020). Compare these estimates to the selected exposure-response function
from Dzierlenga, Crawford, et al. (2020) of -3.2 g per ng/mL. Wikstrom et al., 2019), -45.0
(Starling et al., 2017), and -45.2 (Chu et al., 2020). Compare these estimates to the selected slope
estimate from Negri et al. (2017) of-12.8 g per ng/mL. The four studies considered by EPA/OST
for PFOS report the following slope estimates (in birth weight g per ng/mL serum): -1.1 (Sagiv
et al., 2018), -5.5 (Starling et al., 2017), -8.4 (Wikstrom et al., 2019), and -11.0 (Chu et al.,
2020). Compare these estimates to the selected exposure-response function from Dzierlenga,
Crawford, et al. (2020) of -3.2 g per ng/mL.
EPA reviewed six of the identified meta-analyses of PFAS-low birth weight relationships in
detail. One study, Monroy et al. (2008), presented regression results for body weight versus
maternal PFOA and PFOS concentrations, but the reported slope factors4 were not adjusted for
other covariates. Because of this it was not pursued further. Two of the analyses (Johnson et al.,
2014; Negri et al., 2017) used well-documented systematic review and ROB procedures to
identify relevant studies in the literature. The three other studies did not document ROB
protocols and study quality evaluation criteria (Verner et al., 2015; Dzierlenga, Crawford, et al.,
2020; Steenland et al., 2018). However, as discussed below, there was extensive overlap in the
data sets addressed in the various meta-analyses. Two of the meta-analyses included exposure-
response modeling for both PFOS and PFOA (Verner et al., 2015; Negri et al., 2017), while one
study addressed only PFOS (Dzierlenga, Crawford, et al., 2020) and the remaining two
addressed only PFOA (Johnson et al., 2014; Steenland et al., 2018).
There was relative conformity in the publications evaluated and ultimately selected for use in the
meta-analyses especially amongst the most recent ones, as later authors tended to include all the
studies evaluated in previous studies, adding newer results that had become available (Table
D-l):
• Johnson et al. (2014) conducted random effects meta-analysis based on data from nine
studies (including 4,149 births) published between 2007 and 2012. The authors requested
individual data on PFOA and covariates (variables other than PFAS exposure that may
predict study outcomes) from all authors of the primary studies used in their studies. In
cases where data were available, Johnson et al. (2014) used random effects methods to
estimate covariate-adjusted linear regression coefficients and used these values as inputs
to their meta-analysis. They found that including or excluding studies likely to have high
ROB resulted in only small effects on estimated slope factors for PFOA-birth weight
relationships.
4 When referring to a "slope factor" in this document, EPA is discussing a measure of association between PFAS serum and BW.
Proposed PFAS Rule Economic Analysis
D-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• Verner et al. (2015) included data from all the studies identified by Johnson et al. (2014),
with the exception of results from two studies: Fromme et al. (2010) and Kim et al.
(2011). Verner et al. (2015) excluded these studies because they were based on 50 or
fewer participants.
• Negri et al. (2017) included all the data sets identified by Johnson et al. (2014) plus five
newer data sets (Table D-l). Negri et al. (2017) also included data from an older study
(Monroy et al., 2008) that Johnson et al. (2014) omitted because "BW [birth weight] is
not the dependent model variable."
• Steenland et al. (2018) based their analyses of PFOA-birth weight effects on results from
the same studies in the Negri et al. (2017) meta-analysis (except for one study, Monroy et
al. (2008) plus 10 additional recent epidemiological studies (Table D-l). However,
Steenland et al. (2018) did not conduct a formal ROB evaluation to exclude these studies
based on design or analysis flaws, as was done in prior meta-analyses by Johnson et al.
(2014) and Negri et al. (2017).,5 Dzierlenga, Crawford, et al. (2020) included PFOS-birth
weight data from all the studies identified by Verner et al. (2015), with the exception of
results from Fei et al. (2007), and an additional 22 studies, many of which overlap with
studies evaluated in Steenland et al. (2018). Although Dzierlenga, Crawford, et al. (2020)
did not conduct formal ROB evaluations, the authors examined some study design
aspects by characterizing studies with respect to certain characteristics that might
influence results and evaluating those characteristics in meta-regression analyses.
5 Steenland et al. (2018) noted that ROB analyses have advantages in identifying biases, but stated that "using a quantitative score
of bias as a basis to exclude studies ultimately includes subjective components."
Proposed PFAS Rule Economic Analysis
March 2023
D-3
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table D-l: Data Sources for PFOA/PFOS Meta-Analyses of Birth Weight Effects
PFOA/PFOS-BW Relationship Studies Included in Meta-Analyses for Effects on BW
Study
Johnson et al.
(2014)
Verner et al.
(2015)
Negri et al.
(2017)
Steenland et al.
(2018)
Dzierlenga
(2020)
EPA/OST Review
(PFOA/PFOS) (2021)a
Apelberg et al. (2007)
X
X*
X*
X
X
X
Fei et al. (2007)
X
X*
X*
X
X
Hamm et al. (2010)
X
X*
X*
X
X
X
Washino et al. (2009)
X
X*
X*
X
X
X
Fromme et al. (2010)
X
X
X
Kim et al. (2011)
X
X
X
Whitworth et al. (2012)
X
X*
X*
X
X
X
Maisonet et al. (2012)
X
X*
X*
X
X
X
Chen et al. (2012)
X
X*
X*
X
X
X
Dai'i'nu el al (201 i)
X
X
X
X
l!adi el al. (2(>l<>)
X*
X
X
X
Leiilersj el al. (20 loj
X*
X
X
X
Monrov el al. (2008)
X*
X
X
Robledo el al. (2015) '
X*
X
X
X
Wu et al. (2012)
X
X
Savitz et al. (2012)
x**
X
Callan el al 12t) 1 <• i
X
X
X
(iinails el al (2(>l<>)
X
xd
k\u»n el al (2n|(i)
X
X
l.eeelal <2t) 1 <• i
X
X
X
Waimelal <2<> 1 <• >
X
X
X
\1inaln\a el al (2i> 1 ~i
X
X
Shi el al (2u|"i
X
X
X
Maii/aiKi-Saluacki el al (2d 1 ~i
X
X
X
( lien el al (2(H"i
X
X
X
Siailniu el al (2(>l "i
X
X
xd
Sam\ el al (2(HSi
X
X
xd
\shle\ -\lailiii el al (2(i 1 "i
X
X
l.aiiiil/eii el al (2(H"i
X
X
\1 l.i el al. i:nl"i
X
X
1 .ind el al (2(il")
X
X
Val\ i el al (2(H"i
X
X
Can el al (2(HSi
X
X
Proposed PFAS Rule Economic Analysis
D-4
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table D-l: Data Sources for PFOA/PFOS Meta-Analyses of Birth Weight Effects
PFOA/PFOS-BW Relationship Studies Included in Meta-Analyses for Effects on BW
Study
Johnson et al.
Verner et al.
Negri et al.
Steenland et al.
Dzierlenga
EPA/OST Review
(2014)
(2015)
(2017)
(2018)
(2020)
(PFOA/PFOS) (2021)a
Meng et al. (2018)
X
X
Marks et al. (2019)
X
X
Workman et al. (2019)
X
Xu et al. (2019)
X
Bell et al. (2018)
X
Louis et al. (2018)
X
Gao et al. (2019)
X
Chu et al. (2020)
xd
Hjcrmitslcv et al. (2020)
X
Kashino et al. (2020)
X
Wikstrom et al. (2020)
xd
Abbreviations: BW - birth weight; EPA/OST- U.S. Environmental Protection Agency Office of Science and Technology; PFOA - perfluorooctanoic acid; PFOS -
perfluorooctane sulfonic acid.
Notes:
aEPA/OST evaluation of study quality reflected in blue (high confidence), green (medium confidence) or pink (low confidence) cell shading. EPA/OST literature review
focused on literature published between 2000 and 2020. Studies in this field reflect the studies EPA reviewed to select those that were used for modeling.
* Indicates a data set used for PFOS, as well as PFOA meta-analysis.
** Indicates a data set included only in sensitivity analysis.
m< fIndicates results presented only stratified by sex or location [e.g., Lauritzen et al. (2017)].
d Indicates studies used by EPA/OST for derivation of point of departures (PODs).
Proposed PFAS Rule Economic Analysis
D-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The authors used different techniques to evaluate sources of variability in the meta-analyses. As
expected, random effects models generated results with lower heterogeneity (as measured by the
proportion of between-study variance in the data sets) than fixed effects models. Each of the
meta-analyses reported sensitivity analyses, stratified analyses, or leave-one-out results
(influence analyses) to explore the relative contributions of individual or groups of studies to the
quantitative pooled estimates of PFOA and PFOS effects on birth weight.
Johnson et al. (2014) reported a pooled beta across nine included studies of-18.9 g (95%CI: -
29.8, -7.9) for PFOA per each 1 ng/mL. Johnson et al. (2014) used well-documented meta-
analytical methods: random effects models with inverse variance weighting. In addition, Johnson
et al. (2014) conducted analyses omitting several small studies with relatively high ROB, as well
as one that included a large study (Savitz et al., 2012) that modeled maternal serum levels based
on historical exposures, rather than measured exposures. Johnson et al. (2014) found that
inclusion or exclusion of high-ROB studies and studies based on modeled serum levels resulted
in only a small effect on the estimated slope factor for PFOA-birth weight relationships (Johnson
et al. (2014) reported a pooled beta across nine included studies of-18.9 g (95%CI: -29.8, -7.9)
for PFOA per each 1 ng/mL. Johnson et al. (2014) used well-documented meta-analytical
methods: random effects models with inverse variance weighting. In addition, Johnson et al.
(2014) conducted analyses omitting several small studies with relatively high ROB, as well as
one that included a large study (Savitz et al., 2012) that modeled maternal serum levels based on
historical exposures, rather than measured exposures. Johnson et al. (2014) found that inclusion
or exclusion of high-ROB studies and studies based on modeled serum levels resulted in only a
small effect on the estimated slope factor for PFOA-birth weight relationships (Figure D-l).6
Verner et al. (2015) reported a pooled beta across seven included studies of-5.00 g (95% CI: -
8.92, -1.09) for PFOS and -14.72 g (95% CI: -21.66, -7.78) for PFOA each per each 1 ng/mL. In
addition, Verner et al. (2015) also investigated the potential impact of changing glomerular
filtration rate (GFR), an index of kidney function, on PFAS-birth weight relationships. They
based their analysis on the fact that maternal GFR and blood volume are known to change across
the three trimesters of pregnancy in such a way that the assumed independent effect of GFR on
birth weight, coupled with changes in PFAS excretion rates, could account for part of the birth
weight reduction found in the epidemiological studies of PFAS exposure. In addition to a
standard meta-analysis, they simulated PFOA/PFOS levels in a hypothetical population, using a
pharmacologically based pharmacokinetic (PBPK) model, and evaluated the impact of changes
in GFR on PFAS-associated changes in birth weight across trimesters. The results of the
conventional meta-analysis for the overall effects of PFAS on birth weight were similar to those
derived by Johnson et al. (2014) (Figure D-l). Verner et al. (2015) concluded, however, that a
portion of the observed association may be attributable to confounding by GFR, with the effect
of GFR increasing across trimesters. This suggested that studies which have not controlled for
GFR might overestimate the impact of prenatal exposure to PFAS on fetal growth.
6 Note that this finding may not apply to all meta-analyses, especially if they did not use the exact studies and same
ROB methods as those employed in Johnson et al. (2014).
Proposed PFAS Rule Economic Analysis
D-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
All of the simulations employed different assumptions related to variability in PFOA/PFOS
levels and the strength of GFR impacts on birth weight. The simulated estimated relationships
between PFOA/PFOS and birth weight remained negative for all sample collection times, except
for the initial sampling time (at conception).
Negri et al. (2017) reported a pooled beta across eight included studies of -0.92 g (95% CI: -3.4,
1.6) for PFOS and twelve included studies of-12.8 g (95% CI: -23.2, -2.4) for PFOA each per
each 1 ng/mL. Negri et al. (2017) conducted random effects meta-analyses based on 14 studies.
In addition to the main analysis, Negri et al. (2017) conducted a sensitivity analysis related to
model form (fixed versus random effects), degree of adjustment (full, defined as adjustment for
infant sex, gestational age, maternal age, pre-pregnancy body mass index, education, parity,
and smoking, versus partial, which includes only some of these covariates), and location of
populations (America, Asia, and Europe). They also ran separate analyses for studies in which
the time of blood sampling varied (1st and 2nd trimester, 3rd trimester, and cord blood), to
further investigate the potential impacts of time of blood sampling as a proxy for changes in
GFR. Negri et al. (2017) found that the degree of adjustment had relatively little effect on the
magnitude of estimated slopes for PFOA and PFOS. The pooled PFOA/PFOS effect estimates
(i.e., beta coefficients) for studies in which sampling occurred late in pregnancy reported birth
weight decreases larger magnitude than for those where sampling occurred in the first two
trimesters, but the results were quite uncertain due to the small numbers of studies with late-term
sampling.
Steenland et al. (2018) reported a pooled beta across twenty-four included studies of-10.5 g
(95% CI: -16.7, -4.4) for PFOA per each 1 ng/mL. Steenland et al. (2018) conducted a random
effects meta-analysis based on 24 studies. In addition, they estimated PFOA slope factors
separately for studies of maternal and cord blood and for studies where PFOA serum levels were
measured in the first trimester versus any time later in pregnancy (Figure D-l). The slope factor
from the main analysis was significantly negative and similar in magnitude to that derived by
Negri et al. (2017). Coefficients for maternal blood were slightly smaller in magnitude than in
studies where cord blood was sampled, but still negative. The coefficient for the nine data sets
where blood PFOA was measured during the first trimester was small in magnitude
(-3.3 g per ng/mL), but not significant.
The most recent meta-analysis from Dzierlenga, Crawford, et al. (2020) reported a pooled beta
across thirty-two included studies of -3.2 g (95% confidence interval: -5.1, -1.3) for PFOS per
each 1 ng/mL. The study conducted a random effects meta-analysis based on 32 results from 29
studies. The authors of the analysis estimated a slope of-3.2 g birth weight per ng PFOS/mL
(95%) confidence interval: -5.1, -1.3) with significant moderate heterogeneity (12 = 58%).
Sensitivity analyses suggested that the results are sensitive to timing of blood samples. Among
those with blood measurements before or early in pregnancy, however, PFOS was inversely
associated with birth weight (-1.35, 95% confidence interval: -2.33, -0.37), and for the later
pregnancy group, the association was -7.17 (95% confidence interval: -10.93, -3.41).
Proposed PFAS Rule Economic Analysis
D-7
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Estimate
Mean (g per ng/mL)
Lower CI
Upper CI
Number of Studies
Heterogeneity 12
p-value
Johnson el al. (2014)
PFOA - Main
-18.9
-29.8
-7.9
9
38
0.12
—
PFOA - High ROB study included
-15.4
-26.5
-4.3
10
72
0
Verner etal. (2015)
PFOA - Main
-14.7
-21.7
-7.8
7
-
>0.05
PFOA - Adjusted for GFR
-7.9
-9.4
-6.4
7
-
-
¦
Negri et al. (2017)
PFOA - Main
-12.8
-23.2
-2.4
12
52.9
0.016
—
PFOA - First/second trimester
-10.5
-23.6
2.6
6
-
-
PFOA - Third trimester
-20
-52.1
12.1
2
•
-
PFOA - Cord Blood
-35.3
-101
30.7
4
-
-
<¦—« [-
Steenland et al. (2018)
PFOA - Main
-10.5
-16.7
-4.4
24
63
<0.0001
PFOA - First Trimester
-3.3
-9.6
3
7
68
<0.0001
4
PFOA - Second/Third trimester
-17.8
-25
-10.6
17
29
0.13
—
PFOA - Include Savitz (2012)
-1
-2.4
0.4
25
-
-
•
Verner et al. (2015)
PFOS - Main
-5
-8.9
-1.09
7
-
<0.05
•-
PFOS - Adjusted for GFR
-1.5
-1.8
-1.1
7
-
-
¦
Negri et al. (2017)
PFOS - Main
-0.92
-3.4
1.6
8
74.3
<0.001
•
PFOS - First/second trimester
0.6
-1.4
2.5
5
-
-
•
PFOS - Third trimester
-4
-16.3
8.2
2
-
-
PFOS - Cord Blood
-11.3
-17.4
-5.2
1
-
-
—
Dzierlenga et al. (2020)
PFOS - Main
-3.2
-5.1
-1.3
32
58
0
•
PFOS - Before or early in pregnancy
-1.35
-2.33
-0.37
10
5
0.4
¦
PFOS - Later pregnancy
-7.17
-10.39
-3.41
22
55
0.001
I—!—I—I—I—I—I—I—I—I
-50-40-30-20-10 0 10 20 30 40
Figure D-l: Results and Confidence Limits from PFOA, PFOS Meta-Analyses:
Changes in BW (grams) per Change in Serum PFAS Levels (ng/mL)
Proposed PFAS Rule Economic Analysis D-8 March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D.3 Exposure-Response Functions Based on Epidemiological
Studies
EPA selected the exposure-response result for PFOA from the main analysis reported by
Steenland et al. (2018) for use in the risk assessment from exposure to PFOA and benefits
analysis of reducing PFOA in PWS even though this study did not use a systematic ROB
analysis of the studies included in the meta-analysis. Although Negri et al. (2017) employed a
systematic ROB analysis for the studies included in the meta-analysis and showed moderate
heterogeneity among studies (I2 = 38%)7, EPA did not select it because the study is less recent
and includes fewer studies than Steenland et al. (2018). The Agency selected the main (random
effects) analysis from Steenland et al. (2018) because it is the most recent meta-analysis on
PFOA-birth weight and included the largest number of studies. The pooled beta estimate for
PFOA effects on birth weight in Steenland et al. (2018) is -10.5 g (95% confidence interval: -
16.7; -4.4) birth weight per ng serum PFOA/mL based on 24. The Agency also uses the 95%
confidence limits of -16.7 and -4.4 g birth weight per ng PFOA/mL as lower and upper bound
slope estimates for a sensitivity analysis. The pooled mean estimate (g birth weight per ng
PFOA/mL) for all studies is in the midrange of the results for the early, middle, and late blood
sampling results (Figure D-l).
EPA selected the exposure-response result for PFOS from the most recent meta-analysis of 32
observations from 29 publications reported by Dzierlenga, Crawford, et al. (2020) for use in the
risk assessment from exposure to PFOS and benefits of reducing PFOS in PWS.8 The Agency
chose the main analysis from Dzierlenga, Crawford, et al. (2020) because it considered the
largest number of recent studies, the heterogeneity among studies was moderate (I2 = 58%), and
sensitivity analyses suggested an inverse relationship with birth weight. Additionally, sensitivity
analyses suggested that the results were not particularly sensitive to timing of blood samples,
consistent with the early pregnancy subgroup analysis result. Dzierlenga, Crawford, et al. (2020)
also examined study quality aspects by characterizing studies with respect to certain
characteristics.9 that might influence results and examining those in meta-regression analyses.
712 represents the proportion of total variance in the estimated model due to inter-study variation; a value of 38 percent is
considered "moderate", suggesting that the studies are not seriously inhomogeneous and that a pooled model (meta-analysis)
is appropriate.
8 Although Negri et al. (2017) also estimated an exposure-response slope for PFOS effects on BW based on eight studies,
the analysis includes a slope factor derived from the Maisonet et al. (2012) study that was given as (positive) 5.77 (95%
confidence limits = 2.01, 9.53). However, in the original Maisonet et al. (2012) study, the relationship between maternal PFOS
and female infant BW was reported as being negative; it appears that there was a transcription error in the Negri et al. (2017)
analysis.8 An sensitivity analysis from Negri et al. (2017) that excluded the Maisonet et al. (2012) study resulted in a pooled
estimate of -2.0 g BW per ng/mL PFOS, which is similar in magnitude to the estimate reported by Dzierlenga, Crawford, et al.
(2020). Also, although the estimated slope factor for PFOS effects from Verner et al. (2015), based on seven studies, included the
slope factor from Maisonet et al. (2012) as (negative) -5.77 g BW per ng PFOS/mL (95% confidence limits -9.53, -2.01),
Dzierlenga, Crawford, et al. (2020) includes a larger number of studies, many of which were published more recently than those
considered in Negri et al. (2017) and Verner et al. (2015) (32 results from 29 studies conducted from 2007 to 2019, compared to
seven and eight studies considered in Negri et al., 2017 and Verner et al., 2015, respectively, that were conducted from 2007
to 2016).
9 For example, the quality of evidence was characterized as low for the BW-PFOS associations when the timing of blood draw
was before or early in pregnancy.
Proposed PFAS Rule Economic Analysis
D-9
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
EPA reanalyzed the pooled estimate from this study after determining that the original
Dzierlenga, Crawford, et al. (2020) pooled estimate included a duplicated estimate from Chen et
al. (2017). EPA reran the analysis excluding the duplicated estimate to obtain a slope of-3.0 g
birth weight per ng PFOS/mL with the same heterogeneity (I2 = 58%) as the prior estimate (p-
value for heterogeneity <0.001).
Proposed PFAS Rule Economic Analysis
D-10
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix E. Effects of Reduced Birth Weight on
Infant Mortality
This appendix summarizes EPA's analysis of the relationship between infant mortality and birth
weight. This relationship is fundamental in estimating benefits from changes in birth weight
among infants whose mothers were exposed to PFOA or PFOS during or prior to pregnancy.
EPA developed a cross-sectional model to quantify this relationship based on the most recent
2016/17 and 2017/18 Centers for Disease Control (CDC) Period Cohort Linked Birth-Infant
Death Data files.
E.l Birth Weight-Mortality Relationship
Low birth weight (LBW), defined as weight at birth <2,500 grams, is recognized as a significant
predictor of infant mortality (McCormick, 1985; World Health Organization, 2014).
The majority of infants born with LBW are premature, but other gestational factors such as
maternal hypertensive disorders and anemia can result in full-term infants who are born at LBW
(Joyce et al., 2012). Many of the top 10 causes of infant mortality are factors associated with
preterm birth, including LBW (Jacob, 2016). Advances in U.S. prenatal and neonatal care and
successes in public health initiatives, such as those designed to decrease maternal smoking, have
increased LBW survival rates and reduced the prevalence of LBW infants (Callaghan et al.,
2017; Singh et al., 2019). To quantify potential mortality impacts from changes in infant birth
weight resulting from changes in maternal PFOA and PFOS exposure via drinking water, robust
data supporting a relationship between incremental changes in infant birth weight and mortality
risk are needed.
A number of epidemiological studies in the U.S. have reported relationships between birth
weight and mortality. However, most of these studies evaluate relationships between infant
mortality and birth weight above or below various birth weight thresholds (e.g., Mclntire et al.,
1999; Lau et al., 2013). EPA identified only two studies that show statistically significant
relationships between incremental changes in birth weight and infant mortality that can be
leveraged for PFOS/PFOA health impact modeling: Ma et al. (2010) and Almond et al. (2005).
Ma et al. (2010) used 2001 National Center for Health Statistics/National Vital Statistics System
(NCHS/NVSS) linked birth/infant death data for singleton and multiple birth infants among
subpopulations defined by sex and race/ethnicity to estimate a regression model assessing the
associations between 14 key birth outcome measures, including birth weight, and infant
mortality. They found notable variation in the relationship between birth weight and mortality
across race/ethnicity subpopulations, with odds ratios for best-fit birth weight-mortality models
ranging from 0.8-1 per 100 gram (g) birth weight change. Almond et al. (2005) used 1989-1991
NCHS linked birth/infant death data for multiple birth infants to analyze relationships between
birth weight and infant mortality within birth weight ranges. For their preferred model, they
reported coefficients in deaths per 1,000 births per 1 g increase in birth weight that range from -
0.420 to -0.002.
However, the data used in these studies (Almond et al., 2005 and Ma et al., 2010) are old (1989-
1991 and 2001, respectively). Given the significant decline in infant mortality over the last 30
years (discussed in Section E.2 below), and changes in other maternal and birth characteristics
Proposed PFAS Rule Economic Analysis
E-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
that are likely to influence infant mortality (e.g., average maternal age and rates of maternal
smoking), the birth weight-mortality relationship estimates from Almond et al. (2005) and Ma et
al. (2010) are likely to overestimate benefits of birth weight changes. Moreover, Almond et al.
(2005) focused on multiple birth infants to analyze relationships between birth weight and infant
mortality.
LBW is determined by two main processes: duration of gestation and rate of fetal growth
(Institute of Medicine, 1985; Quah, 2016). Thus, infants can be LBW because they are born
preterm or are born small for gestational age, which is a proxy for intrauterine growth
retardation. Researchers have found that birth weight and gestational age are closely associated
but not perfectly correlated (e.g., Kiely et al., 1994; Mathews, 2013). A study by Almond et al.
(2005) found that gestational age is an important determinant of birth weight as it explains over
half of the overall variance in birth weight among a pooled sample of twins. Moreover, multiple
studies suggest that, when available, both birth weight and gestational age should be included
when predicting infant mortality odds (Almond et al., 2005; Ma et al., 2010; Ray et al., 2017).
Cole et al. (2010) developed a logistic regression model showing that gestational age and birth
weight z-score.10 were the strongest predictors of survival among very preterm infants. Ma et al.
(2010) predicted infant mortality by combining birth weight and gestational age variables to
distinguish between the two major causes of LBW. Ray et al. (2017) used modified Poisson
regression to show that singleton infants who are born preterm and small for gestational age have
a higher risk of neonatal death than infants born preterm alone.
The CDC indicated that the mortality rate among multiples is very high for reasons that are often
unrelated to birth weight and recommended that a model based on singletons may provide a
more representative relationship between birth weight and infant mortality (Communication with
Horon, 2020). Studies of birth weight-specific infant mortality among singletons and multiples
suggest that, due to differences in intrauterine growth restriction, prematurity rates, and zygosity,
analyses that examine perinatal outcomes should be stratified by plurality (Russell et al., 2003;
Cooke, 2010). Furthermore, singleton infants represent the majority of U.S. births (96% of
infants born in 2016 and 2017). Following CDC's recommendations, EPA developed cross-
sectional models to estimate a relationship between birth weight at four distinct gestational age
categories and infant mortality based on the most recently available 2016-2018 NCHS/NVSS
data and focusing on singleton infants. To identify variation in the birth weight-mortality
relationship across race/ethnicity subpopulations, EPA estimated separate relationships for non-
Hispanic Black, non-Hispanic White, and Hispanic subpopulations.
In developing the singleton models, EPA used similar variables and partitioning techniques as
detailed in Ma et al. (2010). Specifically, EPA developed separate models for different
race/ethnicity categories and interacted birth weight with gestational age. Ma et al. (2010) found
that key predictors of infant mortality include birth weight, Apgar score,.11 and gestational age.
Ma et al. (2010) developed multivariate logistic regression models for gender- and race-specific
subpopulations.12 to assess associations of various combinations of birth weight, gestational age,
10 Z-scores describe how far from the mean a given data point is.
11 Apgar score refers to a metric indicating the health of a newborn. The score, which ranges from 0 to 10, is based on skin color,
heart rate, reflexes, muscle tone, and breathing rate/effort.
12 Separate models were fit for non-Hispanic white girls, non-Hispanic white boys, non-Hispanic black girls, non-Hispanic black
boys, Mexican girls, and Mexican boys.
Proposed PFAS Rule Economic Analysis
E-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
fetal growth rate, and Apgar scores with four mortality outcomes (infant mortality, early
neonatal, late neonatal, and post-neonatal mortality). In addition to these covariates, Ma et al.
(2010) automatically selected covariates such as parental characteristics (e.g., maternal age and
education), maternal risk factors (e.g., smoking), and child characteristics (e.g., birth order)
based on predictive power. Ma et al. (2010) showed that the baseline rates of each birth outcome
differ by both race/ethnicity and postnatal period. Model results indicated that birth weight is a
stronger predictor of infant mortality among the non-Hispanic Black subpopulation compared to
the non-Hispanic White and Hispanic subpopulations.
E.2 Basis for Updated Birth Weight-Mortality Relationship
There has been a notable decline in U.S. infant mortality rates during the two decades since
analyses reported in Ma et al. (2010) and Almond et al. (2005). In the last 30 years, overall infant
mortality rates have declined steadily (ICF, 2020).13 The infant mortality rate in 2018 was
5.67 per 1,000 live births, while the infant mortality rate in 1991 was 8.6 per 1,000 live births.
Except for infants born with birth weight lower than 500 grams, for whom mortality rates have
not changed considerably, mortality rates for infants with birth weight greater than 500 grams are
decreasing and converging on a low rate..14
Given a decline in infant mortality in the birth weight categories lower than 1,500 g, a unit
change in birth weight is likely to produce less of an impact on the probability of mortality in
2016-2018 compared to 1989-1991 (the years evaluated in Almond et al., 2005) or 2001 (the
year evaluated in Ma et al., 2010). Despite recent declines in U.S. infant mortality, disparities in
infant mortality experience continue to exist across race/ethnicity subpopulations (Osterman et
al., 2015). Recent research indicates that infant mortality is consistently highest among Black
infants (both Hispanic and non-Hispanic), while non-Hispanic White and Hispanic White infants
have the lowest mortality rates (Rice et al., 2017; Rowley et al., 2012; Collins Jr et al., 2009).
In addition to the decline in infant mortality in LBW categories, other maternal and birth
characteristics that are likely to influence infant mortality have evolved over time. Almond et al.
(2005) provided sample means for birth and maternal characteristics for singletons based on the
1989 NCHS/NVSS Linked Natality-Mortality Detail file. EPA provides similar statistics for
singletons from the 2016-2018 NCHS/NVSS Period/Cohort Linked Birth-Infant Death Data
Files.15 that demonstrate how birth and mortality characteristics have changed over time.
Table E-l shows a subset of the 1989 sample means among singletons born to non-Hispanic
Black and non-Hispanic White mothers from Almond et al. (2005) Table II and the same
statistics derived from the 2016-2018 data. The comparison shows that teen pregnancy rates,
pregnancy among mothers with less than a high school education, and maternal smoking during
pregnancy have decreased since 1989. While mean and median birth weight has decreased
13 CDC publishes National Vital Statistics Reports that summarize mortality trends over time (e.g. Kochanek et al., 2019) and
provides detailed tables of infant mortality trends by race and age at death in annual Health, United States reports (National
Center for Health Statistics, 2019).
14 EPA assembled summary statistics on infant mortality by BW category provided in the documentation for 1983-2018 Linked
Infant Birth-Death Detail Files. These files are published on the online data portal by NCHS/NVSS:
https://www.cdc.gov/nchs/data_access/vitalstatsonline.htm
15 The 2016-2018 NCHS/NVSS Period/Cohort Linked Birth-Infant Death Data Files represent two separate datasets.
The 2016/2017 data includes infants born in 2016 and follows their mortality experience for one year (through the end of 2017).
The 2017/2018 data includes infants born in 2017 and follows their mortality experience through the end of 2018.
Proposed PFAS Rule Economic Analysis
E-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
slightly for singleton infants, the 1-year infant mortality rate has decreased by about 42%.
Possible explanations for this trend may include advancements in prenatal and postnatal care
(e.g., advances in infection control practices and the use of intubation to reduce infant lung
injury; Callaghan et al., 2017) as well as positive effects of public health education (e.g., reduced
smoking during pregnancy; Singh et al., 2019).
In addition to a decreasing 1-year mortality rate, Table E-l shows a decrease in the fraction of
infants with congenital anomalies and a decrease in median gestational age. The decrease in
gestational age is supported by analysis from Donahue et al. (2010), who found that gestational
age among full-term singletons in the United States decreased by more than two days from
1990-2005.
Table E-l: Comparison of Sample Means for Singletons between the 1989
Natality-Mortality Detail File and the combined 2016-2018 Period/Cohort Linked
Birth-Infant Death Data Files
Variable
Sample Meansabc
1989
2016-2018 (% Change)
Sample size
2,655,977
4,212,764
Infant deaths (per 1000 live births)
Within 1 year of birth (infant mortality)
8.46
4.94 (-42%)
Within 28 days (neonatal)
4.99
2.94 (-41%)
28 days to 1 year (postneonatal)
3.49
2.00 (-43%)
Fraction of dead with birth weight < 2500 g
Infant mortality
0.570
0.592 (+4%)
Within 24-hour mortality
0.890
0.285 (-68%)
Neonatal mortality
0.760
0.463 (-39%)
Postneonatal mortality
0.300
0.129 (-57%)
Infant birth weight (g)
Mean
3,369
3,313 (-2%)
Median
3,402
3,345 (-2%)
5th percentile
2,410
2,390 (-1%)
Fraction LB W (<2500 g)
0.061
0.065 (+7%)
Gestational age (in weeks)
Mean
39
39 (0%)
Median
40
39 (-3%)
5th percentile
35
35 (0%)
Characteristics of birth
5-minute Apgar score (0-10)
8.97
8.79 (-2%)
Fraction male
0.512
0.512(0%)
Fraction congenital anomalyd
0.019
0.001 (-93%)
Mother's demographic characteristics
Fraction Black
0.195
0.193 (-1%)
Proposed PFAS Rule Economic Analysis E-4 March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-l: Comparison of Sample Means for Singletons between the 1989
Natality-Mortality Detail File and the combined 2016-2018 Period/Cohort Linked
Birth-Infant Death Data Files
Variable
Sample Meansabc
1989
2016-2018 (% Change)
Fraction high school dropout
0.184
0.085 (-54%)
Fraction college graduate
0.187
0.451 (+141%)
Age
26.3
28.6 (+9%)
Fraction teenager
0.129
0.049 (-62%)
Fraction 30+
0.289
0.444 (+54%)
Fraction married
0.736
0.595 (-19%)
Mother's risk factors
Number of prenatal visits
11.2
11.5 (+3%)
Fraction smoke during pregnancy
0.212
0.100 (-53%)
Abbreviations BW - birth weight; LBW - low birth weight.
Notes:
aThe data are restricted to non-Hispanic Black and White mothers born in the United States, as reported in Almond et al.
(2005) Table II.
bThe 1989 data summary in Almond et al. (2005) included anemia of mother, assisted ventilation (<30 minutes) and assisted
ventilation (>= 30 minutes), which are not included in the 2016-2018 NCHS/NVSS dataset. The 2016-2018 NCHS/NVSS
dataset does include assisted ventilation and assisted ventilation (6 hours), but these variables are not necessarily comparable
to the assisted ventilation variables included in the 1989 NCHS/NVSS dataset. Similarly, 1989 data summary in Almond et al.
(2005) included "pregnancy-associated hypertension" which is further split up into "gestational hypertension" and
"hypertension eclampsia" in the 2016-2018 NCHS/NVSS dataset. Due to differences in variable definitions among the data,
EPA excludes hypertension.
cRecords with "Unknown" or "Not Stated" values not included in the 2016-2018 summary.
"tongenital anomalies among the 1989 and 2016-2018 data are not directly comparable due to differences in the congenital
anomalies included in this metric between the datasets. The 1989 dataset includes the following congenital anomalies:
Anencephalus, spina bifida/meningocele, hydrocephalus, other central nervous system anomalies, heart malformations, other
circulatory /respiratory anomalies, rectal atresia/stenosis, trachea-esophageal fistula/esophageal atresia,
omphalocele/gastroschisis, other gastrointestinal anomalies, malformed genitalia, renal agenesis, other urogenital anomalies,
cleft lip/palate, Polydactyly, club foot, diaphragmatic hernia, other musculoskeletal/integumental anomalies, down's
syndrome, other chromosomal anomalies, and other congenital anomalies. The 2016-2018 dataset includes the following
congenital anomalies: anencephaly, meningomyelocele/spina bifida, cyanotic congenital heart disease, congenital
diaphragmatic hernia, omphalocele, gastroschisis.
The remainder of this appendix summarizes the development of regression models implemented
using newer data.
E.3 Development of the Analytical Dataset
E.3.1 Do to Sources
This analysis relies on Period/Cohort Linked Birth-Infant Death Data Files published by
NCHS/NVSS from the 2017 period/2016 cohort and the 2018 period/2017 cohort..16 Each dataset
includes files linking all infant deaths during the period and cohort years to information from
corresponding birth certificates and separate files consisting of all births occurring during the
period. The data include all infants under 1 year of age in the U.S. or its territories (Centers for
16 https://www.cdc.gov/nchs/data access/vitalstatsonline.htm
Proposed PFAS Rule Economic Analysis
E-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Disease Control and Prevention, 2017f). This analysis excludes multiple birth infants. In addition
to infant birth and mortality information, the data include details on maternal characteristics
(e.g., mother's education, marital status, and age category), maternal risk factors (e.g., smoking
status), and pregnancy and birth characteristics (e.g., gestational age, infant birth weight,
presence of congenital anomalies, and birth order).
E.3.2 Dataset Development
EPA combined the infant birth and death files using the SAS code examples from the user guides
accompanying the datasets to create user-created cohort files, which follow the birth cohorts for
an entire year to ascertain their mortality experience (Centers for Disease Control and
Prevention, 2017f, 2018). At this stage, EPA also selected variables of interest for the regression
analysis. These variables include maternal demographic and socioeconomic characteristics,
maternal risk and risk mitigation factors, and infant birth characteristics. EPA included several
variables used in Ma et al. (2010) as well as additional variables to augment the set of covariates
included in the regression analyses. Variable selection was informed by literature on the leading
causes of infant mortality (e.g., Ahrens et al., 2017; Mishra et al., 2018; Centers for Disease
Control and Prevention, 2020a, 2020b; Ely et al., 2020).
E.3.3 identification of infant Mortality Risk Factors
To identify infant mortality risk factors for inclusion in the regression analyses, EPA relied on
multiple data sources, including key risk factors identified by the CDC and prior studies of the
relationship between infant mortality and various maternal and birth characteristics. Although
risks to infant mortality include conditions related to infant and maternal health, demographic
and socioeconomic characteristics also contribute to infant mortality outcomes. Based on the
studies EPA reviewed, infant mortality risk factors generally fall within three general categories
described below:
• Birth Characteristics:
o Birth Weight and Gestational Age: The CDC identifies preterm birth and LBW as
leading causes of infant death in the United States (Ely et al., 2020). The majority
of infant deaths in 2018 occurred among infants born preterm (gestational age <
37 weeks; Ely et al., 2020). Previous studies of the relationship between birth
weight and infant mortality identify birth weight and gestational age as important
predictors of infant mortality (e.g., Almond et al., 2005; Ma et al., 2010).
o Other Infant Birth Characteristics: Studies of leading causes of infant mortality
suggest that birth order plays a significant role in infant mortality outcomes.
Higher birth order is linked to risk of injury and may be indicative of other
socioeconomic factors (Ahrens et al., 2017; Mishra et al., 2018). Another
substantive predictor of infant mortality is five-minute Apgar score (Almond et
al., 2005; Ma et al., 2010). Birth defects, such as the presence of congenital
anomalies, also contribute to infant mortality (Ely et al., 2020).
Proposed PFAS Rule Economic Analysis
E-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• Maternal Risk and Risk Mitigation Factors: Many causes of infant death are
exacerbated by tobacco use, substance use, and stress (Centers for Disease Control and
Prevention, 2020a). CDC guidance suggests that regular prenatal care visits.17 lead to
detection of infant mortality risk factors (e.g., hypertension).
• Maternal Demographic and Socioeconomic Characteristics: Infant birth outcomes are
influenced by demographic and socioeconomic factors such as maternal race/ethnicity,
age, education, and marital status (Ma et al., 2010). Infant mortality rates vary for
mothers of different ages, with the lowest mortality rates among mothers age 30-34 and
highest mortality rates among teen mothers and mothers over 40 in 2018 (Ely et al.,
2020). Singh et al. (2019) found that the risk of 1-year mortality in 2016 was 3.7 times
greater for mothers with less than 12 years of education than for mothers with 16 or more
years of education. Marital status also influences the risk of infant mortality—studies
show that the risk of infant mortality increases when one parent is absent (Ngui et al.,
2015; Alio et al., 2011). In 2018, the non-Hispanic Black subpopulation had the highest
infant mortality rate at 10.8 deaths per 1,000 live births, while Hispanic and non-Hispanic
White subpopulations experienced much lower rates of infant mortality (4.9 and 4.6
deaths per 1,000 births, respectively; Ely et al., 2020).
While maternal risk variables such as hypertension, diabetes, and infection lead to premature
birth, LBW, and reduced motor function, birth-related factors such as Apgar score, birth weight,
and gestational age likely account for these risks (Backes et al., 2011; Centers for Disease
Control and Prevention, 2016c; M. Li et al., 2017). Given that birth weight impacts on infant
mortality are the focus of our analysis, selected covariates do not include maternal risk factors,
such as maternal hypertension, diabetes, and infection, whose mortality influence pathway is
primarily through birth weight, gestational age, and Apgar score..18
E.4 Development of Variables
The dependent variable (BIRTHMORT) is a binary variable indicating whether the infant died
within one year of birth. Covariates included in the regression analyses fall under three
categories:
• Birth characteristics (denoted with BIRTH prefix)
• Maternal risk and risk mitigation factors (denoted with MRF prefix)
• Maternal demographic and socioeconomic characteristics (denoted with MDEM prefix)
Table E-2 provides a detailed description of all variables included in the singleton regression
analysis and the corresponding variables from the NCHS/NVSS data used to develop the
variables. EPA estimated different regression models for three race/ethnicity subpopulations:
Non-Hispanic Black, non-Hispanic White, and Hispanic. Infants whose mothers fall into these
race/ethnicity subpopulations are identified using the MRACEHISP variable from the
NCHS/NVSS data.
17 While prenatal care visits fall under the maternal risk and risk mitigation factors category, it could also be considered a
maternal demographic and socioeconomic characteristic indicative of access to care.
18 Pearson correlation tests indicated significant relationships between these variables (p-values < 5%).
Proposed PFAS Rule Economic Analysis
E-7
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The focus of EPA's analysis is the relationship between birth weight and infant mortality.
However, Ma et al. (2010) noted that the practice of specifying regression models that assume
that every 1-gram increase in birth weight has the same effect on infant mortality outcome
(regardless of gestational age or LBW status of the infant) has been challenged..19 Following
researchers who emphasize the importance of examining birth outcomes from the perspective of
combined birth weight and gestational age variables (Solis et al., 2000; Powers et al., 2006), Ma
et al. (2010) found that models with birth weight-gestational age interaction variables had higher
predictive power than models that only used birth weight and gestational age separately.
Following best practices from the health economic literature (e.g., Solis et al., 2000; Powers et
al., 2006; Ma et al., 2010), EPA interacted continuous birth weight with four gestational age
category indicator variables (extremely pre-term, very pre-term, moderately pre-term, term as
defined by the World Health Organization, 2018) to account for the heterogeneity in birth weight
impact with respect to the gestational age of the infant. EPA expected that birth weight effects
would be highest for extremely pre-term infants and lowest for full-term infants.
In addition to the set of birth weight-gestational age category interaction variables, EPA added
variables for other infant birth characteristics (birth order, birth year, sex, Apgar score,
congenital anomaly indicator), maternal risk and risk mitigation factors (smoker status,
categorized number of prenatal care visits), and maternal demographic and socioeconomic
characteristics (education, age, marital status). These variables control for additional factors
beyond birth weight and gestational age that contribute to the probability of infant mortality..20
EPA included categorized Apgar score variables based on analysis from Ma et al. (2010), who
found that Apgar scores, separated into low (0-3), medium (4-6), and high (7-10) categories,
were the strongest predictor of infant mortality among race/ethnicity-specific models. Further,
the 2016-2018 NCHS/NVSS data show that Apgar scores are significantly higher for non-
Hispanic White infants than for non-Hispanic Black infants. Ma et al. (2010) also found that the
inclusion Apgar scores in models predicting infant mortality significantly improved goodness of
fit. EPA also included a variable indicating whether the infant was born in 2016 or 2017
(BIRTH_YR_2016) as a control to determine whether there are any significant differences
between the 2016 and 2017 NCHS/NVSS datasets that are not readily captured by other
covariates.
Table E-2: Variables Used in Singleton Mortality Regression Analysis
Variable
Variable
Type
Variable Definition
Basis for Variable in
NCHS/NVSS Dataset
Dependent Variable
BIRTH MORT
Binary
Binary variable indicating whether
the infant died within one year of
birth
DODYY
19 Ma et al. (2010) indicate that birth weight effects vary according to the position on the distribution of birth weight
(they characterize the birth weight-mortality distribution as a reverse J-shaped distribution).
20 EPA also explored adding additional maternal risk factor variables, including maternal hypertension, diabetes, and infection,
based on CDC's identified infant mortality risk factors (see Section E.3.1.2). However, the inclusion of these variables in our
models produced counterintuitive results and they were eliminated from the covariate set.
Proposed PFAS Rule Economic Analysis
E-8
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-2: Variables Used in Singleton Mortality Regression Analysis
Variable
Variable
Type
Variable Definition
Basis for Variable in
NCHS/NVSS Dataset
Covariates
Birth Weight and GA
BIRTH BW I EXT PRETER
M
Discrete/
Continuous
Continuous BW (in grams) if
gestational age is <=28 weeks
(extremely preterm), 0 if otherwise
BRTHWGT,
COMBGEST
BIRTH BW I VER PRETER
M
Discrete/
Continuous
Continuous BW (in grams) if
gestational age is >28 weeks and
<=32 weeks (very preterm), 0 if
otherwise
BRTHWGT,
COMBGEST
BIRTH BW I MOD PRETER
M
Discrete/
Continuous
BW (in grams) if gestational age is
>32 weeks and <=37 weeks
(moderately preterm), 0 if otherwise
BRTHWGT,
COMBGEST
BIRTHBWITERM
Discrete/
Continuous
Continuous BW (in grams) if
gestational age is >37 weeks (term),
0 if otherwise
BRTHWGT,
COMBGEST
Other Infant Birth Characteristics3
BIRTH MALE
Binary
Binary variable indicating that the
infant is male
SEX
BIRTHCONANOM
Binary
Binary variable indicating that the
infant experienced one or more of the
following congenital anomalies:
anencephaly,
meningomyelocele/spina bifida,
cyanotic congenital heart disease,
congenital diaphragmatic hernia,
omphalocele, gastroschisis
CA ANEN,
CA MNSB,
CA CCHD,
CA CDH,
CA OMPH,
CAGAST
BIRTH_APGAR_0_3
Binary
Binary variable indicating that the
five-minute Apgar score is between 0
and 3. Five-minute Apgar score
indicates the health of a newborn
based on skin color, heart rate,
reflexes, muscle tone, and breathing
rate/effort.
APGAR5
BIRTH_APGAR_4_6
Binary
Binary variable indicating that the
five-minute Apgar score is between 4
and 6. Five-minute Apgar score
indicates the health of a newborn
based on skin color, heart rate,
reflexes, muscle tone, and breathing
rate/effort.
APGAR5
BIRTHYR2016
Binary
Binary variable indicating whether
the infant was born in 2016. If 0, the
infant was born in 2017.
N/A; based on CDC
dataset
BIRTHBOCatl
Binary
Binary variable indicating that the
infant has one sibling (second-born)
LBOREC
BIRTH_BOCat2
Binary
Binary variable indicating that the
infant has two or more siblings
(third- or later-born)
LBOREC
Proposed PFAS Rule Economic Analysis
E-9
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-2: Variables Used in Singleton Mortality Regression Analysis
Variable
Variable
Type
Variable Definition
Basis for Variable in
NCHS/NVSS Dataset
Maternal Risk and Risk Mitigation FactorsM
MRFNOPRECARE
Binary
Binary variable indicating that the
mother had no prenatal care visits
PREVIS
MRF19PRECARE
Binary
Binary variable indicating that the
mother had 1 to 9 prenatal care visits
PREVIS
MRF 16 ORMORE PRECAR
E
Binary
Binary variable indicating that the
mother had 16 or more prenatal care
visits
PREVIS
MRFSMOKE
Binary
Binary variable indicating that, if
maternal smoking status is known,
the mother was a smoker
CIGREC
Maternal Demographic and Socioeconomic Characteristicse d
MDEMINOHS
Binary
Binary variable indicating that the
mother's education is known and that
the mother did not graduate high
school or obtain a GED
MEDUC
MDEMICOLLEGEPLU S
Binary
Binary variable indicating that the
mother's education is known and that
the mother attended college or higher
education
MEDUC
MDEMAGETEEN
Binary
Binary variable indicating that the
mother's age is <20
MAGER
MDEMAGEAD V_3 5 40
Binary
Binary variable indicating that the
mother's age is >34 and <= 40
MAGER
MDEM_AGE_ADV_40plus
Binary
Binary variable indicating that the
mother's age is >40
MAGER
MDEM I MARRIED
Binary
Binary variable indicating that the
mother's marital status is known and
that the mother is married
DMAR
Abbreviations: BW - birth weight; NCHS - National Center for Health Statistics; NVSS - National Vital Statistics System.
Notes:
Reference categories for binary variables in the other infant birth characteristics category include female infants, infants who
did not experience a congenital anomaly, infants with Apgar scores from 7 to 10, infants born in 2017, and infants who have
no siblings.
bReference categories for binary variables in the maternal risk and risk mitigation factors category include mothers who had
10 to 15 prenatal care visits and mothers who do not smoke.
Reference categories for binary variables in the maternal demographic and socioeconomic characteristics category include
mothers who went to high school but who did not attend any college, mothers aged 25 to 34, and mothers whose marital status
is unknown or single.
dThe maternal age (MDEM_AGE) variables are split into three categories to show effects associated with teen mothers,
mother's aged 35 to 40, and mothers over the age of 40 with respect to the reference case of mother's aged 20 to 34. This is to
reflect differences in infant mortality rates associated with different maternal age groups. In 2018, the CDC indicated that total
mortality rates were highest for infants of mothers under age 20, while infants of mother's age 30-34 had the lowest mortality
rates (Ely et al., 2020). Infant mortality rates increased among infants born to older mothers, especially those over age 40 (Ely
et al., 2020).
Of the available singleton data, 0.8% had no race information. These records are excluded from
consideration. For regression modeling, records with incomplete or missing data (specified as
"Unknown" or "Not Stated" in the raw NCHS/NVSS data) for any of the covariates listed in
Proposed PFAS Rule Economic Analysis
E-10
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-2 were excluded from the analytical dataset. Records with incomplete or missing
covariate information account for 8.5% of the non-Hispanic Black records, 6.5% of the non-
Hispanic White records, and 7.0% of the Hispanic records (for a combined total of 7.0% of all
records). EPA did not attempt to fill in these data gaps using imputations or assumptions,
because records with missing data constituted less than 10% of all records. The resulting sample
sizes are: 981,212 for the non-Hispanic Black subpopulation, 3,644,499 for the non-Hispanic
White subpopulation, 1,646,713 for the Hispanic subpopulation.
E.5 Summary Statistics
Table E-3 presents maternal and infant characteristics of the study population, including number
and proportion of the sample associated with different age ranges, gestation weeks, races and
ethnicities, educational attainment, marital status, number of prenatal care visits, and whether or
not the mother smoked during pregnancy. Sample statistics indicate that the majority of mothers
are between ages 20 and 33, have full-term pregnancies, are non-Hispanic White, graduated high
school, had more than ten prenatal care visits, and did not smoke during pregnancy.
Proposed PFAS Rule Economic Analysis
E-ll
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-3: Maternal and Infant Characteristics of the Study Population
Description
N
Proportion (%)
Age
<20 years
343,784
5.48
20-33 years
4,606,124
73.43
34-39 years
1,138,646
18.15
40+ years
183,870
2.93
Gestation Week
<=28
43,654
0.70
>28 and <=32
80,408
1.28
>32 and <=37
106,8585
17.04
>37
5,079,777
80.99
Race/Ethnicity
Non-Hispanic White
3,644,499
58.10
Non-Hispanic Black
981,212
15.64
Hispanic
1,646,713
26.25
Education
No high school or GED
871,274
13.89
Graduated high school
2,963,900
47.25
Attended college3
2,437,250
38.86
Marital Status
Married
3,504,095
55.87
Unmarried
2,768,329
44.13
Number of Prenatal Care Visitsb
None
100,231
1.60
1-9
1,519,825
24.23
10-15
4,066,046
64.82
16+
586,322
9.35
Smoking During Pregnancy
Yes
455,758
7.27
No
5,816,666
92.73
Apgar Score
Apgar score between 0 and 3
32,518
0.52
Apgar score between 4 and 6
82,762
1.32
Apgar score between 7 and 10
6,157,144
98.16
Notes:
a Refers to mothers who obtained an associate's degree or more. Mothers who obtained some college credit but not a degree are
categorized in the "Graduated high school" field.
b Number of prenatal care visits in the study population range from 0 to 98.
Proposed PFAS Rule Economic Analysis
E-12
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.6 Estimation Methods
EPA fit the logistic regression model using Stata 15.1 (StataCorp, 2013a). The model is fit to
three different race/ethnicity singleton subpopulations (non-Hispanic Black, non-Hispanic White,
and Hispanic)21 as there are known disparities in the prevalence of LBW by race and ethnicity
(Collins Jr et al., 2009; Rice et al., 2017; Rowley et al., 2012; Ratnasiri et al., 2018). Coefficients
of non-linear regression models with a binary outcome indicate direction of the effect that
covariates have on outcome probability. That is, negative coefficients indicate that the
probability of mortality decreases as the covariate increases, while positive coefficients indicate
that the probability of mortality increases as the covariate increases.
In this analysis, EPA reported the results of regression modeling using both odds ratios 22 and
marginal effects. While the odds ratio is an effect metric commonly reported in epidemiological
research, the impact of a marginal change in the covariate on the probability of the outcome (i.e.,
the marginal effect) is easier to interpret. The magnitude of this marginal effect depends on all
estimated coefficients of the model as well as specific values of all the covariates included in the
model. When estimating marginal effects, EPA used actual observed values for the covariates
rather than using covariate means.23 For non-birth weight-gestational age variables, EPA
estimated marginal effects based on covariate values from all observations included in the
models. For birth weight-gestational age variables, EPA estimated marginal effects based on
covariate values from the subset of observations falling within each gestational age category (see
N columns for sample size used for each marginal effect calculation).24
Section E.5 presents EPA's preferred models. These models had the best fit and offered most
intuitive results, in terms of variable sign and significance. EPA estimated additional model
specifications prior to the final models, including models with the infant birth weight categories
used in Almond et al. (2005) and a separate continuous gestational age variable, models with
different specifications for maternal age, and models with different combinations of maternal
risk factors. EPA does not believe that exclusion of maternal risk factor variables creates omitted
variable bias, given that their effects are accounted for using more direct newborn health state
variables such as Apgar score. The additional model specifications that EPA tested prior to
determining the final model form resulted in marginal effects estimates that were inconsistent
with scientifically expected directionality of their effects.
21 EPA did not develop a model for other race subpopulations because doing so for each individual race/ethnicity or combinations
of all "other" races would suffer from effects of low sample size (i.e., odds ratios and marginal effects that lack significance).
22 The natural exponent of the logistic regression coefficient is a ratio of odds of the outcome when the value of the predictor
variable is changed by a certain amount relative to the odds of the outcome using the baseline value of the predictor variable.
The odds are the ratio of the probability that the outcome of interest occurs to the probability that the outcome of interest does not
occur
23 EPA calculated marginal effects using the "margins, dydx(*)" command in Stata (StataCorp, 2013b). EPA used the default as
observed option.
24 EPA estimated BW-gestational age category-specific marginal effects using subsets of data that contain infants with BW in the
corresponding gestational age category to account for correlations between gestational age and other variables included in the
model. For example, infants in the preterm gestational age categories have lower Apgar score on average.
Proposed PFAS Rule Economic Analysis
E-13
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.7 Results and Discussion
E.7.1 Mortality Regression Models
Overall, the sign and significance of covariates in the regression models align with expectations
based on previous literature. Table E-4 presents odds ratios and marginal effects (in terms of
deaths per 1,000 births) for the non-Hispanic Black, non-Hispanic White, and Hispanic models.
A marginal effect estimate represents the effect of a 1-unit change in a given covariate on the
infant mortality rate per 1,000 births. Pseudo R2 values are approximately 40%, which is in line
with previous literature..25 The Agency notes that the estimated models are potentially subject to
omitted variable bias from other sources, such as income level, but EPA does not have adequate
information to evaluate the impacts of this bias on the marginal birth weight-mortality
relationship. The following subsections discuss the effects of regression model covariates on the
probability of infant mortality.
E.7.1.1 Birth Characteristics
The results for the birth weight-gestational age variables match literature-based expectations. In
all three models, the coefficients and marginal effects for birth weight among different
gestational age categories are negative and statistically significant (p<0.01). Negative marginal
effect values for the birth weight- gestational age categories indicate that a 1-gram birth weight
increase is associated with decreases in the infant mortality rate per 1,000 births, ranging from -
0.20 (extremely preterm) to -0.005 (term) for the non-Hispanic Black population, from -0.12 to -
0.002 for the non-Hispanic White population, and from -0.15 to -0.002 for the Hispanic
population. The magnitude of birth weight marginal effect is lower in gestational age categories
corresponding to longer gestation, indicating that the probability of mortality decreases as both
gestational age and birth weight increase.
Determining the magnitude of the mortality probability decrease is straightforward using
marginal effects. For example, using marginal effects from the non-Hispanic Black model, for
extremely preterm infants a 100 g birth weight increase would translate to 20 fewer infant deaths
per 1000 births in this gestational age category or a 2% decrease in the probability of mortality
within one year of birth.26 The same birth weight increase at a higher gestational age would still
decrease mortality risk but to a lesser extent. A 100 g birth weight increase for a non-Hispanic
Black infant in the moderately pre-term category would translate to only 1 fewer infant death per
1000 births or a 0.1% decrease in the probability of mortality within one year of birth.
Figure E-l shows variability of marginal effects for birth weight among different gestational age
categories across race/ethnicity subpopulations, with larger magnitudes estimated for the non-
Hispanic Black subpopulation compared to those estimated for the non-Hispanic White
subpopulation or Hispanic subpopulation, indicating that LBW increases the probability of
mortality within the first year more so among non-Hispanic Black infants than among non-
Hispanic White and Hispanic infants. This pattern is more pronounced for the extremely preterm
infants and very preterm infants.
25 Ma et al. (2010) reported a Pseudo R2 value of approximately 27%.
26 The implied decrease in probability of death is calculated as (100 g)*(marginal effect in terms of deaths per 1,000 births
per g)/(l,000 births) and multiplied by 100 to obtain a percentage: [(100 g)*(-0.19440/1000)] *(100) = -1.94%.
Proposed PFAS Rule Economic Analysis
E-14
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
0.00
-0.05
c
03 CD
CD to
CD
CuO
ru
-C
u
-0.25
Non-Hispanic Black Non-Hispanic White Hispanic
>
— /
§
O '' y '
A'
A
f-
1 w
~1
fc 5 -0.10 / /' /
a. cq
/
/
/
-a ro /
/
c "0.15
^2 .E
.E M
28 weeks and <=32 weeks),
moderately preterm (>32 weeks and <=37 weeks), and term (>37 weeks). Related covariates in the regression model include
BIRTH_BW_I_EXT_PRETERM, BIRTH_BW_I_VER_PRETERM, BIRTH_BW_I_MOD_PRETERM, BIRTH_BW_I_TERM.
Data based on the 2016/17 and 2017/18 CDC Period Cohort Linked Birth-Infant Death Data Files obtained from NCHS/NVSS.
For the birth order variables (BIRTH BOCatl, BIRTH_BOCat2), the reference category is first-
born children. Across all three models, odds ratios and marginal effects for these variables are
large and significant (p<0.01). Effects for BIRTH_BOCat2 are larger than for BIRTH BOCatl,
which is consistent with research indicating that second- or later-born infants have increasingly
higher probabilities of mortality compared to first-borns (Mishra et al., 2018; Ahrens et al.,
2017). Coefficients and marginal effects for variables indicating male infants (BIRTH MALE)
and infants with congenital anomalies (BIRTH CONANOM) indicate that the probability of
mortality increases when the infants are male and when infants experience at least one congenital
anomaly. The effect of calendar birth year was not statistically different from zero at a 5%
significance level.
Proposed PFAS Rule Economic Analysis
E-15
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Marginal effects for the birth characteristics variables also vary by race/ethnicity. For example,
the marginal effects for the BIRTHBOCatl variables indicate that, relative to first-born infants,
the infant mortality rate per 1,000 births increases by 1.13, 0.90, and 0.59 for second-born non-
Hispanic Black, non-Hispanic White, and Hispanic infants, respectively.27 Compared to the non-
Hispanic White and Hispanic subpopulations, 5-minute Apgar score has a stronger association
with infant mortality among the non-Hispanic Black subpopulations. The marginal effects for the
BIRTHCONANOM variables indicate that, relative to infants without any congenital
anomalies, the infant mortality rate per 1,000 births increases by 18.82, 8.99, and 9.66 for non-
Hispanic Black, non-Hispanic White, and Hispanic infants with congenital anomalies,
respectively.
E.7.1.2 Maternal Risk and Risk Mitigation Factors
The probability of infant mortality varies among certain maternal risk or risk mitigation factors.
The probability of infant mortality increases for mothers who smoke or mothers without a high
school diploma. Maternal smoking increases the infant mortality rate per 1,000 births by 1.34,
0.47, and 0.57 for non-Hispanic Black, non-Hispanic White, and Hispanic infants, respectively.
The probability of infant mortality decreases for mothers with a college education or higher.
Relative to mothers with a high school education, the infant mortality rate per 1,000 births
decreases by 1.29, 0.82, and 0.27 for non-Hispanic Black, non-Hispanic White, and Hispanic
infants born to mothers with a college education or higher, respectively. Relative to the 10 to 15
prenatal care visit category, which is most common in the data (See Table E-3), the probability
of infant mortality increases with zero visits, 1 to 9 visits, and 16 or higher visits. Marginal
effects indicate that having no prenatal care visits increases the infant mortality rate per 1,000
births by 3.03, 0.95, and 0.91 for non-Hispanic Black, non-Hispanic White, and Hispanic infants,
respectively.
E.7.1.3 Maternal Demographic and Socioeconomic Characteristics
Results for the maternal demographic and socioeconomic characteristic variables vary by
race/ethnicity and largely match EPA's expectations. The education variables serve as proxies
for socioeconomic status, and results among all three models indicate that, relative to mothers
with a high school diploma, the probability of infant mortality increases for mothers without a
high school diploma and decreases for mothers with a college education or higher. Maternal
education effects on infant mortality probability vary by race/ethnicity. For example, relative to
mothers with a high school education, the infant mortality rate per 1,000 births decreases by
1.29, 0.82, and 0.27 for non-Hispanic Black, non-Hispanic White, and Hispanic infants born to
mothers with a college education or higher, respectively.
The maternal age variables align with available infant mortality statistics showing the highest
infant mortality rates when mothers are under age 20 and elevated rates when mothers are over
40 (Ely et al., 2020). Compared to mothers aged 20 to 34 years, probability of infant mortality is
higher for mothers younger than 20 years, lower for mothers aged 35 to 40 years, and higher for
mothers older than 40 years. Relative to infants born to mothers aged 20 to 34 years, infants born
to mothers younger than 20 years experience 0.79, 0.61, and 0.68 additional infant deaths per
27 The implied decrease in probability of death is calculated as (marginal effect in terms of deaths per 1,000 births)/(1,000 births)
and multiplied by 100 to obtain a percentage. Example calculation using the marginal effects for BIRTH_BOCatl from the non-
Hispanic Black model: (1.19100/1000)*(100) = 0.119%.
Proposed PFAS Rule Economic Analysis
E-16
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1,000 births in non-Hispanic Black, non-Hispanic White, and Hispanic subpopulations,
respectively. The decreased death probability for mothers aged 35 to 40 might be capturing
effects of the financial stability of mothers in this age group.
Negative and significant coefficients and marginal effects among all models for the mother's
marital status variable, MDEM I MARRIED, indicate that the risk of infant mortality decreases
among infants with two parents, consistent with studies indicating that paternal involvement
reduces the probability of infant mortality (Ngui et al., 2015; Alio et al., 2011). Compared to
infants born to mothers who are not married or mothers whose marital status is unknown, infants
born to married mothers experience 0.35, 0.51, and 0.30 fewer deaths per 1,000 births for non-
Hispanic Black, non-Hispanic White, and Hispanic subpopulations, respectively.
Proposed PFAS Rule Economic Analysis
E-17
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-4: Odds Ratios and Marginal Effects for the Non-Hispanic Black, Non-Hispanic White, and Hispanic Mortality
Regression Models
Variable
Odds Ratios (95% CI)a b
Marginal Effects (Deaths per 1,000 Births (95% CI)a c
Black
White
Hispanic
Black
White
Hispanic
BIRTHBWIEXTPRETERM
0.99817
(0.99802,
0.99832)
0.99866
(0.99855,0.99878)
0.99835
(0.99817,0.99853)
-0.20400
(-0.21910,-
0.18890)
-0.12160
(-0.13080,-
0.11240)
-0.15260
(-0.1677,-0.13750)
BIRTHBWIVERPRETERM
0.99816
(0.99804,
0.99827)
0.9985
(0.99842,0.99858)
0.99846
(0.99835,0.99858)
-0.04580
(-0.04820, -
0.04340)
-0.03290
(-0.03430, -
0.03140)
-0.03290
(-0.0351,-0.03070)
BIRTHBWIMODPRETERM
0.99852
(0.99846,
0.99857)
0.99867
(0.99863,0.99872)
0.99856
(0.99849, 0.99862)
-0.01030
(-0.01080, -
0.00985)
-0.00677
(-0.00702, -
0.00652)
-0.00626
(-0.00659, -
0.00592)
BIRTHBWITERM
0.99856
(0.99851,
0.99860)
0.99865
(0.99861,0.99868)
0.99849
(0.99844, 0.99855)
-0.00453
(-0.00472, -
0.00434)
-0.00228
(-0.00236, -
0.00221)
-0.00219
(-0.00229, -
0.00208)
BIRTHBOCatl
1.20078
(1.12406,
1.28272)
1.37498
(1.30875, 1.44458)
1.23256
(1.14005, 1.33256)
1.13170
(0.72263, 1.54080)
0.90320
(0.76267, 1.04370)
0.59091
(0.37013,0.81170)
BIRTH_BOCat2
1.43158
(1.34271,
1.52634)
1.66176
(1.57927,1.74859)
1.36704
(1.26426, 1.47818)
2.21920
(1.81950,2.61890)
1.44050
(1.29450, 1.58650)
0.88360
(0.66192, 1.10530)
BIRTH_APGAR_0_3
19.89802
(18.35772,
21.56734)
43.36705
(40.67038,
46.24253)
45.87636
(41.39996,
50.83677)
18.49800
(17.92800,
19.06800)
10.69200
(10.46100,
10.92300)
10.81300
(10.466, 11.15900)
BIRTH_APGAR_4_6
3.8631
(3.54196,
4.21336)
5.92239
(5.54208,6.32880)
6.86084
(6.16310, 7.63750)
8.35950
(7.79370, 8.92530)
5.04500
(4.83850,5.25150)
5.44270
(5.1129,5.7726)
BIRTH MALE
1.28589
(1.22265,
1.35240)
1.29367
(1.24351,1.34583)
1.19405
(1.12581, 1.26643)
1.55530
(1.24280, 1.86790)
0.73028
(0.61753, 0.84304)
0.50123
(0.33447, 0.66798)
BIRTHCONANOM
20.95317
(16.73647,
26.23226)
23.81106
(21.33609,
26.57338)
30.45195
(25.31381,
36.63302)
18.81800
(17.39300,
20.24300)
8.99150
(8.65400, 9.32900)
9.65470
(9.096, 10.21300)
BIRTHYR2016
1.04910
(0.99784,
1.10298)
1.01725
(0.97816,1.05791)
0.97538
(0.91965, 1.03449)
0.29646
(-0.01339,0.60632)
0.04852
(-0.06265,0.15968)
-0.07045
(-0.23671,0.09582)
MRFNOPRECARE
1.63300
(1.46647,
1.81844)
1.39979
(1.24374,1.5754)
1.37859
(1.19240, 1.59383)
3.03350
(2.36630, 3.70070)
0.95389
(0.61828, 1.28950)
0.90736
(0.49675, 1.31800)
Proposed PFAS Rule Economic Analysis
E-18
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-4: Odds Ratios and Marginal Effects for the Non-Hispanic Black, Non-Hispanic White, and Hispanic Mortality
Regression Models
Variable
Odds Ratios (95% CI)a b
Marginal Effects (Deaths per 1,000 Births (95% CI)a c
Black
White
Hispanic
Black
White
Hispanic
MRF19PRECARE
1.37775
(1.29674,
1.46382)
1.34652
(1.28399,1.41209)
1.17236
(1.09445, 1.25582)
1.98210
(1.60560,2.35870)
0.84385
(0.70831,0.97940)
0.44942
(0.25471,0.64414)
MRF_ 160RM0REPREC ARE
1.12520
(1.00220,
1.26329)
1.12394
(1.04280,1.21139)
1.35485
(1.20490, 1.52345)
0.72964
(0.013350,1.44590)
0.33139
(0.11875,0.54403)
0.85827
(0.52611, 1.19040)
MRFSMOKE
1.24139
(1.13425,
1.35866)
1.17977
(1.11549,1.24776)
1.22117
(1.02459, 1.45549)
1.33750
(0.77763, 1.89740)
0.46889
(0.30933, 0.62846)
0.56471
(0.06794, 1.06150)
MDEMINOHS
1.05467
(0.97987,
1.13519)
1.10367
(1.03289,1.17930)
1.02742
(0.95914, 1.10056)
0.32924
(-0.12598,0.78447)
0.27977
(0.09167, 0.46788)
0.07644
(-0.11791,0.27079)
MDEMICOLLEGEPLU S
0.81232
(0.75874,
0.86969)
0.7478
(0.71366,0.78357)
0.90822
(0.83434, 0.98863)
-1.28570
(-1.70930,-
0.86211)
-0.82429
(-0.95807, -0.6905)
-0.27208
(-0.51214,-
0.03202)
MDEMAGETEEN
1.13705
(1.02800,
1.25767)
1.24116
(1.13208,1.36077)
1.27144
(1.13883, 1.41948)
0.79446
(0.17048, 1.41840)
0.61279
(0.35157, 0.87402)
0.67869
(0.36668,0.99071)
MDEMAGEADV 3 5 40
0.90639
(0.83721,
0.98130)
0.85079
(0.80231,0.90220)
0.95193
(0.87380, 1.03704)
-0.60792
(-1.0992,-0.11665)
-0.45831
(-0.62493, -
0.29170)
-0.13923
(-0.38131,0.10286)
MDEM_AGE_ADV_40plus
1.37377
(1.17433,
1.60708)
0.96251
(0.83754,1.10613)
1.2633
(1.07379, 1.48624)
1.96430
(0.99358,2.93490)
-0.10838
(-0.50285, 0.28609)
0.66055
(0.20117, 1.11990)
MDEM_I MARRIED
0.94432
(0.88719,
1.00513)
0.83555
(0.79827, 0.87458)
0.89883
(0.84382, 0.95743)
-0.35439
(-0.74074,0.03196)
-0.50957
(-0.63965, -
0.37949)
-0.30144
(-0.48028, -
0.12260)
# Model Observations
981,212
3,644,499
1,646,713
Pseudo R2
0.389
0.357
0.416
Abbreviations: CI - confidence intervals.
Notes:
Confidence intervals and significance testing do not include adjustments for multiple comparisons.
bLogistic regression models and ORs estimated using the "logit" likelihood function in Stata 15.1.
cMarginal effects estimated using the "margins, dydx(*)" command in Stata 15.1 with the default observed option. Fornon-BW-GA variables, EPA estimated marginal effects
based on covariate values from all observations in the models. For BW-GA variables, EPA estimated marginal effects based on covariate values from the subset of observations
falling within each GA category (see Supplementary Table 3).
Proposed PFAS Rule Economic Analysis
E-19
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-4: Odds Ratios and Marginal Effects for the Non-Hispanic Black, Non-Hispanic White, and Hispanic Mortality
Regression Models
Variable
Odds Ratios (95% CI)a b
Marginal Effects (Deaths per 1,000 Births (95% CI)a c
Black
White
Hispanic
Black
White
Hispanic
BIRTHBWIEXTPRETERM
0.99817
(0.99802,
0.99832)
0.99866
(0.99855,0.99878)
0.99835
(0.99817,0.99853)
-0.20400
(-0.21910,-
0.18890)
-0.12160
(-0.13080,-
0.11240)
-0.15260
(-0.1677,-0.13750)
BIRTHBWIVERPRETERM
0.99816
(0.99804,
0.99827)
0.9985
(0.99842,0.99858)
0.99846
(0.99835,0.99858)
-0.04580
(-0.04820, -
0.04340)
-0.03290
(-0.03430, -
0.03140)
-0.03290
(-0.0351,-0.03070)
BIRTHBWIMODPRETERM
0.99852
(0.99846,
0.99857)
0.99867
(0.99863,0.99872)
0.99856
(0.99849, 0.99862)
-0.01030
(-0.01080, -
0.00985)
-0.00677
(-0.00702, -
0.00652)
-0.00626
(-0.00659, -
0.00592)
BIRTHBWITERM
0.99856
(0.99851,
0.99860)
0.99865
(0.99861,0.99868)
0.99849
(0.99844, 0.99855)
-0.00453
(-0.00472, -
0.00434)
-0.00228
(-0.00236, -
0.00221)
-0.00219
(-0.00229, -
0.00208)
BIRTHBOCatl
1.20078
(1.12406,
1.28272)
1.37498
(1.30875,1.44458)
1.23256
(1.14005, 1.33256)
1.13170
(0.72263, 1.54080)
0.90320
(0.76267, 1.04370)
0.59091
(0.37013,0.81170)
BIRTH_BOCat2
1.43158
(1.34271,
1.52634)
1.66176
(1.57927,1.74859)
1.36704
(1.26426, 1.47818)
2.21920
(1.81950,2.61890)
1.44050
(1.29450, 1.58650)
0.88360
(0.66192, 1.10530)
BIRTH_APGAR_0_3
19.89802
(18.35772,
21.56734)
43.36705
(40.67038,
46.24253)
45.87636
(41.39996,
50.83677)
18.49800
(17.92800,
19.06800)
10.69200
(10.46100,
10.92300)
10.81300
(10.466, 11.15900)
BIRTH_APGAR_4_6
3.8631
(3.54196,
4.21336)
5.92239
(5.54208,6.32880)
6.86084
(6.16310, 7.63750)
8.35950
(7.79370, 8.92530)
5.04500
(4.83850,5.25150)
5.44270
(5.1129,5.7726)
BIRTH MALE
1.28589
(1.22265,
1.35240)
1.29367
(1.24351,1.34583)
1.19405
(1.12581, 1.26643)
1.55530
(1.24280, 1.86790)
0.73028
(0.61753, 0.84304)
0.50123
(0.33447, 0.66798)
BIRTHCONANOM
20.95317
(16.73647,
26.23226)
23.81106
(21.33609,
26.57338)
30.45195
(25.31381,
36.63302)
18.81800
(17.39300,
20.24300)
8.99150
(8.65400, 9.32900)
9.65470
(9.096, 10.21300)
BIRTHYR2016
1.04910
(0.99784,
1.10298)
1.01725
(0.97816, 1.05791)
0.97538
(0.91965, 1.03449)
0.29646
(-0.01339,0.60632)
0.04852
(-0.06265,0.15968)
-0.07045
(-0.23671,0.09582)
MRFNOPRECARE
1.63300
(1.46647,
1.81844)
1.39979
(1.24374,1.5754)
1.37859
(1.19240, 1.59383)
3.03350
(2.36630, 3.70070)
0.95389
(0.61828, 1.28950)
0.90736
(0.49675, 1.31800)
Proposed PFAS Rule Economic Analysis
E-20
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table E-4: Odds Ratios and Marginal Effects for the Non-Hispanic Black, Non-Hispanic White, and Hispanic Mortality
Regression Models
Variable
Odds Ratios (95% CI)a b
Marginal Effects (Deaths per 1,000 Births (95% CI)a c
Black
White
Hispanic
Black
White
Hispanic
MRF19PRECARE
1.37775
(1.29674,
1.46382)
1.34652
(1.28399,1.41209)
1.17236
(1.09445, 1.25582)
1.98210
(1.60560,2.35870)
0.84385
(0.70831,0.97940)
0.44942
(0.25471,0.64414)
MRF_ 160RM0REPREC ARE
1.12520
(1.00220,
1.26329)
1.12394
(1.04280, 1.21139)
1.35485
(1.20490, 1.52345)
0.72964
(0.013350,1.44590)
0.33139
(0.11875,0.54403)
0.85827
(0.52611, 1.19040)
MRFSMOKE
1.24139
(1.13425,
1.35866)
1.17977
(1.11549, 1.24776)
1.22117
(1.02459, 1.45549)
1.33750
(0.77763, 1.89740)
0.46889
(0.30933, 0.62846)
0.56471
(0.06794, 1.06150)
MDEMINOHS
1.05467
(0.97987,
1.13519)
1.10367
(1.03289,1.17930)
1.02742
(0.95914, 1.10056)
0.32924
(-0.12598,0.78447)
0.27977
(0.09167, 0.46788)
0.07644
(-0.11791,0.27079)
MDEMICOLLEGEPLU S
0.81232
(0.75874,
0.86969)
0.7478
(0.71366,0.78357)
0.90822
(0.83434, 0.98863)
-1.28570
(-1.70930,-
0.86211)
-0.82429
(-0.95807, -0.6905)
-0.27208
(-0.51214,-
0.03202)
MDEMAGETEEN
1.13705
(1.02800,
1.25767)
1.24116
(1.13208,1.36077)
1.27144
(1.13883, 1.41948)
0.79446
(0.17048, 1.41840)
0.61279
(0.35157, 0.87402)
0.67869
(0.36668,0.99071)
MDEMAGEADV 3 5 40
0.90639
(0.83721,
0.98130)
0.85079
(0.80231,0.90220)
0.95193
(0.87380, 1.03704)
-0.60792
(-1.0992,-0.11665)
-0.45831
(-0.62493, -
0.29170)
-0.13923
(-0.38131,0.10286)
MDEM_AGE_ADV_40plus
1.37377
(1.17433,
1.60708)
0.96251
(0.83754,1.10613)
1.2633
(1.07379, 1.48624)
1.96430
(0.99358,2.93490)
-0.10838
(-0.50285, 0.28609)
0.66055
(0.20117, 1.11990)
MDEM_I MARRIED
0.94432
(0.88719,
1.00513)
0.83555
(0.79827, 0.87458)
0.89883
(0.84382, 0.95743)
-0.35439
(-0.74074,0.03196)
-0.50957
(-0.63965, -
0.37949)
-0.30144
(-0.48028, -
0.12260)
# Model Observations
981,212
3,644,499
1,646,713
Pseudo R2
0.389
0.357
0.416
Abbreviations: CI - confidence intervals.
Notes:
Confidence intervals and significance testing do not include adjustments for multiple comparisons.
bLogistic regression models and ORs estimated using the "logit" likelihood function in Stata 15.1.
cMarginal effects estimated using the "margins, dydx(*)" command in Stata 15.1 with the default observed option. For non-BW-GA variables, EPA estimated marginal effects
based on covariate values from all observations in the models. For BW-GA variables, EPA estimated marginal effects based on covariate values from the subset of observations
falling within each GA category (see Supplementary Table 3).
Proposed PFAS Rule Economic Analysis
E-21
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
E.7.2 Comparison to Prior Studies
EPA's evaluation of the relationship between birth weight and infant mortality differs from those
used in prior literature in terms of included covariates, model specification, and sample
characteristics. In terms of modeling approach, our analysis is closest to the one used by Ma et
al. (2010), who also find that birth weight and GA are important predictors of infant mortality
risk and that the effects of birth weight on infant mortality vary by race/ethnicity. However,
methodological differences between Ma et al. (2010) and our work, summarized in Table E-5,
prevent us from making direct comparisons of birth weight-infant mortality effect magnitudes.
Even in the absence of methodological differences, EPA expects that results would differ from
those reported by older studies due to changes in infant mortality, maternal and birth
characteristics, and maternal demographic over the past 30 years (see Table E-l).
Table E-5: Comparison of Ma et al. (2010) and the EPA Analysis
Analysis Component
Ma et al. (2010)
EPA
Year(s) of NCHS/NVSS
Data
2001
2016-2018
Data Sample
Singletons and multiples
Singletons only
Race/Ethnicity Models
Non-Hispanic Black, non-Hispanic
White, Mexican
Non-Hispanic Black, non-Hispanic
White, Hispanic
Birth Weight-Gestation
Specification3
Birth weight (100 g increment),
gestational age (weeks), and birth weight
x gestational age (continuous product of
birth weight and gestational age)
Birth weight interacted with four
gestational age categories (extremely
preterm, very preterm, moderately
preterm, and term)
Other Covariatesb
Categorized APGAR score (low: 0-3 and
medium: 4-6, with high: 7-10 as
reference category), maternal age,
maternal education, marital status,
whether mother was born in U.S.,
whether father was unreported on birth
certificates, prenatal care,
tobacco/alcohol use during pregnancy,
and birth order
Categorized Apgar score (low: 0-3
and medium: 4-6, with high: 7-10 as
reference category), categorized
number of prenatal care visits (None,
1-9,16+, with ,10-15 as reference
category), maternal education, maternal
age, marital status, smoker status, sex,
presence of congenital anomalies, birth
year, birth order (see Table E-2)
Abbreviations: NCHS - National Center for Health Statistics; NVSS - National Vital Statistics System.
Notes:
aAlthough Ma et al. (2010) tested several different models, EPA focuses on one of their highest-performing model forms,
Model 12, in which the interaction term between gestational age and birth weight is almost always significant.
bEPA notes that Ma et al. (2010) did not report coefficients for a number of maternal and birth characteristics (i.e., maternal
age, maternal education, marital status, whether mother was born in U.S., whether father was unreported on birth certificates,
prenatal care, tobacco/alcohol use during pregnancy, and birth order) or discussed these variables in detail.
E.8 Limitations and Uncertainties
Table E-6 summarizes limitations and sources of uncertainty associated with the estimated
relationship between infant birth weight and mortality.
Proposed PFAS Rule Economic Analysis
E-22
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table E-6: Limitations and Uncertainties in the Analysis of the Birth Weight-Mortality
Relationship
Uncertainty/Assumption
Notes
Transcription errors may be present in the
NCHS/NVSS dataset
Infant birth and death records are compiled based on hand-
written forms and tabulated for use in the NCHS/NVSS
dataset.
The models do not directly account for maternal
socioeconomic status and other potentially
important factors that contribute to LBW and
infant mortality.
Though review of the infant mortality literature suggests that
socioeconomic status is an indicator of infant mortality (Ma
et al., 2010; Ely et al., 2020), the NCHS/NVSS does not
have a variable that would account for individual
socioeconomic status of the mother (e.g., household income)
or even community-level socioeconomic status (e.g., median
income at the county- or state-level). EPA tested a variable
for hospital payment source for delivery that specifies those
who use Medicaid, but model results that included this
variable did not match expectations (variable coefficient was
not significant for all race/ethnicity subpopulations, mixture
of negative and positive coefficients depending on
race/ethnicity subpopulation). Thus, the variable was
excluded from our models. The maternal education, maternal
age, and marital status variables serve as rough proxies for
socioeconomic status in our models. Other factors, such as
indicators of parental support networks (e.g., access to paid
care or grandparents that live nearby) may contribute to the
relationship between birth weight and infant mortality, but
such information is not publicly available at the individual
infant scale.
The analysis relies only on singleton data to
develop relationships between birth weight and
infant mortality.
Because singletons represent the majority of U.S. births
(96% of infants born in 2016 and 2017), EPA does not
expect this to be a significant limitation. In order to address
this limitation, a separate model would be required because
multiples are often born at smaller birth weight than
singleton infants, the mortality rate among multiples is often
higher than singletons for reasons often unrelated to birth
weight (Horon, 2020), and the sample size of multiples in
the 2016-2018 NCHS/NVSS data is likely not adequate to
represent the relationship between birth weight and
mortality.
EPA does not model birth weight-mortality
impacts for infants who fall into race categories
other than non-Hispanic White, non-Hispanic
Black, and Hispanic
While the NCHS/NVSS data specifies additional race
categories, developing models for each individual race or
even a combination of all "other" races would suffer from
effects of low sample size, including coefficient and
marginal effects that lack significance. All combined, the
"other" race/ethnicity subpopulation would have a sample
size that is at least 30 percent smaller than any one of the
non-Hispanic White, non-Hispanic Black, and Hispanic
race/ethnicity models.
Abbreviations: LBW - low birth weight; NCHS - National Center for Health Statistics; NVSS - National
Vital Statistics System.
Proposed PFAS Rule Economic Analysis
E-23
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix F. Serum Cholesterol Dose Response
Functions
This appendix describes EPA's literature review to identify studies to estimate relationships
between cholesterol levels and serum per- and polyfluoroalkyl substances (PFAS) for inclusion
in a meta-analysis of these relationships. This approach has been peer reviewed by EPA's
Science Advisory Board; input provided by that organization has been considered in finalizing
this analysis (U.S. EPA, 2022). Statistical analyses that combine the results of multiple studies,
such as meta-analyses, are widely applied to investigate the dose-specific relationship between
contaminant levels and associated health effects. Such analyses are suitable for economic
assessments because they can improve precision and statistical power (Engels et al., 2000;
Deeks, 2002; Riicker et al., 2009). This appendix also provides details on the meta-data
development, results of the meta-analysis, and limitations and uncertainties associated with the
estimated relationships. EPA used the estimated relationships to estimate cardiovascular disease
(CVD) risk reduction associated with exposure to PFAS mediated by changes in serum
cholesterol markers.
F.l Data Sources
EPA relied on two literature review efforts to identify potential sources of exposure-response
information for the effect of PFAS on serum cholesterol, lipids, and lipoproteins: A literature
review built on the one conducted by the Agency for Toxic Substances and Disease Registry
(ATSDR) in the development of their Toxicological Review Public Comment Draft (ATSDR,
2018), which included literature through mid-2017.
The most recent systematic review of the newly published epidemiological literature for PFAS
performed by EPA included literature from 2013 to 2020 (U.S. EPA, 2023a; U.S. EPA, 2023b).
The relationships between exposure to PFAS and serum total cholesterol (TC) and high-density
lipoprotein cholesterol (HDLC) identified based on these literature reviews allowed EPA to
generate inputs for the Pooled Cohort Atherosclerotic Cardiovascular Disease (ASCVD) risk
model (Goff et al., 2014).28'.29
F.l.l Literature Review and Studies identification for the
Meta-Analysis
Two reviewers independently screened references retrieved from the literature search by title and
abstract, and then reviewed relevant studies in full text. EPA evaluated studies identified during
the search according to the following criteria prior to inclusion in the meta-analysis to ensure
validity, consistency, and applicability. Briefly, of interest were studies conducted on adults in
the general population, evaluating the outcomes of TC and HDLC, and the exposures of PFOA
and PFOS. Because EPA evaluates CVD risk among a general population of adults aged 40 to
28 The ASCVD model relies on the following inputs: demographic information, smoking and diabetes status,
serum TC, and HDLC.
29 Note that EPA evaluated HDLC effects as part of a sensitivity analysis (see Appendix K). EPA did not model the effects of
PFOA/PFOS changes on EtDLC levels in the overall benefits analysis because evidence of an association between PFOA/PFOS
and HDLC effects is uncertain (U.S. EPA, 2023a; U.S. EPA, 2023b).
Proposed PFAS Rule Economic Analysis
F-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
89, studies performed on specific population subsets, such as occupational populations, were not
considered for inclusion in the meta-analysis due to the potential for greater levels of exposure to
PFOA and PFOS in these populations compared to the general population.
Applicability: EPA evaluated each study to determine whether it estimated the association
between exposure to PFOA or PFOS (measured in serum or plasma) and a quantitative measure
of TC or HDLC in general populations (age 20 and older). Of the 39 studies identified as part of
the ATSDR-based literature review that provided information on the relationship between
exposure to PFAS and TC and HDLC levels, 9 were general population studies. Of the 41 studies
identified as part of the EPA/OST literature review that provided information on the relationship
between exposure to PFAS and TC and HDLC levels, 14 were general population studies.
These studies.30 were further evaluated for inclusion in the meta-analysis.
Research methods and study details: EPA evaluated each study to determine whether it reported
numbers of participants, quantitative effect estimates (beta coefficients), measures of effect
estimate variance (95% confidence intervals [CIs], standard errors [SEs], or standard deviations
[SDs]). EPA retained studies with missing measures of effect estimate variance but with reported
p-values for differences. For such studies, EPA used the approach in the Cochrane Handbook for
Systematic Reviews (Higgins et al., 2019) to calculate SDs or SEs. Briefly, the approach
estimates the SEs using the correspondence between the p-value and the t-statistic, with degrees
of freedom equal to the difference between the sample size and the number of parameters in the
model that provided the effect estimate. Then the SE is obtained by dividing the effect estimate
by the t-statistic.
Additional exclusion criteria: EPA also excluded studies that reported data only for pregnant
women, infants, or children. Although there is some evidence that PFAS exposure is associated
with cardiometabolic impairment in children and younger adults (Rappazzo et al., 2017), EPA
did not extract data from these studies because lipid levels are known to change during
pregnancy from pre-pregnancy levels, and the relationships between lipid profiles at early life
stages are not as well defined as they are at later life stages. Another frequent reason for study
exclusion was the reporting of only relative risks or odds ratios for hypercholesteremia or
hyperlipidemia; results in this form could not be used to estimate continuous exposure-response
relationships.
F.1.2 Assessment of Study Applicability to the M eta-An a lysis
Figure F-l presents a flow diagram of the studies reviewed as part of the ATSDR-based and
EPA/OST-based literature reviews and the selection of studies retained for inclusion in the
meta-analysis. Using the study inclusion criteria described in Section F.l.l, EPA retained
14 studies for use in the meta-analysis. Of these, five were identified as part of the ATSDR
literature review (Chateau-Degat et al., 2010; Fisher et al., 2013; Fu et al., 2014; Nelson et al.,
2010; Steenland et al., 2009), seven were identified from the EPA systematic review (Dong et
al., 2019; Fan et al., 2020; Jain et al., 2019; Y. Li et al., 2020; C. Y. Lin et al., 2020; P.-I. D. Lin
30 Of the general population studies identified as part of the EPA/OST literature review, five overlapped with studies
identified as part of the ATSDR-based literature review.
Proposed PFAS Rule Economic Analysis
F-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
et al., 2019; Yang et al., 2018), and two were identified in both literature reviews (He et a!.,
2018; Liu etaL 2018).
Legend:
Literature
Review Basis
\
Retained Studies
Excluded Studies |
Final Retained
Studies
ATSDR-Based
Literature Review"
EPA/OST-Based
Literature Review'
N otes:
ATSDR = Agency for T owe Substances and Disease Registry, EPA = Environmental Protection Agency, OST = Office of Science and
Technology, PFAS = per- and potyfluoroalkyl substances, TC = Total Cholesterol, HDLC = high-density lipoprotein cholesterol
""Included literature thro ugh mid-2017.
"included literature published from 2016 to 2020.
"For example, studies based on occupational data or data only for pregnant women, infants, or children.
dSome studies did not include the estimates required for meta-analysis calculations. For example, certain studies did not report effect
estimates or interquartile ranges.
"Of these studies, Bare based on data from the United States and 6 are based on data outside of the United States.
Figure F-l: Diagram of Literature Retained for Use in the Meta-Analysis and Data
Sources.
Table F-l summarizes the 14 studies that were identified in the ATSDR-based and EPA
literature review that EPA used to derive slope estimates for PFOA and PFOS associations with
serum TC and HDLC levels..31 Six of the studies that EPA retained for use in the meta-analysis
were based on PFAS and serum lipid measurements from the U.S. general population (National
Health and Nutrition Examination Survey [NHANES]) (Dong et al., 2019; Fan et al., 2020; He et
al., 2018; Jain et al., 2019; Liu et al., 2018; Nelson et al., 2010); there were also general
31 For this effort, EPA focused on PFOA and PFOS, since these are by far the most well-studied perfluorinated
compounds.
Proposed PFAS Rule Economic Analysis
F-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
population studies from Canada (Fisher et al., 2013), Sweden (Y. Li et al., 2020), Taiwan (Yang
et al., 2018; C. Y. Lin et al., 2020), and Henan Province, China (Fu et al., 2014). Chateau-Degat
et al. (2010) reported on the relationship between PFOS and serum lipids in a Canadian Inuit
population. EPA also retained the results from a study of a highly exposed population in the
United States (the C8 Health Project cohort) (Steenland et al., 2009) and from a study using
participants in a U.S. diabetes prevention program (P.-I. D. Lin et al., 2019).
EPA excluded two general population studies identified in the ATSDR-based literature review
(Eriksen et al., 2013; Seo et al., 2018) and two general population studies identified based on the
agency's systematic review (Convertino et al., 2018; Huang et al., 2018) that were inadequate for
use in the meta-analysis because they did not include the estimates required for meta-analysis
calculations. For example, EPA excluded the studies identified in the ATSDR literature review
from the meta-analysis because the authors did not report either the effect estimates (Seo et al.,
2018) or interquartile ranges (Eriksen et al., 2013) needed for calculations..32 Similarly, EPA
excluded the studies identified as part of the agency's systematic review because they involved a
Phase 1 controlled trial with modeled exposures in cancer patients dosed with ammonium
perfluorooctanoate (Convertino et al., 2018) or reported effect estimates (Spearman correlation
coefficients) that were not suitable for use in the meta-analysis (Huang et al., 2018). EPA also
considered the longitudinal study by Fitz-Simon et al. (2013) of adults participating in the C8
Health Project who were not taking cholesterol-lowering medication and who were examined
twice, with an average of 4.4 years between examinations. In subjects whose serum PFOA levels
halved between examinations, there was a decrease of an average of 1.65% (95% confidence
interval: 0.32%, 2.97%) for TC and 1.33% (-0.21%), 2.85%) for HDLC. In subjects whose serum
PFOS levels halved between examinations, there were similar decreases, although larger in
magnitude and variability: a decrease of an average of 3.20% (95% confidence interval: 1.63%,
4.76%) for TC and 1.28% (-0.59%, 3.12%) for HDLC. However, given the nature of the results,
the effect estimates from this study were inadequate for inclusion in the meta-analysis.
32 Efforts to contact the study authors for the missing data were unsuccessful at the time of this report.
Proposed PFAS Rule Economic Analysis
F-4
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table F-l: Studies Selected for Inclusion in the Meta-Analyses
Author and Year
Title
Cholesterol and PFAS Relationship
Evaluated
TC HDLC
PFOA PFOS PFOA PFOS
Medications
Association of Perfluorooctanoic Acid and
Participants using lipid-lowering
Steenland et al., 2009Perfluorooctane Sulfonate With Serum Lipids Among
X
X
X
X
medications were excluded
Adults Living Near a Chemical Plant
Chateau-Degat et al..
2010 a'd
Nelson et al.. 2010E
Fisher et al., 2013s
Fu et al.. 2014*
He et al., 2018°
Liu et al., 2018°
Effects of Perfluorooctanesulfonate Exposure
on Plasma Lipid Levels in the Inuit Population
of Nunavik (Northern Quebec)
Exposure to Polyfluoroalkyl Chemicals and
Cholesterol, Body Weight, and Insulin Resistance in the
General U.S. Population
Do Perfhioroalkyl Substances Affect Metabolic
Function and Plasma Lipids?—Analysis of the 2007-
2009, Canadian Health Measures Survey (CHMS)
Cycle 1
Associations Between Serum Concentrations of
Perfhioroalkyl Acids and Serum Lipid Levels in a
Chinese Population
PFOA is Associated with Diabetes and Metabolic
Alteration in US Men: National Health and Nutrition
Examination Survey 2003-2012
Association Among Total Serum Isomers of
Perfluorinated Chemicals, Glucose Homeostasis, Lipid
Profiles, Serum Protein and Metabolic Syndrome in
Adults: NHANES, 2013-2014
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Use of lipid-lowering medication
considered in statistical analysis
Participants using lipid-lowering
medications were excluded
Participants using lipid-lowering
medications were excluded
Not taken into consideration
Not taken into consideration
Use of lipid-lowering
medication considered in
statistical analysis
Proposed PFAS Rule Economic Analysis
F-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table F-l: Studies Selected for Inclusion in the Meta-Analyses
Author and Year
Title
Cholesterol and PFAS Relationship
Evaluated
TC HDLC Medications
PFOA PFOS PFOA PFOS
Yang et al., 2018b
Association of Serum Levels of Perfluoroalkyl
Substances (PFASs) With the Metabolic Syndrome
(MetS) in Chinese Male Adults: A Cross-Sectional
Study
Not taken into consideration
X X
Dong et al., 2019b
Using 2003-2014 U.S. NHANES Data to Determine
the Associations Between Per- and Polyfluoroalkyl
Substances and Cholesterol: Trend and Implications
X
X
X
Participants using lipid-lowering
medications were excluded
Jain et al.. 2019b
Roles of Gender and Obesity in Defining Correlations
Between Perfluoroalkyl Substances and
Lipid/Lipoproteins
Use of lipid-lowering
X X X X medication considered in
statistical analysis
Per- and Polyfluoroalkyl Substances and Blood Lipid
P.-I. D. Lin et al., 2019b Levels in Pre-Diabetic Adults—Longitudinal Analysis
of the Diabetes Prevention Program Outcomes Study
Fanet al.. 2020b
Serum Albumin Mediates the Effect of Multiple Per-
and Polyfluoroalkyl Substances on Serum Lipid Levels
X
X
X
X
X X X X
Participants using lipid-lowering
medications were excluded
Not taken into consideration
Y. Li et al., 2020b
Associations Between Perfluoroalkyl Substances and
Serum Lipids in a Swedish Adult Population With
Contaminated Drinking Water
X X X X
Not taken into consideration
C. Y. Lin etal., 2020b
The Association Between Total Serum Isomers of Per-
and Polyfluoroalkyl Substances, Lipid Profiles, and the
DNA Oxidative/Nitrative Stress Biomarkers in Middle-
Aged Taiwanese Adults
X
X
Not taken into consideration
Proposed PFAS Rule Economic Analysis
F-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table F-l: Studies Selected for Inclusion in the Meta-Analyses
Cholesterol and PFAS Relationship
Evaluated
Author and Year
Title
TC
HDLC
Medications
PFOA PFOS PFOA PFOS
Abbreviations: PFAS - per-and polyfluoroalkyl substances; PFOA - perfluorooctanoic acid; PFOS - perfluorooctanesulfonic acid; TC - total cholesterol; HDLC - high-density
lipoprotein cholesterol.
Notes: Study quality reflected in green (medium confidence) or pink (low confidence) cell shading.
aStudies identified based on ATSDR literature review.
bStudies identified based on EPA literature review.
cStudies available in both assessments.
dStudies available in PFOA and/or PFOS health effects support documents (U.S. EPA, 2016a, 2016b).
Proposed PFAS Rule Economic Analysis F-7 March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
F.2 Meta-Analysis
Based on the study inclusion criteria discussed in Section F.l.l, EPA included 14 studies in the
meta-analysis. Of these 14 studies, 11 were used to develop exposure-response relationships for
serum PFOA and TC, 13 were used to develop exposure-response relationships for serum PFOA
and HDLC, 12 studies were used to develop exposure-response relationships for serum PFOS
and TC, and 13 studies were used to develop exposure-response relationships for serum PFOS
and HDLC (Table F-l). EPA conducted four separate meta-analyses: one analysis for each
combination of chemical (PFOA or PFOS) and health outcome (TC or HDLC).
All studies were evaluated for risk of bias, selective reporting, and sensitivity as applied in
developing EPA's Toxicity Assessments and Proposed Maximum Contaminant Level Goals for
PFOA and PFOS in Drinking Water (U.S. EPA, 2023a; U.S. EPA, 2023b). Briefly, the main
considerations specific to evaluating the quality of studies on serum lipids included use of
medications, fasting, and potential for reverse causality. Because lipid-lowering medications
strongly affect serum lipid levels, studies that did not account for the use of lipid-lowering
medications by restriction, stratification, or adjustment were rated as deficient in the participant
selection domain. For TC and HDLC measurements, fasting is not likely to introduce
measurement error because the serum levels of the lipids considered change minimally after a
meal (Mora, 2016). Measuring PFOS and serum lipids concurrently was considered adequate in
terms of exposure assessment timing. Given the long half-life of PFOA and PFOS (Ying Li et
al., 2018), current blood concentrations are expected to correlate well with past exposures.
Furthermore, although reverse causation due to hypothyroidism (Dzierlenga, Allen, et al., 2020)
or enterohepatic cycling of bile acids (Fragki et al., 2021) has been suggested, there is not yet
clear evidence to support these reverse causal pathways.
Based on these considerations, of the 14 studies, ten were medium confidence in ROB
evaluations, with only four deemed low confidence (Fu et al., 2014; He et al., 2018; Yang et al.,
2018; Y. Li et al., 2020). These low confidence studies had deficiencies in participant selection,
outcome assessment, or confounding domains. None of these studies considered use of lipid-
lowering medications in the selection process or in the statistical analyses. Additional details on
the ROB evaluations are available in ICF (2021).
F.3 Extraction of Slope Values for TC and HDLC
If studies reported linear slope relationships (change in serum TC or HDLC in mg/dL per ng/mL
change in serum PFOA/PFOS), EPA extracted these values, along with their confidence limits,
directly as reported by the study authors. If results from multiple models with different
adjustments for confounders were reported within a single study, either the most adjusted results
or the main model results as presented by the study authors were selected. When studies
provided results for both untransformed and log-transformed PFOA/PFOS, EPA used
untransformed PFOA/PFOS to reduce bias due to back-transformations of effect estimates. For
studies that provided results only for log-transformed PFOA/PFOS (five studies) or log-
transformed outcomes (two studies), or log-transformed both PFOA/PFOS and outcomes (two
studies), EPA approximated the results for an untransformed analysis using the approach
outlined by Rodriguez-Barranco et al. (2017) and Dzierlenga, Crawford, et al. (2020). When not
reported, EPA assumed that the natural logarithm was the basis of the transformation. An
independent EPA reviewer evaluated the extracted slope values for quality assurance.
Proposed PFAS Rule Economic Analysis
F-8
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
F.4 Methods and Key Assumptions
The summary measure of association was a beta coefficient relating changes in TC or HDLC in
mg/dL to increases in serum or plasma33 PFOA or PFOS in ng/mL. EPA conducted random-
effects meta-analyses using the DerSimonian et al. (1986) approach, which uses weights based
on the inverse of the variance of the coefficient of each study plus the addition of an extra
component of variance between studies. When studies reported beta coefficients by quartiles
(e.g., He et al., 2018), EPA estimated a linear coefficient using a weighted linear regression of
the midpoints of the quartiles and the reported beta coefficients, using the inverse of standard
errors as the regression weights.
EPA assessed between-study heterogeneity using Cochran's Q test (Cochran, 1954) and the I2
statistic (Higgins et al., 2003). EPA developed forest plots to display the results. EPA developed
funnel plots and performed an Egger regression on the estimates of effect size to assess potential
publication bias (Begg et al., 1994; Egger et al., 1997; Egger et al., 2008). Because back-
transformations of effect estimates with log-transformed outcomes or exposures could introduce
bias and could be a source of heterogeneity, EPA also conducted sub-analyses by type of model
that provided the study-specific effect estimate (e.g., only including studies that reported linear
associations [six studies] or linear-log associations [five studies]).
If publication bias was observed, EPA conduced sensitivity analyses using trim-and-fill methods
(Duval et al., 2000a, 2000b) to estimate the number of missing studies and predict the impact of
the hypothetical "missing" studies on the pooled effect estimate. To investigate sources of
heterogeneity, EPA conducted several sensitivity analyses:
EPA evaluated the impact of using other estimation methods for the between-study variance
(tau2) besides the DerSimonian et al. (1986) approach, such as restricted maximum likelihood
(Raudenbush, 2009) or Sidik et al. (2005).
• To assess potential impact of a single study on the overall effect estimate, EPA conducted
leave-one-out meta-analyses.
• To assess potential impact of study quality on the overall effect estimate, EPA conducted
sensitivity analyses excluding the four studies considered to have higher ROB.
• To assess the impact of using multiple regression coefficients from the same study
(which are correlated), EPA excluded a study that contributed four effect estimates
(gender- and obesity-specific) for each analysis, which also accounted for most of the
weight in the overall pooled beta coefficient (Jain et al., 2019). EPA also conducted a
sensitivity analysis using a single pooled estimate from the four study-specific estimates.
• EPA also assessed the impact of non-U. S or Canadian general population studies in
sensitivity analyses excluding studies conducted in China (Fu et al., 2014), Taiwan (Yang
et al., 2018; C. Y. Lin et al., 2020), or Sweden (Y. Li et al., 2020), the Canadian Inuit
population study (Chateau-Degat et al., 2010), and the U.S. high-exposure community
study (Steenland et al., 2009).
Six studies that EPA retained for use in the meta-analysis were based on PFAS and serum lipid
measurements using data from overlapping NHANES cycles: Dong et al. (2019) used data from
33 PFOA or PFOS concentrations is serum or plasma were treated interchangeably.
Proposed PFAS Rule Economic Analysis
F-9
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
2003-2014, while He et al. (2018) used 2003-2012 data; Jain et al. (2019) used 2005-2014 data;
Fan et al. (2020) used 2011-2014 data; Liu et al. (2018) used 2013-2014; and Nelson et al.
(2010) used data from 2003-2004. Although the datasets and models were not exactly the same
in all NHANES-based studies, to avoid estimate dependency issues due to overlapping
populations in the meta-analysis, EPA also performed a sensitivity analysis including only the
data from the study covering the broadest range of NHANES cycles (2003-2014) (Dong et al.,
2019).
EPA performed statistical analyses using the software STATA, version 16.1 (StataCorp, 2019),
with the combine, meta esize, meta set, meta summarize, metainf meta funnel, meta bias, and
meta trimfill packages (Palmer et al., 2016). Results of the meta-analyses are presented in Table
F-2 and Table F-3. Overall, there is a high degree of heterogeneity when all studies are
combined. Excluding Jain et al. (2019) did not significantly reduce the heterogeneity; however
restricting analyses to studies reporting linear or linear-log associations did reduce heterogeneity
in most cases.
F.4.1 Slope Estimation for PFOA
When including the six studies reporting linear associations, there was a statistically significant
positive increase in TC of 1.57 (95% confidence interval: 0.02, 3.13) mg/dL per ng/mL serum
PFOA (p-value=0.048,12=87%). The association for HDLC and PFOA was positive (0.11; 95%
CI: -0.22, 0.43) but not statistically significant (Table F-2, Figure F-2). Adjusting for possible
publication bias through funnel plots and trim-and-fill analysis suggested the imputation of two
additional studies for HDLC and PFOA with a smaller effect (-0.01, 95% confidence interval: -
0.42, 0.41). For TC and PFOA, the pooled associations did not change when adjusting for
possible publication bias (Figure F-3). However, methods to assess heterogeneity and publication
bias have limitations in small sample-size meta-analyses, thus these results should be interpreted
cautiously (von Hippel, 2015).
Proposed PFAS Rule Economic Analysis
F-10
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table F-2: Results for PFOA Meta-Analyses
Number of
Group
Outcome
Studies/
Number of
Estimates
Beta
(mg/dL per ng/mL)
95% CIs
p-value
Qa
p-value
for Q
I2
Tau2
All Studies
TC
HDLC
11/14
13/17
0.003
0.001
-0.001
-0.001
0.006
0.004
0.177
0.291
123.68
54.74
<0.001
<0.001
89.49
70.77
0
0
Linear Models Only
TC
HDLC
4
5
1.574
0.105
0.018
-0.219
3.130
0.428
0.048
0.526
23.43
14.01
< 0.001
0.007
87.19
71.45
1.910
0.069
Sensitivity Analyses
All lower risk of bias studies
TC
HDLC
8/11
9/13
0.003
0.002
-0.003
-0.002
0.008
0.005
0.321
0.290
88.86
28.34
<0.001
0.005
88.75
57.65
0
0
Exclude Jain et al. (2019)
TC
HDLC
10
12/13
0.004
0.001
-0.002
-0.003
0.010
0.006
0.179
0.500
82.04
50.18
<0.001
<0.001
89.03
76.09
0
0
Exclude non-US/Canada and
TC
8/11
0.002
-0.003
0.006
0.496
55.65
<0.001
82.03
0
high exposure studies
HDLC
8/11
0.001
-0.003
0.005
0.647
26.17
0.004
61.79
0
All studies, pooled Jain et al.
TC
11
0.003
-0.002
0.008
0.183
91.42
<0.001
89.06
0
(2019)
HDLC
13/14
0.001
-0.002
0.004
0.412
53.07
<0.001
75.51
0
All studies, no NHANES
TC
6
0.017
-0.033
0.067
0.505
21.56
0.001
76.9
0.001
overlap
HDLC
8/9
0.0030
0.0029
0.0031
<0.001
4.12
0.844
0
0
Linear models only, no
TC
lb
1.480
0.180
2.780
0.026
0.00
NA
NA
NA
NHANES overlap
HDLC
2
0.185
-0.897
1.249
0.773
1.29
0.26
22.61
0.29
Linear-log models only
TC
HDLC
3/6
5/9
0.002
0.001
-0.004
-0.003
0.007
0.006
0.594
0.490
31.56
13.56
<0.001
0.094
84.16
41.01
0
0
P.-I. D. Lin etal. (2019)
TC
HDLC
1
1
1.632
-0.131
-0.841
-0.370
2.422
0.107
>0.05
>0.05
0.00
0.00
NA
NA
NA
NA
NA
NA
Dong et al. (2019)
TC
HDLC
1
1
1.480
-0.025
0.180
-0.443
2.780
0.393
0.026
>0.05
0.00
0.00
NA
NA
NA
NA
NA
NA
Abbreviations: CI - confidence interval; HDLC - high-density lipoprotein cholesterol; TC - total cholesterol; PFOA- Perfluorooctanoic Acid.
Notes:
aQ statistics for heterogeneity. Tau2 is the between-studies variance. I2 represents the proportion of total variance in the estimated model due to inter-study variation.
bData from Dong et al. (2019) Statistics for heterogeneity do not apply when only one study is used.
Proposed PFAS Rule Economic Analysis
F-ll
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
TC
Effect estimate
95% CI
Weight
Nelson et al. (2010)
He et al. (2018)
Dong et al. (2019)
Fan et al. (2020)
Overall
Study
1.2200 [
0.0016 [
1.4800 [
6.7400 [
0.0400,
0.0001,
0.1800,
3.2551,
2.4000] 27.74
0.0030] 33.01
2.7800] 26.83
10.2249] 12.43
1.5737 [ 0.0177, 3.1297]
0 5 10
Beta (95% confidence interval)
HDLC
Effect estimate
Weight
95% CI
(%)
-0.1200 [
-0.4050,
0.1650]
25.55
1.5855 [
-1.1600,
4.3310]
1.40
-0.0007 [
-0.0022,
0.0008]
31.46
-0.3000 [
-0.7100,
0.1100]
21.27
0.6500 [
0.0150,
1.2850]
1463
2.2300 [
0.9700,
3.4900]
5.69
0.1495 [ -0.1826,
0.4817]
Nelson etal. (2010)
Fu etal. (2014)
He et al. (2018)
Dong et al. (2019)
Dong et al. (2019)
Fan et al. (2020)
Overall
-2 0 2 4
Beta (95% confidence interval)
Figure F-2: Forest Plots Showing the Beta Coefficients
Relating PFQA Concentrations to TC and HDLC in Each Study Reporting
Linear Associations, and Pooled Estimates After Random-Effects Meta-Analysis.
Proposed PFAS Rule Economic Analysis
F-12
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
CD
"O
CU '
"O
a
03
CO
Funnel plot
•
••
TC
-2
Funnel plot
o -
<
HDLC
•
/ i
• \
O ^ -
/
/
\
q3
•
•
~o
\
CTJ
\
"O
\
c
\
CTJ
\
CO -
•
LO _
-4
-2
Effect size
0
Effect size
Pseudo 95% CI
Estimated 9Dl
Studies
Pseudo 95% CI • Observed studies
Estimated 0Dl * Imputed studies
Figure F-3: Filled-in Funnel Plots to Evaluate Publication Bias of the PFOA
and TC (Left) or HDLC (Right) Association in Studies Reporting Linear Associations.
Note: The funnel plot shows individual studies included in the analysis according to random-effect beta estimates (x-axis) and the standard error of each study-specific
beta (y-axis). Hie red vertical line indicates the pooled estimate for all studies combined and the gray lines indicate pseudo 95% confidence limits around the pooled estimate.
Number of observed studies: 4 (TC) and 6 (HDTC).
Proposed PFAS Rule Economic Analysis
F-13
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Study
Steenland (2009)
Nelson (2010)
Fisher (2013)
Fu(2014)
He (2018)
Liu (2018)
Dong (2019)
Jain (2019) f no
Jain (2019) fo
Jain (2019) m no
Jain (2019) m o
Lin (2019)
Fan (2020)
Li (2020)
Overall
(a) TC and PFOA Serum Relationship
2.5 5.0 7.5
Beta (95% confidence interval)
10.0
Study
Steenland (2009)
Nelson (2010)
Fisher (2013)
Fu (2014)
He (2018)
Liu (2018)
Yang (2018)
Dong (2019)
Jain (2019) f no
Jain (2019) fo
Jain (2019) m no
Jain (2019) m o
Lin (2019)
Fan (2020)
Li (2020)
Lin (2020) br PFOA
Lin (2020) lin PFOA
Overall
-2.5 0.0 2.5 5.0
Beta (95% confidence interval)
Effect Estimate
[95% CI]
Weight
0.0060
[0.0059,
0.0060]
18.5895
1.2200
[0.0400,
2.4000]
0.0009
0.0504
[-0.2375,
0.3383]
0.0157
0.2447
[-0.5493,
1.0386]
0.0021
0.0016
[0.0001,
0.0030]
18.0390
2.4008
[0.6889,
4.1126]
0.0004
1.4800
[0.1800,
2.7800]
0.0008
0.0019
[-0.0022,
0.0059]
15.0655
-0.0004
[-0.0047,
0.0039]
14.7342
-0.0011
[-0.0036,
0.0015]
17.0060
0.0058
[0.0014,
0.0102]
14.5987
1.6317
[0.8413,
2.4221]
0.0021
6.7400
[3.2551,
10.2249]
0.0001
0.0064
[-0.0181,
0.0309]
1.9450
0.0025
[-0.0011,
0.0061]
Serum Relationship
Effect Estimate
[95% Cll
Weiqht
0.0030
[0.0029,
0.0031]
22.7126
-0.1200
[-0.4050,
0.1650]
0.0082
0.0004
[-0.0708,
0.0716]
0.1309
1.5855
[-1.1600,
4.3310]
0.0001
-0.0007
[-0.0022,
0.0008]
21.1506
0.8304
[0.2907,
1.3701]
0.0023
2.2240
[-2.4463,
6.8943]
0.0000
-0.0253
[-0.4438,
0.3933]
0.0038
0.0027
[-0.0021,
0.0075]
12.7084
0.0029
[-0.0045,
0.0104]
7.8535
-0.0006
[-0.0041,
0.0028]
16.0751
0.0016
[-0.0031,
0.0062]
13.0741
-0.1313
[-0.3697,
0.1072]
0.0117
2.2300
[0.9700,
3.4900]
0.0004
0.0027
[-0.0061,
0.0115]
6.2645
-0.5205
[-3.8363,
2.7952]
0.0001
0.1765
[-0.2432,
0.5961]
0.0038
0.0014
[-0.0012,
0.0040]
Figure F-4: Forest Plots Showing the Beta Coefficients Relating TC and HDLC to PFOA
Concentrations in Each Study, and Pooled Estimates After Random-Effects Meta-Analysis.
Abbreviations: f- females; m - males; o - obese; no - non-obese
Proposed PFAS Rule Economic Analysis
F~14
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
CD
TD
CU "
"O
C.
03
CO
Funnel plot
• /
J
/
TC
%
-10
-5
0
Effect size
10
Pseudo 95% CI
Estimated 0DL
Observed studies
Imputed studies
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
FA.2 Slope Estimation for PFOS
When including the five studies reporting linear associations, there was a positive increase in TC
of 0.08 (95% CI: -0.01, 0.16) mg/dL per ng/mL serum PFOS (p-value=0.064,12=84%) that was
significant at the 0.10 level. The association for PFOS and HDLC was positive but not
statistically significant (Table F-3, Figure F-6). Adjusting for possible publication bias through
funnel plots and trim-and-fill analysis suggested the imputation of additional studies; however,
the magnitude or significance of the pooled associations did not change significantly (Figure
F-7).
When all studies were combined (12 studies, 15 results), EPA observed a borderline statistically
significant positive increase in TC of 0.066 (95% CI: -0.001, 0.132) mg/dL per ng/mL serum
PFOS (p-value=0.055,12=100%) (Table F-3, Figure F-8). Adjusting for possible publication bias
through funnel plots and trim-and-fill analysis suggested the imputation of three additional
studies for TC and five for HDLC; however, the pooled effect estimates did not change
significantly (Figure F-9). EPA observed similar results in leave-one-out analyses, sensitivity
analyses restricted to U.S or Canadian general population studies, and analyses excluding Jain et
al. (2019), estimates. Similar results were observed when the analysis excluded the overlapping
NHANES studies. When the analysis excluded the higher ROB studies, the association was
significantly positive with an increase in in TC of 0.09 (95% CI: 0.01, 0.17) mg/dL per ng/mL
serum PFOS (p-value=0.047).
The pooled estimate based on the studies reporting linear associations was 0.08 (95% CI: -0.01,
0.16) and significant at the 0.10 level (p-value=0.064) and there is evidence supporting a positive
and significant relationship between PFOS and TC: EPA/OST's review of 41 recent
epidemiological studies showed positive associations between PFOS and TC in the general
population and the meta-analysis performed with all studies combined showed a positive
increase in TC per ng/mL serum PFOS that was significant at the 0.10 level. Given this weight of
evidence, the large degree of heterogeneity in the pooled associations when all data were
included, and the likelihood of bias that back-transformation of effect estimates with log-
transformed outcomes or exposures could introduce (and difficulty with estimating the
directionality of this bias towards or away from the null), EPA relied on the results from analyses
restricted to studies reporting similar models, favoring the pooled slope (from the six studies
reporting linear associations) of 0.08 mg/dL TC and 0.05 mg/dL HDLC per ng/mL serum PFOS
for interpretability and use in the CVD risk reduction analysis.34
34 EPA characterizes uncertainty surrounding this estimate as described in Appendix M.
Proposed PFAS Rule Economic Analysis
F-16
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table F-3: Results for PFOS Meta-Analyses
Group
Outcome
N Studies/
Number of
Estimates
Beta
(mg/dL per ng/mL)
95% CIs
p-value
Qa
p-value
for Q
I2
Tau2
All
TC
12/15
0.066
-0.001
0.132
0.055
630000
<0.001
100
0.012
/111 ulUUlto
HDLC
14/19
0.0003
-0.001
0.001
0.631
158.85
<0.001
88.67
0
Linear Models Only
TC
5
0.079
-0.005
0.162
0.064
25.84
< 0.001
84.52
0.004
HDLC
6/7
0.050
-0.005
0.105
0.074
31.69
< 0.001
81.06
0.003
Sensitivity Analyses
All lower risk of bias
TC
9/12
0.086
0.001
0.170
0.047
450000
<0.001
100
0.016
studies
HDLC
10/15
0.001
-0.001
0.002
0.606
84.54
<0.001
83.44
0
Exclude Jain et al. (2019)
TC
11
0.114
0.012
0.217
0.028
510000
<0.001
100
0.019
13/15
HDLC
-0.002
-0.002
0.001
0.778
126.90
<0.001
88.97
0
Exclude non-US/Canada
TC
8/11
0.001
-0.0004
0.001
0.301
34.71
<0.001
71.20
0
and high exposure studies
HDLC
8/11
0.001
-0.0002
0.001
0.165
13.12
<0.001
23.76
0
All studies, pooled Jain et
TC
12
0.094
0.010
0.179
0.029
590000
<0.001
100
0.015
al. (2019)
HDLC
14/16
-0.0001
-0.0014
0.0013
0.943
157.53
<0.001
90.48
0
All studies, no NHANES
TC
7
0.109
-0.016
0.234
0.088
120000
<0.001
100
0.022
overlap
HDLC
9/11
-0.001
-0.002
0.002
0.642
94.82
<0.001
89.45
0
Linear models only, no
TC
2b
0.192
-0.162
0.546
0.288
6.88
0.009
85.46
0.057
NHANES overlap
HDLC
3/4
0.078
0.001
0.155
0.048
7.32
0.062
59.03
0.003
Linear-log models only
TC
3/6
0.0003
-0.0003
0.001
0.342
8.33
0.139
39.99
0
HDLC
5/9
0.001
-0.001
0.002
0.270
15.74
0.046
49.18
0
P.-I. D. Lin et al. (2019)
TC
1
0.132
-0.005
0.269
>0.05
0.00
NA
NA
NA
1
0.00
NA
NA
NA
HDLC
-0.021
-0.062
0.020
>0.05
Dong et al. (2019)
TC
1
0.40
0.13
0.67
<0.01
0.00
NA
NA
NA
1
0.00
NA
NA
NA
HDLC
0.014
-0.084
0.110
>0.05
Abbreviations: HDLC- High-Density Lipoprotein Cholesterol; TC- Total Cholesterol; PFOS- Perfluorooctanesulfonic Acid.
Notes:
aQ statistics for heterogeneity. Tau2 is the between-studies variance. I2 represents the proportion of total variance in the estimated model due to inter-study variation.
Abbreviations: CI - confidence interval; HDLC - High-Density Lipoprotein Cholesterol; NHANES - National Health and Nutrition Examination; TC - Total Cholesterol.
Notes:
aQ statistics for heterogeneity. Tau2 is the between-studies variance.
bData from Dong et al. (2019) and Chateau-Degat et al. (2010).
Proposed PFAS Rule Economic Analysis
F-17
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Effect estimate
Weight
Study
TC
95% CI
(%)
Chateau-Degat (2010)
¦
0.0348 [ -0.0049, 0.0745]
38.68
Nelson (2010)
0.2700 [ 0.0550, 0.4850]
11.05
He (2018)
¦
0.0008 [ 0.0003, 0.0012]
42.42
Dong (2019)
—¦
0.4000 [ 0.1300, 0.6700]
7.74
Fan (2020)
3.8500 [ 1.2750, 6.4250]
0.10
Overall
0.0786 [ -0.0045, 0.1617]
"T 1 1 r
0 2 4 6
Beta (95% confidence interval)
Study hdlc
Chateau-Degat (2010) f
Chateau-Degat (2010) m
Nelson (2010)
Fu (2014)
He (2018)
Dong (2019)
Fan (2020)
Overall
0.1624 [
0.0664,
0.2584]
14.78
0.0619 [
0.0254,
0.0984]
24.24
0.0200 [
-0.0500,
0.0900]
18.80
2.5909 [
-0.6767,
5.8584]
0.03
0.0002 [
-0.0004,
0.0008]
27.19
0.0135 [
-0.0836,
0.1107]
14.62
1.2400 [
0.3200,
2.1600]
0.35
0.0498 [
-0.0048,
0.1045]
0 2 4 6
Beta (95% confidence interval)
Figure F-6: Forest Plots Showing the Beta Coefficients Relating TC
and HDLC to PFOS Concentrations in Each Study Reporting Linear Associations,
and Pooled Estimates After Random-Effects Mela-Analysis.
Abbreviations: f - females; m - males; o - obese; no - non-obese
Proposed PFAS Rule Economic Analysis
F-18
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
o -
fc
QJ
¦S
CTJ
~a
c
f0
Funnel plot
%
TC
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(a) TC and PFOS Serum Relationship
Study
Steenland (2009)
Chateau-Degat (2010)
Nelson (2010)
Fisher (2013)
Fu (2014)
He (2018)
Liu (2018)
Dong (2019)
Jain (2019) f no
Jain (2019) fo
Jain (2019) m no
Jain (2019) m o
Lin (2019)
Fan (2020)
Li (2020)
Overall
0 2 4 6
Beta (95% confidence interval)
Effect Estimate
[95% CI]
Weight
0.2091
[0.2086,
0.2095]
9.3960
0.0348
[-0.0049,
0.0745]
9.0953
0.2700
[0.0550,
0.4850]
4.7663
0.0077
[-0.0669,
0.0824]
8.4110
0.1252
[-0.8085,
1.0590]
0.4863
0.0008
[0.0003,
0.0012]
9.3960
0.2116
[-0.4377,
0.8609]
0.9530
0.4000
[0.1300,
0.6700]
3.7111
0.0004
[-0.0006,
0.0014]
9.3959
0.0009
[-0.0003,
0.0021]
9.3958
-0.0003
[-0.0009,
0.0004]
9.3960
0.0006
[-0.0004,
0.0016]
9.3959
0.1318
[-0.0052,
0.2688]
6.7386
3.8500
[1.2750,
6.4250]
0.0670
0.0003
[-0.0008,
0.0014]
9.3958
0.0655
[-0.0014,
0.1324]
Study
Steenland (2009)
Chateau-Degat (2010) f
Chateau-Degat (2010) m
Nelson (2010)
Fisher (2013)
Fu (2014)
He (2018)
Liu (2018)
Yang (2018)
Dong (2019)
Jain (2019) f no
Jain (2019)fo
Jain (2019) m no
Jain (2019) m o
Lin (2019)
Fan (2020)
Li (2020)
Lin (2020) br PFOS
Lin (2020) lin PFOS
Overall
(b) HDLC and PFOS Serum Relationship
Effect Estimate
[95% CI]
Weight
-0.0015
[-0.0016,
-0.0014]
16.4599
0.1624
[0.0664,
0.2584]
0.0122
0.0619
[0.0254,
0.0984]
0.0844
0.0200
[-0.0500,
0.0900]
0.0230
-0.0030
[-0.0293,
0.0233]
0.1613
2.5909
[-0.6767,
5.8584]
0.0000
0.0002
[-0.0004,
0.0008]
15.6444
0.1578
[-0.0801,
0.3958]
0.0020
0.1855
[-1.5301,
1.9010]
0.0000
0.0135
[-0.0836,
0.1107]
0.0120
0.0010
[-0.0004,
0.0025]
12.5475
0.0015
[-0.0005,
0.0034]
10.5739
0.0000
[-0.0008,
0.0008]
14.9834
0.0007
[-0.0006,
0.0019]
13.2682
-0.0208
[-0.0620,
0.0203]
0.0664
1.2400
[0.3200,
2.1600]
0.0001
0.0001
[-0.0003,
0.0005]
16.1448
-2.0986
[-3.9343,
-0.2628]
0.0000
0.0977
[0.0144,
0.1809]
0.0163
0.0003
[-0.0008,
0.0013]
Beta (95% confidence interval)
Figure F-8: Forest Plots Showing the Beta Coefficients Relating PFOS Concentrations to
TC and HDLC in Each Study, and Pooled Estimates After Random-Effects Meta-Analysis.
Abbreviations: f - females; m - males; o - obese; no - non-obese.
Proposed PFAS Rule Economic Analysis
F~20
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
F.4.3 Sensitivity Analyses
EPA considered two studies for use in single-study sensitivity analyses to understand the impact
of using the estimates from the meta-analyses in the CVD risk reduction modeling output. These
analyses are described in greater detail in Appendix K.
Using data from NHANES (2003-2014) on 8,948 adults, Dong et al. (2019) reported significant
increases in TC: 1.48 (95% CI: 0.18, 2.78) mg/dL per ng/mL serum PFOA and 0.40 (95% CI:
0.13, 0.67) mg/dL per ng/mL PFOS (Table F-2). For HDLC the associations were of -0.03 (95%
CI: -0.44, 0.39) mg/dL per ng/mL PFOA and 0.01 (95% CI: -0.08, 0.11) mg/dL per ng/mL
PFOS. The results were adjusted for age, gender, race, family income index, body mass index,
waist circumference, physical activities, diabetes status, smoking status, and number of alcoholic
drinks per day. Participants using lipid-lowering medications were excluded. As part of
developing EPA's Toxicity Assessments and Proposed Maximum Contaminant Level Goals for
PFOA and PFOS in Drinking Water, EPA considered this medium quality study for estimating
point of departure for potential use in toxicity value derivation (U.S. EPA, 2023a; U.S. EPA,
2023b).
The P.-I. D. Lin et al. (2019) study included participants in a clinical trial of the effect of lifestyle
modifications on pre-diabetes. This study included 888 pre-diabetic adults who were recruited
from 27 medical centers in the US during 1996-1999. The study considered both cross-sectional
(baseline) and prospective assessments, with the results showing evidence of an association
between PFOA and increased TC and hypertriglyceridemia. Each doubling of plasma PFOA
concentration at baseline was associated with 6.1 mg/dL (95% CI: 3.1, 9.0) increase in TC. The
results were adjusted for 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. Participants using lipid-lowering medications were
excluded. The results from the longitudinal analysis were not considered because they were not
presented in a format amenable for dose-response analyses. The study provides another line of
evidence to support associations with TC among adults with pre-diabetes and comparable plasma
PFAS concentrations to the U.S. general population.
F.4.4 Limitations and Uncertainties
Table F-4 summarizes limitations and sources of uncertainty associated with the estimated serum
cholesterol dose-response functions. The effects of these limitations and sources of uncertainty
on estimates of risk reduction and benefits evaluated in the PFAS National Primary Drinking
Water Regulation (NPDWR) are uncertain.
Proposed PFAS Rule Economic Analysis
F-22
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table F-4: Limitations and Uncertainties in the Analysis of the Serum Cholesterol Dose
Response Functions
Uncertainty/Assumption
Notes
All of the studies included in the meta-analysis, except
one (P.-I. D. Lin et al., 2019), are cross-sectional
designs with various design or methodologic
limitations. The cross-sectional nature of designs could
raise concerns about reverse causality.
Measuring PFOA or PFOS and serum lipids
concurrently, as was the case in cross-sectional
designs, was considered adequate in terms of exposure
assessment timing. Given the long half-lives of PFOA
and PFOS (with median half-lives of 2.7 and 3.5
years, respectively; Ying Li et al., 2018), current blood
serum concentrations are expected to correlate well
with past exposures. Furthermore, although reverse
causality due to reverse causation due to
hypothyroidism (Dzierlenga, Allen, et al., 2020) or
enterohepatic cycling of bile acids (Fragki et al., 2021)
has been suggested, there is not yet clear evidence to
support these reverse causal pathways. Regarding
methodology, several NHANES-based studies (Dong
et al., 2019; He et al., 2018) did not clearly report
whether sampling weights were used in the analyses to
account for the complex sampling design (as is the
norm in such survey-based studies).
Some NHANES-based studies used data from
overlapping NHANES cycles.
Using study results with overlapping years of data
could result in double counting certain data and may
introduce uncertainty in the meta-analysis estimates.
Dong et al. (2019) used data from 2003-2014, while
He et al. (2018) used data from 2003-2012; Jain et al.
(2019) used data from 2005-2014; Fan et al. (2020)
used data from 2011-2014; Liu et al. (2018) used data
from 2013-2014; and Nelson et al. (2010) used data
from 2003-2004. A sensitivity analysis excluding the
overlapping NHANES studies supported the main
findings.
Studies used a variety of statistical models for
estimating the associations of interest (including
NHANES-based studies).
Most studies provided measurements of PFOA and
PFOS in serum, except in three studies that used
measurements in plasma (Chateau-Degat et al., 2010;
Fisher et al., 2013; P.-I. D. Lin et al., 2019).
Distribution of PFAS to plasma is chain-length
dependent, and within human blood fractions, PFOS
and PFOA accumulate to the highest levels in plasma,
followed by whole blood and serum. Typically, the
study-specific estimated associations are rescaled
when the study-specific measurements are in whole
blood, but in common practice serum and plasma-
based associations are not rescaled.
Including these studies in meta-analyses introduces
uncertainty in the estimates.
Proposed PFAS Rule Economic Analysis
F-23
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table F-4: Limitations and Uncertainties in the Analysis of the Serum Cholesterol Dose
Response Functions
Uncertainty/Assumption
Notes
Existing approaches are limited in their ability to
evaluate statistical heterogeneity and the potential for
publication bias
EPA performed statistical evaluations to assess
sources of heterogeneity in effect estimates, and to
evaluate potential for publication bias. However, the
approaches for evaluating heterogeneity and
publication bias are sometimes limited in their ability
to do so. Evaluating statistical heterogeneity in meta-
analyses with a small number of studies is limited by
the potential that the I2 statistic can be imprecise and
biased, and thus results should be interpreted
cautiously (von Hippel, 2015).a In evaluating
publication bias, the funnel plot asymmetry is a
subjective assessment and is recommended only when
at least 10 studies are included in the meta-analysis
(Higgins et al., 2021). Furthermore, the Egger
regression test and Begg's rank tests for publication
bias (Begg et al., 1994; Egger et al., 1997; Egger et al.,
2008) may suffer from inflated type I error and limited
power in certain situations, especially when there is a
high degree of heterogeneity (L. Lin et al., 2018).
Finally, the small number of studies reporting slopes
from similar models limits the power of the meta-
analysis.
Abbreviations: NHANES-The National Health and Nutrition Examination Survey; PFOA- Perfluorooctanoic acid;
PFOS- Perfluorooctanesulfonic acid.
Note:
aI2 represents the percentage of variation across studies that is due to heterogeneity rather than chance.
Proposed PFAS Rule Economic Analysis
F-24
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix G. CVD Benefits Model Details and Input
Data
This appendix provides details of the CVD model linking changes in TC, HDLC, and systolic
blood pressure to changes in incidence of first hard CVD events in populations exposed to
PFOA/ PFOS through drinking water. These approaches have been peer reviewed by EPA's
Science Advisory Board; input provided by that organization has been considered in finalizing
this analysis (U.S. EPA, 2022). As discussed in the SAB in-person meetings and the final report
(U.S. EPA, 2022), SAB members and the formal report considered the approaches taken in this
document, including using the life table approach and ASCVD model, to be reasonable and valid
approaches for estimating reduced CVD cases associated with reduced PFOA and PFOS.
TC and HDLC were linked to serum PFOA and serum PFOS, as described in Appendix F.
However, evidence of an association between PFOA and PFOS and HDLC effects was
inconclusive (U.S. EPA, 2023a; U.S. EPA, 2023b); therefore, EPA modeled HDLC effects only
as part of a sensitivity analysis (see Appendix K). The relationship between BP and serum PFOS
among those not using hypertensive medications is discussed in Section 6.5 of the Economic
Analysis. First hard CVD events included in the model are non-fatal myocardial infarction (MI),
non-fatal ischemic stroke (IS), and coronary heart disease (CHD) deaths. The model also
captures post-acute CVD mortality experienced by the first non-fatal MI or IS survivors within 6
years of the initial event.
G.l Model Overview and Notation
The CVD model is designed to estimate a time series of hard CVD event incidence for a
population cohort characterized by sex, race/ethnicity, birth year, and age at the beginning of the
evaluation period (i.e., 2023), and birth year-, age- and sex-specific TC, HDLC, and BP level
time series estimated upstream. The first hard CVD event incidence estimates are generated
using the Pooled Cohort ASCVD model (Goff et al., 2014), whose predictors include age,
cholesterol levels, blood pressure, smoking status, and diabetes status. For those ages 40-80, the
ASCVD model predicts the 10-year probability of a hard CVD event—non-fatal MI, fatal and
non-fatal IS, or CHD death—to be experienced by a person without a prior history of MI, IS,
congestive heart failure, percutaneous coronary intervention, coronary bypass surgery, or atrial
fibrillation. EPA models post-acute CVD mortality for survivors of the first MI or IS at ages 45-
65 using race/ethnicity- and sex-specific estimates at 1-year and 5-year follow-up from Thom et
al. (2001). For survivors of the first MI or IS at age 66 or older, EPA models post-acute CVD
mortality using estimates at 1- to 6-year follow-ups from S. Li et al. (2019).
The CVD model integrates the ASCVD model predictions and post-acute CVD mortality
estimates in the series of recurrent calculations that produce a life table estimate for the
population cohort of interest (e.g., non-Hispanic White females aged 70 years at the beginning of
the evaluation period). For each PWS, EPA evaluates population cohorts defined by a
combination of birth year and age in or after 2023 (i.e., pairs of (2023,0), (2022,1), (2021,2),
(1938,85+) and pairs of (2024,0), (2025,0), ... , (2065,0)), sex (males and females), and
race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, Other). In addition to the
standard life table components, such as the annual number of all-cause survivors and deaths for
Proposed PFAS Rule Economic Analysis
G-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
all ages, for ages 40+, the CVD model estimates the number of surviving persons with and
without a history of hard CVD events, the number of persons experiencing hard CVD events at a
given age, and deaths from CVD and non-CVD causes at a given age.
Figure G-l summarizes the main types of CVD model calculations for a population cohort age 0
at the start of the evaluation period..35 The CVD model calculations are identical across the
race/ethnicity and sex demographic subgroups but use subgroup-specific coefficients.36 For
cohorts born prior to or in 2023, the CVD model is initialized using the PWS-, age-,
race/ethnicity-, and sex-specific number of persons estimated to be alive in 2021. For cohorts
born after 2023, the CVD model is initialized using the PWS-, race/ethnicity-, and sex-specific
number of persons aged 0 estimated to be alive in 2021. PWS- and sex, race/ethnicity-, and age-
specific population details are included in Appendix B. Once the model is initialized, the
following types of calculations occur for each year within the simulation period:
• Recurrent standard life table calculations that rely on the all-cause age-specific annual
mortality rates to evaluate the number of deaths among persons of a specific integer age
and the number of survivors to the beginning of the next integer age. These calculations
are executed whenever the current cohort age is in the 0-39 range. They are represented
by the green segments of the timeline shown in Figure G-l.
• Recurrent life table calculations that separately track subpopulations with and without a
history of hard CVD events, including estimation of the number of annual CVD and non-
CVD deaths (in either subpopulation), as well as the number of annual post-acute CVD
deaths experienced by survivors of the first hard CVD events that occurred, at most, 5
years ago. These calculations are executed whenever the current cohort age is over age
40.37 These calculations are represented by the red segment of the timeline in Figure G-l.
Figure G-2 further illustrates the year-specific calculations required for explicit tracking
of subpopulations with and without a hard CVD event history.
35 This initial population cohort age is chosen because it allows for the illustration of the full set of calculation types
used in the CVD model.
36 There are different ASCVD model coefficients for non-Hispanic White and non-Hispanic Black males and
females. The figure shows the generalized approach of the CVD model.
37 People 85 years or older are treated as a single cohort in the model. The mortality rates for this cohort are assumed to be the
average mortality rate for those aged 85-100 years. EPA also relied on serum PFOA/PFOS values at age 85 for the 85+ cohort.
Proposed PFAS Rule Economic Analysis
G-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Model initialization
with estimated
location- and
race/ethnicity-specific
male/female age 0
population Recurrent standard life table
calculations using age-,
race/ethnicity-, and
sex-specific all-cause
annual mortality rates
„ A
V
Modeled population is
split into CVD and non-CVD
subpopulations using
age-, race/ethnicity-, and sex-specific
CVD prevalence data
A
r
Recurrent life table calculations
with explicit treatment of CVD population,
using age- and sex-specific CVD prevalence,
cause-specific annual mortality rates,
ASCVD model-based CVD incidence,
and post-acute CVD event mortality
Figure G-l: Overview of Life Table Calculations in the CVD Model.
Note: The figure illustrates the model for population cohort age 0 at the beginning of the evaluation period (i.e., calendar year 2023). Hie model is initialized using an age 0
PWS-specific population (see Appendix B for PWS population details).
Proposed PFAS Rule Economic Analysis
G-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Figure G-2 provides additional information on the post-acute CVD mortality estimation. Each
person included in the surviving current age-specific incident CVD subpopulation38
(corresponding to the group F result in Figure G-2) is tracked for 5 additional years to estimate
the number of CVD deaths occurring in that timeframe. The recurrent estimates rely on age-
specific non-CVD mortality, estimated based on CDC life table data and age- and sex-specific
annual CVD mortality rates, and age- and post-acute CVD mortality, estimated based on Thom
et al. (2001) and S. Li et al. (2019).
38 For example, persons who experienced their first non-fatal MI or IS at age 70 and survived through the first post-
event year.
Proposed PFAS Rule Economic Analysis
March 2023
G-4
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Non-CVD population eligible for
the first hard CVD event in the
following year (A-B-C-D).
First hard CVD event deaths (C)
First hard CVD event survivors (D)
CVD event survivor population (D) is adjusted for post-acute
excess mortality in year 0 since the initial event (E)*
First hard CVD event
survivors at the end of
first post-event year (F)
Post-acute excess deaths
among CVD survivors in
first post-event year (E)
Living subpopulation without prior history
of CVD events.
Note:
* Estimated number of CVD events is an input to
the monetization step.
Deaths occurring at the current integer age
Living subpopulation that experienced first
hard CVD at the current integer age
Current-year calculations
Calculations occurring in years 1-5
following the first hard CVD event
Figure G-2: CVD Model Calculations Tracking CVD
and Non-CVD Subpopiilations for a Specific Current Age of Cohort.
Proposed PFAS Rule Economic Analysis
G-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table G-l summarizes the data elements and notation of the CVD model..39 The CVD model
elements fall into four categories: indices, data, quantities computed upstream, and internally
computed quantities. Information sources and computational notes for the model elements
identified as "data" are fully described in Section G.5. Changes in the modeled biomarker levels
(Att>,a,s,t) are a birth year, age, sex, and calendar year-specific quantities computed upstream for
the regulatory alternatives as described in Section 6.5 of the Economic Analysis..40 Section G.2
describes the estimation of first hard CVD event incidence and post-acute CVD mortality, which
are internally computed quantities. Derivation of the remaining internally computed quantities
for the baseline life table is given in Section G.3.1 and Section G.3.2, while derivation of those
quantities for the regulatory alternative life table is given in Section G.3.3.
Table G-l: CVD Life Table Model Elements and Notation Summary
Model Element
Element Type
Definition
b
s
r
f
V
b,a,s,r,max (0,b)
h),a,s,r,t
d-b,a,s,r,t
7Tn c r
b,a,s,r,t,p
Index
Index
Index
Index
Index
Index
Index
Index
Index
Data
Internally computed
quantity
Internally computed
quantity
Data
Internally computed
quantity
Current integer age, A = {0,1,2,... ,99}. The life table model
assumes that all persons are born on January 1.
Current calendar year, t = 0 marks the beginning evaluation
period, t = T marks the end of evaluation period
Calendar birth year, B = {—T, ...,0,1, ...,T — 40}
Sex, 5 = {male, female}
Race/Ethnicity, R = {non — Hispanic White, non —
Hispanic Black, other}
First hard non-fatal CVD event type,
F = {non — fatal MI, non — fatal IS}
Population type: CVD - population with a history of hard CVD
events; OTH - non-CVD population
Cause of death: CVD - cardiovascular disease death; OTH - death
from causes other than CVD
Number of years elapsed since first hard CVD event,
K = {0,1,2,3,4,5}
Living population of age a, sex s, and race/ethnicity r, born in
year b. at the beginning of the evaluation period for the cohort:
t = max (0, b)
Living population born in year b. of sex s and race/ethnicity r,
at the beginning of integer age a and calendar year t
Number of all-cause deaths in population born in year b, of sex s
and race/ethnicity r, at integer age a and calendar year t
Prevalence rate of persons with past experience of hard CVD
events at age a, sex s, and race/ethnicity r
Living population born in year b, of type p, sex s, and
race/ethnicity r, at the beginning of integer age a and calendar
year t. Note that lb,o,s,r,t,cvd = 0. 'c- EPA assumes that people
who have just been born do not have CVD history by definition.
39 SafeWater was programmed for maximal computational efficiency and SafeWater performs a series of pre-calculations to
reduce model runtime. Therefore, the specific equations in the SafeWater code differ from the equations in this Appendix, but the
end result is mathematically consistent.
40 Total cholesterol change for the baseline life table calculations is 0 by definition.
Proposed PFAS Rule Economic Analysis
G-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table G-l: CVD Life Table Model Elements and Notation Summary
Model Element
Element Type
Definition
1b,a,s,r,t,V,c
Qa,s,r
tfa,s,r,c
Atb,a,s,t
i-b,a,s,r,t iP^-b,a,s,t)
Ya,s,r,f
Pb,a,s,r
fta,s,r,f,k
%b,a,s,r,t
Xb,a,s,r,t
^-b,a,s,r,f,t,0
n-b,a,s,r,f,t,k
mb,a,s,r,t,0
m-b,a,s,r,t,k
An,
b,a,s,r,f,t
Internally computed
quantity
Data
Data
Quantity computed
upstream
Internally computed
quantity
Data
Internally computed
quantity
Data
Internally computed
quantity
Internally computed
quantity
Internally computed
quantity
Internally computed
quantity
Internally computed
quantity
Internally computed
quantity
Internally computed
quantity
Number of deaths from cause c in population born in year b. of
type p, sex s, and race/ethnicity r, throughout integer age a and
calendar year t; deaths from cardiovascular causes occur only in
the CVD population (i.e., d
b,a,s,r,t, OTH.CVD
= 0)
General population probability of all-cause death at integer age a,
sex s, race/ethnicity r
General population probability of death from cause c at integer age
a, sex s, race/ethnicity r
A 3-tuple of modeled changes in TC/HDLC/BP for population
born in year b, of sex s, age a , in calendar year t. Each element
of the 3-tuple is set to 0 for baseline calculations for all three
biomarkers. Additionally, the change in BP is set to 0 for persons
using antihypertensive medications regardless of whether the
baseline or the regulatory alternative is evaluated.
Incidence rate of first hard CVD events for persons born in year b.
of sex s and race/ethnicity r at age a and calendar year t; this rate
is computed using the ASCVD model.
Share of first non-fatal hard CVD event type / among all first hard
CVD events at age a , sex s, race/ethnicity r
Rate of CVD deaths in CVD population born in year b. alive at the
beginning of age a, for sex s and race/ethnicity r
Probability of post-acute CVD death in age a, sex s, and
race/ethnicity r CVD population who experienced first type /
non-fatal hard CVD event k integer years ago
Incident CVD population born in year b, of sex s and
race/ethnicity r, at the beginning of integer age a and calendar
year t
Calibration factor for the incident CVD population born in year b.
of sex s and race/ethnicity r, at the beginning of integer age a and
calendar year t
Uncalibrated number of living age a, sex s, and race/ethnicity r
persons born in year b. whose first type / non-fatal hard CVD
event occurred 0 years ago, corresponding to calendar year t
Number of living age a, sex s, and race/ethnicity r persons born in
year b. whose first type / non-fatal hard CVD event occurred k
years ago, corresponding to calendar year t
Uncalibrated number of CVD deaths among those born in year b.
age a, sex s, and race/ethnicity r persons whose first hard CVD
event occurred 0 years ago, corresponding to calendar year t
Number of CVD deaths among those born in year b. age a, sex s,
and race/ethnicity r persons whose first hard CVD event occurred
k years ago, corresponding to calendar year t
Difference between regulatory alternative and baseline number of
persons born in year b, of sex s and race/ethnicity r, whose first
type / non-fatal hard CVD event occurred at age a, corresponding
to calendar year t
Proposed PFAS Rule Economic Analysis
G-7
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table G-l: CVD Life Table Model Elements and Notation Summary
Model Element
Element Type
Definition
Am,
b.a.s.r.t
AN,
'ft
AM,
Internally computed
quantity
Internally computed
quantity
Internally computed
quantity
Difference between calendar year t regulatory alternative and
baseline number of CVD deaths among age a, sex s, and
race/ethnicity r persons born in year b, who experienced their first
hard CVD event during calendar years t - 5, t - 4, ... t
Difference between regulatory alternative and baseline number of
persons whose first type / non-fatal hard CVD event occurred
during calendar year t
Difference between regulatory alternative and baseline number of
year t CVD deaths among persons whose first hard CVD event
occurred during calendar years t - 5, t - 4, ... t
Abbreviations: ASCVD - atherosclerotic cardiovascular disease; BP - blood pressure; CVD - cardiovascular disease;
HDLC - high-density lipoprotein cholesterol; TC - total cholesterol.
G.2 Hard CVD Event Incidence Estimation
In this section, EPA describes the process for estimating the probability of the first hard CVD
event ib,a,s,r,t{^xb,a,s,t) using the ASCVD model (Section G.2.1); the prevalence of persons with
a history of hard CVD events na s r (Section G.2.2); the distribution of first hard CVD events by
type, including the share of non-fatal first hard CVD events Ya,s,r,f (Section G.2.3); and post-
acute CVD mortality rates Ha,s,r,f,k. within 6 years of the initial event (Section G.2.4).
G.2.1 Probability of the First Hard CVD Event
The first hard CVD event incidence estimates are generated by the Pooled Cohort ASCVD
model (Goff et al., 2014). The ASCVD model is commonly used in clinical practice to estimate
CVD risk for those aged 40-80 years. The ASCVD model predicts the 10-year probability of a
hard CVD event—fatal and non-fatal MI, fatal and non-fatal IS, or CHD death—to be
experienced by a person without a prior history of MI, IS, congestive heart failure, percutaneous
coronary intervention, coronary bypass surgery, or atrial fibrillation.
Four large longitudinal community-based epidemiologic cohort studies have been combined to
develop a geographically and racially diverse dataset used for the ASCVD model estimation:
(1) the Atherosclerosis Risk in Communities Study (Williams, 1989), (2) the Cardiovascular
Health Study (Fried et al., 1991), (3) the Coronary Artery Risk Development in Young Adults
Study (Friedman et al., 1988), and (4) the Framingham Original and Offspring Cohort Study
(Mahmood et al., 2014). Note that there are several other studies whose design is similar to the
one used in Goff et al. (2014), including D'Agostino et al. (2001), D'Agostino et al. (2000),
D'Agostino et al. (2008), D'Agostino et al. (1994), Pencina et al. (2009), Pencina et al. (2011),
Wilson et al. (1998), and Uno et al. (2011). Except for Uno et al. (2011), who also used the
Breast Cancer Survival Study (Chang et al., 2005), including D'Agostino et al. (2001),
D'Agostino et al. (2000), D'Agostino et al. (2008), D'Agostino et al. (1994), Pencina et al.
(2009), Pencina et al. (2011), Wilson et al. (1998), and Uno et al. (2011). Except for Uno et al.
(2011), who also used the Breast Cancer Survival Study (Chang et al., 2005), all of these studies
used the Framingham cohort study data that are not as diverse as the data used to estimate the
ASCVD model.
Proposed PFAS Rule Economic Analysis
G-8
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table G-2 shows the ASCVD model coefficient estimates used in the analysis. The predictors of
the ASCVD model include age, TC and HDLC concentrations, BP, current smoking, diagnosed
diabetes, and whether the subject is undergoing treatment for high BP. The model has been fit
separately to four population subgroups: non-Hispanic White females, non-Hispanic Black
females, non-Hispanic White males, and non-Hispanic Black males. EPA applied sex-specific
model coefficients for non-Hispanic Blacks to estimate CVD risk in Hispanic and non-Hispanic
other race population subgroups based on validation of the ASCVD model against published
statistics as described in Section G.4.
Table G-2: ASCVD Model Coefficients
Model Coefficient
Variable Name
Non-Hispanic
Non-Hispanic Black
Non-Hispanic
Non-Hispanic Black
White Females
Females*
White Males
Males*
Ln Age (y)
-29.799
17.114
12.344
2.469
Ln Age, squared
4.884
-
-
-
Ln Total Cholesterol (mg/dL)
13.54
0.94
11.853
0.302
Ln Age x Ln Total Cholesterol
-3.114
-
-2.664
-
LnHDL-C (mg/dL)
-13.578
-18.92
-7.99
-0.307
Ln Age x LnHDL-C
3.149
4.475
1.769
-
Ln Treated Systolic BP (mm Hg)
2.019
29.291
1.797
1.916
Ln Age x Ln Treated Systolic BP
-
-6.432
-
-
Ln Untreated Systolic BP (mm
Hg)
1.957
27.82
1.764
1.809
Ln Age x Ln Untreated Systolic
BP
-
-6.087
-
-
Current Smoker (l=Yes, 0=No)
7.574
0.691
7.837
0.549
Ln Age x Current Smoker
-1.665
-
-1.795
-
Diabetes (l=Yes, 0=No)
0.661
0.874
0.658
0.645
Mean (Coefficient x Value),
^s,r fis,r
-29.18
86.61
61.18
19.54
ASCVD Baseline Survival, Ss r
0.9665
0.9533
0.9144
0.8954
Abbreviations: ASCVD - atherosclerotic cardiovascular disease; BP - blood pressure; HDLC - high-density lipoprotein
cholesterol.
Note:
* Based on the results of ASCVD model validation exercises (Section G.4), the models for non-Hispanic Black males and
females are applied to other ethnic groups.
Source: Goffetal. (2014), Table A
In order to be used for risk estimation, the ASCVD model needs to be parameterized using
values of the predictors shown in Table G-2 that are appropriate for the current age, sex, and
race/ethnicity of the cohort being evaluated. As shown in Table G-l, current age, sex, and
race/ethnicity are easily accessible indices of the CVD model. In turn, baseline values for the
other ASCVD model predictors come from several public health surveys implemented by the
Centers for Disease Control and Prevention, as detailed in Section G.5.
Proposed PFAS Rule Economic Analysis
G-9
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
To compute the 10-year probability of the first hard CVD event for a birth year b, sex s and
race/ethnicity r cohort at age a, EPA uses the ASCVD risk equation (Goff et al., 2014, Table
G-5) adjusted to express the type of scenario being evaluated (i.e., baseline or regulatory
alternative):
Equation G-l:
D ( at A 1 c exP (ln(Tas,r+AT^a,s,t) [PT,s,r^"PaT,s,r'^(.cO\^"x-T,a,s,r P-T,s,r~%s,r Ps,r)
Kb,a,s,r,t-.t+9\i^l-b,a,s,t) ~ 1 Js,r
where
Rb,a,s,r,t:t+9(^Tb,a,s,t) probability of the first hard CVD event to occur between years t
and t + 9 for a birth year b, sex s / race/ethnicity r person whose age at
time t is a. Rb,a,s,r,t-.t+9(.Q) represents baseline 10-year first hard CVD
event risk, whereas Rb,a,s,r,t-.t+9(,&Tb,a,s,t) expresses regulatory alternative
risk consistent with a birth year b-, age a-, sex s-, calendar year t-specific
change in the baseline TC/HDLC/BP levels Arb a s
ASCVD baseline CVD event-free survival rate at 10 years, consistent with
the sex s and race/ethnicity r of the cohort being evaluated (see parameter
estimates in Table G-2);
a vector of baseline inputs for TC, HDLC, and BP consistent with the
current age a, sex s, and race/ethnicity r of the cohort being evaluated
(see Section G.5);
a vector of ASCVD model coefficients for the log-TC, log-HDLC, log-BP
predictors, consistent with the sex s and race/ethnicity r of the cohort
being evaluated (see parameter estimates in Table G-2);
a vector of ASCVD model coefficient for the interaction between
log-current age and log-TC, log-HDLC, log-BP predictor, consistent with
the sex s and race/ethnicity r of the cohort being evaluated (see parameter
estimates in Table G-2);
X-r,a,s,r'P-T,s,r inner product of the ASCVD model coefficient vector (excluding
TC, HDLC, and BP-related coefficients) and a vector of baseline input
values (excluding TC, HDLC, and BP-related inputs), consistent with the
current age a, sex s, and race/ethnicity r of the cohort being evaluated
(see parameter estimates in Table G-2 and Section G.5); and
%s,r' Ps,r inner product of the ASCVD model coefficient vector and a vector
of average input values in the ASCVD estimation dataset (see parameter
estimates in Table G-2).
Pr.s.r
Proposed PFAS Rule Economic Analysis
G-10
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
To obtain the annual probability of the first hard CVD event, EPA adjusts Rb,a,s,r,t-.t+9{^xb,a,s,t)
as follows:
Equation G-2:
1
^¦b,a,s,r,t(^^b,a,s,t} 1 — Rb,a,s,r,t.t+9^T-bia,s,t)^
where
ib,a,s,r,t{kxb,a,s,t) probability of the first hard CVD event to occur in year t for a
birth year b, sex s / race/ethnicity r person whose age at time t is a; and
Rb,a,s,r,t:t+9(^Tb,a,s,t) probability of the first hard CVD event to occur between years t
and t + 9 for a birth year b, sex s / race/ethnicity r person whose age at
time t is a.
G. 2.2 Prevalence of Post Hard CVD Events
Because the population evaluated for the first hard CVD event estimation excludes those with a
history of hard CVD events, model inputs require information on the baseline prevalence of the
past hard CVD event history in the U.S. population. EPA used the Medical Expenditure Panel
Survey (MEPS) 2010-2017 data to estimate the prevalence of persons with a prior experience of
hard CVD events, including MI, stroke, and other acute CHD events. MEPS is a nationally
representative survey of the U.S. civilian non-institutionalized population implemented by the
Agency for Healthcare Research and Quality (AHRQ). The survey has an overlapping panel
design, tracking individuals for, at most, two years and interviewing participants, at most, six
times. MEPS collects demographic, socioeconomic, and health status information on the first
interview and in each subsequent interview asks about medical events experienced between the
current and the previous interview (generally 4-5 months), as well as changes in employment
status, health insurance coverage, and so forth. Section G.5 provides additional information on
MEPS public use files that have been used in this analysis.
The prevalence of persons with a prior experience of hard CVD events has been estimated by
dividing the number person-years in MEPS interview rounds with a reported history of MI,
stroke, or other CHD by the total number of person-years in subpopulations defined by sex and
round-specific age. The estimated ratios have been adjusted for MEPS complex survey design.
Table G-3 shows the resulting estimates of sex-, race/ethnicity-, and age category-specific
prevalence of persons with prior experience of hard CVD events, along with 95% confidence
intervals that reflect sampling uncertainty. Compared with the prevalence estimates for females,
the estimated prevalence is higher for males in all age categories and for all CVD event
categories. Among adults aged 65 or older, estimated MI, other CHD, and overall prevalence is
highest for non-Hispanic White males, while stroke prevalence is highest among non-Hispanic
Black males. Regardless of the age category, the estimated prevalence of an MI history is higher
for males, while the prevalence of a stoke history is higher for females. The prevalence of other
CHD event history is approximately three to 10 times higher compared with the prevalence of an
MI or stroke history.
Proposed PFAS Rule Economic Analysis
G-ll
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-3: Estimated Past Hard CVD Event Prevalence per 100,000
Sex
Age
(years)
Race/
Ethnicity
MI
Stroke
Other CHD
Overall
632
495
5,709
6,292
Males
18-44
NH White
(410-855)
(317-673)
(5,072-6,346)
(5,620-6,965)
5,099
3,314
15,439
17,963
45-64
NH White
(4,569-5,629)
(2,804-3,823)
(14,523-16,355)
(16,930-18,995)
16,477
11,002
41,600
47,465
65 or older
NH White
(15,088-17,865)
(9,956-12,047)
(40,040-43,161)
(45,831^19,099)
436
614
3,886
4,667
Males
18-44
NH Black
(146-726)
(304-924)
(2,998-4,773)
(3,651-5,684)
4,786
5,316
12,261
16,590
45-64
NH Black
(3,928-5,644)
(4,222-6,409)
(10,801-13,720)
(14,898-18,282)
13,768
18,908
30,307
42,090
65 or older
NH Black
(11,218-16,319)
(16,185-21,631)
(26,724-33,891)
(38,368—45,812)
480
180
3,065
3,417
Males
18-44
Hispanic
(293-667)
(75-285)
(2,479-3,651)
(2,816-4,019)
4,299
3,010
9,979
12,584
45-64
Hispanic
(3,383-5,214)
(2,225-3,796)
(8,640-11,318)
(11,045-14,124)
14,071
8,254
25,866
30,548
65 or older
Hispanic
(11,569-16,573)
(6,031-10,477)
(22,420-29,313)
(26,960-34,136)
347
342
3,262
3,669
Males
18-44
NH Other
(122-572)
(75-610)
(2,330-4,194)
(2,695-4,643)
4,338
2,693
11,339
13,638
45-64
NH Other
(3,012-5,665)
(1,791-3,595)
(9,033-13,645)
(11,118-16,158)
12,256
12,354
30,516
36,932
65 or older
Other
(9,167-15,344)
(8,911-15,798)
(25,051-35,982)
(31,240^2,624)
439
830
6,262
6,954
Females
18-44
NH White
(278-600)
(608-1,052)
(5,528-6,997)
(6,223-7,685)
2,199
3,127
15,496
17,925
45-64
NH White
(1,841-2,557)
(2,595-3,659)
(14,522-16,469)
(16,791-19,059)
7,510
10,055
31,861
37,538
65 or older
NH White
(6,686-8,335)
(9,098-11,011)
(30,278-33,445)
(35,913-39,162)
393
1,092
4,628
5,612
Females
18-44
NH Black
(204-582)
(783-1,402)
(3,917-5,338)
(4,847-6,378)
3,484
6,491
15,292
19,596
45-64
NH Black
(2,808-4,160)
(5,640-7,343)
(13,915-16,670)
(17,981-21,210)
8,803
14,188
29,296
38,073
65 or older
NH Black
(7,130-10,476)
(12,304-16,071)
(26,441-32,151)
(35,102—41,045)
313
717
3,690
4,363
Females
18-44
Hispanic
(171-454)
(469-965)
(3,182-4,199)
(3,808-4,918)
2,597
3,627
10,335
12,777
45-64
Hispanic
(1,947-3,248)
(2,864-4,391)
(9,066-11,604)
(11,361-14,193)
7,513
9,469
23,149
29,186
65 or older
Hispanic
(5,953-9,073)
(7,385-11,554)
(20,350-25,948)
(26,206-32,167)
722
383
4,569
4,884
Females
18-44
NH Other
(123-1,320)
(90-675)
(3,181-5,957)
(3,502-6,266)
1,292
2,770
11,098
13,148
45-64
NH Other
(710-1,874)
(1,679-3,860)
(8,978-13,218)
(10,758-15,538)
4,150
7,321
19,001
23,463
65 or older
NH Other
(2,557-5,742)
(5,054-9,589)
(15,308-22,694)
(19,638-27,288)
Proposed PFAS Rule Economic Analysis
G-12
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-3: Estimated Past Hard CVD Event Prevalence per 100,000
Sex ,Age, I?^aCC^ MI Stroke Other CHD Overall
(years) Ethnicity
Abbreviations: MI - myocardial infarction (ICD9=410 or MIDX=1); NH - non-Hispanic; Other CHD - other coronary heart
disease (ICD9=413,414,427,428 or CHDDX=1, ANGIDX=1, OHRTDX=l); Stroke (ICD9=433,434,435,436 or STRKDX=1);
95% confidence interval shown in parentheses below the point estimate.
Source: EPA analysis based on MEPS, 2010-2017
G.23 Distribution of Fotol and Non-Fatal First Hard CVD Events
The ASCVD model predicts the risk of a composite hard CVD event (i.e., MI, IS, or CHD
death). However, modeling requires separate tracking of morbidity and mortality for life table
calculation purposes. In addition, acute-phase mortality and morbidity valuation depends on the
endpoint (i.e., MI or IS). Therefore, EPA used MEPS 2010-2017 data to estimate the distribution
of first hard CVD events by type of condition (i.e., MI, stroke, and other CHD). EPA estimated
the incidence of first hard CVD events by dividing the number of person-years in MEPS
interview rounds with reported new occurrences of MI, stroke, or other CHD by the number of
person-years in MEPS interview rounds without resorted prior experience of CVD events, in
subpopulations defined by race/ethnicity, sex and round-specific age. EPA adjusted the estimated
ratios for MEPS complex survey design. Distribution of CVD events by condition type was
calculated based on the estimated condition-specific incidence rates.
Table G-4 shows the resulting estimates of sex-, race/ethnicity-, and age category-specific first
hard CVD event incidence, along with 95% confidence intervals that reflect sampling
uncertainty. The table also shows the distribution of first hard CVD events by event type. In
males, 15% to 17% of first hard CVD events are Mis, whereas 13% to 20% of first hard CVD
events are strokes. In females, 8% to 12% of first hard CVD events are Mis, whereas 17% to
28% of first hard CVD events are strokes. The shares of Mis and strokes increase with age for
both sexes. Among adults aged 65 or older, estimated MI, stroke, other CHD, and overall
incidence are highest for non-Hispanic White males and females.
Table G-4: Estimated First Hard CVD Event Incidence and Distribution by CVD
Event Type
Sex
Age (years)
Race/ Ethnicity
MI
Stroke
Other CHD
Overall
82
57
454
540
Males
18-44
NH White
(29-135)
(3-110)
(299-609)
(375-705)
356
333
1,536
2,048
45-64
NH White
(225-486)
(194-471)
(1,213-1,859)
(1,678-2,417)
1,326
2,001
6,233
8,125
65 or older
NH White
(679-1,973)
(1,248-2,754)
(5,035-7,431)
(6,651-9,598)
23
81
363
447
Males
18-44
NH Black
(-3-49)
(4-159)
(156-570)
(227-668)
235
805
1,039
1,862
45-64
NH Black
(64-407)
(399-1,211)
(676-1,401)
(1,339-2,385)
319
765
2,332
3,273
65 or older
NH Black
(-1-639)
(76-1,454)
(1,217-3,447)
(1,926-4,621)
Males
18-44
Hispanic
52
40
135
212
Proposed PFAS Rule Economic Analysis
G-13
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-4: Estimated First Hard CVD Event Incidence and Distribution by CVD
Event Type
Sex
Age (years)
Race/ Ethnicity
MI
Stroke
Other CHD
Overall
(6-99)
(-4-83)
(55-214)
(111-313)
276
421
735
1,142
45-64
Hispanic
(72-479)
(2-839)
(419-1,052)
(625-1,659)
951
816
2,747
3,915
65 or older
Hispanic
(285-1,618)
(349-1,283)
(1,432-4,061)
(2,440-5,390)
72
85
121
278
Males
18-44
NH Other
(-70-215)
(-54-223)
(35-207)
(63-493)
830
548
1,513
2,537
45-64
NH Other
(171-1,489)
(39-1,057)
(643-2,383)
(1,356-3,718)
665
1,232
2,940
4,251
65 or older
NH Other
(-14-1,343)
(431-2,033)
(1,496-4,383)
(2,506-5,997)
56
135
492
646
Females
18-44
NH White
(-21-134)
(54-216)
(317-668)
(437-856)
140
407
1,423
1,865
45-64
NH White
(56-225)
(193-620)
(1,109-1,737)
(1,490-2,240)
831
2,102
4,271
6,294
65 or older
NH White
(533-1,130)
(1,498-2,705)
(3,461-5,081)
(5,358-7,231)
96
57
487
597
Females
18-44
NH Black
(1-191)
(5-108)
(279-695)
(360-834)
196
530
1,168
1,754
45-64
NH Black
(74-318)
(247-812)
(793-1,543)
(1,285-2,223)
382
1,607
3,383
4,546
65 or older
NH Black
(8-756)
(762-2,453)
(2,221-4,545)
(3,179-5,913)
38
78
308
392
Females
18-44
Hispanic
(-24-100)
(25-131)
(130-487)
(190-595)
145
308
664
1,065
45-64
Hispanic
(33-257)
(76-541)
(393-936)
(699-1,432)
992
1,321
2,610
4,456
65 or older
Hispanic
(215-1,768)
(611-2,031)
(1,670-3,550)
(3,348-5,564)
47
315
315
Females
18-44
NH Other
(-46-141)
Omitted
(42-589)
(42-589)
201
399
759
1,297
45-64
NH Other
(-6-409)
(74-724)
(259-1,259)
(627-1,967)
576
1,328
2,689
4,349
65 or older
NH Other
(-43-1,195)
(381-2,276)
(1,234-4,144)
(2,463-6,234)
Abbreviations: MI - myocardial infarction (ICD9=410 or MIDX=1); NH - non-Hispanic, Stroke (ICD9=433,434,435,436
or STRKDX=1); Other CHD - other coronary heart disease (ICD9=413,414,427,428 or CHDDX=1, ANGIDX=1,
OHRTDX=l); 95% confidence interval shown in parentheses below the point estimate.
Proposed PFAS Rule Economic Analysis
G-14
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The ASCVD model predicts the risk of first MI (fatal and non-fatal), IS (fatal and non-fatal), or
other fatal CHD within the next 10 years. Notably, other non-fatal CHD events are not included
among the CVD event types predicted by the ASCVD model (Goff et al., 2014). Because MEPS
data do not have sufficient information to estimate acute-phase CVD event mortality, EPA used
AHRQ's Healthcare Cost and Utilization Project (HCUP) data on hospital mortality to allocate
CVD events into fatal and non-fatal categories. Section G.5 provides additional information on
the in-hospital mortality data.
Table G-5 shows sex- and age category-specific probability of in-hospital CVD event death
based on HCUP 2017 inpatient data (Agency for Healthcare Research and Quality, 2017a).
Probability of an in-hospital death is highest for MI events (4.64%), followed by IS events
(4.01%), and then other CHD events (1.07%). This probability grows with age across all CVD
event types and is higher for females when compared with males.
Table G-5: Probability of Hospital Death for a Hard CVD Event
Category
MI (%)
IS (%)
Other CHD (%)
Overall
4.65
4.01
1.07
Age (years)
18-44
1.43
1.91
0
45-64
2.60
2.46
0.67
65-84
5.42
3.88
1.23
85 or older
9.80
7.29
3.14
Sex
Males
4.41
3.71
1.01
Females
5.04
4.30
1.20
Abbreviations: IS - ischemic stroke (ICD10=I63); MI - myocardial infarction (ICD10=I21); Other CHD - other coronary heart
disease (ICD10=I20,122-125).
Source: HCUP 2017 (Agency for Healthcare Research and Quality, 2017a)
EPA combined estimates in Table G-4 and Table G-5 to derive the ASCVD event distribution
over the following event types: non-fatal MI, non-fatal IS, and fatal CVD events (i.e., fatal MI,
fatal IS, and other fatal CHD events). Table G-6 shows the final sex-, race/ethnicity-, and age
category-specific estimates of the ASCVD event distribution needed as the CVD model input.
For males, the share of non-fatal MI events is 22% to 58%, the share of non-fatal IS events is
39% to 77%), and the share of fatal CVD events is 2% to 13%. For females, the share of non-fatal
MI events is 16% to 62%, the share of non-fatal IS events is 36% to 76%, and the share of fatal
CVD events is 2% to 14%. The shares of non-fatal MI decrease with age, whereas the share of
fatal CVD events increase with age. Shares of non-fatal MI are generally highest among non-
Hispanic White males, while shares of non-fatal IS are highest among non-Hispanic Black
males. Among non-Hispanic White females, shares of non-fatal IS are highest for those aged 45-
64 years. Among non-Hispanic Black females, shares of non-fatal IS are highest for those aged
65-84 years. Among females aged 65 or older, shares of non-fatal MI are highest in the Hispanic
population.
Proposed PFAS Rule Economic Analysis
G-15
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-6: Estimated Distribution of Fatal and Non-Fatal First Hard CVD Events
Sex
Age (years)
Race/Ethnicity
Non-Fatal MI
(%)
Non-Fatal IS
(%)
Fatal CVD
Event (%)
Males
18-44
NH White
58
40
1.5
45-64
NH White
50
47
3.7
65-84
NH White
37
57
6.2
85 or older
NH White
34
53
13
Males
18-44
NH Black
22
77
1.7
45-64
NH Black
22
75
2.9
65-84
NH Black
27
66
6.4
85 or older
NH Black
25
62
13
Males
18-44
Hispanic
56
42
1.5
45-64
Hispanic
38
59
3
65-84
Hispanic
50
44
6.1
85 or older
Hispanic
47
41
12
Males
18-44
NH Other
46
53
1.6
45-64
NH Other
58
39
3.1
65-84
NH Other
33
62
5.8
85 or older
NH Other
30
58
12
Females
18-44
NH White
29
69
1.9
45-64
NH White
24
71
4.6
65-84
NH White
26
67
6.5
85 or older
NH White
24
63
13
Females
18-44
NH Black
62
36
1.7
45-64
NH Black
26
70
3.9
65-84
NH Black
18
76
6.7
85 or older
NH Black
16
70
14
Females
18-44
Hispanic
32
66
1.9
45-64
Hispanic
31
65
3.8
65-84
Hispanic
40
54
6.4
85 or older
Hispanic
37
51
12
Females
18-44
NH Other
45
53
1.8
45-64
NH Other
32
64
3.6
65-84
NH Other
28
66
6.5
85 or older
NH Other
26
61
13
Abbreviations: Fatal CVD - includes fatal MI, fatal IS, and fatal other coronary heart disease events; IS - ischemic stroke;
MI - myocardial infarction; NH - non-Hispanic.
G.2.4 Post-Acute CVD Mortality
Persons who have experienced non-fatal MI and non-fatal IS events have elevated post-acute
CVD mortality and morbidity (Roger et al., 2012). EPA identified four studies that examined
risk factors for secondary hard CVD events. These studies differ in terms of outcomes tracked
(e.g., recurrent MI, recurrent IS, angina, heart failure, CVD, and all-cause death), conditioning
event definition (e.g., MI, IS, CHD), and the length of follow-up for which statistics are reported
(e.g., 1-year follow-up, 5-year follow-up). The data used to estimate the risks of secondary CVD
events differ with respect to average age, sex, and share of individuals who are White among the
participants:
Proposed PFAS Rule Economic Analysis
G-16
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• Data used in Kannel et al. (1999) and D'Agostino et al. (2000) come from the
Framingham Heart Survey (Mahmood et al., 2014) and represent White males
and females approximately age 60.
• Data used in Thom et al. (2001) are from the pooled Atherosclerosis Risk in
Communities Study (Williams, 1989), Cardiovascular Health Study (Fried et al., 1991),
and Framingham Original and Offspring Cohort Study (Mahmood et al., 2014).
This pooled dataset offers representation for Black males and females, in addition to
White males and females, and captures persons aged 45 or older.
• Beatty et al. (2015) used two predominantly White male datasets developed based on the
Heart and Soul Study (Whooley et al., 2008) and the PEACE trial (PEACE Trial
Investigators, 2004), capturing persons aged 67 years and 64 years, on average,
respectively.
• S. Li et al. (2019) used data for 2008 and 2012 and two types of conditioning events
(i.e., MI and IS) to assess the risk of secondary events in four large Medicare cohorts:
survivors of the first MI in 2008, survivors of the first IS in 2008, survivors of the first
MI in 2012, and survivors of the first IS in 2012..41 These data represent older
populations (age 80, on average) and are not limited to a particular race/ethnicity or sex.
Of the studies that assessed risk factors for secondary hard CVD events, only three focused on
developing a risk prediction model (Beatty et al., 2015; D'Agostino et al., 2000; Kannel et al.,
1999) and only two have changes in cholesterol levels and systolic blood pressure as a primary
predictors (Beatty et al., 2015; D'Agostino et al., 2000). In these two studies, TC, HDLC, and BP
levels do not appear to significantly increase the risk of recurrent CVD events, although
D'Agostino et al. (2000) identified statistically significant relationships between the ratio of TC
to HDLC and probability of recurrent CVD events. Beatty et al. (2015) concluded that
precautionary measures and medication taken by patients who had suffered from a primary CVD
event may decrease the initial risk factors (i.e., TC, HDLC, BP) and may be a reason for the lack
of correlation between secondary CVD events and the modeled biomarkers.
In sum, studies focusing on secondary CVD events point to an elevated risk of these events
among survivors of the first hard CVD event. However, the link between these risks and TC,
HDLC, and BP levels is less clear, with limited supporting evidence coming from decades-old
data evaluated by D'Agostino et al. (2000). Therefore, the CVD model relies on the same
secondary hard CVD event rates to estimate secondary hard CVD event incidence under baseline
and regulatory alternatives. Specifically, EPA focuses on post-acute CVD mortality as the
secondary event of interest, because other non-fatal secondary CVD events are captured in the
available unit values for first non-fatal MI and IS (see, e.g., O'Sullivan et al., 2011). EPA
selected estimates in Thom et al. (2001) to model post-acute CVD mortality for survivors of MI
or IS at ages 40-65, because Thom et al. (2001) is the only study that analyzed this age group.
EPA selected estimates in S. Li et al. (2019) to model post-acute CVD mortality for survivors of
MI or IS at ages 66-89, because cohorts analyzed in S. Li et al. (2019) are the largest and most
representative of the U.S. population compared with the cohorts analyzed by other studies.
41 Note that relative to other studies with sample sizes of, at most, 10,000, the sizes of these cohorts are 20,000,
on average.
Proposed PFAS Rule Economic Analysis
G-17
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
G.2.5 Survivors of the first hard CVD event at ages 40-65
EPA used estimates of all-cause post-acute mortality for MI survivors at the 1- and 5-year
follow-ups from Thom et al. (2001) to model post-acute CVD mortality for survivors of
non-fatal MI and non-fatal IS events at ages 45-65. While EPA was unable to identify
comparable post-acute mortality statistics for non-fatal IS, an analysis of the Medicare
population by S. Li et al. (2019) suggests that post-acute MI mortality is a reasonable
approximation for post-acute IS mortality.42
Table G-7 shows estimated all-cause probability of death following first non-fatal MI by age
category, race/ethnicity, and sex from Thom et al. (2001), as reported in Roger et al. (2012).
These estimates are based on the analysis of pooled data from the Atherosclerosis Risk in
Communities Study (Williams, 1989), the Cardiovascular Health Study (Fried et al., 1991), and
the Framingham Original and Offspring Cohort Study (Mahmood et al., 2014). The estimates are
available only for non-Hispanic Whites and non-Hispanic Blacks.
Table G-7: Post-Acute All-Cause Mortality After the First Myocardial Infarction
Age Group
(years)
Race/Ethnicity
Follow-Up Period
(years)
Probability of All-Cause Death (%)
Males Females
45-64
Non-Hispanic White
1
5
9
45-64
Non-Hispanic Black
1
14
8
65 or older
Non-Hispanic White
1
25
30
65 or older
Non-Hispanic Black
1
25
30
45-64
Non-Hispanic White
5
11
18
45-64
Non-Hispanic Black
5
22
28
65 or older
Non-Hispanic White
5
46
53
65 or older
Non-Hispanic Black
5
54
58
Abbreviations: MI - myocardial infarction (ICD9=410; ICD10=I21).
Source: Thom etal. (2001)
Table G-8 shows estimated probabilities of post-acute CVD mortality after the first MI. EPA
derived these probabilities by adjusting all-cause post-acute mortality probabilities reported in
Table G-7 for the ages 45-64 group43 to exclude the probability of death from non-CVD causes.
Section G.5 provides details on an estimation of integer age-, race/ethnicity- and sex-specific
probability of death from non-CVD causes based on the U.S. Life Tables, 2017 (Arias et al.,
2019) and CVD death rates, 1999-2019 (Centers for Disease Control and Prevention, 2020c).
The last two columns of Table G-8 show annual race/ethnicity- and sex-specific post-acute CVD
death probabilities used by the CVD model in estimation of secondary mortality in years 1-5
following the first non-fatal MI or IS that occurred at ages 45-65. EPA used post-acute mortality
data for non-Hispanic Whites to estimate mortality effects for the other race/ethnicity groups.
42 For those aged 65 or older, S. Li et al. (2019) have estimated the probability of death within 1 year after a
non-fatal IS to be 32.07% and the probability of death within 1 year after a non-fatal MI to be 32.09%.
43 EPA applies post-acute mortality probabilities estimated for ages 45-64 to the survivors of first MI or IS,
ages 45-65, because the magnitude of the annual death probability at age 65 is closer to the average annual
death probability for ages 45-64 than to the average annual death probability for ages 66-99.
Proposed PFAS Rule Economic Analysis
G-18
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-8: Post-Acute Mortality After the First Myocardial Infarction
All-Cause Death Probability Non-CVD Death Probability
Integer Year (%) (%)b CVD Death Probability (%)c
MP
Males
Females Males
Females
Males
Females
All Races/Ethnicitiesd
0
5.6
8.8 0.56
0.38
5.0
8.4
1
1.5
2.7 0.60
0.41
0.93
2.3
2
1.5
2.7 0.65
0.44
0.88
2.3
3
1.5
2.7 0.70
0.48
0.83
2.3
4
1.5
2.7 0.75
0.51
0.78
2.2
Non-Hispanic White®
0
5.0
9.0
-
4.5
8.6
1
1.5
2.3
-
0.91
1.9
2
1.5
2.3
-
0.86
1.9
3
1.5
2.3
-
0.82
1.9
4
1.5
2.3
-
0.76
1.8
Non-Hispanic Black
0
14
8.0
-
12
7.7
1
2.0
5.0
-
1.2
4.3
2
2.0
5.0
-
1.1
4.2
3
2.0
5.0
-
1.1
4.1
4
2.0
5.0
-
1.0
4.1
Abbreviations: CVD - cardiovascular disease; MEPS - Medical Expenditure Panel Survey; MI - myocardial infarction
(ICD9=410; ICD 10=121).
Notes:
aPost-acute death probabilities at 1- and 5-year follow-ups in Table G-9 are converted to the integer year-specific post-acute
death probabilities by assuming that the annual death probabilities in years 1^1 are identical. This assumption is supported
by data in S. Ti et al. (2019), who report post-acute death probabilities at 1-, 2-, 3-, 4-, 5-, and 6-year follow-ups.
bReported annual probability of non-CVD death is a weighted average of life table age-specific probabilities for ages 45-64.
The weights are the sex-specific age distribution of the first MI survivor population, estimated using MEPS 2010-2017 data.
Tor all race/ethnicity categories, CVD death probability is the difference between all-cause death probability and non-CVD
death probability. For the non-Hispanic White and non-Hispanic Black race/ethnicity categories, EPA obtained the estimates
by multiplying the corresponding all-cause post-acute death probability with the all-race/ethnicity ratio of post-acute CVD
death probability to all-cause post-acute death probability.
dRace/Ethnicity-specific data for the ages 45-64 group in Table G-9 are pooled using a sex-specific race/ethnicity distribution
of the first MI survivor population, estimated using MEPS 2010-2017 data.
ePost-acute CVD death probability for non-Hispanic Whites is used to estimate mortality effects for the other race/ethnicity
groups.
Sources: Thom etal. (2001); U.S. Life Tables, 2017 (Arias etal., 2019); CVD death rates, 1999-2019 (Centers for Disease
Control and Prevention, 2020c)
Proposed PFAS Rule Economic Analysis
G-19
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
G.2.6Survivors of the first hard CVD event at ages 66+
EPA used the results in S. Li et al. (2019) to estimate the number of post-acute CVD deaths for
survivors of the first MI and IS events, aged 66 years or older at the time of the initial event.
Table G-9 summarizes the key results in S. Li et al. (2019) that are used to parameterize the
CVD model and the results of adjustments that EPA made to incorporate CVD mortality
information in the model. First, EPA estimated CVD death probabilities by subtracting non-CVD
death probabilities from all-cause post-acute mortality probabilities reported in S. Li et al.
(2019). EPA derived the sex- and age-specific non-CVD mortality rates from U.S. Life Tables,
2017 (Arias et al., 2019); CVD death rates, 1999-2019 (Centers for Disease Control and
Prevention, 2020c); and U.S. Life Tables Eliminating Certain Causes of Death, 1999-2000
(Arias et al., 2013). EPA has averaged age- and sex-specific non-CVD death probabilities for
those age 66 or older using the demographic characteristics of the MI and IS cohorts analyzed by
S. Li et al. (2019). Second, EPA calculated CVD mortality probability as the difference between
the all-cause death probability and the non-CVD death probability. Third, EPA calculated CVD
mortality rate multipliers as a ratio of CVD mortality probability to the non-CVD death
probability. EPA combined these multipliers (reported in Table G-9 for MI and IS survivors)
with age-, sex-, and race/ethnicity-specific non-CVD death rates to obtain post-acute CVD
mortality rates for each cohort included in the analysis.
Proposed PFAS Rule Economic Analysis
G-20
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table G-9: Post-Acute CVD Mortality Following the First Myocardial Infarction and First
Ischemic Stroke in the Population Aged 66 Years or Older
MI Survivors IS Survivors
Follow-
up
Period
(years)
All-Cause Death
Probability (%)a
Non-CVD Death
Probability (%)b
CVD Death
Probability (%)c
CVD Mortality Rai
Multiplier11
All-Cause Death
Probability (%)a
Non-CVD Death
Probability (%)b
CVD Death
Probability (%)c
CVD Mortality Rai
Multiplier11
0
32
4.3
27
6.4
32
4.5
28
6.1
1
16
4.6
11
2.5
15
4.8
9.9
2.07
2
15
4.9
9.6
1.9
16
5.2
10
2.1
3
14
5.2
9.04
1.7
15
5.5
9.8
1.8
4
14
5.6
8.6
1.5
15
5.9
8.9
1.5
5
14
5.9
8.04
1.4
14
6.2
8.03
1.3
Abbreviations: CVD - cardiovascular disease; IS - ischemic stroke (ICD9=433,434; ICD10=I63); MI - myocardial infarction
(ICD9=410; ICD10=I21).
Notes:
Tor MI, the follow-up year specific all-cause death probability is from S. Li et al. (2019) reported data for the 2008 MI survivor
cohort (N=26,46). For IS, the follow-up year specific all-cause death probability is from S. Li et al. (2019) reported data for the
2008 IS survivor cohort (N= 17,566).
bNon-CVD annual mortality rate is based on U.S. Life Tables 2017 (Arias et al., 2019); CVD death rates, 1999-2019 (Centers for
Disease Control and Prevention, 2020c); and U.S. Life Tables Eliminating Certain Causes of Death, 1999-2000 (Arias et al.,
2013) for those age 66 or older. The annual age- and sex-specific death probabilities were averaged using S. Li et al. (2019) MI/IS
survivor cohort demographic characteristics.
cPost-acute CVD death probability rate is estimated by subtracting the non-CVD annual death probability from the all-cause post-
acute death probability.
dThe CVD mortality rate multiplier is defined as the difference between all-cause death probability and non-CVD death
probability divided by the non-CVD death probability. The CVD model combines the baseline rate multiplier with race/ethnicity-,
age-, and sex-specific non-CVD baseline death rates to obtain mortality rates that are appropriate for the race/ethnicity, age, and
sex of each cohort included in the analysis.
Sources: Li etal. (2019); U.S. Life Tables, 2017 (Arias etal., 2019); CVD death rates, 1999-2019 (Centers for Disease Control
and Prevention, 2020c); U.S. Life Tables Eliminating Certain Causes of Death, 1999-2000 (Ariasetal., 2013).
G.3 Detailed CVD Model Calculations
Table G-10 provides a guide to sections containing the recurrent CVD model calculations
applicable under conditions defined by initial cohort age, current cohort age, and estimation type.
Estimation types include baseline estimation, regulatory alternative estimation, and risk
reduction estimation. Note that standard life table calculations for current cohort ages 0-39 in
Section G.3.1 apply to both the baseline and regulatory alternative estimation types. The CVD
risk reduction estimation equations in Section G.3.5 apply to ages 40+, for which the model
explicitly estimates the number of first hard CVD events and the number of post-acute CVD
deaths for survivors of the first hard CVD event.
Proposed PFAS Rule Economic Analysis
G-21
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table G-10: A Mapping of CVD Model Calculations by Initial Cohort Age, Current
Cohort Age, and Estimation Type
Initial
Current Cohort Age (years)
Cohort
Age
0-39
40-65
66+
(years)
Baseline Estimation
0-39
Section G.3.1
Section G.3.2, Section G.3.4
Section G.3.2, Section G.3.4
40-85+
-
Section G.3.2, Section G.3.4
Section G.3.2, Section G.3.4
Regulatory Alternative Estimation
0-39
Section G.3.1
Section G.3.3, Section G.3.4
Section G.3.3, Section G.3.4
40-85+
-
Section G.3.3, Section G.3.4
Section G.3.3, Section G.3.4
Risk Reduction Estimation
0-39
-
Section G.3.5
Section G.3.5
40-85+
-
Section G.3.5
Section G.3.5
Abbreviations: CVD - cardiovascular disease.
G.3.1 Baseline Recurrent Calculations Without Explicit
Treatment of the CVD Population
The number of deaths occurring in year t is estimated using the number of persons alive at the
start of the year, lb,a,s,r,t, ar|d all-cause annual probability of death, qasr:
Equation G-3:
d-b,a,s,r,t ~ 1a,s,r ' Ib,a,s,r,t
The number of persons surviving to the start of the next year is calculated as the difference
between the number of persons alive at the start of the year, lb,a,s,r,t, ar|d the number of deaths
estimated to occur during the year, dbia,s,r,t-
Equation G-4:
lb,a+l,s,r,t+l ~ lb,a,s,r,t ~ db,a,s,r,t
G.3.2 Baseline Recurrent Calculations with Explicit Treatment
of the CVD Population
The population of persons alive at the start of year t, lb,a,s,r,t, is split into CVD and non-CVD
subpopulations using externally estimated age-, race/ethnicity-, and sex-specific CVD
prevalence, na s r :
Equation G-5:
Proposed PFAS Rule Economic Analysis
G-22
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
lb,a,s,r,t,CVD — a,s,r ' Ib,a,s,r,t
Equation G-6:
lb,a,s,r,t,OTH (l ^a,s,r) ' ^b,a,s,r,t
The year t number of non-CVD deaths in the CVD and non-CVD subpopulations is estimated
by applying the annual age-, race/ethnicity-, and sex-specific probability of non-CVD death,
qa,s,r,oth> to the number of persons alive at the start of the year in each subpopulation
(lb,a,s,r,t,cvd and lb,a,s,r,t,othX respectively:
Equation G-7:
db,a,s,r,t,CVD,OTH — la,s,r,OTH ' ^b,a,s,r,t,CVD
Equation G-8:
d-b,a,s,r,t, OTH,OTH — la,s,r,OTH ' ^b,a,s,r,t,OTH
The year t number of CVD deaths in the CVD subpopulation is estimated by applying the annual
CVD death probability, qa,s,r,cvd> to the total population alive at the start of the year, lbA,s,r,t^ net
of deaths from other causes, qa,s,r,OTH^ estimated to occur during the year:
Equation G-9:
d-b,a,s,r,t,CVD,CVD — Ra,s,r,CVD ' (l — 9a,s,r,OTH) ' ^b,a,s,r,t
The number of persons surviving to the start of the next year is estimated as:
Equation G-10:
lb,a+l,s,r,t+l ~ lb,a,s,r,t ~ ^-b,a,s,r,t,CVD,CVD — ^b,a,s,r,t, OTH,OTH — ^-b,a,s,r,t, CVD, OTH
The uncalibrated number of persons experiencing their first hard CVD event in year t is
estimated by applying the baseline annual probability of first hard CVD event, ib,a,s,r,t(0). to the
start-of-the-year number of persons in the non-CVD subpopulation, lb,a,s,r,t,oth> net of non-CVD
deaths, db a s r t 0TH 0TH. The ASCVD model applies to ages 40-80 and predicts a 10-year
probability of the first hard CVD event. However, EPA uses the ASCVD model to estimate 10-
year probability of the first hard CVD event for adults ages 81+ years. For those in 85+ age
group, EPA uses age 85 as the input to ASCVD model at the start of the evaluation period.
Finally, EPA uses the externally estimated share of non-fatal first hard CVD events, Ya,s,rj-. ar|d
same-year post-acute CVD mortality probability, fJ.a,s,r,f,o, to compute the number of persons
surviving their first hard type / CVD event in year t:
Equation G-ll:
Proposed PFAS Rule Economic Analysis
G-23
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
^b,a,s,r,f,t,0 ~ Ma,s,r,/,o) ' Ya,s,r,f ' ib,a,s,r,t(Sty ' \}b,a,s,r,t,OTH a,s,r,t,OTH,OTHy
EPA uses the externally estimated share of fatal first hard CVD events, 1 — £/ef Ya.s.rj, and
same-year post-acute CVD mortality probability, fJ.a,s,r,f,o,t0 compute the uncalibrated number
of year t deaths in the incident CVD population at baseline:
Equation G-12:
^-b,a,s,r,t, 0 — [l + HfeF^a,s,r,f,0 l) " Ya,s,r,f\ ' ib,a,s,r,t(®) 1 (jb,a,s,r,t, OTH ^-b, a, s,r,t, OTH, OTh)
For calibration purposes, EPA calculated the incident CVD population size, xb a s r ^ that is
consistent with the reported CVD prevalence rates, 7ra sr and na+1iS,r, and cause-specific
mortality rates, qa,s,r,cvd and qa,s,r,oth:
Equation G-13:
Xb,a,s,r,t ~ TCa+l,s,r^b,a+l,s,r,t+l ~ ^-b,a,s,r,t,CVD ^b,a,s,r,t,CVD,CVD ^b,a,s,r,t,CVD,OTH
EPA used the incident CVD population size to estimate a calibration factor for scaling raw
ASCVD model-based results:
Equation G-14:
_ 3Cb,a,s,r,t
Xb,a,s,r,t — y ^ , ~
ZjfeFnb,a,s,r,f,t,0 mb,a,s,r,t, 0
Proposed PFAS Rule Economic Analysis
G-24
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Using the estimated calibration factor, EPA adjusted the raw number of persons surviving their
first hard type / CVD event in year t, nbia,s,r,f,t, o> ar|d the raw number of year t deaths in the
incident CVD population at baseline, inbaiSrti0, to ensure that EPA does not project a larger
number of incident events than is consistent with the CVD prevalence statistics and mortality
rates:
Equation G-15:
^¦b,a,s,r,f,t,0 ~ 77U77. (1, Xb,a,s,r,t) ' ^-b,a,s,r,f,t,0
Equation G-16:
^-b,a,s,r,t, 0 — 77U77. (1, Xb,a,s,r,t) ' b,a,s,r,t,0
Finally, EPA uses the overall number of year t CVD deaths, dbasrtcVDCVD, net of the number
of deaths in the incident CVD population, mb a s r 10, and the size of CVD population alive at the
start of the year, lb,a,s,r,t,cwd> to estimate the baseline CVD death rate in the prevalent CVD
population. This quantity is needed to support regulatory alternative estimation:
Equation G-17:
Pb,a,s,r ~ (db,a,s,r,t,CVD,CVD ~ 777b,a,s,r,t,o)/^b,a,s,r,t,CVD
G.33 Regulatory Alternative Recurrent Calculations with Explicit
Treatment of the CVD Population
If current cohort age a is equal to the initial cohort age, the sizes of CVD and non-CVD
subpopulations at the start of year 0 are calculated using externally estimated CVD prevalence,
7ra s r, and the initial population size, lbA,s,r,t- If, however, the current cohort age a is greater than
the initial cohort age, then the sizes of CVD and non-CVD subpopulations at the start of year t
are the same as the end-of-year t — 1 CVD and non-CVD subpopulation sizes. That is, the CVD
and non-CVD populations are computed in a recurrent manner.
Equation G-18 :
. (^a,s,r ' Ib,a,s,r,t if CL TflCLx(jX t, 40)
b,a,s,r,t,cvD - j a_x s r f_1 if a > max(a - t, 40)
Equation G-19:
, _ K1 ~ na,s,r) ¦ h,a,s,r,t if a = max(a - t, 40)
'¦b,a,s,r,t,OTH — 1 , ^ r ± a r\\
I <-b,a~i,s,r,t-i,0TH if a, > max(a t, 40)
Proposed PFAS Rule Economic Analysis
G-25
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The year t number of non-CVD deaths in CVD and non-CVD subpopulations is estimated by
applying the annual age-, race/ethnicity-, and sex-specific probability of non-CVD death,
qa,s,r,oth> to the number of persons alive at the start of the year in each subpopulation,
respectively:
Equation G-20:
d-b,a,s,r,t,CVD,OTH ~ 1a,s,r,OTH ' Ib,a,s,r,t,CVD
Equation G-21:
d-b,a,s,r,t,OTH,OTH ~ 1a,s,r,OTH ' Ib,a,s,r,t,OTH
The uncalibrated number of fatal and non-fatal first hard CVD events under the regulatory
alternative is estimated using the same equations (i.e., Eq. G-l 1 and Eq. G-12) as the ones used
for the baseline scenario, except for the non-zero difference between regulatory alternative and
baseline total cholesterol AT^a^t:
Equation G-22:
W-b,a,s,r,f,t,0 (l f^a,s,r,f,o) ' Ya,s,r,f ' ^-b,a,s,r,t(^^b,a,s,t) ' (jb,a,s,r,t,OTH ~ db,a,s,r,t,OTH,OTH)
Equation G-23:
1Tlb,a,s,r,t, 0 — [l + HfeF^a.s.r.f,0 l) " Ya,s,r}f \ ' ib,a,s,r,t(^-b,a,s,t) ' (j-b,a,s,r,t,OTH
db,a,s,r,t, OTH.OTh)
These estimates are used in combination with the baseline calibration factor, Xb,a,s,r,t, ar|d
EPA-estimated regulatory alternative incident CVD population size, xb a s r t :
Equation G-24:
Xb,a,s,r,t ~ Xb,a,s,r,t(^lfeF^-b,a,s,r,f,t,0 1Tlb,a,s,r,t,o)
Proposed PFAS Rule Economic Analysis G-26 March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Using the estimated baseline calibration factor, Xb,a,s,r,t, EPA adjusted the raw number of
persons surviving their first hard type / CVD event in year t, nbia,s,r,f,t,o> ar|d the raw number of
year t deaths in the incident CVD population, inbaiSrti0:
Equation G-25:
^¦b,a,s,r,f,t,0 ~ Tflin (1, Xb,a,s,r,t) ' ^-b,a,s,r,f,t,0
Equation G-26:
^-b,a,s,r,t, 0 — TYlifl (1, Xb,a,s,r,t) ' b,a,s,r,t,0
The number of CVD deaths at age a during year t is estimated as the sum of the number of
deaths among those whose CVD event history began before age a, pbia,s,r ' h,a,s,r,t, ar|d the
number of deaths among those who experienced their first CVD event at age a, mb a s r 1
The number of deaths among those whose CVD event history began before age a is the product
of the baseline CVD death rate in the CVD subpopulation, pbia,s,r-. and the size of the CVD
subpopulation at the start of year t, lbia,s,r,t-
Equation G-27:
d-b,a,s,r,t,CVD,CVD ~ Pb,a,s,r ' Ib,a,s,r,t ^b,a,s,r,t,0
Finally, the following recurrent equations are used to compute the sizes of total, CVD, and non-
CVD populations surviving through to the beginning of year t + 1:
Equation G-28:
lb,a+l,s,r,t+l = lb,a,s,r,t ~ ^fc.a.s.r.t.CVD.CVD — ^i),a,s,r,t,Or//,Ora — ^fc.aAr.t.CVD.OTT/
Equation G-29:
lb,a+l,s,r,t+l,CVD = h,a,s,r,t,CVD + xb,a,s,r,t ~ ^fc.a.s.r.t.CVD.CVD — ^i),a,s,r,t,Cra,OT//
Equation G-30:
lb,a+l,s,r,t+l,OTH ~ Ib,a,s,r,t,OTH ~ xb,a,s,r,t ~ db,a,s,r,t,OTH,OTH
G.3.4 Recurrent Estimation of Post-Acute CVD Mortality
Survivors of the first type / non-fatal hard CVD event at age a in year t, nb a s r j 10,
are followed for five future years (i.e., k = 1,2,3,4,5) to evaluate post-acute CVD mortality.
EPA estimates the number of post-acute CVD deaths among survivors of a first hard CVD event
in year k since the initial event at age a, mb,a+k,s,r,t+k,k-. by (1) adjusting the number of those
who survived k — 1 years after the initial event, nb a+k_liS rjit+k_lik_1, for non-CVD mortality
using externally estimated non-CVD mortality rate, qa+k,s,r,OTH-.-. (2) multiplying the result by
Proposed PFAS Rule Economic Analysis
G-27
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
externally estimated post-acute CVD mortality rate, /J.a+k,s,rj,kar|d (3) summing over the first
hard CVD event type /:
Equation G-31:
^-b,a+k,s,r,t+k,k ~ / \p-a+k,s,r,f,k ' (l — Ra+k,s,r,OTH) ' W-b,a+k-l,s,r,f,t+k-l,k-l\
t—'fEF
EPA estimates the number of survivors of type / first hard CVD event in year k since the initial
event at age a, nba+kiSrjit+kk, by adjusting the number of those who survived k — 1 years after
the initial event, nba+k_liSrjit+k_lk_1, for mortality using externally estimated non-CVD
mortality rate, qa+k,s,r,OTH-. ar|d post-acute CVD mortality using rate, l^a+k,s,r,f,k-
Equation G-32:
W-b,a+k,s,r,f,t+k,k (l fta+k,s,r,f,k) ' (l — Ra+k,s,r,OTH) ' ^b,a+k-l,s,r,f,t+k-l,k-l
G. 3.5 Risk Reduction Coiculotions
Assuming that the regulatory alternative is associated with a lower incidence of first hard CVD
events (via lower total cholesterol levels due to lower serum PFAS), at the end of time period t,
the number of avoided type / non-fatal first hard CVD events in the sex s and race/ethnicity r
cohort born in year b and currently age a is estimated as:
Equation G-33:
_ ^Baseline Scenario _ Regulatory Alternative
nnb,a,s,r,f,t ~ 'Lb,a,s,r,f,t,0 nb,a,s,r,f,t,0
The number of avoided year t CVD deaths in the first hard CVD population in the sex s and
race/ethnicity r cohort born in year b and currently age a years is:
Equation G-34:
b,a,s,r,t ~ / (j
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Total number of avoided CVD deaths in the first hard CVD population in year t is:
Equation G-36:
AMt- — ^AtH^ a,s,r,t
^—'aEA.bEB 'seS *—'rER
G.4 ASCVD Model Validation
The validation analysis described herein relied on methodology implemented in R software and
differs slightly from SafeWater MCBC methods. Specifically, SafeWater performs a set of pre-
calculations to maximize computational efficiency and, as such, the order of analytical steps
across R and SafeWater models differs; however, results across models are mathematically
consistent. Furthermore, the R-based model version treats each integer age cohort between 85
and 99 separately, implements the CVD calculations for those aged 40-89 years only, and applies
the ASCVD model-based annual incidence at age 80 years to ages 81-89 because the ASCVD
model has been fit to those aged 40-80 years and predicts the 10-year probability of the first
CVD event.
EPA generated life table CVD model results for race/ethnicity subpopulations under different
assumptions regarding the applicability of ASCVD coefficients for non-Hispanic Whites and
non-Hispanic Blacks to Hispanic and non-Hispanic other subpopulations. CVD model inputs are
summarized in Table G-12. The size of each subpopulation cohort was estimated using the 2020
U.S. population size and nationally representative age / sex / race/ethnicity distribution from the
American Community Survey, 2017 (U.S. Census Bureau, 2017). EPA evaluated the alignment
among age-, sex-, and race/ethnicity-specific CVD incidence prediction using the ASCVD model
and age-, sex-, and race/ethnicity-specific CVD incidence prediction calculated by the CVD
model on the basis of race-, sex-, and age-specific prevalence of persons with a history of CVD
events based on MEPS 2010-2017 (see Section G.2.2); U.S. Life Tables, 2017 (Arias et al.,
2019); and CVD death rates, 1999-2019 (Centers for Disease Control and Prevention, 2020c).
For each race/ethnicity, sex, and age combination, EPA first computed the ratio of CVD
incidence based on reported data and incidence based on the ASCVD model. EPA then
computed the absolute value of the deviation of this ratio from 1 and averaged the results over
age using population weights for each sex and race/ethnicity subpopulation. Table G-l 1 reports
the resulting alignment metrics for each combination of subpopulation and ASCVD model
coefficient set. Results show that the ASCVD model coefficients for the non-Hispanic Black
model are more consistent with data on CVD prevalence and mortality for Hispanic and non-
Hispanic other race subpopulations than the ASCVD model coefficients for the non-Hispanic
White model.
Proposed PFAS Rule Economic Analysis
G-29
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-ll: Summary of ASCVD Model Validation
Alignment of ASCVD Model Predictions with Prevalence and
Mortality Statistics"
Sex
Race/Ethnicity
ASCVD Model Coefficients
Estimated in Non-Hispanic
White Sample
ASCVD Model Coefficients
Estimated in Non-Hispanic
Black Sample
Non-Hispanic White
0.64
-
Males
Non-Hispanic Black
-
0.22
Hispanic
0.44
0.23
Non-Hispanic Other
0.57
0.18
Non-Hispanic White
2.00
-
Females
Non-Hispanic Black
Hispanic
1.53
1.37
0.90
Non-Hispanic Other
1.44
1.07
Note:
Alignment is represented by the population-weighted absolute value of age-specific |R - 11 within each sex and race/ethnicity
subpopulation, where R is the race/ethnicity-, age-, and sex-specific ratio of CVD incidence computed from reported data and
incidence computed from the ASCVD model.
G.5 CVD Model Inputs
Table G-12 summarizes the inputs and data sources used in the CVD model, including survey
health data, model coefficients, Centers for Disease Control and Prevention life tables,
hospitalization data, and mortality incidence data.
Table G-12: Summary of Inputs and Data Sources Used in the CVD Model
Data Element
Modeled Variability
Data Source
Notes
Percentage of
population with
high blood
pressure
Age: 10-year age
groups (ages 40-79)
Sex: males, females
Race/Ethnicity: non-
Hispanic White, non-
Hispanic Black, non-
Hispanic other,
Hispanic
NHANES 2011-
2016 (Centers for
Disease Control
and Prevention,
2013b, 2015a,
2015b, 2016b,
2017b, 2017c)
EPA used the percentage of population with
high blood pressure in 10-year age groups to
estimate the number of exposed individuals with
high blood pressure who are exposed to
PFOA/PFOS in drinking water. The blood
pressure measurement NHANES datasets from
2011-2016 were combined with corresponding
respondent-specific demographic profile,
medical questionnaire, and blood pressure
questionnaire datasets to summarize the
percentage of the non-CVD population that has
high blood pressure for each age-, sex-,
and race-specific stratum.
Proposed PFAS Rule Economic Analysis
G-30
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-12: Summary of Inputs and Data Sources Used in the CVD Model
Data Element
Modeled Variability
Data Source
Notes
Percentage of
population
receiving blood
pressure
treatment
Age: 10-year age
groups (ages 40-79)
Sex: males, females
Race/Ethnicity: non-
Hispanic White, non-
Hispanic Black, non-
Hispanic other,
Hispanic
NHANES 2011-
2016 (Centers for
Disease Control
and Prevention,
2013b, 2015a,
2015b, 2016b,
2017b, 2017c)
To determine the percentage of the population
with controlled high blood pressure, the
percentage of the populations per age group and
sex who have high blood pressure was
multiplied by the percentage of the populations
per age group and sex who received treatment
for high blood pressure. The blood pressure
measurement NHANES datasets from
2011-2016 were combined with corresponding
respondent-specific demographic profile,
medical questionnaire, and blood pressure
questionnaire datasets to summarize the
percentage of the non-CVD population that is
being treated for having high blood pressure for
each age-, sex-, and race-specific stratum.
Treated,
untreated, and
normal systolic
blood pressure
measurements
Age: age groups 40-
59, 60+
Sex: males, females
Race/Ethnicity: non-
Hispanic White, non-
Hispanic Black, non-
Hispanic other,
Hispanic
Treatment status:
controlled,
uncontrolled-high,
uncontrolled-normal
NHANES 2011-
2016 (Centers for
Disease Control
and Prevention,
2013b, 2015a,
2015b, 2016b,
2017b, 2017c)
The blood pressure measurement NHANES
datasets from 2011-2016 were combined with
corresponding respondent-specific demographic
profile, medical questionnaire, and blood
pressure questionnaire datasets to summarize the
percentage of the non-CVD population that is
being treated for having high blood pressure for
each treatment status-, age-, sex-, and
race-specific stratum.
Baseline total
cholesterol level
Age: 10-year age
groups (ages 40-79)
Sex: males, females
Race/Ethnicity: non-
Hispanic White, non-
Hispanic Black, non-
Hispanic other,
Hispanic
NHANES 2011-
2016 (Centers for
Disease Control
and Prevention,
2013b, 2015a,
2015b, 2016b,
2017b, 2017c)
The total cholesterol NHANES datasets from
2011-2016 were combined with corresponding
respondent-specific demographic profile and
medical questionnaire datasets to summarize
weighted average total cholesterol levels in
mg/dL for each age-, sex-, and race-specific
stratum in the non-CVD population.
Baseline high
density
lipoprotein
cholesterol level
(HDLC)
Age: 10-year age
groups (ages 40-79)
Sex: males, females
Race/Ethnicity: non-
Hispanic White, non-
Hispanic Black, non-
Hispanic other,
Hispanic
NHANES 2011-
2016 (Centers for
Disease Control
and Prevention,
2013a, 2015a,
2015b, 2016a,
2017a, 2017c)
The HDLC NHANES datasets from 2011-2016
were combined with corresponding respondent-
specific demographic profile and medical
questionnaire datasets to summarize weighted
average HDLC levels in mg/dL for each age-,
sex-, and race-specific stratum in the non-CVD
population.
Proposed PFAS Rule Economic Analysis
G-31
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-12: Summary of Inputs and Data Sources Used in the CVD Model
Data Element
Modeled Variability
Data Source
Notes
Smoking
prevalence
Age: 10-year age
groups (ages 40-79)
Sex: males, females
Smoking status:
fraction of smokers
NHANES 2011-
2016 (Centers for
Disease Control
and Prevention,
2013d, 2015a,
2015b, 2015d,
2017c, 2017e)
The percentage of smokers and non-smokers in
each stratum were used as inputs in the ASCVD
model, providing results similar to using binary
variables representing that an individual is either
a smoker or a non-smoker and further stratifying
the sample. The smoking NHANES datasets
from 2011-2016 were combined with
corresponding respondent-specific demographic
profile and medical questionnaire datasets to
summarize the percentage of the non-CVD
population that smokes for each age-, sex-,
and race-specific stratum.
Diabetes
prevalence
Age: 10-year age
groups (ages 40-79)
Sex: males, females
Diabetes status:
fraction of diabetics
NHANES 2011-
2016 (Centers for
Disease Control
and Prevention,
2013c, 2015a,
2015b, 2015c,
2017c, 2017d)
The percentage of the population with and
without diabetes in each stratum were used as
inputs in the ASCVD model, providing results
similar to using binary variables representing
that an individual has or does not have diabetes
and further stratifying the sample. The diabetes
NHANES datasets from 2011-2016 were
combined with corresponding respondent-
specific demographic profile and medical
questionnaire datasets to summarize the
percentage of the non-CVD population that has
diabetes for each age-, sex-, and race-specific
stratum.
ASCVD model
coefficients
Sex: males, females
Race: non-Hispanic
White, non-Hispanic
Black
Goffetal. (2014),
Table A
For modeling purposes, the Hispanic
subpopulation was assigned coefficients
estimated for the non-Hispanic White
subpopulation. The model applies to ages
40-89. ASCVD regressors include age, TC,
HDLC, treated systolic BP, untreated systolic
BP, smoking status, and diabetes status.
Annual all-cause
death probability
Sex: males, females
Age: integer ages 0 ...
100
Race/Ethnicity: all,
non-Hispanic White,
non-Hispanic Black,
Hispanic
U.S. Life Tables,
2017 (Arias et al.,
2019)
The quantity used in modeling is qx (i.e., the
probability of dying between ages x and x + 1).
Life table data for the non-Hispanic other race
category are not available; for subsequent
modeling, all-race life tables are used for
this category.
Proposed PFAS Rule Economic Analysis
G-32
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-12: Summary of Inputs and Data Sources Used in the CVD Model
Data Element
Modeled Variability
Data Source
Notes
Annual non-
CVD death
probability for
age 90+
Sex: males, females
Age: integer ages 90 ...
100
Race/Ethnicity: all,
non-Hispanic White,
non-Hispanic Black,
Hispanic
U.S. Life Tables,
2017 (Arias et al.,
2019); U.S. Life
Tables Eliminating
Certain Causes of
Death, 1999-2000
(Arias et al., 2013)
Annual non-CVD death probability is estimated
by multiplying qx from the 2017 U.S. life tables
by the sex-specific ratio of non-CVD qx to
all-cause qx from 1999-2000 U.S. life tables
eliminating certain causes. Life table data for the
non-Hispanic other race category are not
available; for subsequent modeling, all-race life
tables are used for this category. The 1999-2000
U.S. life tables eliminating certain causes are not
race/ethnicity-specific; the U.S. general
population ratios of non-CVD qx to all-cause qx
were applied to all race/ethnicity categories.
The 1999-2000 U.S. life tables eliminating
certain causes are abridged and report 5-year
rates. The corresponding 5-year ratios are
applied to all individual years within the
5-year range.
Annual non-
CVD death
probability for
ages 40+
Sex: males, females
Age: integer ages 40 ...
89
Race/Ethnicity: non-
Hispanic White, non-
Hispanic Black, non-
Hispanic other,
Hispanic
U.S. Life Tables
2017 (Arias et al.,
2019); CVD death
rates, 1999-2019
(Centers for
Disease Control
and Prevention,
2020c)
Annual non-CVD death probability is estimated
by multiplying qx from 2017 U.S. life tables by
the ratio of non-CVD qx to all-cause qx.
The non-CVD qx estimate was obtained for each
integer age by sex combination as the difference
between all-cause qx from U.S. 2017 life tables
and CVD qx from CDC 1999-2019 cause-
specific mortality rates. U.S. 2017 life table data
for the non-Hispanic other race category are not
available; life tables for the U.S. general
population are used for this category.
CVD prevalence
Sex: males, females
Age: age groups 18-
44, 45-64, 65+
Race/Ethnicity: non-
Hispanic White, non-
Hispanic Black, non-
Hispanic other,
Hispanic
Condition: MI, IS,
other CHD, MI + IS +
other CHD conditions
combined
MEPS 2010-2017
(Agency for
Healthcare
Research and
Quality, 2011,
2012a, 2012b,
2013a, 2013b,
2014a, 2014b,
2015a, 2015b,
2016a, 2016b,
2017b, 2017c,
2018, 2019a,
2019b, 2019c)
MEPS longitudinal files were used to obtain
survey weights, design variables, and
information on cardiovascular conditions
(including age at diagnosis) that began prior to
the start date for the survey panel. MEPS
medical conditions files were used to obtain
information on the newly diagnosed conditions
of interest. Specifically, MI events were
identified using ICD9=410 or MIDX=1,
stroke events were identified using
ICD9=433,434,435,436 or STRKDX=1,
other CHD were identified using
ICD9=413,414,427,428 or CHDDX=1,
ANGIDX=1, OHRTDX= 1. CVD prevalence
was estimated based on persons whose condition
started at an age prior to the age at which the
MEPS round interview was conducted.
Proposed PFAS Rule Economic Analysis
G-33
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-12: Summary of Inputs and Data Sources Used in the CVD Model
Data Element
Modeled Variability
Data Source
Notes
CVD incidence
in the non-CVD
population
Sex: males, females
Age: age groups 18-
44, 45-64, 65+
Race/Ethnicity: non-
Hispanic White, non-
Hispanic Black, non-
Hispanic other,
Hispanic
Condition: MI, IS,
other CHD
MEPS 2010-2017
(Agency for
Healthcare
Research and
Quality, 2011,
2012a, 2012b,
2013a, 2013b,
2014a, 2014b,
2015a, 2015b,
2016a, 2016b,
2017b, 2017c,
2018, 2019a,
2019b, 2019c)
MEPS longitudinal files were used to obtain
survey weights, design variables, and
information on cardiovascular conditions
(including age at diagnosis) that began prior to
the start date for the survey panel. MEPS
medical conditions files were used to obtain
information on the newly diagnosed conditions
of interest. Specifically, MI events were
identified using ICD9=410 or MIDX=1,
stroke events were identified using
ICD9=433,434,435,436 or STRKDX=1,
other CHD were identified using
ICD9=413,414,427,428 or CHDDX=1,
ANGIDX=1, OHRTDX= 1. CVD incidence was
estimated based on persons whose condition
started at an age that was the same as the age at
which the MEPS round interview was
conducted.
In-hospital death
probability for
CVD events
Sex: males, females
Age: age groups 18-
44, 45-64, 65-84, 85+
Condition: MI, IS,
other CHD
HCUP2017
(Agency for
Healthcare
Research and
Quality, 2017a)
Hospital death probabilities were estimated from
condition-specific hospitalizations identified
using the following ICD10 codes: ICD 10=121
for MI, ICD 10=163 for IS, and ICD 10=120,122-
125 for other CHD. HCUP reports death
probabilities separately by sex or within age
groups. EPA estimated age group- and sex-
specific hospital death probabilities by assuming
that male/female relative risk does not vary
across age groups.
1-year, 2-year,
3-year, 4-year,
and 5-year all-
cause mortality
incidence in MI
survivors ages
40-64
Sex: males, females
Race: all
Age: age groups 40-65
Condition: MI
Thometal. (2001);
MI incidence
based on the
MEPS 2010-2017
analysis, U.S. Life
Tables, 2017
(Arias et al., 2019)
Thom et al. (2001) sex- and race-specific
estimates for 1-year follow-up and 5-year
follow-up all-cause mortality for ages 45-64 MI
survivors are as reported in Roger et al. (2012)
(the text of the original report is not accessible).
Thom et al. (2001) generated separate estimates
for non-Hispanic White and non-Hispanic Black
persons. To derive sex-specific all-race/ethnicity
estimates, EPA used MEPS-based
race/ethnicity- and sex-specific MI incidence for
ages 45-64 and assumed that non-Hispanic
White mortality estimates apply to other
race/ethnicity categories. To derive 2-year,
3-year, and 4-year all-cause post-MI mortality
incidence, EPA further assumed that the annual
probability of death between 1-year follow-up
and 5-year follow-up was constant. Finally, EPA
assumed that the resulting estimates apply to
ages 40-44 MI survivors and age 65 MI
survivors.
Proposed PFAS Rule Economic Analysis
G-34
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table G-12: Summary of Inputs and Data Sources Used in the CVD Model
Data Element
Modeled Variability
Data Source
Notes
1-year, 2-year,
3-year, 4-year,
5-year, and 6-
year all-cause
mortality
incidence in MI
survivors and IS
survivors age
65+
Sex: all
Race: all
Age: age group 65+
Condition: MI, IS
S. Lietal. (2019)
S. Li et al. (2019) estimates based on 2008 MI
and 2008 IS Medicare cohorts (see Figure 1 of
the paper) were used. Note that these estimates
are neither race- nor sex-specific.
1-year, 2-year,
3-year, 4-year,
and 5-year CVD
mortality
incidence in MI
survivors ages
40-65
Sex: males, females
Race: non-Hispanic
White, non-Hispanic
Black,
Age: age groups 40-65
Condition: MI
Thometal. (2001);
MI incidence
based on the
MEPS 2010-2017
analysis, U.S. Life
Tables, 2017
(Arias et al., 2019);
CVD death rates
1999-2019
(Centers for
Disease Control
and Prevention,
2020c)
EPA used estimated annual age- and sex-
specific non-CVD death probability (estimated
as described above) to calculate the probability
of non-CVD death within the next 1, 2, 3, 4, and
5 years. These probabilities were averaged over
ages 45-64 using MI incidence-based weights
estimated from MEPS 2010-2017 (estimated as
described above). EPA then subtracted these
estimates from 1-, 2-, 3-, 4-, and 5-year sex-
specific all-cause mortality incidence in MI
survivors ages 45-64 (estimated as described
above) to obtain 1-, 2-, 3-, 4-, and 5-year CVD
mortality incidence. Based on this result, EPA
estimated the sex-specific ratios of CVD
mortality to all-cause mortality in MI survivors
1, 2, 3, 4, and 5 years after the initial event.
These ratios were applied to non-Hispanic
White and non-Hispanic Black all-cause post-
Mi mortality reported in Thom et al. (2001) to
obtain post-acute CVD mortality estimates for
these races. The other race/ethnicity categories
used in modeling were assigned post-acute CVD
mortality rates for non-Hispanic Whites. Finally,
EPA assumed that the resulting estimates
applied to ages 40-44 MI survivors and to age
65 MI survivors.
1-year, 2-year,
3-year, 4-year,
5-year, and 6-
year CVD
mortality
incidence in MI
survivors and IS
survivors ages
65+
Sex: male, female
Race: all
Age: ages 66 ... 89
Condition: MI, IS
S. Lietal. (2019);
U.S. Life Tables,
2017 (Arias et al.,
2019); CVD death
rates, 1999-2019
(Centers for
Disease Control
and Prevention,
2020c); U.S. Life
Tables Eliminating
Certain Causes of
Death, 1999-2000
(Arias et al., 2013)
EPA used estimated annual age- and sex-
specific non-CVD death probability (estimated
as described above) to calculate the probability
of non-CVD death within the next 1, 2, 3, 4, 5,
and 6 years. These results were averaged using
S. Li et al. (2019) 2008 MI/IS cohort age and
sex characteristics. In conjunction with all-cause
post-MI/IS mortality estimates from S. Li et al.
(2019), these estimates were used to estimate the
ratio of CVD mortality to the general population
non-CVD mortality 1, 2, 3, 4, 5, and 6 years
after the initial MI/IS event. The sex- and age-
specific probabilities of CVD death 1, 2, 3, 4, 5,
and 6 years after the initial MI/IS event were
estimated by applying these ratios to sex- and
age-specific non-CVD mortality probabilities.
Proposed PFAS Rule Economic Analysis
G-35
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table G-12: Summary of Inputs and Data Sources Used in the CVD Model
Data Element
Modeled Variability Data Source
Notes
Abbreviations: ASCVD - atherosclerotic cardiovascular disease; CHD - coronary heart disease; CVD - cardiovascular disease;
HCUP - Healthcare Cost and Utilization Project; IS - ischemic stroke; MEPS - Medical Expenditure Panel Survey;
MI - myocardial infarction; NCHS - National Center for Health Statistics; NHANES - National Health and Nutrition
Examination Survey; PFOA - perfluorooctanoic acid; PFOS - perfluorooctanesulfonic acid.
Proposed PFAS Rule Economic Analysis
G-36
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix H. Cancer Benefits Model Details and
Input Data
This appendix details the cancer life table approach, the data used to estimate reduced RCC cases
resulting from changes in exposure to PFOA via drinking water, and the data used to estimate
reduced bladder cancer cases resulting from changes in exposures to disinfection byproducts
(DBPs) via drinking water. This appendix also provides baseline kidney and bladder cancer
statistics.
H.l Details on the Cancer Life Table Approach
This appendix details the life table calculations used to estimate reduced cancer cases among
population cohorts affected by reductions in PFAS and co-occurring contaminant levels at PWS
following implementation of drinking water treatment technologies.
The life table is a metric designed to represent the longevity of people from a certain population.
The inputs to the life table are the age-specific probability of death and the initial population size
(e.g., the retail population served at a given PWS). Based on this information, the life table
computes the number of persons surviving to a specific age, the number of deaths occurring at a
given age, the number of person-years lived at a given age, the number of person-years lived
beyond a given age, and age-specific life expectancy. The details of standard life table
calculations can be found in Anderson (1999). EPA has previously used life table approaches in
regulatory analyses, including the analysis of lead-associated health effects in the 2015 Benefit
and Cost Analysis for the Effluent Limitations Guidelines, Standards for the Steam Electric
Power Generating Point Source Category (U.S. EPA, 2015), and PM: 5-related health effects in
revisions to the National Ambient Air Quality Standards for ground-level ozone (U.S. EPA,
2008). Other examples of use of a life table approach among federal agencies include EPA's
analysis of Benefits and Costs of the Clean Air Act from 1990 to 2020 (U.S. EPA, 201 la) and
the Occupational Safety and Health Administration (OSHA) assessment of lifetime excess lung
cancer, nonmalignant respiratory disease mortality, and silicosis risks from exposure to
respirable crystalline silica (81 FR 16285, March 25, 2016; OSHA, 2010).
To estimate the health effects of changes in exposures to cancer-causing pollutants, the health
risk model tracks evolution of two populations over time - the cancer-free population and the
population living with cancer.44 These two populations are modeled for both the baseline annual
exposure scenario and for the regulatory alternative annual exposure scenario. Populations in the
baseline and regulatory alternative exposure scenarios are demographically identical, but they
differ in the pollutant levels to which they are exposed. EPA assumes that the population is
exposed to baseline pollutant levels prior to technology implementation year (i.e., change in a
given pollutant equals 0) and to alternative pollutant levels that reflect the impact of treatment
implementation under the regulatory alternative. All PWSs with baseline PFAS exceedances are
assumed to upgrade their treatment by 2026 to comply with the proposed regulation. To capture
these effects while being consistent with the remainder of the benefit framework, EPA modeled
44 When referring to the "cancer-free" population, EPA is referring to the population that is free of the specific type of cancer
modeled in this analysis, rather than the population that is free of all cancers.
Proposed PFAS Rule Economic Analysis
H-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
changes in health outcomes resulting from changes in exposure over an evaluation period that
starts in 2023 and ends in 2104.45
The model tracks all-cause mortality and cancer experience for a set of model populations
defined by sex, location (if modeled), birth year B = 1938, ...,2023,2024, ...,2104,
and age attained by 2023 (for those alive in 2023), which is denoted by A = 0,1,2,3,... 85 +.46
Each model population is followed from age 0 in year B to age min (100,2104 — B) in year
min (B + 100,2104), using a one-year time step. For cohorts born prior to or in 2023, the model
is initialized using the location- (if modeled), age-, race/ethnicity- (if modeled), and sex-specific
number of persons estimated to be alive in 2021. For cohorts born after 2023, the model is
initialized using the location- (if modeled), race/ethnicity-, and sex-specific number of persons
age 0 estimated to be alive in 2021. Location- and sex, race/ethnicity-, and age-specific
population details are included in Appendix B.
Below, EPA provides a list of variables included in the health risk model (Table H-l) and
describes the process for quantifying the evolution of model population defined by B and A
under baseline exposure assumptions.47 EPA omits sex and location-specific indices because
calculation steps do not differ across sexes and locations. EPA then describes the process for
quantifying the evolution of the population under regulatory alternative exposures. Finally, EPA
describes the process for estimating the total calendar year y-specific health benefits. EPA
aggregates benefits estimates over all model populations (([B, A) =
{(1938,85+),..., (2023,0), (2024,0),..., (2104,0)}).
Table H-l: Health Risk Model Variable Definitions
Variable Definition
a
Current age or age at cancer diagnosis
xa
A person's lifetime pollutant exposure under the regulatory alternative by age a
Za
A person's lifetime baseline pollutant exposure by age a
LRa
Lifetime risk of cancer per person within age interval [0, a) under the baseline conditions
IRa
Age-specific baseline annual cancer incidence rate per person
B
Birth year
A
Age in 2023 (years) for those alive in 2023, 0 for those born after 2023
P
Number of affected persons of age A in 2023 or persons aged 0 born after 2023
y
Calendar year
x„ „
A person's lifetime pollutant exposure under the regulatory alternative by age a given that this
a,y
age occurs in year y
za,y
A person's lifetime baseline pollutant exposure by age a given that this age occurs in year y
lc=0,a,y(za,y)
The baseline number of cancer-free living individuals at the beginning of age a given that this
age occurs in year y
45 Although benefits of lagged changes in lifetime cancer risk after 2104 may be attributed to changes in
contaminant exposure during the analysis period, EPA did not model effects beyond this period.
46 Note that those born after the start of the evaluation period in 2023 (i.e., during 2023 -2104) are always tracked
starting from age 0. As with the CVD model, those aged 85 years or older at the start of the analysis are treated as a
single cohort, with mortality statistics averaged over ages 85-100 years and serum PFOA/PFOS set at values
corresponding to age 85 years at the beginning of evaluation.
47 SafeWater was programmed for maximal computational efficiency and SafeWater performs a series of pre-calculations to
reduce model runtime. Therefore, the specific equations in the SafeWater code differ from the equations in this Appendix, but the
end result is mathematically consistent.
Proposed PFAS Rule Economic Analysis
H-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-l: Health Risk Model Variable Definitions
Variable Definition
dc=o av(zav) The baseline number of deaths among cancer-free individuals at age a given that this age occurs
in year y
lc= ] ay(zay) The baseline number of new cancer cases at age a given that this age occurs in year y
qa Probability of a general population all-cause death at age a
Ta Share of cancer deaths among all-cause deaths at age a
Ya Baseline probability of a new cancer diagnosis at age a
k Cancer duration in years
s Cancer stage (localized, regional, distant, unstaged)
Ss=s a Age-specific share of new stage s cancers
h=s a y o (za y) The baseline number of new stage s cancers occurring at age a given that this age occurs in year
y
rs=s a k Relative survival rate k years after stage s cancer occurrence at age a
qs=s,a,k Stage-specific probability of death in the cancer population whose cancer was diagnosed at age
a and they lived k years after the diagnosis. Current age of these individuals is a + k.
ds=s a y o (za ) The baseline number of deaths in the stage s cancer population in the year of diagnosis (i.e.,
when k = 0), given the current age a and the corresponding year y.
h=s a y k (za y-k) The baseline number of individuals living with the stage s cancer in the k-th year after diagnosis
in year y, given the cancer diagnosis at age a and the cumulative exposure through to that age
and year y - k.
ds=s a y k (za y_k) The baseline number of deaths among those with the stage s cancer in the k-th year after
diagnosis in year y, given the cancer diagnosis at age a and the cumulative exposure through to
that age and year y-k.
es=s ayk(zay-k) The baseline number of excess cancer deaths (i.e., the number of deaths in the cancer population
over and above the number of deaths expected in the general population of the same age) among
those with the stage s cancer in the k-th year after diagnosis in year y, given the cancer
diagnosis at age a and the cumulative exposure through to that age and year y-k.
LRa y(zay) Recursive estimate of the lifetime risk of cancer within age interval [0, a) under the baseline
conditions, given that age a occurs in year y
RR (xa y, za y) Relative risk of cancer by age a given that this age occurs in year y, baseline exposure za y and
regulatory alternative exposure xa y
LRa y (xa y) Recursive estimate of the lifetime risk of cancer within age interval [0, a) under the regulatory
alternative, given that age a occurs in year y
NCB,A,y,s The incremental number of new stage s cancer cases in year y for the model population (B,A).
LCB,A,y,s The incremental number of individuals living with stage s cancer in year y for the model
population (B,A).
EDgAy The incremental number of excess in stage s cancer population in year y for the model
population (B,A).
H.l.l Evolution of Model Population (B,A) under Baseline
Pollutant Exposure
Given a model population (B,A), for each current age a and calendar year y, the following
baseline exposure zay = Baseline Pollutant( y_a+(- dependent quantities are computed:
lc=o,a,y(za,y): The number of cancer-free living individuals at the beginning of age a,
in year y;
d-c=o,a,y(za,y)- The number of deaths among cancer-free individuals aged a during the year y;
Proposed PFAS Rule Economic Analysis
H-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
lc=i,a,y(za,y): The number of new cancer cases among individuals aged a during the
year y.
To compute each quantity above, EPA makes assumptions about the priority of events that
terminate a person's existence in the pool of cancer-free living individuals. These events are
general population non-cancer deaths that occur with probability 48 q a(l — ra) and new cancer
diagnoses that occur with probability ya, which is approximated by age-specific annual cancer
incidence rate lRa. In the model, EPA assumes that the new cancer diagnoses occur after general
population non-cancer deaths and use the following recurrent equations for ages a > 0:49
Equation H-l:
lc=0,a,y(za,y) — ^C=0,a-l,y-l(za-l,y-l) — ^C=0,a-l,y-l(za-l,y-l) — ^C=l,a-l,y-l(za-l,y-l)
Equation H-2:
dc=0,a,y(za,y) ~ — Ta~) ' lc=Q,a,y(za,y)
Equation H-3:
lc=l,a,y(za,y) — Ya ' (j-C=0,a,y(za,y) ~ ^C=0,a,y(^a,y)^
To initiate each set of recurrent equations for those alive in 2023, EPA estimates the number of
cancer-free individuals at age a = 0, denoted by ^c=o,o,y-^(zo,y-;i)> that is consistent with the
number of affected persons of age A in 2023, denoted by P. To this end, Equation H-l,
Equation H-2, and Equation H-3 are estimated as find /c=o,o,y-^(zo,y-4) = P/Tli=o(.l ~ Ri)
where P = 1-c=o,a,2023(za,2023)- To initiate each set of recurrent equations for those born after
2023, EPA uses the PWS-, race/ethnicity-, sex, and scenario-specific number of persons who
died in the previous year of the analysis, thereby ensuring that the size of the modeled population
remains constant throughout the analysis period.
48 The model does not index the general population death rates using the calendar year, because the model relies
on the most recent static life tables.
49 EPA notes that this is a conservative assumption that results in a lower bound estimate of the regulatory
alternative impact (with respect to this particular uncertainty factor). An upper bound estimate of the regulatory
alternative impact can be obtained by assuming that new cancer diagnoses occur before general population deaths.
In a limited sensitivity analysis performed as part of the Benefit and Cost Analysis for Proposed Revisions to the
Effluent Limitations Guidelines and Standards for the Steam Electric Power Generating Source Category (U.S. EPA,
2019), EPA found that estimates generated using this alternative assumption were approximately 5 percent larger
than the estimates assuming that new cancer diagnoses occur after general population deaths.
Proposed PFAS Rule Economic Analysis
H-4
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Consistent with available cancer survival statistics, EPA models mortality experience in the
cancer populations lc=i,a,y{za,y) as dependent on the age-at-onset a, disease duration k, and
cancer stage s (e.g., localized, regional, distant, unstaged). Given each age-specific share of new
cancer cases lc=i,a,y{za,y) ar|d age-specific share of new stage s cancers 8s=sa, EPA calculates
the number of new stage s cancers occurring at age a in year y:
Equation H-4:
^¦S=s,a,y,o(za,y) ~ &S=s,a ' ^C=l,a,y(^a,y)
For a model population (B,A) and cancer stage s, EPA separately tracks min (85,2104 — B) —
A + 1 new stage-specific cancer populations from age-at-onset a to age min (85,2104 — 5)..50
Next, a set of cancer duration /c-dependent annual death probabilities is derived for each
population from available data on relative survival rates.51 rs=s a k and general population annual
death probabilities qa+k as follows:
Equation H-5:
~ 1 rS=s,a,k+1 ^
Rs=s,a,k ~ -L „ v Ra+kJ
rS=s,a,k
EPA estimates deaths in the cancer population in the year of diagnosis (i.e., when k = 0)
as follows:
Equation H-6:
ds=s,a,y,o(Za,y) ~ Rs=s,a,0 ' ^¦S=s,a,y,o(za,y)
In years that follow the initial diagnosis year (i.e., k > 0), EPA uses the following recurrent
equations to estimate the number of people living with cancer and the annual number of deaths
in the cancer population:
Equation H-7:
^S=s,a,y,k(za,y-k) — ^¦S=s,a,y,k-l(za,y-k) ~ dS=s,a,y,k-l(^a,y-k)
Equation H-8:
d-s=s,a,y,k(za,y-k) — Rs=s,a,k ' ^¦S=s,a,y,k(za,y-k)
50 In total, there are 4 ¦ (min (85,2104 — B) — A + 1) new cancer populations being tracked for each model
population.
51 Note that rs=s a k is a multiplier that modifies the general probability of survival to age a + k to reflect the fact
that the population under consideration has developed cancer k years ago.
Proposed PFAS Rule Economic Analysis
H-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Because the Agency is interested in cancer-related deaths rather than all deaths in the cancer
population, EPA also tracks the number of excess cancer population deaths {i.e., the number of
deaths in the cancer population over and above the number of deaths expected in the general
population of the same age). The excess deaths are computed as:
Equation H-9:
^S=s,a,y,k(za,y-k) ~ Rs=s,a,k ' ^¦S=s,a,y,k(za,y-k) ~ tfa+k ' ^¦S=s,a,y,k(za,y-k)
H. 1.2 Evolution of Model Population (B,A) under the Regulatory
Alternative Pollutant Exposure
Under the baseline conditions when the change in contaminant levels is zero (i.e., before 2026),
EPA approximates the annual cancer probability ya by age-specific annual cancer incidence rate
IRa. EPA computes the pollutant-dependent annual new cancer cases under the regulatory
alternative conditions, lc=i,a,y(,xa,y)> m three steps. First, EPA recursively estimates LRay(zay\
the lifetime risk of cancer within age interval [0, a) under the baseline conditions:
Equation H-10:
LRa,y{Za,y) ~ ~a(zo T) ' ^-'=0 a ^ ^ an<^ ^^O.y-Ai^O.y-A) ~ 0
Second, the result of Equation H-10 is combined with the relative risk estimate RR{xay,zay\
associated with each cancer type:
Equation H-ll:
^^a,y— RR(jXa,y> za,y)^^a,y(^a,y)
This results in a series of lifetime cancer risk estimates under the regulatory alternative. Third,
EPA computes a series of new annual cancer case estimates under the regulatory alternative as
follows:
Equation H-12:
lc=l,a,y(xa,y) ~ (j-,^a+l,y+l{xa+l,y+l) ~ LRaiy(Xa,y)^ ' lc=0,0,y-A(z0,y-A)
H. 13 Health Effects and Benefits Attributable to the Regulatory
Alternatives
To characterize the overall impact of the regulatory alternatives in a given year y, for each model
population defined by (B,A), sex, and location, EPA calculates three quantities: the incremental
number of new stage s cancer cases {NCA y s), the incremental number of individuals living with
stage s cancer {LCA y s), and the incremental number of excess deaths in the cancer population
{EDA y). The formal definitions of each of these quantities are given below:
Proposed PFAS Rule Economic Analysis
H-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Equation H-13:
NCB A,y,s = [0 < y — max (2023,5) + A < min (85,2104 — 5) ] ¦
(js=s,y-max (2023,B)+A,y,0 (^y-max (2023,5) +A,y) ~ ^S=s,y-max (2023,B)+A,0 (-^y-max (2023,B)+A,y}^
Equation H-14:
Z100
[0 < y — max (2023, 5) + A + k < min (85,2104 — B) ]
k=1
(js=s,y-max (2023,B)+A-k,y,k^y-max (2023,B)+A-k,y-k)
~ ls=s,y-max (2023,B)+A-k,y,k (-^y-max (2023,B)+A-k,y-k}^
Equation H-15:
Z100
[0 < y — max (2023,5) + A + k
k=0
< min (85,2104
-5)]
%=s,y-max (2023,B)+4-fc,y,fc(^y-max (2023,B)+A-k,y-k)
ses
~ &S=s,y-max (2023,B)+A-k,y,k (-^y-max (2023,B)+A-k,y-k}^
These calculations are carried out to 2104.
H.2 Cancer Life Table Model Input Data
As noted in Section 6.6.2 of the Economic Analysis, EPA relied on data sources including EPA
SDWIS, age-, race/ethnicity- and sex-specific population from U.S. Census Bureau (2020) (See
Appendix B), the Surveillance, Epidemiology, and End Results (SEER) program database
(National Cancer Institute), and the Centers for Disease Control and Prevention (CDC) National
Center for Health Statistics (NCHS) to characterize sex-, race/ethnicity- and age group-specific
general population mortality rates and cancer incidence rates used in model simulations. Table
H-2 summarizes these data sources; Appendix B provides details on the population size
estimates.
Table H-2: Summary of Data Sources Used in Cancer Lifetime Risk Models
Data Element
Modeled Variability
Data Source
Notes
Cancer incidence
rate (IR) per
100,000 persons
Age at diagnosis: 1-year
groups (ages 0 to 100)
Sex: males, females
Cancer type: Kidney
Cancer; Urinary Bladder
(Invasive & In Situ)
Cancer
Race/ethnicity: All, non-
Hispanic White, non-
Surveillance,
Epidemiology, and End
Results (SEER) 21 cancer
incidence rates by age, sex,
and race at diagnosis for
2014-2018 (Surveillance
Research Program -
National Cancer Institute,
2020b)
Distinct SEER 21 IR data were
available forages 0, 1-4, 5-9, 10-
14, 15-19, 20-24, 25-29, 30-34,
35-39, 40-44, 45-49, 50-54, 55-
59, 60-64, 65-69, 70-74, 75-79,
80-84, 85+. EPA assumed that
the same IR applies to all ages
within each age group. EPA
assumed that non-Hispanic
Black iRs can be approximated
Proposed PFAS Rule Economic Analysis H-7 March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-2: Summary of Data Sources Used in Cancer Lifetime Risk Models
Data Element
Modeled Variability
Data Source
Notes
Hispanic Black, Hispanic,
non-Hispanic Other
by Black iRs. EPA assumed that
non-Hispanic Other iRs can be
approximated by all race iRs.
General
population
probability of
death
Age: 1-year groups (ages
0 to 100)
Sex: males, females
Race/ethnicity: All, non-
Hispanic White, non-
Hispanic Black, Hispanic,
non-Hispanic Other
Centers for Disease
Control and Prevention
(CDC)/National Center for
Health Statistics (NCHS)
United States Life Tables,
2017 (Arias etal., 2019)
EPA used race/ethnicity-, age-
and sex-specific probabilities of
dying within the integer age
intervals. EPA assumed that
non-Hispanic Other data can be
approximated by all race data.
Share of cancer
deaths among all-
cause deaths
Age at diagnosis: 1-year
groups (ages 0 to 100)
Sex: males, females
Cancer type: Kidney
Cancer; Urinary Bladder
(Invasive & In Situ)
Cancer
Race/ethnicity: All, non-
Hispanic White, non-
Hispanic Black, Hispanic,
non-Hispanic Other
Underlying Cause of
Death, 1999-2019 on CDC
WONDER Online
Database (Centers for
Disease Control and
Prevention, 2020c)
EPA calculated share of cancer
deaths among all-cause deaths
by race/ethnicity, age and sex by
dividing the number of cancer
deaths during 1999-2019 with
the number of all-cause deaths
during 1999-2019.
Share of bladder
cancer incidence
at specific cancer
stage
Age at diagnosis: 1-year
groups (ages 0 to 100)
Sex: males, females
Cancer stage: localized,
regional, distant,
unstaged
Cancer type: Urinary
Bladder (Invasive & In
Situ) Cancer
SEER 21 distribution of
bladder cancer incidence
over stages by age and sex
at diagnosis for 2008-2018
(Surveillance Research
Program - National Cancer
Institute, 2020b)
Distinct SEER 21 data were
available for ages 0-15, 15-39,
40-64, 65-74, 75+. EPA assumed
that the same cancer incidence
shares by stage apply to all ages
within each age group.
Share of kidney
cancer incidence
at specific cancer
stage
Age at diagnosis: 1-year
groups (ages 0 to 100)
Sex: males, females
Cancer stage: localized,
regional, distant,
unstaged
Cancer type: Kidney
Cancer
Race/ethnicity: All, non-
Hispanic White, non-
Hispanic Black, Hispanic,
non-Hispanic Other
SEER 21 distribution of
kidney cancer incidence
over stages by
race/ethnicity, age and sex
at diagnosis for 2008-2018
(Surveillance Research
Program - National Cancer
Institute, 2020b)
Distinct SEER 21 data were
available for ages 0-15, 15-39,
40-64, 65-74, 75+. EPA assumed
that the same cancer incidence
shares by stage apply to all ages
within each age group. EPA
assumed that non-Hispanic
Black data can be approximated
by Black data. EPA assumed
that non-Hispanic Other data can
be approximated by all race data.
Relative bladder
cancer survival by
cancer stage
Age at diagnosis: 1-year
groups (ages 0 to 100)
Sex: males, females
Duration: 1-year groups
(durations 0 to 100 years)
Cancer stage: localized,
regional, distant,
unstaged
Cancer type: Urinary
SEER 18 relative bladder
cancer survival by age at
diagnosis, sex, cancer stage
and duration with
diagnosis for 2000-2017
(Surveillance Research
Program - National Cancer
Institute, 2020a)
Distinct SEER 18 data were
available for ages at diagnosis
0-14, 15-39, 40-64, 65-74, 75+.
EPA assumed that the same
cancer relative survival patterns
apply to all ages within each age
group. SEER 18 contained data
on relative survival among
persons that had bladder cancer
forO, 1,2,3,4, 5,6, 7, 8, 9, and
Proposed PFAS Rule Economic Analysis
H-8
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-2: Summary of Data Sources Used in Cancer Lifetime Risk Models
Data Element
Modeled Variability
Data Source
Notes
Bladder (Invasive & In
Situ) Cancer
10 years. For disease durations
longer than 10 years EPA
applied 10-year relative survival
rates.
Relative kidney
cancer survival by
cancer stage
Age at diagnosis: 1-year
groups (ages 0 to 100)
Sex: males, females
Duration: 1-year groups
(durations 0 to 100 years)
Cancer stage: localized,
regional, distant,
unstaged
Cancer type: Kidney
Cancer
Race/ethnicity: All, non-
Hispanic White, non-
Hispanic Black, Hispanic,
non-Hispanic Other
SEER 18 relative kidney
cancer survival by
race/ethnicity, age at
diagnosis, sex, cancer stage
and duration with
diagnosis for 2000-2017
(Surveillance Research
Program - National Cancer
Institute, 2020a)
Distinct SEER 18 data were
available for ages at diagnosis
0-14, 15-39, 40-64, 65-74, 75+.
EPA assumed that the same
cancer relative survival patterns
apply to all ages within each age
group. EPA assumed that
non-Hispanic Black data can be
approximated by Black data.
EPA assumed that non-Hispanic
Other data can be approximated
by all race data. SEER 18
contained data on relative
survival among persons that had
kidney cancer for 0, 1, 2, 3, 4, 5,
6, 7, 8, 9, and 10 years. For
disease durations longer than
10 years EPA applied 10-year
relative survival rates.
Abbreviations: CDC - Centers for Disease Control and Prevention; EPA - U.S. Environmental Protection Agency; IR -
Incidence Ratio; NCHS - National Center for Health Statistics; SEER - Surveillance, Epidemiology, and End Results.
H.3 Baseline Kidney Cancer Statistics
Table H-3 provides baseline kidney cancer incidence data used in the life-table model. Kidney
cancer incidence rates per 100,000 range from 0.25 to 44 for females and from 0.16 to 96 for
males. Kidney cancer incidence rates are highest for men in their 60s, 70s, and 80s, ranging from
62 per 100,000 to 96 per 100,000. Localized kidney cancers comprise 37%-84% of all kidney
cancer incidence, whereas regional kidney cancers comprise 8.0%-34%, distant kidney cancers
comprise 6.0%-26%, and unstaged kidney cancers comprise 1,7%-l 1% of all kidney cancer
incidence. Table H-4 provides baseline kidney cancer incidence data by race/ethnicity used in
the life-table model.
Proposed PFAS Rule Economic Analysis
H-9
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-3: Summary of Baseline Kidney Cancer Incidence Data Used in the Model
Females
Males
Percent of Incidence in Stage
Percent of Incidence in
Stage
Age
Incidence
per 100K
Localized
Regional
Distant
Unstaged
Incidence per
100K
Localized
Regional
Distant
Unstaged
<1
1.6
37
34
26
3.1
1.9
43
33
21
3.3
1-4
2.0
37
34
26
3.1
1.8
43
33
21
3.3
5-9
0.82
37
34
26
3.1
0.53
43
33
21
3.3
10-14
0.25
37
34
26
3.1
0.18
43
33
21
3.3
15-19
0.27
84
8.0
6.0
1.9
0.16
81
10
7.7
1.7
20-24
0.60
84
8.0
6.0
1.9
0.51
81
10
7.7
1.7
25-29
1.1
84
8.0
6.0
1.9
1.3
81
10
7.7
1.7
30-34
2.7
84
8.0
6.0
1.9
3.5
81
10
7.7
1.7
35-39
4.7
84
8.0
6.0
1.9
7.2
81
10
7.7
1.7
40-44
7.8
77
11
10
1.8
14
70
14
13
2.1
45-49
11
77
11
10
1.8
22
70
14
13
2.1
50-54
16
77
11
10
1.8
33
70
14
13
2.1
55-59
22
77
11
10
1.8
47
70
14
13
2.1
60-64
29
77
11
10
1.8
62
70
14
13
2.1
65-69
37
71
14
13
2.9
81
67
16
14
3.2
70-74
41
71
14
13
2.9
91
67
16
14
3.2
75-79
44
59
12
17
11
96
57
16
17
9.3
80-84
40
59
12
17
11
84
57
16
17
9.3
85+
33
59
12
17
11
68
57
16
17
9.3
Proposed PFAS Rule Economic Analysis
H-10
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table H-4: Summary of Race/Ethnicity-Specific Baseline Kidney Cancer Incidence Data
Used in the Model
Age
Females
Incidence
per 100K
Percent of Incidence in Stage
13
(J
©
-J
es
C
©
'5x
$
c
es
©X
es
C
P
Males
Incidence
per 100K
Percent of Incidence in
Stage
13
u
o
-J
a
s
e
"Si
(2
s
a
<1
1.4
38
33
27
2.5
2.5
40
35
22
1-4
2.2
38
33
27
2.5
2.1
40
35
22
5-9
38
33
27
2.5
0.54
40
35
22
10-14
0.2
38
33
27
2.5
0.19
40
35
22
15-19
0.32
87
7.8
1.7
85
8.9
20-24
0.52
87
7.8
1.7
0.46
85
8.9
25-29
1.2
87
7.8
1.7
1.5
85
8.9
30-34
2.9
87
7.8
1.7
85
8.9
35-39
4.9
87
7.8
1.7
7.7
85
8.9
40-44
76
12
10
1.6
14
70
15
13
45-49
12
76
12
10
1.6
23
70
15
13
50-54
16
76
12
10
1.6
35
70
15
13
55-59
22
76
12
10
1.6
48
70
15
13
60-64
28
76
12
10
1.6
62
70
15
13
65-69
37
70
14
13
2.7
82
66
17
14
70-74
40
70
14
13
2.7
94
66
17
14
75-79
46
58
13
17
11
99
58
16
17
80-84
41
58
13
17
11
89
58
16
17
85+
33
58
13
17
11
72
58
16
17
<1
34
39
23
3.6
40
34
22
1-4
2.4
34
39
23
3.6
1.7
40
34
22
5-9
0.88
34
39
23
3.6
0.58
40
34
22
10-14
34
39
23
3.6
40
34
22
15-19
75
8.3
14
2.7
68
12
17
20-24
0.84
75
8.3
14
2.7
0.78
68
12
17
25-29
1.1
75
8.3
14
2.7
1.5
68
12
17
30-34
2.4
75
8.3
14
2.7
3.4
68
12
17
35-39
3.8
75
8.3
14
2.7
8.1
68
12
17
40-44
7.4
7.9
2.4
15
76
9.5
45-49
11
7.9
2.4
26
76
9.5
50-54
16
7.9
2.4
38
76
9.5
55-59
23
7.9
2.4
54
76
9.5
60-64
38
7.9
2.4
79
76
9.5
65-69
46
78
8.6
10
3.8
95
74
11
70-74
49
78
8.6
10
3.8
94
74
11
75-79
47
67
8.2
14
11
103
63
10
17
80-84
46
67
8.2
14
11
82
63
10
17
85+
37
67
8.2
14
11
61
63
10
17
Proposed PFAS Rule Economic Analysis
H-ll
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table H-4: Summary of Race/Ethnicity-Specific Baseline Kidney Cancer Incidence Data
Used in the Model
Race/Ethnicity
Age
Females
Males
Incidence
per 100K
Percent of Incidence in Stage
Incidence
per 100K
Percent of Incidence in
Stage
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Hispanic
<1
-
35
35
27
3
1.6
50
28
20
1.7
1-4
1.7
35
35
27
3
1.7
50
28
20
1.7
5-9
0.57
35
35
27
3
0.51
50
28
20
1.7
10-14
-
35
35
27
3
-
50
28
20
1.7
15-19
-
84
8.6
5.4
1.7
-
79
11
8.2
2.1
20-24
0.63
84
8.6
5.4
1.7
0.47
79
11
8.2
2.1
25-29
1
84
8.6
5.4
1.7
0.92
79
11
8.2
2.1
30-34
2.8
84
8.6
5.4
1.7
3
79
11
8.2
2.1
35-39
5.9
84
8.6
5.4
1.7
6.4
79
11
8.2
2.1
40-44
9.2
76
12
10
2.1
13
67
15
15
2.4
45-49
13
76
12
10
2.1
20
67
15
15
2.4
50-54
19
76
12
10
2.1
30
67
15
15
2.4
55-59
24
76
12
10
2.1
45
67
15
15
2.4
60-64
34
76
12
10
2.1
62
67
15
15
2.4
65-69
42
69
14
14
2.9
83
66
16
15
3.6
70-74
46
69
14
14
2.9
91
66
16
15
3.6
75-79
45
59
12
17
12
96
54
18
19
9
80-84
39
59
12
17
12
79
54
18
19
9
85+
35
59
12
17
12
70
54
18
19
9
Other
<1
1.6
37
34
26
3.1
1.9
43
33
21
3.3
1-4
2
37
34
26
3.1
1.8
43
33
21
3.3
5-9
0.82
37
34
26
3.1
0.53
43
33
21
3.3
10-14
0.25
37
34
26
3.1
0.18
43
33
21
3.3
15-19
0.27
84
8
6
1.9
0.16
81
10
7.7
1.7
20-24
0.6
84
8
6
1.9
0.51
81
10
7.7
1.7
25-29
1.1
84
8
6
1.9
1.3
81
10
7.7
1.7
30-34
2.7
84
8
6
1.9
3.5
81
10
7.7
1.7
35-39
4.7
84
8
6
1.9
7.2
81
10
7.7
1.7
40-44
7.8
77
11
10
1.8
14
70
14
13
2.1
45-49
11
77
11
10
1.8
22
70
14
13
2.1
50-54
16
77
11
10
1.8
33
70
14
13
2.1
55-59
22
77
11
10
1.8
47
70
14
13
2.1
60-64
29
77
11
10
1.8
62
70
14
13
2.1
65-69
37
71
14
13
2.9
81
67
16
14
3.2
70-74
41
71
14
13
2.9
91
67
16
14
3.2
75-79
44
59
12
17
11
96
57
16
17
9.3
80-84
40
59
12
17
11
84
57
16
17
9.3
85+
33
59
12
17
11
68
57
16
17
9.3
Proposed PFAS Rule Economic Analysis
H-12
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table H-5 shows relative kidney cancer survival rates.52 by sex, age group at diagnosis, cancer
stage, and the number of years post diagnosis. The relative kidney cancer survival ranges from
3.2% to 100%, and generally decreases as the number of years post-diagnosis increases. The
table also shows the absolute survival probability, averaged over the age range for which the
relative survival data were available; these probabilities are a product of general population
survival probability and the relative kidney cancer survival probability by sex, age group at
diagnosis, and the number of years post-diagnosis. The life-table model uses derived absolute
survival probabilities to model all-cause mortality experience in kidney cancer populations for
the baseline scenario and the regulatory alternatives. Table H-6 provides kidney cancer survival
rates by race/ethnicity used in the life-table model. Finally, Table H-7 shows all-cause and
kidney cancer mortality rates used in the life-table model. Kidney cancer deaths represent <1%
of all-cause mortality among females and <2% of all-cause mortality among males. Table H-8
provides all-cause and kidney cancer mortality rates by race/ethnicity used in the life-table
model.
52 Relative kidney cancer survival rate is the probability of being alive K years after diagnosis at age A divided by the general
probability to survive K years for a person alive at age A without such a diagnosis.
Proposed PFAS Rule Economic Analysis
March 2023
H-13
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-5: Summary of Relative and Absolute Kidney Cancer Survival Used in the Model
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by
Stage (Percent)
Absolute Survival (Average) by
Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival (Average) by
Stage (Percent)
¦a
.a
13
u
o
-J
CS
S
o
"Si
C
es
"S
a
¦a
0>
#J3
a
%
s
P
¦a
0>
.a
13
u
o
-J
CS
S
o
"Si
£
C
es
"S
5
¦a
0>
ex
a
%
s
P
¦a
0>
.a
13
u
o
-J
CS
S
o
"Si
£
C
es
"S
Q
¦a
ex
a
%
s
P
¦a
.a
13
o
o
-J
CS
S
o
"Si
£
fi
es
"S
Q
¦a
a
%
s
P
Ages
<15
1 year
99
99
92
100
99
98
91
99
99
99
88
-
99
98
88
-
Ages
<15
2 years
98
97
86
100
98
97
85
99
99
96
79
-
98
95
78
-
Ages
<15
3 years
98
95
83
96
97
94
82
96
97
95
76
-
96
95
75
-
Ages
<15
4 years
97
94
81
92
97
93
81
92
97
95
74
-
96
94
73
-
Ages
<15
5 years
97
93
80
92
96
93
79
92
97
94
73
-
96
93
72
-
Ages
<15
6 years
96
93
79
92
95
93
79
92
96
94
72
-
95
93
71
-
Ages
<15
7 years
95
93
79
87
95
92
79
86
96
94
71
-
95
93
70
-
Ages
<15
8 years
95
93
78
87
95
92
78
86
96
94
70
-
95
93
69
-
Ages
<15
9 years
95
93
78
87
95
92
78
86
96
92
69
-
95
91
68
-
Ages
<15
10
years
95
93
78
87
95
92
78
86
96
92
69
-
95
90
68
-
Ages
15-39
1 year
99
93
50
90
99
92
49
89
99
92
42
91
97
90
41
89
Ages
15-39
2 years
99
85
32
83
98
84
31
82
99
85
27
84
97
83
26
83
Ages
15-39
3 years
98
80
24
77
97
79
24
76
98
78
20
83
96
76
19
81
Ages
15-39
4 years
98
75
21
77
97
74
21
76
98
74
15
83
95
72
14
81
Proposed PFAS Rule Economic Analysis
H-14
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-5: Summary of Relative and Absolute Kidney Cancer Survival Used in the Model
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by
Stage (Percent)
Absolute Survival (Average) by
Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival (Average) by
Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages
15-39
5 years
97
73
16
77
96
72
16
76
97
71
12
79
94
69
12
77
Ages
15-39
6 years
97
72
15
77
96
71
15
76
96
69
10
72
93
67
10
70
Ages
15-39
7 years
97
71
14
77
95
70
14
76
95
68
9
69
92
65
9
67
Ages
15-39
8 years
96
70
13
77
95
69
13
76
95
66
8
66
92
64
7
64
Ages
15-39
9 years
96
69
13
77
94
68
12
76
94
65
8
66
91
62
7
63
Ages
15-39
10
years
95
69
13
77
93
68
12
76
94
65
8
66
90
62
7
63
Ages
40-64
1 year
99
91
43
73
94
87
40
70
99
92
46
78
90
84
42
71
Ages
40-64
2 years
98
85
28
67
92
80
26
63
97
86
31
69
89
78
28
63
Ages
40-64
3 years
97
80
21
64
91
75
19
60
96
81
23
64
87
73
20
58
Ages
40-64
4 years
96
77
17
61
89
72
15
57
95
77
18
61
85
69
16
54
Ages
40-64
5 years
95
74
14
60
88
69
13
55
94
74
14
58
83
65
13
51
Ages
40-64
6 years
94
71
12
56
87
66
11
52
92
71
12
55
81
62
11
48
Ages
40-64
7 years
93
69
11
55
85
63
10
50
91
68
11
52
79
58
9
45
Ages
40-64
8 years
92
66
10
52
83
60
9
47
90
65
9
50
77
55
8
43
Proposed PFAS Rule Economic Analysis
H-15
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-5: Summary of Relative and Absolute Kidney Cancer Survival Used in the Model
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by
Stage (Percent)
Absolute Survival (Average) by
Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival (Average) by
Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages
40-64
9 years
91
64
9
50
82
57
8
45
89
63
9
48
75
53
7
40
Ages
40-64
10
years
90
63
8
50
80
56
7
44
87
60
8
45
72
50
6
38
Ages
65-74
1 year
98
89
38
66
90
82
35
61
98
90
41
67
87
80
37
60
Ages
65-74
2 years
97
82
24
58
88
75
22
53
97
84
26
60
84
73
23
52
Ages
65-74
3 years
95
76
17
53
85
68
16
47
95
78
19
54
80
66
16
45
Ages
65-74
4 years
94
73
14
49
82
64
12
43
94
74
15
48
77
60
13
39
Ages
65-74
5 years
92
69
11
47
79
59
9
40
92
70
12
44
73
55
10
35
Ages
65-74
6 years
90
66
10
46
75
55
8
38
91
67
10
42
69
52
8
32
Ages
65-74
7 years
88
63
8
44
72
51
7
36
89
65
9
37
65
48
7
27
Ages
65-74
8 years
87
61
8
39
68
48
6
31
87
63
8
37
61
44
6
26
Ages
65-74
9 years
85
57
7
35
65
43
5
27
86
61
8
34
58
41
5
23
Ages
65-74
10
years
83
53
6
34
60
39
5
25
85
57
7
32
54
37
4
20
Ages
75+
1 year
92
78
22
49
47
40
11
25
94
83
28
52
46
41
14
26
Ages
75+
2 years
91
71
12
38
46
35
6
19
93
77
17
45
44
37
8
21
Proposed PFAS Rule Economic Analysis
H-16
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-5: Summary of Relative and Absolute Kidney Cancer Survival Used in the Model
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by
Stage (Percent)
Absolute Survival (Average) by
Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival (Average) by
Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages
75+
3 years
89
66
9
32
43
32
5
16
92
74
12
38
42
34
5
17
Ages
75+
4 years
88
61
7
29
41
29
4
13
89
70
9
32
39
31
4
14
Ages
75+
5 years
86
57
6
25
39
26
3
11
88
67
7
27
36
28
3
11
Ages
75+
6 years
84
54
5
24
36
24
2
10
87
62
6
23
34
24
2
9
Ages
75+
7 years
81
51
5
22
34
21
2
9
85
60
6
20
31
22
2
7
Ages
75+
8 years
78
50
5
19
31
20
2
8
82
57
5
19
28
20
2
7
Ages
75+
9 years
74
47
4
18
28
18
1
7
81
55
4
17
26
17
1
5
Ages
75+
10
years
72
42
3
18
25
15
1
6
79
52
4
16
23
15
1
5
Proposed PFAS Rule Economic Analysis
H-17
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
%
0>
B
Females
Males
"8
-=
*45
o
s
ex
a
5
"ea
H
S.
P
1
!£
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
cs
a>
ex
"o
u.
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages <15
1 year
100
98
95
-
99
98
94
-
99
99
92
-
98
99
92
-
Ages <15
2 years
99
98
90
-
98
98
90
-
99
95
88
-
98
95
88
-
Ages <15
3 years
98
94
85
-
98
94
85
-
96
95
85
-
95
95
85
-
Ages <15
4 years
98
94
85
-
97
93
85
-
96
95
84
-
95
95
84
-
Ages <15
5 years
98
93
83
-
97
92
82
-
96
94
84
-
95
94
83
-
Ages <15
6 years
98
93
83
-
97
92
82
-
96
94
83
-
95
94
82
-
Ages <15
7 years
97
93
83
-
96
92
82
-
96
93
82
-
95
92
81
-
Ages <15
8 years
97
93
83
-
96
92
82
-
96
93
82
-
95
92
81
-
£
Ages <15
9 years
97
93
83
-
96
92
82
-
96
91
82
-
95
90
81
-
C
es
o.
Ages <15
10
years
97
93
83
-
96
92
82
-
96
91
82
-
95
90
81
-
M
=
o
Z
Ages 15-
39
1 year
100
97
58
-
99
96
58
-
99
91
52
96
97
89
51
94
Ages 15-
39
2 years
99
91
38
-
98
90
38
-
99
87
33
84
97
85
33
82
Ages 15-
39
3 years
99
85
27
-
98
84
27
-
99
83
25
84
96
81
24
82
Ages 15-
39
4 years
99
82
21
-
97
81
21
-
98
78
18
84
96
76
18
82
Ages 15-
39
5 years
98
80
18
-
97
79
18
-
97
77
14
84
95
75
14
81
Ages 15-
39
6 years
98
77
18
-
96
76
18
-
97
75
13
79
94
73
13
77
Proposed PFAS Rule Economic Analysis
H-18
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 15-
39
7 years
98
76
17
-
96
74
16
-
96
73
10
79
93
70
10
77
Ages 15-
39
8 years
97
74
17
-
96
73
16
-
96
72
7.2
79
92
69
7
77
Ages 15-
39
9 years
97
74
17
-
95
73
16
-
95
72
7.2
79
91
69
7
76
Ages 15-
39
10
years
96
74
17
-
94
73
16
-
94
72
7.2
79
91
69
6.9
76
Ages 40-
64
1 year
99
92
44
71
94
87
42
67
99
93
47
77
91
85
43
70
Ages 40-
64
2 years
98
85
28
65
93
80
26
61
98
87
32
69
89
79
29
63
Ages 40-
64
3 years
97
80
22
63
91
75
20
59
96
82
24
65
87
74
21
58
Ages 40-
64
4 years
96
77
17
61
90
72
16
57
95
78
18
61
85
70
16
54
Ages 40-
64
5 years
96
74
15
60
88
69
14
55
94
75
15
57
83
66
13
50
Ages 40-
64
6 years
95
71
13
57
87
65
12
52
93
72
13
54
81
63
11
48
Ages 40-
64
7 years
94
69
11
55
86
62
10
50
92
69
11
52
79
59
10
45
Ages 40-
64
8 years
93
66
10
51
84
59
8.6
46
91
67
10
49
78
57
8.2
42
Ages 40-
64
9 years
92
64
8.6
51
82
57
7.7
45
90
64
OO
00
49
76
54
7.4
41
Ages 40-
64
10
years
91
63
8.1
50
80
56
7.2
44
88
61
7.9
46
73
51
6.5
38
Proposed PFAS Rule Economic Analysis
H-19
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 65-
74
1 year
98
89
38
65
91
83
35
60
98
91
42
65
87
81
37
58
Ages 65-
74
2 years
97
82
24
58
88
75
22
52
97
85
26
59
84
73
23
51
Ages 65-
74
3 years
96
77
18
50
86
69
16
45
96
79
20
52
81
67
17
44
Ages 65-
74
4 years
95
75
14
46
83
65
13
40
94
75
15
47
78
61
13
38
Ages 65-
74
5 years
93
70
11
44
79
60
9.4
38
93
71
13
44
74
57
10
35
Ages 65-
74
6 years
91
67
9.3
42
76
56
7.7
35
91
69
11
43
70
53
8.2
33
Ages 65-
74
7 years
89
64
7.9
39
72
52
6.4
32
89
67
9.2
39
66
49
6.7
29
Ages 65-
74
8
years
87
61
7.2
36
68
48
5.6
28
87
65
8.5
38
62
46
6
27
Ages 65-
74
9 years
85
57
6.4
34
65
43
4.9
26
86
63
7.9
35
58
43
5.4
24
Ages 65-
74
10
years
82
54
6
33
60
39
4.4
24
85
61
6.9
31
55
39
4.4
20
Ages 75+
1 year
92
79
21
47
47
40
11
24
94
83
28
52
47
41
14
26
Ages 75+
2 years
92
72
12
37
46
36
5.9
18
94
77
17
45
44
37
8.2
21
Ages 75+
3 years
90
67
9
31
44
32
4.3
15
93
74
12
38
42
34
5.4
17
Ages 75+
4 years
89
63
6.9
28
42
29
3.2
13
91
71
9
32
39
31
3.9
14
Ages 75+
5 years
87
59
5.2
24
39
27
2.3
11
89
69
7.3
27
37
29
3
11
Proposed PFAS Rule Economic Analysis
H-20
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 75+
6 years
85
56
4.2
23
37
24
1.8
10
89
64
6.4
23
35
25
2.5
8.9
Ages 75+
7 years
84
54
4.1
22
35
22
1.7
9.1
86
61
6.1
21
32
22
2.2
7.7
Ages 75+
8 years
82
52
4.1
19
32
21
1.6
7.6
84
58
5.9
20
28
20
2
7
Ages 75+
9 years
77
49
3.1
17
29
18
1.2
6.4
83
56
4.6
17
26
18
1.4
5.5
Ages 75+
10
years
75
44
2.9
17
26
15
1
6
82
55
3.8
16
24
16
1.1
4.7
Non-Hispanic Black
Ages <15
1 year
99
99
92
-
97
97
91
-
99
96
81
-
97
95
80
-
Ages <15
2
years
99
96
88
-
97
95
87
-
99
94
69
-
97
93
68
-
Ages <15
3 years
97
91
86
-
96
90
85
-
99
94
64
-
97
93
63
-
Ages <15
4 years
95
89
81
-
94
88
80
-
99
94
64
-
97
93
63
-
Ages <15
5 years
91
89
78
-
90
88
77
-
99
92
64
-
97
90
63
-
Ages <15
6 years
91
89
78
-
90
88
77
-
97
92
64
-
95
90
62
-
Ages <15
7 years
91
89
78
-
90
88
77
-
97
92
64
-
95
90
62
-
Ages <15
8 years
91
89
78
-
90
88
77
-
97
92
59
-
95
90
58
-
Ages <15
9 years
91
89
78
-
90
88
77
-
97
92
59
-
95
90
58
-
Ages <15
10
years
91
89
78
-
90
88
77
-
97
92
59
-
94
90
58
-
Ages 15-
39
1 year
98
83
34
-
97
81
34
-
96
86
29
-
93
84
28
-
Proposed PFAS Rule Economic Analysis
H-21
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 15-
39
2 years
98
77
20
-
96
76
20
-
95
70
15
-
92
67
15
-
Ages 15-
39
3 years
96
74
16
-
95
73
16
-
93
57
12
-
90
55
12
-
Ages 15-
39
4 years
95
70
14
-
94
69
14
-
92
51
9.4
-
89
49
9
-
Ages 15-
39
5 years
95
70
10
-
93
69
10
-
91
47
7.8
-
88
45
7.5
-
Ages 15-
39
6 years
94
70
10
-
93
69
10
-
90
41
5.9
-
86
40
5.6
-
Ages 15-
39
7 years
93
70
10
-
91
69
10
-
89
41
5.9
-
85
39
5.6
-
Ages 15-
39
8 years
92
70
10
-
90
69
10
-
89
41
5.9
-
84
39
5.6
-
Ages 15-
39
9 years
92
70
10
-
90
68
10
-
87
37
5.9
-
82
35
5.6
-
Ages 15-
39
10
years
90
70
10
-
88
68
10
-
87
37
5.9
-
82
35
5.6
-
Ages 40-
64
1 year
98
87
33
71
91
81
31
66
98
83
33
79
86
73
29
69
Ages 40-
64
2 years
96
78
23
64
88
71
21
58
96
77
19
67
84
67
17
58
Ages 40-
64
3 years
95
72
16
59
86
66
14
53
95
70
13
62
81
60
11
53
Ages 40-
64
4 years
93
68
12
53
84
62
11
47
93
66
8.6
57
79
56
7.3
48
Ages 40-
64
5 years
92
66
11
50
82
59
9.4
45
92
64
6.8
56
76
53
5.7
47
Proposed PFAS Rule Economic Analysis
H-22
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 40-
64
6
years
91
62
9.5
48
80
55
8.4
42
90
60
6.2
54
74
50
5.1
44
Ages 40-
64
7 years
90
59
9.5
48
78
52
8.3
42
89
57
6
53
71
46
4.8
43
Ages 40-
64
8 years
89
57
9
48
77
49
7.8
41
87
51
5.7
52
69
40
4.5
41
Ages 40-
64
9 years
87
54
9
41
74
46
7.7
35
86
49
4.9
48
67
38
3.8
37
Ages 40-
64
10
years
87
52
9
41
73
44
7.6
35
84
48
4.9
43
64
37
3.7
32
Ages 65-
74
1 year
96
80
34
70
87
72
31
63
97
82
32
78
82
69
27
66
Ages 65-
74
2 years
95
74
21
58
83
65
19
51
95
76
20
70
78
62
16
57
Ages 65-
74
3 years
92
66
14
54
79
57
12
46
94
68
13
59
74
54
10
47
Ages 65-
74
4 years
90
58
10
52
75
49
8.7
43
92
64
10
56
69
48
7.7
42
Ages 65-
74
5 years
88
57
8.2
52
72
46
6.6
42
92
59
7.9
51
66
43
5.7
37
Ages 65-
74
6 years
86
56
7.4
52
68
44
5.9
41
91
59
6.2
37
63
41
4.3
25
Ages 65-
74
7 years
84
56
5.6
52
64
43
4.3
39
90
57
5.9
31
59
38
3.9
21
Ages 65-
74
8 years
83
56
5.6
35
61
41
4.2
26
88
52
5.2
28
55
32
3.3
18
Ages 65-
74
9 years
80
50
5.6
27
57
35
4
19
87
48
3.7
28
51
28
2.2
17
Proposed PFAS Rule Economic Analysis
H-23
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 65-
74
10
years
80
47
5.6
27
54
32
3.8
18
85
48
2
28
47
27
1.1
16
Ages 75+
1 year
90
65
20
58
47
34
11
30
87
73
27
49
43
36
13
24
Ages 75+
2 years
88
60
14
41
44
30
6.9
20
87
59
19
43
41
27
00
00
20
Ages 75+
3 years
85
53
10
32
41
26
4.8
15
87
54
10
37
38
24
4.5
17
Ages 75+
4 years
83
47
8.9
29
39
22
4.2
14
82
48
8.4
27
34
20
3.5
11
Ages 75+
5 years
80
43
8.3
24
36
19
3.7
11
80
41
6.4
27
32
16
2.5
11
Ages 75+
6 years
75
36
8.3
21
32
16
3.6
9
78
40
6.4
24
29
15
2.4
00
00
Ages 75+
7 years
69
35
8.3
19
28
14
3.4
7.9
73
38
5.2
14
25
13
1.8
4.7
Ages 75+
8 years
64
35
8.3
19
25
13
3.2
7.5
71
38
3.7
14
22
12
1.2
4.3
Ages 75+
9
years
61
31
8.3
19
22
11
3
7.1
70
38
3.7
-
20
11
1.1
-
Ages 75+
10
years
60
30
4.8
19
20
10
1.6
6.7
70
36
3.7
-
19
9.4
1
-
Hispanic
Ages <15
1 year
98
99
90
-
98
99
89
-
100
100
85
-
99
99
84
-
Ages <15
2 years
98
97
79
-
98
97
78
-
98
98
69
-
98
98
68
-
Ages <15
3 years
98
97
77
-
98
97
77
-
98
98
67
-
97
98
66
-
Ages <15
4 years
98
96
74
-
98
95
73
-
96
98
60
-
96
98
60
-
Ages <15
5 years
98
96
74
-
98
95
73
-
95
98
58
-
94
98
57
-
Ages <15
6 years
97
96
72
-
96
95
71
-
93
98
58
-
93
98
57
-
Ages <15
7 years
97
94
72
-
96
93
71
-
93
98
58
-
93
98
57
-
Ages <15
8 years
97
94
72
-
96
93
71
-
93
98
58
-
92
98
57
-
Proposed PFAS Rule Economic Analysis
H-24
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages <15
9 years
97
94
72
-
96
93
71
-
93
95
58
-
92
94
57
-
Ages <15
10
years
97
94
72
-
96
93
71
-
93
95
58
-
92
94
57
-
Ages 15-
39
1 year
99
89
53
-
99
88
53
-
99
93
49
-
98
92
48
-
Ages 15-
39
2 years
99
79
34
-
98
78
33
-
99
86
35
-
98
85
34
-
Ages 15-
39
3 years
98
72
23
-
97
71
23
-
99
74
25
-
98
73
25
-
Ages 15-
39
4 years
98
66
23
-
97
65
23
-
99
73
20
-
97
72
20
-
Ages 15-
39
5 years
98
66
14
-
97
65
14
-
98
71
19
-
96
70
18
-
Ages 15-
39
6 years
97
66
11
-
96
65
11
-
97
70
15
-
95
68
14
-
Ages 15-
39
7 years
96
66
11
-
95
65
11
-
96
70
15
-
94
68
14
-
Ages 15-
39
8 years
96
66
11
-
95
65
11
-
96
64
15
-
94
62
14
-
Ages 15-
39
9 years
96
66
11
-
95
65
11
-
96
60
15
-
93
58
14
-
Ages 15-
39
10
years
96
66
11
-
95
65
11
-
96
60
15
-
93
58
14
-
Ages 40-
64
1 year
99
91
43
79
95
87
42
76
98
92
46
77
92
86
43
72
Ages 40-
64
2 years
98
86
29
75
94
82
28
72
96
86
31
66
90
80
29
61
Proposed PFAS Rule Economic Analysis
H-25
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 40-
64
3 years
97
82
21
70
93
78
20
67
95
82
24
60
88
76
22
56
Ages 40-
64
4 years
96
80
18
65
91
76
17
62
93
78
19
56
85
72
17
51
Ages 40-
64
5 years
94
78
16
63
89
74
15
60
92
73
16
52
83
67
14
47
Ages 40-
64
6 years
94
75
13
58
88
71
12
55
89
71
13
50
81
64
12
45
Ages 40-
64
7 years
92
70
12
58
87
66
11
54
88
67
11
45
79
60
10
40
Ages 40-
64
8 years
91
68
11
58
85
64
10
54
86
65
10
44
76
57
8.5
39
Ages 40-
64
9 years
90
66
10
54
83
61
9.1
50
86
61
8.9
42
75
53
7.8
37
Ages 40-
64
10
years
89
66
8
54
81
60
7.4
50
83
59
8.3
42
72
51
7.1
37
Ages 65-
74
1 year
98
90
37
62
93
85
35
59
97
92
40
66
88
84
37
60
Ages 65-
74
2 years
97
86
22
53
90
81
21
50
95
85
25
55
85
76
23
49
Ages 65-
74
3 years
95
77
18
53
88
71
16
49
93
78
18
51
81
68
16
45
Ages 65-
74
4 years
94
75
13
49
85
68
12
44
92
71
15
45
79
61
13
38
Ages 65-
74
5 years
93
74
11
44
83
66
10
39
90
65
12
39
75
54
10
32
Ages 65-
74
6 years
91
73
10
44
80
64
9.1
38
88
62
11
32
71
50
8.5
26
Proposed PFAS Rule Economic Analysis
H-26
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 65-
74
7
years
89
69
10
44
76
59
8.2
38
87
59
10
26
68
46
7.8
20
Ages 65-
74
8 years
89
67
10
44
74
56
8
37
84
56
10
26
64
42
7.6
20
Ages 65-
74
9 years
87
64
10
35
71
52
7.8
28
83
54
10
25
61
40
7.3
18
Ages 65-
74
10
years
87
58
7.6
27
69
46
6.1
21
80
45
8.9
25
56
32
6.2
17
Ages 75+
1 year
93
78
25
45
50
42
13
24
93
86
28
43
47
44
14
22
Ages 75+
2 years
90
72
13
35
48
38
6.8
18
91
78
19
32
45
39
9.3
16
Ages 75+
3 years
89
67
7.8
31
46
35
4
16
89
73
15
27
42
35
7.4
13
Ages 75+
4 years
85
60
5.8
25
43
30
2.9
13
86
67
13
20
40
31
6.1
9.3
Ages 75+
5 years
82
56
4.5
21
41
27
2.2
10
83
61
10
16
37
27
4.4
7
Ages 75+
6 years
79
55
3.6
20
38
26
1.7
9.5
82
56
7.3
14
35
24
3.1
6
Ages 75+
7 years
74
47
3.6
13
34
22
1.7
6.1
80
52
6.1
14
32
21
2.5
5.7
Ages 75+
8 years
68
44
3.6
11
31
20
1.6
5.1
75
52
5
10
29
20
1.9
3.7
Ages 75+
9 years
65
40
2.2
10
28
17
1
4.2
73
47
5
10
26
17
1.8
3.5
Ages 75+
10
years
63
33
2.2
5.2
26
14
0.9
2.1
68
43
0
10
23
14
0
3.2
Other
Ages <15
1 year
99
99
92
100
99
98
91
99
99
99
88
-
99
98
88
-
Ages <15
2 years
98
97
86
100
98
97
85
99
99
96
79
-
98
95
78
-
Ages <15
3 years
98
95
83
96
97
94
82
96
97
95
76
-
96
95
75
-
Ages <15
4 years
97
94
81
92
97
93
81
92
97
95
74
-
96
94
73
-
Proposed PFAS Rule Economic Analysis
H-27
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages <15
5 years
97
93
80
92
96
93
79
92
97
94
73
-
96
93
72
-
Ages <15
6 years
96
93
79
92
95
93
79
92
96
94
72
-
95
93
71
-
Ages <15
7 years
95
93
79
87
95
92
79
86
96
94
71
-
95
93
70
-
Ages <15
8 years
95
93
78
87
95
92
78
86
96
94
70
-
95
93
69
-
Ages <15
9 years
95
93
78
87
95
92
78
86
96
92
69
-
95
91
68
-
Ages <15
10
years
95
93
78
87
95
92
78
86
96
92
69
-
95
90
68
-
Ages 15-
39
1 year
99
93
50
90
99
92
49
89
99
92
42
91
97
90
41
89
Ages 15-
39
2 years
99
85
32
83
98
84
31
82
99
85
27
84
97
83
26
83
Ages 15-
39
3 years
98
80
24
77
97
79
24
76
98
78
20
83
96
76
19
81
Ages 15-
39
4 years
98
75
21
77
97
74
21
76
98
74
15
83
95
72
14
81
Ages 15-
39
5 years
97
73
16
77
96
72
16
76
97
71
12
79
94
69
12
77
Ages 15-
39
6 years
97
72
15
77
96
71
15
76
96
69
10
72
93
67
10
70
Ages 15-
39
7 years
97
71
14
77
95
70
14
76
95
68
8.9
69
92
65
8.7
67
Ages 15-
39
8
years
96
70
13
77
95
69
13
76
95
66
7.7
66
92
64
7.4
64
Ages 15-
39
9 years
96
69
13
77
94
68
12
76
94
65
7.7
66
91
62
7.4
63
Proposed PFAS Rule Economic Analysis
H-28
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 15-
39
10
years
95
69
13
77
93
68
12
76
94
65
7.7
66
90
62
7.4
63
Ages 40-
64
1 year
99
91
43
73
94
87
40
70
99
92
46
78
90
84
42
71
Ages 40-
64
2 years
98
85
28
67
92
80
26
63
97
86
31
69
89
78
28
63
Ages 40-
64
3 years
97
80
21
64
91
75
19
60
96
81
23
64
87
73
20
58
Ages 40-
64
4 years
96
77
17
61
89
72
15
57
95
77
18
61
85
69
16
54
Ages 40-
64
5 years
95
74
14
60
88
69
13
55
94
74
14
58
83
65
13
51
Ages 40-
64
6 years
94
71
12
56
87
66
11
52
92
71
12
55
81
62
11
48
Ages 40-
64
7 years
93
69
11
55
85
63
10
50
91
68
11
52
79
58
9.2
45
Ages 40-
64
8 years
92
66
10
52
83
60
8.7
47
90
65
9.3
50
77
55
7.9
43
Ages 40-
64
9 years
91
64
8.6
50
82
57
7.7
45
89
63
8.6
48
75
53
7.2
40
Ages 40-
64
10
years
90
63
8.1
50
80
56
7.2
44
87
60
7.7
45
72
50
6.4
38
Ages 65-
74
1 year
98
89
38
66
90
82
35
61
98
90
41
67
87
80
37
60
Ages 65-
74
2 years
97
82
24
58
88
75
22
53
97
84
26
60
84
73
23
52
Ages 65-
74
3 years
95
76
17
53
85
68
16
47
95
78
19
54
80
66
16
45
Proposed PFAS Rule Economic Analysis
H-29
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
Race/Ethnicity
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 65-
74
4 years
94
73
14
49
82
64
12
43
94
74
15
48
77
60
13
39
Ages 65-
74
5 years
92
69
11
47
79
59
9.4
40
92
70
12
44
73
55
10
35
Ages 65-
74
6 years
90
66
10
46
75
55
8
38
91
67
10
42
69
52
8
32
Ages 65-
74
7 years
88
63
8.1
44
72
51
6.6
36
89
65
9
37
65
48
6.6
27
Ages 65-
74
8 years
87
61
7.7
39
68
48
6
31
87
63
8.5
37
61
44
6
26
Ages 65-
74
9 years
85
57
7
35
65
43
5.3
27
86
61
7.8
34
58
41
5.3
23
Ages 65-
74
10
years
83
53
6.5
34
60
39
4.7
25
85
57
6.8
32
54
37
4.4
20
Ages 75+
1 year
92
78
22
49
47
40
11
25
94
83
28
52
46
41
14
26
Ages 75+
2 years
91
71
12
38
46
35
6.2
19
93
77
17
45
44
37
8.3
21
Ages 75+
3 years
89
66
9.4
32
43
32
4.6
16
92
74
12
38
42
34
5.5
17
Ages 75+
4 years
88
61
7.4
29
41
29
3.5
13
89
70
9.2
32
39
31
4
14
Ages 75+
5 years
86
57
5.9
25
39
26
2.7
11
88
67
7.2
27
36
28
3
11
Ages 75+
6 years
84
54
5
24
36
24
2.2
10
87
62
6.3
23
34
24
2.5
8.9
Ages 75+
7 years
81
51
4.8
22
34
21
2
9
85
60
5.8
20
31
22
2.1
7.5
Ages 75+
8 years
78
50
4.7
19
31
20
1.9
7.7
82
57
5.4
19
28
20
1.9
6.6
Ages 75+
9
years
74
47
3.6
18
28
18
1.4
6.6
81
55
4.3
17
26
17
1.4
5.4
Proposed PFAS Rule Economic Analysis
H-30
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-6: Summary of Race/Ethnicity-Specific Relative and Absolute Kidney Cancer Survival Used in the Model
&
8
Females
Males
"8
-=
*45
©
c
©X
eS
s
"S
H
s.
p
1
!£
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival
(Average)
by Stage (Percent)
cs
©X
"o
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages 75+
10
years
72
42
3.2
18
25
15
1.1
6.2
79
52
3.7
16
23
15
1.1
4.6
Proposed PFAS Rule Economic Analysis
H-31
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-7: Summary of All-Cause and Kidney Cancer Mortality Data Used in the Model
Females
Males
Rate per 100K
Rate per 100K
Age
All-Cause
Kidney Cancer
Percent Kidney Cancer
All-Cause
Kidney Cancer
Percent Kidney Cancer
<1
537
0.04
0.007
646
0.045
0.007
1-4
36
0.06
0.17
44
0.094
0.22
5-9
12
0.11
0.95
15
0.053
0.36
10-14
10
0.05
0.47
12
0.035
0.28
15-19
19
0.05
0.26
34
0.021
0.063
20-24
40
0.03
0.084
112
0.077
0.069
25-29
54
0.08
0.14
142
0.15
0.11
30-34
73
0.11
0.15
159
0.16
0.10
35-39
98
0.15
0.16
185
0.35
0.19
40-44
135
0.31
0.23
229
0.80
0.35
45-49
203
0.7
0.35
323
1.8
0.57
50-54
317
1.3
0.42
508
3.8
0.74
55-59
470
2.6
0.55
784
6.9
0.88
60-64
675
3.8
0.56
1136
11
0.95
65-69
987
6.2
0.63
1593
16
1.02
70-74
1533
8.9
0.58
2304
22
0.94
75-79
2481
13
0.51
3577
30
0.84
80-84
4171
17
0.41
5770
36
0.63
85+
-
-
0.31
-
-
0.49
Proposed PFAS Rule Economic Analysis
H-32
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-8: Summary of Race/Ethnicity-Specific All-Cause and Kidney Cancer Mortality Data Used in the Model
Race/Ethnicity
Age
Females
Males
Rate per 100K
Percent Kidney Cancer
Rate per 100K
Percent Kidney Cancer
All-Cause
Kidney Cancer
All-Cause
Kidney Cancer
Non-Hispanic White
<1
453
0.01
0.002
554
0.04
0.01
1-4
33
0.03
0.08
41
0.08
0.19
5-9
11
0.09
0.8
14
0.06
0.42
10-14
8.5
0.06
0.69
12
0.02
0.21
15-19
19
0.02
0.13
32
0.02
0.05
20-24
41
0.02
0.05
103
0.02
0.02
25-29
57
0.03
0.05
143
0.11
0.08
30-34
80
0.09
0.11
166
0.1
0.06
35-39
106
0.15
0.14
196
0.31
0.16
40-44
143
0.26
0.18
238
0.81
0.34
45-49
211
0.7
0.33
333
1.9
0.58
50-54
324
1.4
0.44
516
3.9
0.76
55-59
472
2.6
0.55
783
7.2
0.91
60-64
668
3.9
0.59
1117
11
0.99
65-69
985
6.5
0.66
1566
17
1.1
70-74
1553
9.1
0.59
2298
22
0.96
75-79
2537
13
0.51
3616
31
0.87
80-84
4282
17
0.41
5894
38
0.64
85+
-
-
0.31
-
-
0.5
Non-Hispanic Black
<1
1042
0.03
0
1249
0.03
0.002
1-4
59
0.09
0.15
70
0.09
0.13
Proposed PFAS Rule Economic Analysis
H-33
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-8: Summary of Race/Ethnicity-Specific All-Cause and Kidney Cancer Mortality Data Used in the Model
Race/Ethnicity
Age
Females
Males
Rate per 100K
Percent Kidney Cancer
Rate per 100K
Percent Kidney Cancer
All-Cause
Kidney Cancer
All-Cause
Kidney Cancer
5-9
18
0.18
1
24
0.03
0.12
10-14
15
0.09
0.59
20
0.09
0.44
15-19
23
0.15
0.63
50
0.03
0.06
20-24
54
0.11
0.2
181
0.32
0.17
25-29
76
0.25
0.32
220
0.39
0.18
30-34
102
0.36
0.35
251
0.57
0.23
35-39
152
0.29
0.19
288
0.81
0.28
40-44
211
0.52
0.25
358
1.2
0.33
45-49
316
0.83
0.26
483
2.4
0.49
50-54
488
1.4
0.29
737
3.9
0.53
55-59
725
3.1
0.42
1175
7.5
0.64
60-64
1049
4.1
0.39
1783
11
0.64
65-69
1457
6.4
0.44
2500
17
0.68
70-74
2065
7.9
0.38
3375
23
0.68
75-79
3073
12
0.4
4751
30
0.64
80-84
4821
17
0.35
6991
34
0.49
85+
-
-
0.31
-
-
0.41
Hispanic
<1
435
0.05
0.01
513
0.07
0.01
1-4
31
0.09
0.29
35
0.1
0.3
5-9
11
0.14
1.3
12
0.05
0.45
10-14
9
0
0
10
0.02
0.18
Proposed PFAS Rule Economic Analysis
H-34
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-8: Summary of Race/Ethnicity-Specific All-Cause and Kidney Cancer Mortality Data Used in the Model
Race/Ethnicity
Age
Females
Males
Rate per 100K
Percent Kidney Cancer
Rate per 100K
Percent Kidney Cancer
All-Cause
Kidney Cancer
All-Cause
Kidney Cancer
15-19
16
0.04
0.25
29
0.02
0.07
20-24
31
0.02
0.07
97
0.07
0.08
25-29
38
0.1
0.28
110
0.09
0.09
30-34
46
0.02
0.05
111
0.12
0.1
35-39
58
0.09
0.15
127
0.27
0.21
40-44
82
0.38
0.46
159
0.67
0.43
45-49
126
0.75
0.6
227
1.7
0.75
50-54
193
1.2
0.6
365
3.5
0.97
55-59
298
2.5
0.85
572
6.2
1.1
60-64
461
3.6
0.79
854
11
1.3
65-69
707
5.6
0.79
1230
16
1.3
70-74
1118
9.8
0.87
1793
23
1.3
75-79
1843
12
0.67
2774
22
0.78
80-84
3174
17
0.54
4463
31
0.69
85+
-
-
0.38
-
-
0.47
Other
<1
409
0.22
0.05
498
0
0
1-4
29
0.14
0.49
39
0.2
0.52
5-9
12
0.07
0.57
14
0.07
0.48
10-14
8.6
0
0
11
0.07
0.63
15-19
15
0.07
0.45
26
0.07
0.26
20-24
30
0
0
72
0
0
Proposed PFAS Rule Economic Analysis
H-35
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-8: Summary of Race/Ethnicity-Specific All-Cause and Kidney Cancer Mortality Data Used in the Model
Race/Ethnicity
Age
Females
Males
Rate per 100K
Percent Kidney Cancer
Rate per 100K
Percent Kidney Cancer
All-Cause
Kidney Cancer
All-Cause
Kidney Cancer
25-29
36
0.05
0.15
81
0.16
0.2
30-34
44
0.05
0.11
85
0.05
0.06
35-39
55
0.15
0.28
103
0.17
0.16
40-44
74
0.16
0.21
132
0.47
0.36
45-49
114
0.39
0.35
192
0.57
0.29
50-54
181
0.68
0.37
310
2.3
0.74
55-59
256
1.4
0.54
451
4.3
0.95
60-64
373
1.7
0.46
641
6.1
0.95
65-69
553
3.6
0.65
940
7.8
0.83
70-74
895
6.1
0.68
1364
11
0.82
75-79
1498
7.3
0.49
2206
18
0.82
80-84
2648
9.6
0.36
3665
15
0.4
85+
-
-
0.25
-
-
0.55
Proposed PFAS Rule Economic Analysis
H-36
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
H.4 Baseline Bladder Cancer Statistics
Table H-9 provides baseline bladder cancer incidence data used in the life-table model. Bladder
cancer incidence rates per 100,000 range from 0.17 to 76 for females and from 0.11 to 357 for
males. Bladder cancer incidence rates are highest for men in their 60s, 70s, and 80s, ranging
from 67 per 100,000 to 357 per 100,000. Localized bladder cancers comprise 66%-90% of all
bladder cancer incidence, whereas regional bladder cancers comprise 4.5%-8.6%, distant bladder
cancers comprise 3.1%-14%, and unstaged bladder cancers comprise 0%-6.8% of all bladder
cancer incidence.
Table H-9: Summary of Baseline Bladder Cancer Incidence Data Used in the Model
Females
Males
Percent of Incidence in Stage
Percent of Incidence in
Stage
Age
Incidence
per 100K
Localized
Regional
Distant
Unstaged
Incidence per
100K
Localized
Regional
Distant
Unstaged
<1
_
77
4.5
14
4.5
_
66
23
11
0
1-4
_
77
4.5
14
4.5
_
66
23
11
0
5-9
_
77
4.5
14
4.5
_
66
23
11
0
10-14
_
77
4.5
14
4.5
_
66
23
11
0
15-19
_
82
8.2
5.1
4.9
0.11
90
4.8
3.1
2.5
20-24
0.17
82
8.2
5.1
4.9
0.30
90
4.8
3.1
2.5
25-29
0.26
82
8.2
5.1
4.9
0.51
90
4.8
3.1
2.5
30-34
0.50
82
8.2
5.1
4.9
1.1
90
4.8
3.1
2.5
35-39
0.89
82
8.2
5.1
4.9
2.1
90
4.8
3.1
2.5
40-44
1.5
83
8.6
6.1
2.7
4.2
85
7.4
4.9
2.5
45-49
2.9
83
8.6
6.1
2.7
00
00
85
7.4
4.9
2.5
50-54
6.6
83
8.6
6.1
2.7
19
85
7.4
4.9
2.5
55-59
11
83
8.6
6.1
2.7
38
85
7.4
4.9
2.5
60-64
18
83
8.6
6.1
2.7
67
85
7.4
4.9
2.5
65-69
29
84
7.9
5.6
2.8
114
86
6.7
4.3
2.9
70-74
43
84
7.9
5.6
2.8
176
86
6.7
4.3
2.9
75-79
58
80
7.1
5.8
6.8
245
85
6.2
4.1
5.2
80-84
71
80
7.1
5.8
6.8
315
85
6.2
4.1
5.2
85+
76
80
7.1
5.8
6.8
357
85
6.2
4.1
5.2
Proposed PFAS Rule Economic Analysis
H-37
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table H-10 shows relative bladder cancer survival rates.53 by sex, age group at diagnosis, cancer
stage, and the number of years post diagnosis. The relative bladder cancer survival ranges from
0% to 100%, and generally decreases as the number of years post-diagnosis increases. The table
also shows the absolute survival probability, averaged over the age range for which the relative
survival data were available; these probabilities are a product of general population survival
probability and the relative bladder cancer survival probability by sex, age group at diagnosis,
and the number of years post-diagnosis. The life-table model uses derived absolute survival
probabilities to model all-cause mortality experience in bladder cancer populations for the
baseline scenario and the regulatory alternative. Finally, Table H-l 1 shows all-cause and bladder
cancer mortality rates used in the life-table model. Bladder cancer deaths <1% of all-cause
mortality among females and <2% of all-cause mortality among males.
53 Relative bladder cancer survival rate is the probability of being alive K years after diagnosis at age A divided by the general
probability to survive K years for a person alive at age A without such a diagnosis.
Proposed PFAS Rule Economic Analysis
March 2023
H-38
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-10: Summary of Relative and Absolute Bladder Cancer Survival Used in the Model
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival (Average) by
Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival (Average) by
Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages
15-39
1
year
98
79
20
90
97
79
20
90
99
85
46
100
97
83
45
98
Ages
15-39
2
years
97
58
4
83
96
57
4
83
99
67
23
97
96
65
22
95
Ages
15-39
3
years
96
47
0
80
95
46
0
79
98
60
14
95
96
58
13
92
Ages
15-39
4
years
95
39
0
80
94
39
0
79
97
58
11
91
95
56
11
89
Ages
15-39
5
years
95
32
0
80
93
32
0
79
96
56
11
91
94
54
11
89
Ages
15-39
6
years
94
28
0
80
93
27
0
79
96
56
9
91
93
54
9
89
Ages
15-39
7
years
94
28
0
80
92
27
0
79
96
56
7
91
93
54
7
88
Ages
15-39
8
years
93
28
0
80
92
27
0
78
95
56
7
91
92
54
7
88
Ages
15-39
9
years
93
28
0
80
91
27
0
78
94
52
5
91
91
51
4
88
Ages
15-39
10
years
93
28
0
80
91
27
0
78
93
52
5
85
90
50
4
82
Ages
40-64
1
year
97
73
34
84
92
69
32
80
98
78
36
85
90
72
33
78
Ages
40-64
2
years
95
53
15
81
90
50
14
76
96
57
16
79
87
52
15
72
Ages
40-64
3
years
94
45
9
77
88
42
9
72
94
48
11
75
85
43
10
67
Ages
40-64
4
years
93
40
7
76
87
37
7
70
93
43
9
73
83
38
8
65
Proposed PFAS Rule Economic Analysis
H-39
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-10: Summary of Relative and Absolute Bladder Cancer Survival Used in the Model
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival (Average) by
Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival (Average) by
Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages
40-64
5
years
92
37
5
74
85
34
5
69
91
40
8
71
81
35
7
63
Ages
40-64
6
years
91
36
5
74
84
33
5
68
90
38
7
68
79
33
7
60
Ages
40-64
7
years
90
34
4
73
82
31
4
66
89
37
7
66
77
32
6
57
Ages
40-64
8
years
89
32
4
71
80
29
4
64
88
36
7
64
75
30
6
54
Ages
40-64
9
years
88
31
4
70
79
28
3
63
87
35
7
61
73
29
6
51
Ages
40-64
10
years
87
31
4
70
77
27
3
62
86
34
7
61
71
28
6
51
Ages
65-74
1
year
95
67
25
72
88
62
24
66
97
74
32
81
86
66
29
72
Ages
65-74
2
years
92
48
11
67
83
44
10
61
94
55
16
75
82
48
13
65
Ages
65-74
3
years
90
38
8
63
80
34
7
57
92
47
11
72
77
39
9
60
Ages
65-74
4
years
88
34
6
60
77
30
5
52
89
42
8
69
73
34
6
56
Ages
65-74
5
years
86
31
5
58
73
26
5
50
88
39
6
66
70
31
5
52
Ages
65-74
6
years
85
28
5
56
71
23
4
47
86
36
6
64
66
27
4
49
Ages
65-74
7
years
84
27
4
54
68
22
3
44
84
34
5
61
62
25
4
45
Ages
65-74
8
years
82
25
4
52
64
20
3
41
82
32
5
57
58
23
4
40
Proposed PFAS Rule Economic Analysis
H-40
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-10: Summary of Relative and Absolute Bladder Cancer Survival Used in the Model
Age at Diagnosis
Follow-Up Time
Females
Males
Relative Survival by Stage
(Percent)
Absolute Survival (Average) by
Stage (Percent)
Relative Survival by Stage
(Percent)
Absolute Survival (Average) by
Stage (Percent)
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Localized
Regional
Distant
Unstaged
Ages
65-74
9
years
81
25
3
51
61
19
2
39
80
30
4
56
54
20
3
38
Ages
65-74
10
years
79
25
3
51
58
18
2
37
79
29
4
56
50
19
3
36
Ages
75+
1
year
86
48
17
39
44
25
9
20
92
60
22
59
45
30
11
29
Ages
75+
2
years
81
36
8
32
40
18
4
16
87
44
10
51
42
21
5
24
Ages
75+
3
years
77
30
6
27
38
15
3
13
84
38
7
45
38
17
3
21
Ages
75+
4
years
76
28
5
24
36
13
2
11
81
35
5
40
35
15
2
17
Ages
75+
5
years
73
26
4
22
33
12
2
10
79
33
5
37
33
14
2
15
Ages
75+
6
years
71
24
4
22
31
11
2
9
76
32
4
34
30
13
2
13
Ages
75+
7
years
69
22
3
20
29
9
1
8
74
29
3
31
27
11
1
11
Ages
75+
8
years
68
21
3
18
27
8
1
7
72
28
3
29
25
10
1
10
Ages
75+
9
years
66
21
2
18
25
8
1
7
70
28
3
26
22
9
1
8
Ages
75+
10
years
65
18
2
18
23
6
1
6
68
28
3
23
20
8
1
7
Proposed PFAS Rule Economic Analysis
H-41
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-ll: Summary of All-Cause and Bladder Cancer Mortality Data Used in the Model
Females
Males
Rate per 100K
Rate per 100K
Age
All-Cause
Bladder Cancer
Percent Bladder Cancer
All-Cause
Bladder Cancer
Percent Bladder Cancer
<1
537
0
0
646
0.0090
0.0014
1-4
36
0
0
44
0
0
5-9
12
0
0
15
0
0
10-14
10
0
0
12
0.0086
0.070
15-19
19
0
0
34
0
0
20-24
40
0.0085
0.021
112
0.012
0.011
25-29
54
0.017
0.030
142
0.020
0.014
30-34
73
0.034
0.046
159
0.046
0.029
35-39
98
0.14
0.14
185
0.19
0.10
40-44
135
0.31
0.23
229
0.52
0.23
45-49
203
0.64
0.31
323
1.4
0.42
50-54
317
1.3
0.40
508
3.1
0.61
55-59
470
2.2
0.48
784
7.1
0.91
60-64
675
4.0
0.60
1136
12
1.1
65-69
987
6.5
0.66
1593
22
1.4
70-74
1533
12
0.77
2304
37
1.6
75-79
2481
22
0.87
3577
70
1.9
80-84
4171
36
0.85
5770
123
2.1
85+
-
-
0.77
-
-
1.9
Proposed PFAS Rule Economic Analysis
H-42
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
H.5 RCC Valuation Data
EPA identified the study selected for use in evaluating potential medical costs avoided as a result
of the PFAS regulatory alternatives, Ambavane et al. (2020), as part of a targeted kidney cancer
valuation literature search. The scope of the search covered cost of illness (COI) and willingness-
to-pay (WTP) literature published in English language peer reviewed sources during 2010-
2021.54 The searches were executed in the Google Scholar article database. EPA reviewed 153
references retrieved by the WTP-oriented searches and the top 348 references retrieved by the
COI-oriented searches.,55
The search did not identify any suitable kidney cancer WTP studies. However, there were seven
additional studies containing COI information. Of those, four were cost-effectiveness studies that
focused only on medication costs. The remaining three studies focused on the overall medical
care costs but had methodological issues that prevented EPA from using them as the basis for
kidney cancer morbidity valuation:
• Hollenbeak et al. (2011) reported 5-year RCC cost estimates based on Medicare data
from early 2000s; however, even after adjusting for medical care price inflation, these
RCC cost estimates were too low relative to the costs reported by more recent cost-
effectiveness studies.
• Bhattacharjee et al. (2017) annual cost estimates were based on the Medical Expenditure
Panel Survey 2002-2011 data for persons experiencing kidney cancer but included
expenditures for conditions other than kidney cancer.
• Mitchell et al. (2020) reported Medicare costs for various first line kidney cancer
treatment types, but not the frequency and duration with which these treatments were
typically applied.
Detailed notes on the 8 studies reviewed by EPA are provided in Table H-12.
54 The query terms used for WTP-oriented and COI-oriented searches are available upon request.
55 EPA applied exclusion-term based automated screening to the raw Google Scholar result sets; exclusion terms are available
upon request. The number of references listed in this document reflect the size of the result sets after the automated screening was
applied. There were 153 references in the WTP-oriented search result set and 1,342 references in the COI-oriented search result
set. The overall budget for manual review was approximately 500 references, with priority given to the WTP-oriented results set.
Therefore, EPA reviewed all 153 references in the WTP-oriented results set and top 348 references in the COI-oriented results
set. The references in the COI-oriented results set were prioritized using Okapi BM25 metric applied to article titles and Google
Scholar ranks.
Proposed PFAS Rule Economic Analysis
H-43
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-12: Studies Reviewed Related to Kidney Cancer Medical Treatment Costs
Study
Reference
Valuation
Target
Focus
Result(s) Type &
Quality
Geographic
Scope & Scale
Population
Datasets
Data
Collection
Year
Methodology and
Other Notes
Ambavane et
al. (2020)
Lifetime
treatment costs
of several
treatment
sequences
(first and
second line
drug costs,
administration
costs, disease
management,
and adverse
effects
management)
Incidence-
based
Accounting for
first and second
line, drug costs +
administration
costs + disease
management costs
per month + single
time AE
management cost
(not accounting for
mean AE
disutility/month) =
$189,594.76/month
+ $48,122; annual
cost = $2.3 million
(without including
monthly disutility).
Dollar values
reported in 2018$.
-26% U.S.;
-35%
Canada/Western
Europe/North
Europe; -39%
rest of world
779,
majority
male and
white with
baseline
median age
of 62 years
Cohort data
from the
CheckMate
214 trial
Not stated
Discrete event
simulation model
estimates lifetime
costs and survival
among patients.
Recent US-based
costs; risk data are
bias toward older
white males and 26%
of trial participants
were from U.S.;
provides costs but
not information on
baseline treatment
frequencies.
Hollenbeak
et al. (2011)
Payments
made by
Medicare for
all-cause
medical
treatments
including
inpatient stays,
emergency
room visits,
outpatient
procedures,
office visits,
home health
visits, durable
medical
Prevalence-
based, by
year since
diagnosis
Mean costs per
patient per month
(PPPM) in the first
year were $3,673
for patients with
RCC. PPPM costs
were higher for
RCC patients with
more advanced
stage (i.e., regional
or distant) disease.
Average
cumulative total
costs for RCC
patients were
$33,605 per patient
USA, individual
scale
4,938
patients
with RCC
and 9,876
non-HMO
noncancer
comparison
group. The
sample was
limited to
non-HMO
patients
aged 65
years or
older who
were
Surveillance,
Epidemiology,
and End
Results
Program
(SEER)-
Medicare
database,
which
combines
tumor registry
data from the
National
Cancer
Institutes
(NCI) SEER
1995-2002
Estimated all-cause
health care costs
associated with RCC
using SEER-
Medicare data. Using
the method of Bang
et al. (2000),
estimated cumulative
costs at 1 and 5 years
by estimating
average costs for
each patient in each
month up to 60
months following
diagnosis. Medicare
population; costs
Proposed PFAS Rule Economic Analysis
H-44
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-12: Studies Reviewed Related to Kidney Cancer Medical Treatment Costs
Study
Reference
Valuation
Target
Focus
Result(s) Type &
Quality
Geographic
Scope & Scale
Population
Datasets
Data
Collection
Year
Methodology and
Other Notes
equipment,
and hospice
care, but
excluding
outpatient
prescription
drugs
in the first year
following
diagnosis and
$59,397 per patient
in the first 5 years
following
diagnosis. Costs
available for first
five years and
separated by stage.
diagnosed
with a first
primary
RCC (SEER
site recode
59, kidney
and renal
pelvis)
between
1995 and
2002
program for
patients who
are covered by
Medicare with
their Medicare
billing records
within 5-years of
diagnosis; data from
2005.
Mitchell et
al. (2020)
Medicare costs
for first-line
and
maintenance
treatment
Cost
accounting-
based
First-line
treatments for
kidney cancer
range from
$30,538 to
$31,190, while
maintenance
treatments range
from $7,722 to
$8,997. These
costs represent the
average monthly
cost of treatment.
USA, individual
scale
Not
specified
Medicare costs
for first-line
and
maintenance
treatments for
cancers with
the highest
incidence in
the US that had
published
National
Comprehensive
Cancer
Network
(NCCN)
Evidence
Blocks as of
December 31,
2018; costs
based on
Medicare
prices from the
January 2019
2018
Calculated Medicare
costs for all first-line
and maintenance
treatments for 30
cancers with the
highest incidence in
the US that had
published National
Comprehensive
Cancer Network
(NCCN) Evidence
Blocks as of
December 31, 2018.
Categorized each
treatment as either
"time-limited" or
"time-unlimited."
For time-unlimited
treatments (all
kidney cancer
treatments fall into
this category),
calculated the
average monthly cost
Proposed PFAS Rule Economic Analysis
H-45
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-12: Studies Reviewed Related to Kidney Cancer Medical Treatment Costs
Study
Reference
Valuation
Target
Focus
Result(s) Type &
Quality
Geographic
Scope & Scale
Population
Datasets
Data
Collection
Year
Methodology and
Other Notes
Medicare ASP
file
of treatment. No
information on
treatment duration.
Bhattachaijee
etal. (2017)
Total
healthcare
expenditure,
which includes
inpatient,
outpatient,
emergency
room,
prescription
drugs, home
health agency,
dental care,
vision care,
and other
expenditures.
The study
included
different
sources of
payment such
as direct
payments from
individuals,
private
insurance,
Medicare,
Prevalence-
based
The annual average
total healthcare
expenditures
($15,078 vs.
$8,182; P<.001)
for adults with
kidney cancer were
significantly higher
compared with
propensity-score-
matched adults
with other forms of
cancer. The
average inpatient
($6755 vs. $1959)
and prescription
drug ($3485 vs.
$1570)
expenditures were
significantly higher
for adults with KC
compared with
matched controls.
Dollar values
reported in 2011$.
USA, individual
scale
Adults aged
21 or older
who did not
die during
the calendar
year of
MEPS data
and had
positive
total
healthcare
expenditures
(N = 541 for
time-
unlimited
treatments,
N = 845 for
time-limited
treatments-
analysis
includes
~30 cancer
types).
Cancer
stage not
specified.
Medical
Expenditure
Panel Survey
2002-2011
Used a retrospective,
cross-sectional,
propensity-score-
matched, case-
control study design
using 2002 to 2011
MEPS data to
determine impacts of
health and functional
status and co-
occurring chronic
conditions.
Developed OLS
regressions on log-
transformed
expenditures for total
and subtypes of
health expenditures.
Calculated
percentage change in
expenditure. Very
small sample of -100
persons; non-
incremental annual
average healthcare
expenditures among
Proposed PFAS Rule Economic Analysis
H-46
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-12: Studies Reviewed Related to Kidney Cancer Medical Treatment Costs
Study
Reference
Valuation
Target
Focus
Result(s) Type &
Quality
Geographic
Scope & Scale
Population
Datasets
Data
Collection
Year
Methodology and
Other Notes
Medicaid,
Workers'
Compensation,
and
miscellaneous
other sources.
All
expenditures
inflated using
medical CPI.
those with RCC that
could include care
for other health
issues; no stage and
no variation by time
since diagnosis;
focus on those with
positive
expenditures.
Wan et al.
(2019)
Compares
cost-
effectiveness
of kidney
cancer
treatments:
nivolumab
plus
ipilimumab vs
sunitinib
Incidence-
based
Provides total cost
of regimen, other
values reported in
ICER/QALY; cost
effectiveness
analysis of two
different
treatments for renal
cell carcinoma
USA, individual
scale
1096
patients
with mRCC
from
clinical trial
modeled to
receive the
drug
CheckMate
214, Centers
for
Medicare &
Medicaid
Services
2018
A Markov model
was developed to
compare the lifetime
cost and
effectiveness of
nivolumab plus
ipilimumab vs
sunitinib in the first-
line treatment of
mRCC using
outcomes data from
the CheckMate 214
phase 3 randomized
clinical trial, which
included 1096
patients with mRCC
(median age, 62
years) and compared
nivolumab plus
ipilimumab vs
sunitinib as first-line
treatment of mRCC.
In the analysis,
Proposed PFAS Rule Economic Analysis
H-47
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-12: Studies Reviewed Related to Kidney Cancer Medical Treatment Costs
Study
Reference
Valuation
Target
Focus
Result(s) Type &
Quality
Geographic
Scope & Scale
Population
Datasets
Data
Collection
Year
Methodology and
Other Notes
patients were
modeled to receive
sunitinib or
nivolumab plus
ipilimumab for 4
doses followed by
nivolumab
monotherapy,
provides costs of
treatment but does
not provide the
frequency with
which these
treatments are
applied in the general
population.
Reinhorn et
al. (2019)
Compares
cost-
effectiveness
of kidney
cancer
treatments:
nivolumab and
ipilimumab
versus
sunitinib
Incidence-
based
Cost effectiveness
analysis of two
different
treatments for renal
cell carcinoma;
study centered on
specific drug cost
and was limited by
data availability
USA, individual
scale
Markov
model-
simulated
population
with each
model cycle
representing
1 month
over a 10-
year time
horizon
CheckMate
214
2017
A Markov model
was developed to
compare the costs
and effectiveness of
nivolumab and
ipilimumab with
those of sunitinib in
the first-line
treatment of
intermediate- to
poor-risk advanced
RCC. Health
outcomes were
measured in life-
years and quality-
adjusted life-years
(QALYs). Drug costs
were based on
Medicare
Proposed PFAS Rule Economic Analysis
H-48
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-12: Studies Reviewed Related to Kidney Cancer Medical Treatment Costs
Study
Reference
Valuation
Target
Focus
Result(s) Type &
Quality
Geographic
Scope & Scale
Population
Datasets
Data
Collection
Year
Methodology and
Other Notes
reimbursement rates
in 2017. Study
extrapolated survival
beyond the trial
closure using
Weibull distribution.
Model robustness
was addressed in
univariable and
probabilistic
sensitivity analyses.
Provides costs of
treatment but does
not provide the
frequency with
which these
treatments are
applied in the general
population
Perrin et al.
(2015)
Compares
cost-
effectiveness
of kidney
cancer
treatments:
everolimus vs
axitinib;
provides costs
per patient
from
simulated data
Incidence-
based
Cost effectiveness
analysis of two
different
treatments for renal
cell carcinoma
USA, individual
scale
Simulated
population
of advanced
RCC
patients
MarketScan
Commercial
Claims and
Encounters and
Medicare
Supplemental
database
2004-2011
A Markov model
was developed to
simulate a cohort of
sunitinib-refractory
advanced RCC
patients and estimate
the cost of treating
patients with
everolimus vs
axitinib. The
following health
states were included:
stable disease
without adverse
events (aEs), stable
disease with aEs,
Proposed PFAS Rule Economic Analysis
H-49
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-12: Studies Reviewed Related to Kidney Cancer Medical Treatment Costs
Study
Reference
Valuation
Target
Focus
Result(s) Type &
Quality
Geographic
Scope & Scale
Population
Datasets
Data
Collection
Year
Methodology and
Other Notes
disease progression
(PD), and death. The
model included the
following resources:
active treatments,
post-progression
treatments, aEs,
physician and nurse
visits, scans and
tests, and palliative
care. Resource
utilization inputs
were derived from a
US claims database
analysis.
Additionally, a 3%
annual discount rate
was applied to costs,
and the robustness of
the model results was
tested by conducting
sensitivity analyses,
including those on
dosing scheme and
post-progression
treatment costs.
Provides costs of
treatment but does
not provide the
frequency with
which these
treatments are
applied in the general
population.
Proposed PFAS Rule Economic Analysis
H-50
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-12: Studies Reviewed Related to Kidney Cancer Medical Treatment Costs
Study
Reference
Valuation
Target
Focus
Result(s) Type &
Quality
Geographic
Scope & Scale
Population
Datasets
Data
Collection
Year
Methodology and
Other Notes
Racsa et al.
(2015)
Compares
cost-
effectiveness
of kidney
cancer
treatments:
two tyrosine
kinase
inhibitors;
provides
original dollar
estimates for
different
medications
Incidence-
based
Cost effectiveness
analysis of two
different
treatments for renal
cell carcinoma
USA, individual
scale
1,438 RCC
patients
aged 19 to
89 years,
with
medical and
pharmacy
insurance
through
commercial
or Medicare
plans
Humana
Research
Database
2009-2012
Study used claims
data to conduct an
observational,
retrospective cohort
study of individuals
aged 19 to 89 years,
with commercial or
Medicare insurance,
advanced RCC, and
at least one
pharmacy claim for
sunitinibor
pazopanib between 1
November 2009 and
31 December 2012.
Treatment
characteristics
(treatment
interruption,
adherence, duration,
and discontinuation),
survival, and costs
were measured up to
12 months. Statistical
models were
adjusted for age,
gender, geographic
region, race, and
RxRisk-Vscore.
Provides costs of
treatment but does
not provide the
frequency with
which these
treatments are
Proposed PFAS Rule Economic Analysis
H-51
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table H-12: Studies Reviewed Related to Kidney Cancer Medical Treatment Costs
Study
Reference
Valuation
Target
Focus
Result(s) Type &
Quality
Geographic
Scope & Scale
Population
Datasets
Data
Collection
Year
Methodology and
Other Notes
applied in the general
population; addresses
a younger
population.
Abbreviations: AE - Adverse Effects; CPI - Consumer Price Index; HMO - Health Maintenance Organization; MEPS - Medical Expenditure Panel Survey; mRCC metastatic
Renal Cell Carcinoma; NCCN - National Comprehensive Cancer Network; NC I- National Cancer Institute; OLS - Ordinary Least Squares; PD - Disease Progression; PPPM
- Per Patient Per Month; QALYs - Quality Adjusted Life Years; RCC - Renal Cell Carcinoma; SEER - Surveillance, Epidemiology, and End Results Program.
Proposed PFAS Rule Economic Analysis
H-52
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix I. Trihalomethane Co-Removal Model
Details and Analysis
1.1 Data Analysis
EPA analyzed Information Collection Rule Treatment Study Database (ICR TSD) data to predict
time-based removal efficacy of total organic carbon (TOC) and four regulated trihalomethanes
(THM4) from pilot and rapid small-scale column tests (RSSCTs). In all, EPA extracted
182 datasets from the ICR TSD database, which included some quarterly RSSCTs and some
long-term pilots. EPA used RSSCT scaling factors identified in the original datasets to scale
predictions to expected full-scale operational time, rather than short duration experimental time.
This appendix focuses on estimates of THM4 production because it forms the basis of potential
reductions in health risks resulting from reducing PFAS levels under various regulatory
scenarios. Note that the same approaches described in this appendix were used to estimate TOC
removal. EPA developed a Python program to standardize the data analysis and produce
graphics. Figure 1-1 shows example data from one study (SystemID 1003, RSSCT) to
demonstrate the approach for estimating THM4 reduction. Each dataset provided influent and
effluent concentrations for TOC and THM4 formation potential for a 10-min empty bed contact
time (EBCT). Most datasets also included 20-min EBCT effluent concentrations. If data was not
available for 20-min EBCT effluent concentrations, then only 10-min EBCT data was included
in the analysis. For all datasets and EBCTs, EPA used a logistic function to estimate the expected
breakthrough curve over time (effluent concentrations vs. time). Since the logistic function is
non-linear, EPA used the Python function scipy.optimize.curvefit to estimate equation
parameters.
Proposed PFAS Rule Economic Analysis
1-1
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
70
60 -
50 -
CI
=1
2 40-
S 30
c
0
u
1 20 H
10 -
0 -
X-
7*7
THM4 Influent
65.87 avg. C,
THM4 lOmin EBCT
THM4 20min EBCT
THM4 10 min Logistic
THM4 20 min Logistic
20
~4Q
60
—i—
80
100
120
140
160
Time (days)
Figure 1-1: Example Breakthrough Curve for THM4
from the ICR Dataset with Logistic Fit Functions Shown.
The logistic function is provided as:
Equation 1-1:
C(t) = Cf(Ae~rt + l)~n+1
where C is effluent concentration, Cf is the final concentration (concentration units), A, r and n
are additional fit parameters and t is time (in days). EPA generated a set of fit parameters for
each of the datasets and EBCTs. The logistic function provides a continuous function throughout
a period and can be used to estimate effective effluent concentrations beyond the original test
period. This assumes that Cf could be estimated effectively and represents the long-term
effective removal after breakthrough (i.e., that an equilibrium removal was achieved). Figure 1-2
shows the projected removal percentage for bed replacement intervals from 30 days (1 month) to
730 days (2 years). Percent removal for each data pair was calculated as:
Equation 1-2:
%Removal = 100 * (1 C^ )
Cinf,avg
where, C(t) is the result of the logistic function over time, and Cinf avg is the average influent
concentration for each species.
Proposed PFAS Rule Economic Analysis 1-2 March 2023
-------
THM4 - Avg. Cone. = 65.87
10 min EBCT
20 min EBCT
0 "T
0
100 200 300 400 500
Replacement Interval (days)
600
700
Figure 1-2: Example Percent Removal Results vs.
Time based on Logistic Plots Shown in Figure 1-1.
Proposed PFAS Rule Economic Analysis
1-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
TOC - 10 min EBCT
27.2±8.8 % removed
TOC - 20 min EBCT
35.5±13.6 % removed
\
THM4 - 10 min EBCT
29.9+15.6 % removed
THM4 - 20 min EBCT
40.0± 18.4 % removed
-i—r
n>
TTTfir
-y~r
t-f j
r-
* ¦ —<—
r
i
y- —
1
-Z
r
/ sr
i
100 200 300 400 500
Replacement Interval (days)
600
700
Figure 1-3: Mean Percentage Removal (Shaded Area ± 1 Standard Deviation)
Proposed PFAS Rule Economic Analysis
1-4
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
The percent removal formula provides a conservative estimate for removal over each EBCT.
EPA assumes that the percent removal at the carbon removal day is the best removal that was
achieved, where breakthrough curves demonstrate that additional removal may be achieved for
earlier portions of the operational carbon life. For longer operational times, this early removal
capacity for each species becomes a diminishingly small percentage of removal percentage.
EPA used the percentage removal at V2 year intervals for V2, 1, 1 V2, and 2 years in the
co-removal benefits analysis. Information about the source water (pre-categorized type from the
ICR, groundwater or surface water) and averages of influent concentrations of TOC, and THM4
were stored with results, which were used during further analyses.
Figure 1-3 represents the mean percentage removal for TOC, THM4 over time with shaded areas
representing mean ±1 standard deviation. Figure 1-4 also shows a probability density function
representation of concentration reduction following treatment after 2 years of carbon operations
(i.e., GAC replacement time). These plots demonstrate the variability in the results.
0.05 -
0.04 -
> 0.03 -
4-1
U)
c
CD
Q
0.02 -
0.01 -
0.00 -
0 50 100 150 200
THM4 (Infl.-Effl.) Concentration (/L/g/l_)
Figure 1-4: Probability Density Function of Concentration Difference
at 2 years of Carbon Life (Subdivided by TOC level).
1.2 Discussion of Other Models
EPA explored another existing model to determine ATHM4 resulting from GAC treatment.
The Water Treatment Plant model uses the ICR TSD data along with other datasets and includes
specific process selection inputs such as GAC units (U.S. EPA, 2001). In contrast with the
logistic model detailed in Section 1.1, the Water Treatment Plant model cannot be run with the
GAC unit in isolation. Within the Water Treatment Plant model, the GAC unit process equation
relies on TOC and ultraviolet
THM4 Density Plot
— All TOC: 10-min (N: 182)
All TOC: 20-min (N: 182)
1-2.0 TOC: 10-min (N
20)
1-2.0 TOC: 20-min (N
20)
2-3.5 TOC: 10-min (N
103)
2-3.5 TOC: 20-min (N
103)
- ¦ - 3.5-5 TOC: 10-min (N
44)
- ¦ - 3.5-5 TOC: 20-min (N
44)
above5 TOC: 10-min (N: 15)
above5 TOC: 20-min (N: 15)
Proposed PFAS Rule Economic Analysis
1-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
absorbance (UVA) changes and does not directly predict THMs. Additional data needed to use
the Water Treatment Plant model include types of chemicals used, dosing concentrations, contact
times, and full process train information, which EPA did not have outside of the DBP ICR for
national scale estimates. Comparing the models, the logistic equations for GAC treatment were
generally in the same form. However, in this analysis EPA fit the THM4 results reported in the
ICR dataset directly whereas the Water Treatment Plant model would need to have simulated all
various treatment trains including GAC to calculate TOC levels (some uncertainty) followed by
a conversion with then another model equation (with more uncertainty) to predict the ATHM4.
While these equations result in the same shape of function to find predictions, the logistic model
approach outlined in Section 1.1 uses a singular step with singular uncertainty that was data
driven.
1.3 THM4 Reduction Results
All systems used free chlorine for the THM4 formation potential experiments in the ICR TSD.
However, the hold time to replicate the distribution system varied based on the typical
disinfectant used in the PWS. Table 1-1 shows the THM4 removal (ATHM4) differences based
on source water type, EBCTs, and disinfectant type of the parent system. Table 1-2 to Table 1-5
shows the ATHM4 differences based on GAC replacement intervals (1/2, 1, 1 V2, and 2 years),
disinfectant type (free chlorine versus chloramine), source water type (ground versus surface
water), and TOC range (1-2.0, 2-3.5, 3.5-5, and above 5 mg/L).
Proposed PFAS Rule Economic Analysis
1-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table 1-1: ICR TSD Predictions for ATHM4 Based on Disinfectant
Disinfectant
Type
Source
Type
Pilot/ RSSCT
Count
ATHM4 with 10 min
EBCT(%)
ATHM4 with 20 min
EBCT(%)
ATHM4 with 10 min
EBCT Qig/L)
ATHM4 with 20 min
EBCT Qig/L)
Chloramine GW
Chloramine, SW
Free Chlorine GW
Free Chlorine SW
21
102
16
43
30.5 ± 10.5
26.6 ± 12.8
34.7 ±24.3
35.40 ± 17.8
29.6 ± 15.3
36.7 ± 14.5
35.3 ± 17.6
54.7 ±20.8
43.0 ±32.2
29.0 ±24.3
18.8 ± 13.5
20.2 ± 17.5
38.1 ±32.2
37.7 ±26.2
18.8 ± 10.7
32.9 ±31.2
Abbreviations: EBCT - Empty Bed Contact Time; GW - Groundwater; RSSCT - Rapid Small-Scale Column Test; SW - Surface Water; THM4 - Four Regulated
Trihalomethanes.
Proposed PFAS Rule Economic Analysis
1-7
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table 1-2: ICR TSD Predictions for ATHM4 for V2 Year GAC Replacement Based on Disinfectant Type, EBCT, and Source
Water Type
Source
Water
Type
TOC
Range
(mg/L)
ATHM4 with 10 min
ATHM4 with 20 min
ATHM4 with 10 min
ATHM4 with 20 min
Disinfectant
Count
EBCT (%Reduction
EBCT(%
EBCT (jig/L
EBCT (jig/L
Type
(N)
±1 Standard
Reduction ± 1
Reduction ± 1
Reduction ± 1
Deviation)
Standard Deviation)
Standard Deviation)
Standard Deviation)
1-2.0
3
38.09 ± 14.59
48.46 ±21.42
16.02 ±6.77
20.42 ±9.85
GW
2-3.5
4
51.61 ± 11.77
70.85 ± 1.40
31.79 ± 18.76
50.07 ±43.63
Vi
year
3.5-5
6
34.84 ±4.41
39.33 ±2.39
34.04 ± 17.05
42.42 ± 27.47
Chloramine
Above 5
8
33.41 ±6.39
34.53 ± 14.62
86.59 ±20.77
84.86 ±30.12
1-2.0
5
33.69 ±27.18
43.68 ±30.09
16.49 ±8.62
22.78 ± 12.69
SW
2-3.5
59
36.87 ± 15.24
57.29 ± 17.23
29.15 ± 17.83
44.57 ±23.77
3.5-5
31
36.11 ± 11.62
52.84 ± 13.91
49.95 ±33.55
72.35 ±41.99
Above 5
7
40.79 ±5.04
51.16 ±8.68
73.81± 20.77
90.92 ±21.64
1-2.0
5
55.33 ±22.41
59.13 ±20.53
28.74 ± 19.06
25.74 ± 12.18
GW
2-3.5
10
33.81 ± 17.98
48.58 ± 19.85
18.95 ±9.83
27.45 ± 12.81
Free chlorine
3.5-5
1
87.56
49.50
41.99
23.73
1-2.0
7
60.83 ± 25.20
84.69 ±25.89
13.91 ±8.54
20.28 ± 12.94
SW
2-3.5
30
49.21 ± 19.68
74.65 ± 15.39
32.04 ±23.71
50.60 ±36.79
3.5-5
6
42.78 ± 10.26
63.53 ± 17.68
30.57 ±24.87
42.46 ±31.69
Abbreviations: EBCT - Empty Bed Contact Time; GAC - granular activated carbon; ICR TSD - Information Collection Rule Treatment Study Database; THM4 - Four
Regulated Trihalomethanes; TOC - total organic carbon.
Proposed PFAS Rule Economic Analysis
1-8
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table 1-3: ICR TSD Predictions for ATHM4 for One Year GAC Replacement Based on Disinfectant Type, EBCT, and
Source Water Type
Disinfectant
Type
Source
Water
Type
TOC
Range
(mg/L)
Count
(N)
ATHM4 with 10
min EBCT
(%Reduction ± 1
Standard Deviation)
ATHM4 with 20
min EBCT
(%Reduction ± 1
Standard Deviation)
ATHM4 with 10
min EBCT (jig/L
Reduction ± 1
Standard Deviation)
ATHM4 with 20
min EBCT (jig/L
Reduction ± 1
Standard Deviation)
1-2.0
3
32.14 ± 14.75
33.55 ± 16.87
13.55 ±6.76
14.16 ±7.68
GW
2-3.5
4
39.39 ± 17.79
55.20 ±7.81
21.38 ±7.40
38.25 ±32.05
3.5-5
6
31.61 ±4.48
32.56 ±3.55
30.76 ± 15.12
33.06 ± 15.17
Chloramine
Above 5
8
31.33 ±6.43
27.57 ± 16.09
81.10 ± 19.88
66.03 ± 35.55
1
year
1-2.0
5
22.40 ± 16.25
33.48 ±23.63
11.13 ± 6.38
17.24 ±9.33
SW
2-3.5
59
29.59 ± 13.50
44.65 ± 15.02
23.82 ± 15.60
34.77 ± 18.39
3.5-5
31
30.88 ± 12.05
42.95 ± 13.96
43.06 ±30.99
58.76 ±35.32
Above 5
7
36.90 ±4.72
42.70 ± 9.72
66.85 ± 19.58
75.13 ± 18.43
1-2.0
5
45.26 ±20.71
48.48 ± 18.62
23.75 ± 16.84
21.17 ± 10.73
GW
2-3.5
10
28.46 ± 17.25
36.76 ± 17.66
16.17 ±9.50
21.35 ± 11.95
Free Chlorine
3.5-5
1
93.04
49.50
44.61
23.73
1-2.0
7
49.44 ±21.75
73.99 ±25.56
11.00 ±6.30
17.02 ±9.75
SW
2-3.5
30
39.04 ± 17.75
61.02 ± 16.94
25.33 ±20.13
41.75 ±34.79
3.5-5
6
36.29 ± 14.08
55.21 ±21.66
26.15 ±20.67
35.33 ±25.67
Abbreviations: EBCT - Empty Bed Contact Time; GAC - granular activated carbon; ICR TSD - Information Collection Rule Treatment Study Database; THM4 - Four
Regulated Trihalomethanes; TOC - total organic carbon.
Proposed PFAS Rule Economic Analysis
1-9
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table 1-4: ICR TSD Predictions for ATHM4 for 1 Vi Year GAC Replacement Based on Disinfectant Type, EBCT, and
Source Water Type
Source
Water
Type
TOC
Range
(mg/L)
ATHM4 with 10
ATHM4 with 20
ATHM4 with 10
ATHM4 with 20
Disinfectant
Count
min EBCT (%
min EBCT (%
min EBCT (jig/L
min EBCT (jig/L
Type
(N)
reduction ± 1
reduction ± 1
reduction ± 1
reduction ± 1
standard deviation)
standard deviation)
standard deviation)
standard deviation)
1-2.0
3
30.17 ± 14.81
27.31 ± 13.19
12.73 ±6.76
11.52 ±6.02
GW
2-3.5
4
35.06 ±20.01
48.68 ± 11.30
17.79 ±5.62
33.79 ±28.67
3.5-5
6
30.54 ±4.61
30.32 ±5.21
29.67 ± 14.51
29.96 ± 11.22
Chloramine
Above 5
8
30.64 ±6.45
25.26 ± 16.63
79.29 ± 19.61
59.80 ±37.61
1 '/2
year
1-2.0
5
18.19 ± 13.29
28.56 ± 19.06
9.21 ±6.28
14.93 ±8.17
SW
2-3.5
59
26.99 ± 13.11
39.59 ± 14.66
21.94 ± 14.98
30.94 ± 16.92
3.5-5
31
29.14 ± 12.31
39.60 ± 14.37
40.78 ±30.26
54.13 ±33.41
Above 5
7
35.61 ±4.79
39.86 ± 10.48
64.55 ± 19.30
69.85 ± 18.23
1-2.0
5
41.91 ±20.19
44.95 ± 17.99
22.10 ± 16.10
19.66 ± 10.25
GW
2-3.5
10
26.68 ± 17.09
32.73 ± 17.45
15.26 ±9.44
19.27 ± 11.88
Free chlorine
3.5-5
1
94.96
49.50
45.53
23.73
1-2.0
7
45.53 ±21.01
68.48 ±25.48
10.02 ±5.61
15.42 ±8.41
SW
2-3.5
30
35.66 ± 17.51
55.85 ± 18.31
23.10 ± 19.09
38.58 ±34.59
3.5-5
6
34.14 ± 15.63
52.45 ±23.08
24.69 ± 19.35
32.96 ±23.79
Abbreviations: EBCT - Empty Bed Contact Time; GAC - granular activated carbon; ICR TSD - Information Collection Rule Treatment Study Database; THM4 - Four
Regulated Trihalomethanes; TOC - total organic carbon.
Proposed PFAS Rule Economic Analysis
1-10
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table 1-5: ICR TSD Predictions for ATHM4 for Two Year GAC Replacement Based on Disinfectant Type, EBCT, and
Source Water Type
Source
Water
Type
TOC
Range
(mg/L)
ATHM4 with 10
ATHM4 with 20
ATHM4 with 10
ATHM4 with 20
Disinfectant
Count
min EBCT (%
min EBCT (%
min EBCT (jig/L
min EBCT (jig/L
Type
(N)
reduction ± 1
reduction ± 1
reduction ± 1
reduction ± 1
standard deviation)
standard deviation)
standard deviation)
standard deviation)
1-2.0
3
29.18 ± 14.84
24.02 ± 11.12
12.31 ±6.75
10.13 ±5.09
GW
2-3.5
4
32.87 ±21.16
45.31 ± 13.18
15.99 ±5.85
31.51 ±27.06
3.5-5
6
30.00 ±4.69
29.20 ± 6.06
29.13 ± 14.21
28.40 ±9.32
Chloramine
Above 5
8
30.30 ±6.47
24.10 ± 16.91
78.37 ± 19.48
56.66 ±38.69
2
1-2.0
5
16.08 ± 12.47
26.09 ± 16.95
8.25 ±6.42
13.76 ±7.67
year
SW
2-3.5
59
25.69 ± 13.10
36.81 ± 14.64
21.00 ± 14.73
28.86 ± 16.36
3.5-5
31
28.27 ± 12.46
37.92 ± 14.65
39.63 ±29.92
51.80 ±32.56
Above 5
7
34.97 ±4.86
38.44 ± 10.92
63.39 ± 19.18
67.20 ± 18.30
1-2.0
5
40.23 ± 19.94
43.17 ± 17.68
21.26 ± 15.73
18.90 ± 10.01
GW
2-3.5
10
25.79 ± 17.03
30.70 ± 17.46
14.79 ±9.42
18.23 ± 11.89
Free chlorine
3.5-5
1
95.92
49.50
46.00
23.73
1-2.0
7
43.57 ±20.76
65.69 ±25.67
9.52 ±5.27
14.61 ±7.76
SW
2-3.5
30
33.97 ± 17.48
53.22 ± 19.21
21.99 ± 18.59
36.97 ±34.54
3.5-5
6
33.06 ± 16.43
51.06 ±23.81
23.95 ± 18.71
31.77 ±22.87
Abbreviations: EBCT - Empty Bed Contact Time; GAC - granular activated carbon; ICR TSD - Information Collection Rule Treatment Study Database; THM4 - Four
Regulated Trihalomethanes; TOC - total organic carbon.
Proposed PFAS Rule Economic Analysis
1-11
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
1.4 Sampling Points from the Fourth Six Year Review Plants
with Granular Activated Carbon Treatment
To examine the Six Year Review 4 (SYR4) four regulated trihalomethanes (THM4) data,
EPA extracted and matched sampling point IDs for the years that represent before and after GAC
treatment. Only sampling point IDs with the same number of samples before and after GAC
treatment were used to determine THM4 averages. To calculate a single location comparison,
EPA selected one sampling point ID for each public water system identification (PWSID). Entry
point (EP) sampling point types were used when available. When unavailable, EPA used the first
sampling point type. Table 1-6 shows an example of sampling point IDs, sampling point types,
and number of samples available for one PWSID in the SYR4 dataset.
Table 1-6: Sampling Point IDs for each PWSID were Extracted and Matched for the
Years that Represent Before/After GAC Treatment (Example: PWSID AL0000577)
Sampling Point ID
Sampling Point Type
# Of Samples (2017,2019)
ATHM4 (jig/L)a
12967
WS
29 (4, 4)
8.5
12970
WS
29 (4, 4)
8.9
12972
WS
29 (4, 4)
8.5
12974
WS
29 (4 ,4)
9.3
12975
EP
32 (4, 4)
5.7
12976
WS
29 (4, 4)
15.8
12977
DS
32 (4, 4)
10.4
12978
WS
28 (4, 4)
9.4
12979
WS
29 (4, 4)
9.8
12980
DS
24 (3, 0)
-
12981
DS
26 (4, 0)
-
12983
DS
26 (4, 0)
-
13022
WS
25 (4, 4)
11.9
13044
DS
6 (0, 4)
-
13089
MR
2(1,0)
-
Abbreviations: DS - distribution system; EP - entry point; MR - point of maximum residence; WS - water system facility point.
Notes:
a ATHM4 was not calculated for sampling point IDs that did not have sample data for the years that represent either before or
after GAC treatment.
Proposed PFAS Rule Economic Analysis
1-12
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix J. Value of a Statistical Life Updating
EPA follows U.S. EPA (2010) to estimate the economic value of avoiding premature mortality.
To obtain a value of a statistical life (VSL) suitable for valuation of mortality risk reductions
during 2023-2104, EPA relies on the base value estimate of $4.8 million ($1990, 1990 income
year), which is the central tendency of the VSL distribution recommended for use in EPA's
regulatory impact analyses (U.S. EPA, 2010). EPA adjusted the base VSL estimate for inflation
and income growth as follows:
Equation J-l:
P2021 ( \
K.2021 = ^1990,1990 " p \y )
' 1990 ^*1990 '
Where:
^,2021 VSL value ($2021) updated for use in evaluation year t, t =
2023 ...2050;
^1990,1990 Base VSL value of $4,800,000 ($1990, 1990 income year);
P2021 Gross Domestic Product (GDP) price deflator index value in 2021;
P1990 GDP price deflator index value in 1990;
Yt Projected income per capita ($2012) in evaluation year t, t =
2023 ...2050;
liggo Historical income per capita ($2012) in 1990;
e VSL income elasticity of 0.4 as recommended by U.S. EPA (U.S. EPA,
2010).
EPA used disposable personal annual income to represent U.S. income per capita. Because the
PFAS analysis spans a future time period 2023-2104, EPA relied on the long-term personal
disposable income projections from the U.S. Energy Information Administration (2021).
The long-term personal income projections are available annually from 2020 to 2050.
EPA's SafeWater model requires a single income growth factor to project the 2023 VSL value
(in $2021) to future years (2024 through 2104). Based on the VSL estimates calculated using
Equation J-l, EPA calculated the compound annual growth rate, CAGR, of VSL values from
2023 to 2050 as follows:
Equation J-2:
,(2050-;
Ut>U,ZUZl
>v7
/ ^2050 2021 \V205°-2023;
CAGR = , ' - 1
2023,2021 >
EPA used the calculated CAGR value to approximate VSL growth during the analysis period
(2023 to 2104) based on the 2021 VSL value estimated using Equation J-l.
Proposed PFAS Rule Economic Analysis
J-13
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Equation J-3:
^t,2021 = ^2023,2021 " (1 + CAGRy 2023
Table J-l summarizes the projected VSL estimates through 2050 and the approximated VSL
estimates through 2104.
Table J-l: Estimated VSL Series
Historical Proiected T „
„ , „ Income Growth
Personal Personal _ , ..
T.- , , ,, Factor (Ratio oi .
Disposable Disposable p pDYp, ,rojec(edvSL Appro™.
Income Per Income Per
Capita Capita ^"tothe ^ } ($2021>
(PDYPP, (PDYPP, Power of 0 4)
$2012) $2012) ;
1990
30,327
-
1
8,929,233
"
2023
-
47,515
1.1967
10,686,015
10,686,015
2024
-
48,311
1.2047
10,757,230
10,752,695
2025
-
49,241
1.2140
10,839,662
10,819,792
2026
-
50,084
1.2222
10,913,510
10,887,308
2027
-
50,939
1.2305
10,987,592
10,955,245
2028
-
51,817
1.2390
11,062,980
11,023,605
2029
-
52,727
1.2476
11,140,284
11,092,393
2030
-
53,711
1.2569
11,223,004
11,161,609
2031
-
54,639
1.2655
11,300,154
11,231,258
2032
-
55,566
1.2741
11,376,490
11,301,341
2033
-
56,512
1.2827
11,453,543
11,371,861
2034
-
57,446
1.2911
11,528,885
11,442,821
2035
-
58,369
1.2994
11,602,639
11,514,225
2036
-
59,277
1.3074
11,674,522
11,586,073
2037
-
60,163
1.3152
11,744,001
11,658,370
2038
-
61,038
1.3228
11,812,021
11,731,119
2039
-
61,925
1.3305
11,880,357
11,804,321
2040
-
62,791
1.3379
11,946,526
11,877,980
2041
-
63,684
1.3455
12,014,214
11,952,098
2042
-
64,620
1.3534
12,084,516
12,026,680
2043
-
65,553
1.3612
12,154,071
12,101,726
2044
-
66,480
1.3688
12,222,510
12,177,241
2045
-
67,418
1.3765
12,291,208
12,253,227
2046
-
68,366
1.3842
12,360,002
12,329,687
2047
-
69,341
1.3921
12,430,265
12,406,624
2048
-
70,351
1.4002
12,502,317
12,484,041
2049
-
71,343
1.4080
12,572,559
12,561,942
Proposed PFAS Rule Economic Analysis
J-14
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table J-l: Estimated VSL Series
Historical Proiected T „
„ , „ Income Growth
Personal Personal _ . f
Disposable Disposable ,^[e(d^p ProJected vSL Approximated
Year IncomePer IncomePer ^ J VSL
Capita Capita pnvpptnthP ($2021)
(PDYPP, (PDYPP, Power of 0 4)
$2012) $2012) ;
2050 - 72,304 1.4156 12,640,072 12,640,328
2051 - - - - 12,719,204
2052 - - - - 12,798,572
2053 - - - - 12,878,435
2054 - - - - 12,958,796
2055 - - - - 13,039,659
2056 - - - - 13,121,027
2057 - - - - 13,202,902
2058 - - - - 13,285,288
2059 - - - - 13,368,188
2060 - - - - 13,451,606
2061 - - - - 13,535,544
2062 - - - - 13,620,006
2063 - - - - 13,704,994
2064 - - - - 13,790,514
2065 - - - - 13,876,566
2066 - - - - 13,963,156
2067 - - - - 14,050,286
2068 - - - - 14,137,960
2069 - - - - 14,226,181
2070 - - - - 14,314,952
2071 - - - - 14,404,278
2072 - - - - 14,494,160
2073 - - - - 14,584,604
2074 - - - - 14,675,612
2075 - - - - 14,767,188
2076 - - - - 14,859,335
2077 - - - - 14,952,057
2078 - - - - 15,045,358
2079 - - - - 15,139,241
2080 - - - - 15,233,710
2081 - - - - 15,328,768
2082 - - - - 15,424,420
2083 - - - - 15,520,668
2084 - - - - 15,617,517
Proposed PFAS Rule Economic Analysis J-15 March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table J-l: Estimated VSL Series
Year
Historical
Personal
Disposable
Income Per
Capita
(PDYPP,
$2012)
Projected
Personal
Disposable
Income Per
Capita
(PDYPP,
$2012)
Income Growth
Factor (Ratio of
Projected PDYPP
to Historical 1990
PDYPP to the
Power of 0.4)
Projected VSL
($2021)
Approximated
VSL
($2021)
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
15,714,970
15,813,032
15,911,705
16,010,994
16,110,903
16,211,435
16,312,594
16,414,385
16,516,810
16,619,875
16,723,583
16,827,939
16,932,945
17,038,606
17,144,927
17,251,912
17,359,564
17,467,887
17,576,887
17,686,567
Acronyms: PDYPP- Personal Disposable Income Per Capita; VSL- Value of a Statistical Life.
Table J-2 summarizes the data employed in updating the values used to monetize reductions in
mortality and morbidity risks in the population exposed to PFOA and PFOS in drinking water.
EPA uses the VSL to monetize reduced mortality benefits and uses the COI to monetize reduced
morbidity benefits. The details on morbidity valuation for birth weight, CVD, RCC, and bladder
cancer analyses are provided in the respective sections of the main document.
Table J-2: Summary of Inputs and Data Sources Used for Valuation
Data Element Modeled Data Source Notes
Variability
IIS FPA
Base VSL None ^qiq as recommended by the U.S. EPA Guidelines
The base value of 4,800,000 ($1990) was used
as recommended by the U.S. EPA (
for Preparing Economic Analyses.
Proposed PFAS Rule Economic Analysis
J-16
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table J-2: Summary of Inputs and Data Sources Used for Valuation
Data Element
Modeled
Variability
Data Source
Notes
VSL income
elasticity
Medical Care
CPI
Employment
Cost Index
GDP Price
Deflator Index
Historical
income per
capita
Projected
income per
capita
None
Time: Annual,
1990..2021
Time: Quarterly,
2001..2021
Time: Annual,
1990..2021
Time: Annual,
1990..2020
Time: Annual,
2020..2050
U.S. EPA,
2010
BLS 2022
(U.S. Bureau of
Labor
Statistics,
2020)
BLS 2022
(Bureau of
Labor
Statistics,
2022)
BEA 2022
(U.S. Bureau of
Economic
Analysis, 2022)
BEA 2021
(U.S. Bureau of
Economic
Analysis, 2021)
U.S. EIA 2021
(U.S. Energy
Information
Administration,
2021)
Income growth adjustments were done using
income elasticity 0.4 per recommendations in
the U.S. EPA Guidelines for Preparing
Economic Analyses.
Medical cost inflation adjustments were done
using annual CPI for medical care (U.S. city
average, all urban consumers, series number
CUUR0000SAM).
Opportunity cost inflation adjustments were
done using quarterly index for total
compensation for all civilian workers in all
industries and occupations (series number
CIS 10100000000001).
VSL inflation adjustments were done using
annual GDP price deflator index.
Disposable personal annual income per capita
(series number A229RC0A052NBEA). Data are
in $2021. The series were converted to constant
$2012 to align with US EIA 2021 projections
using BLS 2020 CPI series.
The U.S. EIA long-term projections focus on
components of potential growth, fiscal balances
and debt accumulation, domestic saving and
investment balances, and external balances are
covered and interest rates consistent with those
projections. The projection horizon is 2050.
EPA used the ratio of projected real disposable
personal income (in constant $2012, series
number 18-AEO2021.55.ref2021-dl 13020a) to
project population size (series number
18-AEQ2021.42.ref2021-dl 13020a).
Abbreviations: BEA - Bureau of Economic Analysis; BLS - bureau of labor statistics; CPI - consumer price index;
EIA - Energy Information Administration; GDP - gross domestic product; VSL - value of a statistical life.
Proposed PFAS Rule Economic Analysis
J-17
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix K. Benefits Sensitivity Analyses
This appendix provides details on the sensitivity analyses implemented by EPA to evaluate the
impact of the exposure-response assumptions in the CVD benefits model and the impact of
Perfluorononanoic Acid (PFNA) inclusion in the birth weight benefits model. Section K. 1
describes hypothetical regulatory alternatives evaluated in the sensitivity analyses. Section K.2
provides details on estimation of blood serum PFOA, PFOS, and PFNA. Section K.3 summarizes
the CVD exposure response scenarios and presents the associated results. Section K.4
summarizes the birth weight dose response scenarios and results. Section K.5 summarizes the
RCC exposure response scenarios and results.
The sensitivity analyses described herein relied on methodology implemented in R software (R
Core Team, 2021) and differ slightly from SafeWater MCBC methods. Specifically, SafeWater
performs a set of pre-calculations to maximize computational efficiency and, as such, the order
of analytical steps across R and SafeWater models differs; however, results across models are
mathematically consistent. The R-based model version treats each integer age cohort between 85
and 99 separately, implements the CVD calculations for those aged 40-89 years only, and applies
the ASCVD model-based annual incidence at age 80 years to ages 81-89 because the ASCVD
model has been fit to those aged 40-80 years and predicts the 10-year probability of the first
CVD event.
K.l Overview of the Hypothetical Exposure Reduction
Table K-l shows the details of the two hypothetical exposure reductions developed by EPA for
the sensitivity analyses. For both alternatives, EPA assumed the same population served size of
100,000 distributed over age-, sex-, and race-ethnicity categories using national-level
demographic data (see Appendix B). Hypothetical exposure reduction 1 assumes a reduction of 1
ppt in PFOA and a reduction of 1 ppt in PFOS. Hypothetical exposure reduction 2 assumes a
reduction of 1 ppt in PFNA,.56 in addition to the reductions specified for hypothetical exposure
reduction 1. Additional sensitivity analysis assumptions (other than those pertaining to the
exposure-response scenarios in Section K.3 and Section K.4), such as evaluation period,
population growth, etc., align with those used in the Economic Analysis. EPA notes that
uncertainty was not characterized for these sensitivity analysis scenarios. All parameters treated
as uncertain in the Economic Analysis were set to their central estimate values (see Appendix L).
EPA notes that relative magnitudes of reductions in PFOA, PFOS, and PFNA may differ from
those evaluated in the Economic Analysis. At entry points where PFOA and PFOS
concentrations exceed their respective proposed MCLs, EPA expects reductions of 1 ppt or
greater. Reductions in PFNA resulting from the proposed NPDWR are not predicted in this EA
given the available occurrence data for PFNA at regulatory thresholds under consideration.
However, multiple data sources, including UCMR 3 and state-collected finished drinking water
data, demonstrate that PFNA has been detected between 0.22 ppt and 94.2 ppt. In UCMR 3,
0.28% of participating systems (14 total) had PFNA detections greater than/equal to the MRL
(20 ppt), while state monitoring efforts showed that the number of systems in each state with
PFNA detections ranged between 0.0% and 13.3%. EPA chose to evaluate unit reductions (i.e., 1
56 Note that the inclusion of PFNA under Alternative 2 was only relevant to BW sensitivity analysis because there is evidence
that PFNA reductions can improve BW. There is a lack of supporting evidence for an impact for CVD and RCC benefits.
Proposed PFAS Rule Economic Analysis
K-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
ppt each) to demonstrate the effects of and make comparisons between unit changes in PFOA,
PFOS, and PFNA exposure (U.S. EPA, 2023c). Caution should be exercised in quantifying the
potential magnitude of change in the national benefits estimates based on the results of these
sensitivity analyses, although conclusions about the directionality of these effects can be
inferred.
Table K-l: Overview of Hypothetical Exposure Reductions
Hypothetical Exposure Reduction
Parameter Description i 2
(PFOA+PFOS, (PFOA+PFOS+PFNA,
Population served at the start of the evaluation period 100,000 100,000
Reduction in PFOA concentration (ppt) 1 1
Reduction in PFOS concentration (ppt) 1 1
Reduction in PFNA concentration (ppt) 0 1
Abbreviations: PFNA - perfluorononanoic acid; PFOA - perfluorooctanoic acid; PFOS - perfluorooctane sulfonic acid.
K.2 Estimation of Blood Serum PFOA, PFOS, and PFNA
EPA used PFOA and PFOS drinking water concentrations as inputs to its Pharmacokinetic (PK)
model to estimate blood serum PFOA and PFOS concentrations for adult males and females. In
this analysis, the Agency used the September 18, 2021 PFOA/PFOS PK model version. See
EPA's Toxicity Assessments and Proposed Maximum Contaminant Level Goals for PFOA and
PFOS in Drinking Water for further information on the PFOA/PFOS model (U.S. EPA, 2023a;
U.S. EPA, 2023b). Application of the PK model in the context of the benefits estimation is
detailed in Section 6.3 of the Economic Analysis.
To estimate blood serum PFNA based on its drinking water concentration, EPA used a first-order
single-compartment model whose behavior was previously demonstrated to be consistent with
PFOA pharmacokinetics in humans (Bartell et al., 2010). Equation K-l-Equation K-4
summarize this model (Bartell, 2003; Bartell, 2017; Lu et al., 2020):
Equation K-l:
W*S
Cm — B + ¦
1000
Equation K-2:
Ct = Cm + (fl-Coo) * e~kt
Equation K-3:
k = Zn(2) /t1/2
Proposed PFAS Rule Economic Analysis K-2 March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Equation K-4:
Where:
Cqo = steady-state serum PFNA concentration (ng/mL);
Ct = serum concentration at time t (ng/mL);
t = time since beginning of / change in the water exposure (days);
B = background serum PFNA concentration (ng/mL). EPA used an estimate of 0.411
ng/mL for 2017-2018 from Centers for Disease Control and Prevention (2022);
W = drinking water PFNA concentration (ppt);
S = steady-state serum/water concentration ratio (unitless);
k = first order elimination rate constant for PFNA from serum (days"1), defined as a
function of half-life in Equation K-3 (Bartell, 2003);
t-i/2 = PFNA half-life in serum (days). Following Lu et al. (2020) model assumptions, EPA
used an estimate of 3.9 years from Zhang et al. (2013) (weighted average estimate), after
converting it to 1,424.5 days;
/ = fraction of PFNA absorbed (unitless). Following Lu et al. (2020) model assumptions,
EPA used 100% absorption;
Q = water intake (L/kg body weight per day). Consistent with assumptions used for serum
PFOA and PFOS, EPA used a water intake of 0.013 L/kg of body weight per day (U.S. EPA,
201 lb) in order to compute the PFNA dose from drinking water sources; and
Vd = volume of distribution (L/kg body weight per day), a proportionality constant relating
the total amount of a chemical in the body to the concentration in plasma (Hoffman et al., 2011).
Following Lu et al. (2020) model assumptions, EPA used an estimate of 0.17 L/kg body weight
from Zhang et al. (2013).
Using this model, EPA evaluated lifetime baseline and lifetime regulatory alternative exposure
scenarios described in Section 6.3 of the Economic Analysis and used the difference between the
two as an input to the downstream analysis of health effects.
K.3 CVD Sensitivity Analyses
CVD sensitivity analyses rely on hypothetical exposure reduction 1 (i.e., 1 ppt reduction in
PFOA and 1 ppt reduction in PFOS) to explore the impact of the following changes in the CVD
exposure-response modeling:
Proposed PFAS Rule Economic Analysis
K-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• The use of single study-based TC effect estimates, rather than EPA meta-analysis-based
effect estimates. To this end, EPA used estimates from a large NHANES study (Dong et
al., 2019) and estimates from a longitudinal study of diabetes prevention program
outcomes study (P.-I. D. Lin et al., 2019);
• Inclusion of HDLC effects from the CVD analysis; and,
• Exclusion of BP effects from the CVD analysis
Table K-2 summarizes the exposure-response scenarios, while Table K-3 provides details on the
slope factors used in this sensitivity analysis.
Table K-2: Overview of CVD Exposure-Response Scenarios
Exposure-Response Scenario Definition
Scenario
Economic Analysis (EA) scenario using EPA meta-analysis for TC, Liao et al.
(2020) for BP, and excluding HDLC impacts.
Scenario using Dong et al. (2019) for TC, Liao et al. (2020) for BP, and
excluding HDLC impacts.
Scenario using P.-I. D. Lin et al. (2019) for TC, Liao et al. (2020) for BP, and
excluding HDLC impacts.
Scenario using EPA meta-analysis for TC and HDLC, and Liao et al. (2020) for
BP.
Scenario using Dong et al. (2019) for TC and HDLC, and Liao et al. (2020) for
BP.
Scenario using P.-I. D. Lin et al. (2019) for TC and HDLC, and Liao et al. (2020)
for BP.
Scenario using EPA meta-analysis for TC and excluding HDLC and BP impacts.
This scenario is most comparable to the U.S. EPA (2021a) analysis implemented
for the SAB review.
Scenario using Dong et al. (2019) for TC and excluding HDLC and BP impacts.
Scenario using P.-I. D. Lin et al. (2019) for TC and excluding HDLC and BP
impacts.
Scenario using EPA meta-analysis for TC and HDLC, and excluding BP impacts.
Scenario using Dong et al. (2019) for TC and HDLC, and excluding BP impacts.
Scenario using P.-I. D. Lin et al. (2019) for TC and HDLC, and excluding BP
impacts.
Abbreviations: BP - blood pressure; CVD - cardiovascular disease; EA - economic analysis; HDLC - high-density
lipoprotein cholesterol; PFOA - perfluorooctanoic acid; PFOS - perfluorooctane sulfonic acid; SAB - Science Advisory
Board; TC - total cholesterol.
1-EA
2-Dong
3-Lin
4-EA (+HDLC)
5-Dong (+HDLC)
6-Lin (+HDLC)
7-EA (-BP)
8-Dong (-BP)
9-Lin (-BP)
10-EA (-BP +HDLC)
11-Dong (-BP +HDLC)
12-Lin (-BP +HDLC
Proposed PFAS Rule Economic Analysis
K-4
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table K-3: Exposure-Response Information for CVD Biomarkers
Source
Contaminant
Linear Slope Estimate (mg/dL per 1 ng/mL)
TC
HDLC
BP
Serum PFOA
1.57
0.11
EPA meta-analysis3
Serum PFOS
(CI95: 0.02,3.13)
0.08
(CI95:-0.01,0.16)
(CI95: -0.22, 0.43)
0.05
(CI95:-0.01, 0.11)
-
Serum PFOA
1.48
-0.03
Dong et al. (2019)
Serum PFOS
(CI95: 0.18, 2.78)
0.40
(CI95: 0.13, 0.67)
(CI95: -0.44, 0.39)
0.01
(CI95:-0.08, 0.11)
-
Serum PFOA
1.63
-0.13
P.-I. D. Lin et al.
(CI95: -0.84, 2.42)
(CI95:-0.37,0.107)
(2019)
Serum PFOS
0.13
(CI95: -0.005,0.27)
-0.02
(CI95: -0.06, 0.02)
-
Liao et al. (2020)
Serum PFOS
0.044
—
(CI95: 0.006,0.083)
Abbreviations: BP - systolic blood pressure; CI95 - 95% CI; CVD - cardiovascular disease; HDLC - high-density lipoprotein
cholesterol; PFOA - perfluorooctanoic acid; PFOS - perfluorooctane sulfonic acid; TC - total cholesterol.
Notes:
a See Section 6.5.2 of the Economic Analysis.
Table K-4 shows the results of the CVD sensitivity analysis. EPA made the following
observations:
• Relative to the annualized CVD benefits estimated using EPA meta-analysis-based slope
factors, using the Dong et al. (2019) slope factors increases the annualized CVD benefits
by 13%, while using the P.-I. D. Lin et al. (2019) slope factors increases the annualized
CVD benefits by 7%.
• Inclusion of HDLC effects decreases annualized CVD benefits by 23%-25% if EPA
meta-analysis slope factors are used. The use of Dong et al. (2019) and the P.-I. D. Lin et
al. (2019) instead of the EPA meta-analysis slope factors decreases annualized benefits
by 2.7%-3.0% and 20.1%-21.9%, respectively. The wide variation in the impact of
HDLC inclusion may be explained by high variance in the slope factor estimates. EPA
notes, however, that none of the PFOA/PFOS-HDLC slope factors are statistically
significant at the 5% level.
• Exclusion of BP effects decreases annualized CVD benefits by 1.8%-2.2% if EPA meta-
analysis slope factors are used. However, estimates decrease by 1,6%-l .9% and 1.7%-
2.0% if the Dong et al. (2019) and the P.-I. D. Lin et al. (2019), respectively, slope
factors are used.
The relative magnitudes of reductions in PFOA and PFOS used in this sensitivity analysis
may differ from those implied by the regulatory alternatives evaluated in the Economic
Analysis. Therefore, the potential magnitude of changes in national CVD benefits due to
alternative TC/HDLC
exposure-response assumptions as well as exclusion of the BP effects may differ from the
ones estimated in this sensitivity analyses.
Proposed PFAS Rule Economic Analysis
K-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table K-4: Summary of CVD Sensitivity Analysis for Hypothetical Exposure Reduction 1 (PFOA+PFOS)
Result Description"
Exposure-Response Scenariob<
<
W
ex
s
o
Q
< hJ
4 §
+
ix. U
S J
p B
3 5
.5 j
« a
+
&
CO
<
w
M /-s
= Q-
0 53
£5 m
1
90
&
CO
I
a.
Sg
< Q
a.
e P
W M pB
J _|_ i _L
¦ +
a.
3§
.3 P
fN +
Average reduction in serum PFOA
concentration (ng/mL)
Average reduction in serum PFOS
concentration (ng/mL)
Average reduction in TC
concentration (mg/dL)
Average reduction in HDLC
concentration (mg/dL)
Average reduction in BP (mmHg)
Non-fatal first MI (total cases
avoided)d
Non-fatal first IS (total cases
avoided)d
CVD deaths (total cases avoided)d
PDV, non-fatal first MI (3%
discount rate, millions $2021)
PDV, non-fatal first IS (3%
discount rate, millions $2021)
PDV, CVD deaths (3% discount
rate, millions $2021)
PDV, total CVD benefits (3%
discount rate, millions $2021)
Annualized CVD benefits (3%
discount rate, millions $2021)
PDV, non-fatal first MI (7%
discount rate, millions $2021)
PDV, non-fatal first IS (7%
discount rate, millions $2021)
PDV, CVD deaths (7% discount
rate, millions $2021)
0.094 0.094 0.094 0.094 0.094 0.094 0.094 0.094 0.094 0.094 0.094 0.094
0.086 0.086 0.086 0.086 0.086 0.086 0.086 0.086 0.086 0.086 0.086 0.086
0.152 0.173 0.164 0.152 0.173 0.164 0.152 0.173 0.164 0.152 0.173 0.164
0.000 0.000 0.000 0.015 -0.002 -0.014 0.000 0.000 0.000 0.015 -0.002 -0.014
0.004 0.004 0.004 0.004 0.004 0.004 0.000 0.000 0.000 0.000 0.000 0.000
2.725 3.096 2.931 1.942 3.200 3.675 2.686 3.057 2.892 1.902 3.161 3.636
4.091 4.649 4.401 3.078 4.783 5.363 4.030 4.587 4.339 3.016 4.722 5.302
0.849 0.966 0.913 0.690 0.986 1.064 0.825 0.941 0.889 0.665 0.962 1.040
0.103 0.117 0.111 0.073 0.121 0.139 0.102 0.115 0.109 0.072 0.119 0.138
0.043 0.049 0.047 0.032 0.051 0.057 0.043 0.049 0.046 0.032 0.050 0.057
4.965 5.627 5.336 3.844 5.777 6.403 4.855 5.517 5.226 3.734 5.666 6.292
5.111 5.793 5.493 3.949 5.948 6.599 4.999 5.681 5.381 3.837 5.836 6.487
0.168 0.191 0.181 0.130 0.196 0.217 0.165 0.187 0.177 0.126 0.192 0.214
0.042 0.048 0.045 0.030 0.049 0.057 0.042 0.047 0.045 0.029 0.049 0.056
0.018 0.020 0.019 0.013 0.021 0.024 0.018 0.020 0.019 0.013 0.021 0.024
2.351 2.654 2.525 1.775 2.733 3.076 2.308 2.611 2.482 1.732 2.690 3.033
Proposed PFAS Rule Economic Analysis
K-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table K-4: Summary of CVD Sensitivity Analysis for Hypothetical Exposure Reduction 1 (PFOA+PFOS)
<
w
PDV, total CVD benefits (7%
discount rate, millions $2021)
Annualized CVD benefits (7%
2.411 2.722 2.590 1.818 2.803 3.157 2.368 2.678 2.546 1.774 2.759 3.113
0.169 0.191 0.182 0.128 0.197 0.222 0.166 0.188 0.179 0.125 0.194 0.219
discount rate, millions $2021)
Abbreviations: PFOA - perfluorooctanoic acid; PFOS - perfluorooctane sulfonic acid; TC - total cholesterol; HDLC - high-density lipoprotein cholesterol; BP -
systolic blood pressure; CVD - cardiovascular disease; EA - economic analysis; SAB - science advisory board; MI - myocardial infarction; IS - ischemic stroke;
PDV - present discounted value.
Notes:
aSee Table K-l
bSee Table K-3
cNegative values refer to increases in a particular result (e.g., the HDTC reduction of -0.002 mg/dL in Scenario 2-Dong refers to an increase in HDTC).
dTotal over the period of analysis.
Proposed PFAS Rule Economic Analysis K-7 March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
K.4 Birth Weight Sensitivity Analyses
Birth weight sensitivity analyses rely on the two hypothetical exposure reductions described in
Table K-l to explore the impact of the following changes in the birth weight exposure-response
modeling:
• Early pregnancy birth weight effects using first trimester estimates from Steenland et
al. (2018) for PFOA and Dzierlenga, Crawford, et al. (2020) for PFOS; and
• Inclusion of PFNA-birth weight effects using estimates from two studies (Lenters et al.,
2016; Valvi et al., 2017), in addition to the PFOA-birth weight and PFOS-birth weight
effects analyzed in the Economic Analysis; inclusion of PFNA-birth weight effects using
estimates from two studies (Lenters et al., 2016; Valvi et al., 2017), in addition to the
PFOA-birth weight and PFOS-BW effects analyzed in the Economic Analysis.
Table K-5 summarizes the exposure-response scenarios, while Table K-6 provides details on the
slope factors used in this sensitivity analysis.
Table K-5: Overview of Birth Weight Exposure-Response Scenarios
Exposure- Scenario Definition
Response
Scenario
Economic Analysis (EA) scenario using Steenland et al. (2018) for PFOA, Dzierlenga,
Crawford, et al. (2020) for PFOS
Scenario using first trimester estimates from Steenland et al. (2018) for PFOA and
Dzierlenga, Crawford, et al. (2020) for PFOS
Scenario using Steenland et al. (2018) for PFOA, Dzierlenga, Crawford, et al. (2020) for
PFOS, Lenters et al. (2016) forPFNA/PFDA
Scenario using Steenland et al. (2018) for PFOA, Dzierlenga, Crawford, et al. (2020) for
PFOS, Valvi et al. (2017) forPFNA
Abbreviations: PFDA - Perfluorodecanoic Acid; PFNA - perfluorononanoic acid; PFOA - perfluorooctanoic acid;
PFOS - perfluorooctane sulfonic acid.
1-EA
2-First Trimester
3-EA+Lenters
4-EA+Valvi
Proposed PFAS Rule Economic Analysis
K-8
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table K-6: Exposure-Response Information for Birth Weight
Source
Linear Slope Estimate
(g birth weight per 1 ng/mL)
Serum PFOA Serum PFOS Serum PFNA
Dzierlenga, Crawford, et al. (2020)
First trimester - Steenland et al. (2018)
First trimester - Dzierlenga, Crawford, et al.
(2020)
Lenters et al. (2016)
Valvi et al. (2017)
Steenland et al. (2018)
Abbreviations: CI95 - 95% confidence interval; PFNA - perfluorononanoic acid; PFOA - perfhiorooctanoic acid;
PFOS - perfhiorooctane sulfonic acid.
Table K-7 shows the results of the birth weight sensitivity analysis. EPA made the following
observations:
• Using early pregnancy study-based dose-response estimates could reduce annualized
benefits by 66%.
• Inclusion of a 1 ppt PFNA reduction could increase annualized birth weight benefits 5.4-
7.7-fold, relative to the scenario that quantifies a 1 ppt reduction in PFOA and a 1 ppt
reduction in PFOS only.
• The range of estimated PFNA-related increases in benefits is driven by the exposure-
response, with smaller estimates produced using the slope factors from Lenters et al.
(2016), followed by Valvi et al. (2017). EPA notes that the PFNA slope factor estimates
are orders of magnitude larger than the slope factor estimates used to evaluate the impacts
of PFOA/PFOS reductions. EPA also notes that the PFNA slope factor estimates are not
precise, with 95% CIs covering wide ranges that include zero (i.e., serum PFNA slope
factor estimates are not statistically significant at 5% level).
The relative magnitudes of reductions in PFOA, PFOS, and PFNA used in this sensitivity
analysis may differ from those implied by the regulatory alternatives evaluated in the Economic
Analysis. Therefore, the potential magnitude of increase in the national birth weight benefits
estimates due to inclusion of PFNA effects may differ from the one estimated in this sensitivity
analyses.
Proposed PFAS Rule Economic Analysis K-9 March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table K-7: Summary of Birth Weight Sensitivity Analysis
Hypothetical Exposure Reduction3 /
Exposure-Response Scenariob
Result Description (PFOA+PFOS) (PFOA+PFOS+PFNA)
1-EA 2-First 3- 4-EA+Valvi
Trimester
EA+Lenters
Average reduction in serum PFOA concentration
0.092
0.092
0.092
0.092
(ng/mL)
Average reduction in serum PFOS concentration
0.083
0.083
0.083
0.083
(ng/mL)
Average reduction in serum PFNA concentration
0.000
0.000
0.139
0.139
(ng/mL)
Total increase in birth weight (g)
1.214
0.415
6.843
9.584
Total number of births affected0
95,263
95,263
95,263
95,263
Total number of surviving births affected0
94,831
94,830
94,833
94,835
Birth weight-related deaths (total cases avoided)0
0.592
0.203
3.324
4.647
PDV, birth weight-related deaths (3% discount
2.337
0.799
12.951
18.091
rate, millions $2021)
PDV, birth weight-related morbidity (3%
0.076
0.026
0.422
0.591
discount rate, millions $2021)
PDV, total birth weight benefits (3% discount
2.414
0.825
13.373
18.682
rate, millions $2021)
Annualized birth weight benefits (3% discount
0.079
0.027
0.440
0.615
rate, millions $2021)
PDV, birth weight-related deaths (7% discount
0.848
0.290
4.612
6.436
rate, millions $2021)
PDV, birth weight-related morbidity (7%
0.028
0.010
0.154
0.215
discount rate, millions $2021)
PDV, total birth weight benefits (7% discount
0.877
0.299
4.766
6.651
rate, millions $2021)
Annualized birth weight benefits (7% discount
0.062
0.021
0.335
0.467
rate, millions $2021)
Abbreviations: PDV - present discounted value; PFNA - perfluorononanoic acid; PFOA - perfluorooctanoic acid; PFOS -
perfluorooctane sulfonic acid.
Notes:
aSee Table K-l
bSee Table K-5
cTotal over the period of analysis.
Proposed PFAS Rule Economic Analysis
K-10
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
K.5 RCC Sensitivity Analyses
RCC sensitivity analyses rely on the first hypothetical exposure reduction described in Table K-l
to explore the impact of the following changes in the RCC exposure-response modeling:
• The use of the serum PFOA central tendency slope from Vieira et al. (2013), as derived
by EPA/OST (U.S. EPA, 2023b); and
• The use of the serum PFOA central tendency slopes from Vieira et al. (2013) excluding a
very high exposure group, as derived by EPA/OST (U.S. EPA, 2023b).
Table K-8 summarizes the exposure-response scenarios, while Table K-9 provides details on the
slope factors used in this sensitivity analysis.
Table K-8: Overview of RCC Exposure-Response Scenarios
Exposure-Response
Scenario
Scenario Definition3
Economic Analysis (EA) scenario using the serum PFOA central tendency slope from
Shearer et al. (2021)
Scenario using the serum PFOA central tendency slope from Vieira et al. (2013)
1-EA
2-Vieira
3-VieiraExcludeHigh
Scenario using the serum PFOA central tendency slope from Vieira et al. (2013),
excluding a very high exposure group
Abbreviations: PFOA - perfluorooctanoic acid; RCC - renal cell carcinoma.
Note:
aAll exposure-response scenarios include the 3.94% PAF-based cap on the magnitude of relative risk reductions, as described
in Section 6.6.
Table K-9: Exposure-Response Information for RCC
Source
Linear Slope Estimate, Serum PFOA
(per 1 ng/mL)
Shearer et al. (2021), as derived by EPA/OST (U.S. EPA, 2023b)
0.00178
(CI95: 0.00005, 0.00352)
Vieira et al. (2013), as derived by EPA/OST (U.S. EPA, 2023b)
0.00007
(CI95: 0.000001, 0.00014)
Vieira et al. (2013) excluding very high exposure group from
0.00025
Vieira et al. (2013), as derived by EPA/OST (U.S. EPA, 2023b)
(CI95: 0.00001, 0.00048)
Abbreviations: CI95 - 95% CI; PFOA - perfluorooctanoic acid; RCC - renal cell carcinoma.
Proposed PFAS Rule Economic Analysis
K-ll
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table K-10 shows the results of the RCC sensitivity analysis. EPA made the following
observations:
• Using the slope factor based on Vieira et al. (2013) could reduce annualized benefits
by 96%;
• Using the slope factor based on Vieira et al. (2013) excluding a very high exposure group
could reduce annualized benefits by 86%.
EPA also notes that the PAF-based cap of 3.94% on the RCC relative risk reductions associated
with a 1 ppt reduction in PFOA is rarely binding for the Economic Analysis scenario presented
below and never binding for the sensitivity analysis scenarios. For larger PFOA reduction
magnitudes, the PAF-based cap could become binding, which would attenuate the differences
across the sensitivity analysis scenarios.
Table K-10: Summary of RCC Sensitivity Analysis
Exposure-Response Scenario
a
Result Description
1-EA
2-Vieira
3- VieiraExcludeHigh
Average reduction in serum PFOA
concentration (ng/mL)
0.088
0.088
0.088
Non-fatal RCC (cases avoided)
9.627
0.377
1.336
RCC-related deaths (cases avoided)13
3.887
0.152
0.539
PDV, Non-fatal RCC (3% discount
rate, millions $2021)
1.569
0.061
0.218
PDV, RCC-related deaths (3%
14.376
0.562
1.994
discount rate, millions $2021)
PDV, total RCC benefits (3%
15.945
0.623
2.212
discount rate, millions $2021)
Annualized RCC benefits (3%
0.525
0.021
0.073
discount rate, millions $2021)
PDV, Non-fatal RCC (7% discount
rate, millions $2021)
0.483
0.019
0.067
PDV, RCC-related deaths (7%
3.961
0.155
0.549
discount rate, millions $2021)
PDV, total RCC benefits (7%
discount rate, millions $2021)
4.444
0.174
0.617
Annualized RCC benefits (7%
0.312
0.012
0.043
discount rate, millions $2021)
Abbreviations: PDV - present discounted value; PFOA - perfluorooctanoic acid; RCC - renal cell carcinoma.
Notes:
aSee Table K-8.
bTotal over the period of analysis.
Proposed PFAS Rule Economic Analysis K-12 March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix L Uncertainty Characterization Details
and Input Data
L.l Cost Analysis Uncertainty Characterization
In addition to occurrence uncertainty, the national cost estimates reflect two additional sources of
uncertainty. The first is the total organic carbon concentration, which affects PFAS treatment
selection and is a factor for the DBP co-benefits analysis. The second is the unit cost curve
selection. The following subsections provide additional details on EPA's approach to modeling
these sources of uncertainty.
L.l.l Total Organic Carbon Concentration Uncertainty
For the national cost analysis, total organic carbon (TOC) is an input to the technology selection
and design equations for granular activated carbon (GAC). Section 5.3.1.1 of the Economic
Analysis provided a description of how TOC affects the decision tree for technology selection.
The process design equations in Section 5.3.1.1.1 show the effect of TOC on the estimation of
bed volumes for GAC.
As noted in Section 4.3.1.1 of the Economic Analysis, there is no national dataset of TOC values
or ranges at PWSs. Some data are available at the system level in periodic data voluntarily
provided by primacy agencies. EPA used the most recent data obtained in response to the ICR
for the fourth Six-Year Review of drinking water regulations. EPA separated the systems into
two groups - those with ground water sources and those with surface water sources - to reflect
expected variations in TOC in different types of source water. Some of the systems provided
TOC values at different facilities. Facilities can include water intakes or wells, treatment
processes, and distribution system entry points. TOC levels at systems that have treatment may
differ pre- and post-treatment.
EPA randomly assigned a TOC level to each entry point from the corresponding ground water or
surface water distribution. EPA retained that value for each of the 4,000 uncertainty simulations.
Thus, EPA's estimates reflect TOC uncertainty across entry points, but not TOC uncertainty
interacted with PFAS uncertainty.
L.l.2 Compliance Technology Unit Cost Curve Selection
Uncertainty
Each WBS model includes an input that determines whether the cost estimate generated is a low,
medium, or high cost estimate (U.S. EPA, 2023d). This input drives the selection of materials for
items of equipment that can be constructed of different materials. For example, a low cost system
might include fiberglass pressure vessels and PVC piping. A high cost system might include
stainless steel pressure vessels and stainless steel piping. This input also drives other model
assumptions that can affect the total cost including assumptions about building quality. High,
medium, and low quality settings affect building costs for substructure, superstructure, exterior
enclosure, interior finishes, and mechanical and electrical services.
Proposed PFAS Rule Economic Analysis
L-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
For every technology, EPA generated cost curves for low-, medium-, and high-cost options.
SafeWater MCBC randomly selects from these cost curves. EPA assigned a triangular
distribution to the cost curve selection: 25% probability for low-cost, 50% probability for
medium-cost, and 25% for high-cost.
L.2 Benefits Analysis Uncertainty Characterization
EPA characterizes sources of uncertainty in its analysis of potential benefits resulting from
changes in PFAS levels in drinking water. The analysis reports uncertainty bounds for benefits
estimated in each category modeled for the proposed rule. Each lower (upper) bound value is the
5th (95th) percentile of the category-specific benefits estimate distribution represented by 4,000
Monte Carlo draws. Table L-l provides the sources of uncertainty that EPA quantified in the
benefits analysis that are specific to this analysis. In addition to these sources of uncertainty,
reported uncertainty bounds also reflect the following upstream sources of uncertainty: baseline
PFAS occurrence (Section 4.4 of the Economic Analysis), affected population size and
demographic composition (Section 4.3 of the Economic Analysis), and the magnitude of PFAS
concentration reduction (Section 4.4 of the Economic Analysis).
Proposed PFAS Rule Economic Analysis
L-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table L-l: Quantified Sources of Uncertainty in Benefits Estimates
Source Description of Uncertainty
TC-serum
PFOA slope
factor; TC-
serum PFOS
slope factor
The slope factors that express the effects of PFOA and PFOS on serum lipid markers are
based on 12 key studies with high-quality data and clearly defined PFAS-lipid level
relationships (see Appendix F). EPA meta-analysis of these studies provides a central estimate
and a standard error estimate for the slope factors. EPA uses a normal distribution with a
mean set at the central slope factor estimate and a standard deviation set at the standard error
estimate for the slope factor to characterize uncertainty surrounding these parameters.
BP-serum PFOS
slope factor
The slope factor that expresses the effects of serum PFOS on systolic BP is from Liao et al.
(2020) - a high confidence study conducted based on U.S. general population data from
NHANES cycles 2003-2012. This study provides a central estimate and a standard error
estimate for the slope factor. EPA uses a normal distribution with a mean set at the central
slope factor estimate and a standard deviation set at the standard error estimate for the slope
factor to characterize uncertainty surrounding this parameter.
BW-serum
PFOA slope
factor; BW-
serum PFOS
slope factor
The slope factors were obtained from meta-analyses of several studies on the subject:
Steenland et al. (2018) for PFOA and an EPA reanalysis of Dzierlenga, Crawford, et al.
(2020) for PFOS.b The meta-analyses provide a central estimate and a standard error estimate
for the slope factors. EPA uses a normal distribution with a mean set at the central slope factor
estimate and a standard deviation set at the standard error estimate for the slope factor to
characterize uncertainty surrounding these parameters.
RCC-serum
PFOA slope
factor
The slope factor that expresses the effects of serum PFOA exposure on lifetime RCC risk is
from Shearer et al. (2021), which estimated a higher slope factor for the impact of PFOA on
RCC than previous estimates (Steenland et al., 2012; Vieira et al., 2013).° This study provides
a central estimate and a standard error estimate for the slope factor. EPA uses a normal
distribution with a mean set at the central slope factor estimate and a standard deviation set at
the standard error estimate for the slope factor to characterize uncertainty surrounding this
parameter.
Bladder cancer-
THM4 slope
factor
The slope factor that expresses the effect of co-occurring THM4 on bladder cancer is from
Regli et al. (2015), who estimated a linear slope factor relating the lifetime bladder cancer risk
associated with lifetime exposure to THM4 concentration in drinking water. This study
provides a central estimate for the slope factor. EPA estimated a standard error for this slope
factor based on the data reported in Regli et al. (2015). EPA uses a normal distribution with a
mean set at the central slope factor estimate and a standard deviation set at the standard error
estimate for the slope factor to characterize uncertainty surrounding this parameter.
RCC PAF to cap
risk reductions
for this endpoint
EPA developed a central tendency estimate and an uncertainty distribution for the PAF values
to cap the relative risk estimates derived from the RCC exposure-response relationship.
Abbreviations: ASCVD -Atherosclerotic Cardiovascular Disease; BW - birth weight; BP - blood pressure; CVD -
cardiovascular disease; PAF - population attributable fraction; PFAS - per
and polyfluoroalkyl substances; PFOA - perfluorooctanoic acid; PFOS - perfluorooctane sulfonic acid; RCC - renal cell
carcinoma; TC - total cholesterol; THM4- four regulated trihalomethanes.
Notes:
aThe slope factors contributing to the CVD benefits analysis include the relationship between total cholesterol and PFOA
and PFOS, the relationship between high-density lipoprotein cholesterol and PFOA and PFOS, and the relationship between
blood pressure and PFOS.
bIn the original Dzierlenga, Crawford, et al. (2020) estimate, the authors duplicated an estimate from Chen et al. (2017) in the
pooled estimate. EPA reran the analysis excluding the duplicated estimate.
CA sensitivity analysis of the RCC slope factor based on alternate estimates from Vieira et al. (2013) and pooled estimates of
studies included in Shearer et al. (2021) and Vieira et al. (2013) is shown in Appendix K.
As described in Section 6.1 of the Economic Analysis, EPA did not characterize the following
sources of potential uncertainty: U.S. population life tables (including standard and cause-
eliminated life tables; See Section 6.1.3 of the Economic Analysis), annual all-cause and health
Proposed PFAS Rule Economic Analysis
L-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
outcome-specific mortality rates, CVD risk model (Goff et al., 2014) predictors (e.g., share of
smokers) estimated from health survey data, prevalence of CVD event history in the U.S.
population, distribution of CVD events by type, the estimated infant mortality-birth weight slope
factor (See Section 6.4.3.1 of the Economic Analysis), state-level distributions of infant births
and infant deaths over discrete birth weight ranges, the 200-g cap on birth weight changes
estimated under the rule, COI estimates for all modeled non-fatal health outcomes, the VSL
reference value, the VSL income elasticity value used for VSL income growth adjustment, and
the gross domestic product per capita projection used to for VSL income growth adjustment (see
Appendix J). EPA expects that the sources listed in Table L-l, in addition to uncertainty
surrounding about estimated PFAS occurrence, affected population size, and the magnitude of
PFAS reduction, account for the largest portion of uncertainty in the benefits analysis.
L.2.1 Exposure-Response Function Uncertainty
Table L-2 presents the central tendency estimates, 95% confidence interval bounds (2.5th and
97.5th quantile), and standard errors for the slope factors used in EPA's assessment of benefits
resulting from the PFAS NPDWR. This table also presents information on the uncertainty
distribution used by EPA to characterize uncertainty for each slope factor.
Proposed PFAS Rule Economic Analysis
L-4
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table L-2: Standard Errors and Distributions for Benefits Model Exposure-Response Slope Factors
Pollutant
Health
Benefits
Health
Exposure-Response Slope Factor
Uncertainty
Data Source
Analysis
Category
Outcome
Central
Estimate
LCB
UCB
Standard
Error
Units
Distribution
CVD
TC
1.57
0.02
3.13
0.79
mg/dT per ng/mL
Normal
EPA meta-analysis based on 12 studies
(see Appendix F)
PFOA
BW
BW
-10.5
-16.7
-4.4
3.14
g per ng/mT
Normal
Steenland et al. (2018)
RCC
RCC
0.00178
0.00005
0.00352
0.00
per ng/mL
Normal
Shearer et al. (2021)
CVD
TC
0.08
-0.01
0.16
0.04
mg/dL per ng/mL
Normal
EPA meta-analysis based on 12 studies
(see Appendix F)
PFOS
BP
0.044
0.006
0.083
0.02
mmHg per ng/mL
Normal
Liao et al. (2020)
BW
BW
-3.0
-4.9
-1.1
0.97
g per ng/mL
Normal
EPA reanalysis of Dzierlenga, Crawford, et al.
(2020)
THM4
Bladder
cancer
Bladder
cancer
0.00427
0.00331
0.00522
0.00
per ug/L
Normal
Reglietal. (2015)
Abbreviations: BW - birth weight; BP - blood pressure; CVD - cardiovascular disease; HDLC - high-density lipoprotein cholesterol; LCB - lower confidence bound,
2.5% quantile; PFAS - per and polyfluoroalkyl substances; PFOA - perfluorooctanoic acid; PFOS - perfluorooctane sulfonic acid; RCC - renal cell carcinoma;
TC - total cholesterol; THM4- four regulated trihalomethanes; UCB - upper confidence bound, 97.5% quantile.
Proposed PFAS Rule Economic Analysis
L-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
L.2.2 Population Attributable Fraction Uncertainty
As described in Section 6.6 of the Economic Analysis and ICF (2022), EPA placed a PAF-based
cap on the estimated RCC risk reductions associated with changes in serum PFOA exposure.
EPA used a log-uniform distribution (also known as reciprocal) to approximate the distribution
of PAF estimates given existing PAF estimates for other specific environmental exposures and
other specific cancers (i.e., nitrate exposure in drinking water and colon cancer). The minimum
of the distribution was set at the smallest identified PAF estimate (0.2%) and the maximum was
set at the largest identified estimated PAF (17.9%). EPA used 3.94% (i.e., the mean of this log-
uniform distribution) for as the central estimate of the PAF-based cap on the RCC relative risk
reductions.
Proposed PFAS Rule Economic Analysis
L-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix M. Environmental Justice
This appendix provides additional detail on EPA's environmental justice (EJ) analysis. This
includes discussion of results from EPA's EJ exposure analysis using EJScreen for PWS service
areas in categories 4 and 5.
M.l Demographic Profile of Category 4 and 5 PWS Service
Areas
Table M-l summarizes the number of PWSs, size of PWSs, and population served for category 4
PWS service areas. There are 311 category 4 PWSs serving a population of 610,463, or 0.18% of
the overall U.S. population; 98% of category 4 PWSs are small systems, serving 546,478 people.
Table M-2 summarizes the demographic profile of category 5 PWS service areas. There are 148
category 5 PWSs serving a population of 578,751, or 0.17% of the overall U.S. population. 97%
percent of category 5 PWSs are small systems, serving 519,924 people.
Table M-3 summarizes the demographic profile for category 4 and 5 PWS service areas
combined and compares it to the demographic characteristics of the overall U.S. population.
Population served by category 4 and 5 PWS service areas account for 0.2% of the U.S.
population. Compared to the overall U.S. population, the population served by category 4 and 5
PWSs have lower percentages of American Indian or Alaska Native populations, Asian and
Pacific Islander populations, Black populations, and Hispanic populations and populations with
income less than twice the poverty level. Category 4 and 5 PWS service areas also have
relatively higher percentages of non-Hispanic White populations and populations with income
above twice the federal poverty level. Among category 4 and 5 PWS service areas, there are no
Native American-owned community water systems.
Proposed PFAS Rule Economic Analysis
M-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
Table M-l: Number of Category 4 PWSs and Population Served by Size and State
MARCH 2023
Population
Total Population Population Served in Served in
Served Small Systems" Medium and
Large Systems
New Jersey 311 98% 610,463 546,478 63,985
Abbreviation: PWS - public water system.
Note:
aSmall systems are defined as serving populations of 10,000 people or less.
„ , ..,, ^ Percent of Small
Number 01 Systems
State PWSs
Table M-2: Number of Category 5 PWSs and Population Served by Size and State
State
Number of Service
Areas
Percent Small Service
Areas
Total Population
Served
Population
Served in Small
Systems"
Population Served in
Medium and Large
Systems
Alabama
3
100%
9,955
9,955
-
Colorado
24
96%
94,604
83,737
10,867
Illinois
15
100%
44,945
44,945
-
Kentucky
8
100%
45,099
45,099
-
Massachusetts
7
86%
42,807
31,044
11,763
Michigan
30
93%
130,011
105,728
24,283
New Hampshire
7
100%
14,795
14,795
-
New Jersey
5
100%
4,177
4,177
-
Ohio
34
100%
123,566
123,566
-
South Carolina
7
86%
42,008
30,094
11,914
Vermont
8
100%
26,784
26,784
-
TOTAL
148
97%
578,751
519,924
58,827
Abbreviation: PWS - public water system.
Note:
aSmall systems are defined as serving populations of 10,000 people or less.
Proposed PFAS Rule Economic Analysis
M-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table M-3: Population Served by Category 4 and 5 PWSs Compared to Percent of U.S. Population by Demographic Group
Results
Race and Ethnicity
Income
American Asian and
Indian or Pacific
Alaska Native Islander
Black
Hispanic
Non-Hispanic
White
Below
Twice the
Poverty
Level
Above
Twice
the
Poverty
Level
Total
Population
Served
Population Served
Percent of Total
Population Served
U.S. Population
Percent by
Demographic Group
Percent Difference
Between Population
Served and U.S.
Population
3,281
0.3%
0.8%
-0.5%
27,098
2.3%
5.8%
-3.5%
68,682
5.8%
12.6%
-6.8%
96,511
8.1%
18.2%
-10.1%
976,596 290,466 898,748 1,189,214
82.1% 24.4% 75.6% 100%
60.1%
22.0%
29.8% 70.2% 326,569,308
-5.4% 5.4%
Proposed PFAS Rule Economic Analysis
M-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
M.2 Exposure Analysis Results
M.2.1 Baseline Scenario
Table M-4 summarizes the population served by category 4 and 5 PWS service areas with PFAS
occurrence above baseline thresholds based on Method 537.1 detection limits. The second set of
rows in Table M-4 summarizes the percentage of the total population served by demographic
group with modeled PFAS occurrence above these baseline thresholds. Percentages are bolded
and italicized when the percentage of the population in a specific demographic group exposed to
modeled PFAS above the baseline threshold is greater than the percentage of the total population
served across all demographic groups exposed to modeled PFAS above this threshold (right-hand
column). Higher percentages indicate higher PFAS exposure for a given demographic group
compared to the percentage of the total population served across all demographic groups.
Notably, PFAS occurrence above the baseline thresholds is higher for Asian and Pacific Islander
across all PFAS analytes compared to the total population served across all demographic groups.
The difference in exposure in PFAS occurrence is even greater when compared to non-Hispanic
White populations, or nearly 10% higher for PFOS and 15% higher for PFOA. PFAS occurrence
above baseline thresholds is higher for Black and Hispanic populations for PFOS, PFHxS, and
PFOA compared to the total population served across all demographic groups. When compared
to non-Hispanic White populations instead of the total population served across all demographic
groups, Black populations face over 10% greater PFOS exposure, while Hispanic populations
face nearly 10% greater PFOA exposure. PFAS occurrence above the baseline threshold is
higher for American Indian or Alaska Native populations for PFOS compared to the total
population served across all demographic groups. However, for other PFAS analytes, exposure
for American Indian or Alaska Native populations is less than or similar to occurrence rates for
the total population served across all demographic groups. PFAS occurrence above the baseline
thresholds is generally lower for populations with income below twice the federal poverty level
compared to occurrence for the total population served across all demographic groups.
Populations with income above the twice the poverty level have comparable but slightly higher
occurrence in comparison to the total population served across all demographic groups.
Table M-5 expands on this analysis, showing average population-weighted PFAS concentrations
across demographic groups in category 4 and 5 PWSs. Cells are highlighted in yellow when the
average concentration for a given demographic group is higher than the average for the total
population served across all demographic groups. These results are very similar to those of Table
M-4, demonstrating again that Asian and Pacific Islander as well as Hispanic populations have
higher average exposure than the total population served across all demographic groups in
category 4 and 5 PWSs.
Proposed PFAS Rule Economic Analysis
M-4
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table M-4: Baseline Scenario: Population Served by Category 4 and 5 PWS Service Areas Above Baseline Thresholds and
as a Percent of Total Population Served
Race and Ethnicity Income
Results
Analyte
American
Indian and
Alaska
Native
Asian and
Pacific
Islander
Black
Hispanic
Non-
Hispanic
White
Below
Twice
the
Poverty
Level
Above
Twice
the
Poverty
Level
Population
Served
PFOS
688
7,690
20,061
23,785
184,981
56,145
181,620
237,765
Population Served
PFHxS
225
3,184
6,581
10,543
63,241
18,648
64,954
83,602
Above Baseline
Threshold
PFHpA
65
4,297
2,666
8,486
85,439
18,059
83,652
101,711
PFOA
724
11,316
23,956
35,861
259,752
74,091
260,928
335,019
Population Served
PFOS
21.0%
28.4%
29.2%
24.6%
18.9%
19.3%
20.2%
20.0%
Above Baseline
PFHxS
6.9%
11.7%
9.6%
10.9%
6.5%
6.4%
7.2%
7.0%
Threshold as a
Percent of
PFHpA
2.0%
15.9%
3.9%
8.8%
8.7%
6.2%
9.3%
8.6%
Total Population
PFOA
22.1%
41.8%
34.9%
37.2%
26.6%
25.5%
29.0%
28.2%
Served
Abbreviations: PFHpA - Perfluoroheptanoic acid; PFHxS - Perfluorohexanesulfonic acid; PFOA - Perfluorooctanoic Acid PFOS - Perfluorooctanesulfonic Acid.
Proposed PFAS Rule Economic Analysis
M-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table M-5: Average PFAS Concentrations (ppt) by Demographic Group in the Baseline, Category 4 and 5 PWS Service
Areas
Race and Ethnicity
Income
PFAS
American
Indian or
Alaska
Native
Asian and
Pacific
Islander
Black
Hispanic
People of
Color3
Non-
Hispanic
White
Below
Twice the
Poverty
Level
Above
Twice the
Poverty
Level
loiai
Population
Served
PFOS
0.94
1.97
1.71
1.93
1.68
1.37
1.41
1.43
1.42
PFHxS
0.25
0.72
0.41
0.57
0.49
0.39
0.38
0.42
0.41
PFHpA
0.06
0.61
0.12
0.36
0.29
0.27
0.17
0.31
0.27
PFOA
0.94
3.59
1.83
2.19
2.13
1.65
1.29
1.88
1.73
Abbreviations: PFHpA - perfluoroheptanoic acid; PFHxS - perfluorohexanesulfonic acid; PFOA - perfluorooctanoic acid; PFOS - perfluorooctanesulfonic acid.
Note:
aThe demographic group people of color includes individuals who identify as Hispanic and/or a race other than White. It is calculated from EJScreen's percent minority
indicator and is non-duplicative across race and ethnicity categories.
Proposed PFAS Rule Economic Analysis
M-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
M.2.2 Hypothetical Regulatory Scenario #1: UCMR 5 MRLs
Table M-6 summarizes the results for populations served by category 4 and 5 PWS service areas
with PFAS occurrence above UCMR 5 MRL values. For this hypothetical regulatory scenario,
EPA assumed that PWSs with PFAS system-level means above the MRL value will reduce
PFAS levels to comply with the proposed rule. The first set of rows in Table M-6 summarizes
population served by category 4 and 5 PWS service areas with modeled PFAS occurrence above
the UCMR 5 MRLs. The second set of rows provides these estimates as a percentage of the total
population served by PWS service areas included in EPA's analysis.
Percentages are bolded and italicized when the percentage of the population in a specific
demographic group with PFAS occurrence above the MRL is greater than the percentage of the
total population served across all demographic groups with PFAS occurrence above the MRL
(right-hand column). Under this hypothetical regulatory scenario, where MCLs are assumed to
be equal to UCMR 5 MRL values, these populations would be expected to experience reductions
in PFAS exposure to below the hypothetical regulatory thresholds.
EPA's EJ exposure analysis shows that PFAS occurrence above the UCMR 5 MRL values is
higher for Asian and Pacific Islander, Black, and Hispanic populations for particular PFAS
analytes in comparison to occurrence over the MRL for the total population served across all
demographic groups. Specifically, PFAS occurrence above the UCMR 5 MRL values is higher
for Asian and Pacific Islander populations for all four PFAS analytes. Notably, for all PFAS
analytes, occurrence above the MRL values for Asian and Pacific Islander populations are
roughly two times the occurrence above the MRL values for the total population served across
all demographic groups. PFAS occurrence above the UCMR 5 MRL values is higher for
Hispanic populations for all four PFAS analytes. In each case, occurrence above the MRL values
for Hispanic populations is more than 1% greater than PFAS occurrence for the total population
served across all demographic groups. PFAS occurrence above the UCMR 5 MRL values is
higher for Black populations for PFOS, PFHxS, and PFOA compared to the total population
served across all demographic groups, with notable differences for PFHxS specifically. PFAS
occurrence above UCMR 5 MRL values is lower for populations with income below twice the
federal poverty level in comparison to PFAS occurrence for the total population served across all
demographic groups.
Table M-7 presents average population-weighted PFAS reductions across demographic groups in
category 4 and 5 PWSs under a hypothetical regulatory scenario where system-level means are
reduced to UCMR 5 MRL values. Cells are highlighted when the average concentration for a
given demographic group is higher than the average for the total population served across all
demographic groups. These results suggest a very similar pattern to the findings of Table M-6.
Nevertheless, Table M-7 suggests that reductions in PFOS and PFHxS are smaller for Black
populations than for the total population served across all demographic groups, despite Black
populations having a higher percent of potentially exposed population in category 4 and 5 PWSs.
Proposed PFAS Rule Economic Analysis
M-7
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
M.2.3 Hypothetical Regulatory Scenario #2:10.0 ppt
Table M-8 summarizes the results of the population served by category 4 and 5 PWS service
areas with PFAS occurrence above 10.0 ppt. For this hypothetical regulatory scenario, EPA
assumed that PWSs with PFAS system-level means above 10.0 ppt will reduce PFAS levels to
comply with the proposed rule. Percentages are bolded and italicized when the percentage of the
population in a specific demographic group with PFAS occurrence above 10.0 ppt is greater than
the percentage of the total population served across all demographic groups with PFAS
occurrence above 10.0 ppt (right-hand column). Under this hypothetical regulatory scenario,
where MCLs are assumed to be equal to 10.0 ppt, these populations would be expected to
experience reductions in PFAS exposure to below the hypothetical regulatory thresholds.
EPA's EJ exposure analysis shows that PFAS occurrence above 10.0 ppt is higher for Asian and
Pacific Islander, Hispanic, and non-Hispanic White populations for particular PFAS.
Specifically, PFAS occurrence above 10.0 ppt is higher for Asian and Pacific Islander
populations for PFOA and PFOS. Exceedances of 10.0 ppt for Asian and Pacific Islander
populations are the highest of any demographic group, with PFOA occurrence in particular being
roughly threefold the occurrence rate for the total population served across all demographic
groups. PFAS occurrence above 10.0 ppt is higher for Hispanic individuals than for the total
population served across all demographic groups for PFOA and PFOS. PFAS occurrence above
10.0 ppt is slightly higher for non-Hispanic White individuals than for the total population across
all demographic groups for PFHxS, with a difference in occurrence of only 0.1%.PFAS
occurrence above 10.0 ppt is generally somewhat lower for populations with income below twice
the federal poverty level in comparison to the occurrence rate for the total population served
across all demographic groups. One exception is for PFHxS occurrence for populations with
income below twice the federal poverty level, which is 0.3% higher than occurrence for the total
population served across all demographic groups.
Table M-9 presents average population-weighted PFAS reductions across demographic groups in
category 4 and 5 PWSs under a hypothetical regulatory scenario where system-level means are
reduced to 10.0 ppt. Cells are highlighted when the average concentration for a given
demographic group is higher than the average for the total population served across all
demographic groups. As in Table M-8, reductions in PFOA and PFOS are larger for Hispanic
populations than the total population served across all demographic groups. Asian and Pacific
Islander populations see the greatest reductions in PFOA of any demographic group. Notably,
Table M-9 demonstrates that individuals with income below twice the poverty level have greater
reductions in PFOS and PFHxS than the total population served across all demographic groups in
category 4 and 5 PWSs.
Proposed PFAS Rule Economic Analysis
M-8
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table M-6: Hypothetical Regulatory Scenario #1: Demographic Breakdown of Population Served by Category 4 and 5 PWS
Service Areas Above UCMR 5 MRL and as a Percent of Total Population Served
Race and Ethnicity Income
Results
Analyte
American
Indian
and
Alaska
Native
Asian and
Pacific
Islander
Black
Hispanic
Non-
Hispanic
White
Below
Twice the
Poverty
Level
Above
Twice the
Poverty
Level
Population
Served
PFOS
144
4,809
9,824
12,443
95,665
25,751
96,851
122,602
Population Served
PFHxS
120
2,623
5,731
7,877
39,317
12,786
42,843
55,629
Above UCMR 5 MRL
PFHpA
5
2,438
753
4,919
31,069
5,171
34,062
39,233
PFOA
370
9,233
12,647
19,592
164,953
39,879
168,584
208,463
Population Served
PFOS
4.4%
17.7%
14.3%
12.9%
9.8%
8.9%
10.8%
10.3%
Above UCMR 5 MRL
PFHxS
3.7%
9.7%
8.3%
8.2%
4.0%
4.4%
4.8%
4.7%
as a Percent of
Total Population
PFHpA
0.2%
9.0%
1.1%
5.1%
3.2%
1.8%
3.8%
3.3%
Served
PFOA
11.3%
34.1%
18.4%
20.3%
16.9%
13.7%
18.8%
17.5%
Abbreviations: PFHpA - Perfluoroheptanoic acid; PFHxS - Perfluorohexanesulfonic acid; PFOA - Perfluorooctanoic Acid PFOS - Perfluorooctanesulfonic Acid.
Proposed PFAS Rule Economic Analysis
M-9
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table M-7: Reductions in Average PFAS Concentrations (ppt) by Demographic Group in a Hypothetical Regulatory
Scenario with Maximum Contaminant Level at the UCMR 5 MRLs, Category 4 and 5 PWS Service Areas
Race and Ethnicity Income
PFAS
American
Indian or
Alaska
Native
Asian and
Pacific
Islander
Black
Hispanic
People of
Color3
White, Non-
Hispanic
Below
Twice the
Poverty
Level
Above
Twice the
Poverty
Level
loiai
Population
Served
PFOS
0.30
0.90
0.65
1.00
0.75
0.68
0.72
0.68
0.69
PFHxS
0.05
0.32
0.10
0.22
0.18
0.17
0.17
0.17
0.17
PFHpA
0.00
0.20
0.02
0.11
0.08
0.07
0.04
0.08
0.07
PFOA
0.26
1.97
0.63
0.97
0.92
0.76
0.48
0.89
0.79
Abbreviations: PFHpA - perfluoroheptanoic acid; PFHxS - perfluorohexanesulfonic acid; PFOA - perfluorooctanoic acid; PFOS - perfluorooctanesulfonic acid.
Note:
aThe demographic group people of color includes individuals who identify as Hispanic and/or a race other than White. It is calculated from EJSCREEN's percent minority
indicator and is non-duplicative across race and ethnicity categories.
Proposed PFAS Rule Economic Analysis
M-10
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table M-8: Hypothetical Regulatory Scenario #2: Demographic Breakdown of Population Served by Category 4 and 5 PWS
Service Areas Above 10.0 ppt and as a Percent of Total Population Served
Race and Ethnicity Income
Results
Analyte
American
Indian and
Alaska
Native
Asian and
Pacific
Islander
Black
Hispanic
Non-
Hispanic
White
Below
Twice the
Poverty
Level
Above
Twice the
Poverty
Level
Population
Served
PFOS
13
1,229
1,008
3,690
28,318
6,444
27,841
34,285
Population Served
PFHxS
6
29
251
223
5,870
2,400
4,008
6,408
Above 10.0 ppt
PFHpA
0
0
0
0
0
0
0
0
PFOA
26
3,145
935
4,838
41,674
5,658
45,205
50,863
Population Served
PFOS
0.4%
4.5%
1.5%
3.8%
2.9%
2.2%
3.1%
2.9%
Above 10.0 ppt as a
Percent of
Total Population
Served
PFHxS
0.2%
0.1%
0.4%
0.2%
0.6%
0.8%
0.4%
0.5%
PFHpA
PFOA
0.0%
0.8%
0.0%
11.6%
0.0%
1.4%
0.0%
5.0%
0.0%
4.3%
0.0%
1.9%
0.0%
5.0%
0.0%
4.3%
Abbreviations: PFHpA - Perfluoroheptanoic acid; PFHxS - Perfluorohexanesulfonic acid; PFOA - Perfluorooctanoic Acid PFOS - Perfluorooctanesulfonic Acid.
Proposed PFAS Rule Economic Analysis
M-ll
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Table M-9: Reductions in Average PFAS Concentrations (ppt) by Demographic Group in a Hypothetical Regulatory
Scenario with Maximum Contaminant Level at 10.0 ppt, Category 4 and 5 PWS Service Areas
Race and Ethnicity Income
PFAS
American
Indian or
Alaska
Native
Asian and
Pacific
Islander
Black
Hispanic
People of
Color3
White, Non-
Hispanic
Below
Twice the
Poverty
Level
Above
Twice the
Poverty
Level
loiai
Population
Served
PFOS
0.21
0.34
0.26
0.55
0.35
0.34
0.42
0.31
0.34
PFHxS
0.01
0.01
0.02
0.02
0.02
0.04
0.05
0.03
0.04
PFHpA
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
PFOA
0.01
0.71
0.09
0.35
0.28
0.26
0.12
0.31
0.26
Abbreviations: PFHpA - perfluoroheptanoic acid; PFHxS - perfluorohexanesulfonic acid; PFOA - perfluorooctanoic acid; PFOS - perfluorooctanesulfonic acid.
Note:
aThe demographic group people of color includes individuals who identify as Hispanic and/or a race other than White. It is calculated from EJScreen's percent minority
indicator and is non-duplicative across race and ethnicity categories.
Proposed PFAS Rule Economic Analysis
M-12
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix N. Supplemental Cost Analyses
Section N.l discusses the approach EPA used to estimate the costs of the rule for PWSs serving
more than 1 million people. Section N.2 discusses the potential impact on national costs if PWSs
must dispose of treatment residuals as hazardous waste. Section N.3 explores the potential
impact of limited PFAS occurrence data on national cost estimates.
N.l Cost Analysis for Very Large Systems
EPA identified 25 PWS that serve more than one million people based on retail population
estimates in SDWIS/Fed. All of these systems are CWS with multiple entry points; most are
surface water systems (see Table N-l).
Table N-l: Characteristics of PWSs Serving a Retail Population Greater than One
Million
PWSID
Name
SDWIS/Fed
Retail
Population
Water
Source
Entry
Points
AZ0407025
Phoenix, City Of
1,579,000
SW
20
CAO110005
East Bay Municipal Utility District
1,405,000
SW
5
CA1910067
Los Angeles-City, Dept. Of Water & Power
4,041,284
SW
11
CA3710020
San Diego - City Of
1,394,515
SW
3
CA4310011
San Jose Water
1,007,514
SW
3
CO0116001
Denver Water Board
1,362,071
SW
3
FL4130871
Miami-Dade Water and Sewer Department - Main System
2,300,000
GW
3
GA1210001
Atlanta
1,089,893
SW
2
IL0316000
Chicago
2,700,000
SW
2
MA6000000
Massachusetts Water Resources Authority
2,550,000
SW
2
MD0150005
Washington Suburban Sanitary Commission
1,800,000
SW
2
MD0300002
Baltimore City
1,600,000
SW
3
M06010716
Missouri American St Louis County St Charles County
1,100,000
SW
4
NC0160010
Charlotte Water
1,093,901
SW
2
NV0000090
Las Vegas Valley Water District
1,502,604
SW
10
NY5110526
Suffolk County Water Authority
1,100,000
GW
236
NY7003493
New York City System
8,271,000
SW
4
OH1801212
Cleveland Public Water System
1,308,955
SW
4
OH2504412
Columbus Public Water System
1,233,879
SW
3
PA1510001
Philadelphia Water Department
1,600,000
SW
3
TX0150018
San Antonio Water System
1,999,472
SW
38
TX0570004
Dallas Water Utility
1,286,380
SW
3
TX1010013
City of Houston
2,221,706
SW
41
TX2270001
City of Austin Water & Wastewater
1,044,405
SW
3
VA6059501
Fairfax County Water Authority
1,074,422
SW
2
Abbreviations: GW - ground water; PWS - public water system; SDWIS/Fed - Safe Drinking Water Information System
Federal Data Warehouse; SW - surface water.
Rather than model treatment costs using the PFAS occurrence values simulated from the MCMC
model, EPA reviewed UCMR3 data and recent system consumer confidence reports to obtain
entry point PFAS values. EPA used these values to determine which entry points at these
systems exceed the MCLs and/or HI for the proposed rule and alternative options.
Proposed PFAS Rule Economic Analysis
N-l
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
PFOA and PFOS levels at multiple entry points for two systems exceeded one or more MCLs for
the proposed option and alternative options (no HI exceedances occurred). EPA used these
reported PFAS values as the baseline occurrence estimates for the cost analysis. EPA applied the
cost estimating methods described in Chapter 5 to these systems to derive point estimate of the
costs to meet each MCL. EPA also estimated the corresponding affected entry point service
population.
System 1 has multiple entry points with existing GAC treatment. EPA determined that meeting
PFAS MCLs would reduce GAC bed life, thereby increasing O&M costs associated with more
frequent media change-outs and disposal. To estimate service population and average flow rates
at these entry points, EPA divided total system population by the number of entry points. This is
the same assumption made for all systems serving less than one million and, therefore,
introduces the same type of uncertainty to the cost analysis.
System 2 also has multiple entry points that would need to be treated to meet the proposed rule
and alternatives. Because source water quality makes GAC infeasible and the high entry point
flow rates result in impractical quantities of IX pressure vessels, EPA assumed that the system
would install RO processes and estimated the removal rates and blending rates, which vary with
baseline PFAS levels across entry points and reduction targets across regulatory options. EPA
obtained the design and average flows data for the existing treatment plants. Based on relative
average flow rates, EPA proportionately allocated the system service population across the entry
points (e.g., allocating 25% of system population to an entry point accounting for 25% of total
system average flow).
N.2 Hazardous Waste Disposal Cost Impacts
The national cost analysis reflects the assumption that PFAS-contaminated wastes are not
considered hazardous wastes. As a general matter, EPA notes that such wastes are not currently
regulated under federal law as a hazardous waste. To address stakeholder concerns, including
those raised during the SBREFA process, EPA conducted a sensitivity analysis with an
assumption of hazardous waste disposal for illustrative purposes only. As part of this analysis,
EPA generated a second full set of unit cost curves that are identical to the curves used for the
national cost analysis with the exception that spent GAC and spent IX resin are considered
hazardous. EPA acknowledges that if federal authorities later determine that PF AS-contaminated
wastes require handling as hazardous wastes, the residuals management costs are expected to be
higher.
For GAC, the national cost analysis assumes the spent media is reactivated off-site under current
RCRA non-hazardous waste regulations. Under this scenario, the WBS model uses a unit cost for
reactivation that includes transportation to the reactivation facility and back to the treatment
plant. To account for losses in the reactivation and replacement process, it also adds the cost of
replacing 30 percent of the spent GAC with virgin media. The hazardous waste sensitivity
analysis assumes spent GAC is disposed off-site as a hazardous waste in a RCRA Subtitle C
landfills and replaced with virgin GAC (i.e., single use operation). Under this scenario, the WBS
model incorporates the cost of hazardous waste disposal, transportation to a hazardous waste
facility 200 miles away, a minimum charge per hazardous waste shipment, and replacement of
100 percent of the spent GAC with virgin media. This scenario provides an upper bound on other
options that might emerge under future air quality regulations that prevent reactivation of PFAS-
Proposed PFAS Rule Economic Analysis
N-2
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
contaminated GAC (i.e., spent GAC must be disposed off-site as a non-hazardous waste and
replaced with virgin GAC) or RCRA hazardous waste regulations (i.e., off-site reactivation
remains feasible, but process wastes require hazardous waste disposal).
For IX, the national cost analysis assumes the spent resin is incinerated off-site under current
RCRA non-hazardous waste regulations. Under this scenario, the WBS model uses a unit cost for
non-hazardous incineration that includes transportation to the incineration facility. The
hazardous waste sensitivity analysis assumes spent resin is incinerated off-site as a hazardous
waste and replaced with virgin resin. Under this scenario, the WBS model incorporates the cost
of hazardous waste incineration, transportation to a hazardous waste facility 200 miles away, and
a minimum charge per hazardous waste shipment. Both scenarios incorporate the cost of
replacing the spent resin with virgin resin. Because hazardous waste incineration costs more than
disposal of spent resin in a hazardous waste landfill this hazardous waste scenario provides an
upper bound on other options that might emerge under future air quality regulations (e.g., off-site
disposal in a non-hazardous waste landfill) or RCRA hazardous waste regulations (e.g., off-site
disposal in a hazardous waste landfill).
The potential impact on PWS treatment costs is shown in Table N-2 for the proposed option. At
a 3 percent discount rate, the annualized cost would be $30M higher (4%) higher if hazardous
waste disposal is required. At a 7 percent discount rate, PWS treatment costs would be $61
million (6%) higher if hazardous waste disposal is required. Note that these estimated costs do
not include the costs associated with the storage, transportation and underground injection of the
brine concentrate residuals from the RO/NF process that could possibly be required under a
PFAS hazardous waste scenario.
Table N-2: Annualized PWS Treatment Cost Associated with Non-Hazardous and
Hazardous Residual Management Requirements, Proposed Option (PFOA and PFOS
MCLs of 4.0 ppt and HI of 1.0) (Million $2021)
3% Discount Rate 7% Discount Rate
5^ Mean 95* 5* Mean 95*
Percentile Percentile Percentile Percentile
Non-Hazardous $6l9 $673 $741 $1,012 $1,101 $1,206
Disposal
Hazardous $661 $703 $769 $1,092 $1,162 $1,262
Disposal
Increase due to $30 $61
Hazardous
Disposal
Note: Percentiles cannot be subtracted.
Proposed PFAS Rule Economic Analysis
N-3
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
N.3 Incremental Treatment Cost of Other PFAS
To illustrate the potential incremental costs for removing PFAS compounds for which national
occurrence data are not available, EPA used a model system approach. Table N-3 shows the
characteristics (population served, number of entry points, and resulting design and average
flows) of the ground water model systems in each of seven size categories. Table N-4 shows the
same characteristics used to simulate surface water system costs. Surface water systems tend to
have fewer entry points and, therefore, higher entry point flow rates and treatment costs.
Table N-3: Model System Characteristics for Ground Water Systems
System Size Category (Population Served)
Parameter
25-500
501-
3,301-
10,001-
50,001-
100,001-
>500,000
3,300
10,000
50,000
100,000
500,000
System Population3
A
110
1,140
5,476
18,660
66,549
163,947
826,664
System Design
Flowb (MGD)
System Average
Flowb (MGD)
B
0.049
0458
2.051
6.617
22.296
56.683
314.128
C
0.012
0.147
0.776
2.841
10.914
28.342
157.064
Entry Points0
D
1
2
2
4
10
12
39
Entry Point Design
Flow (MGD)
Entry Point Average
Flow (MGD)
E = B/D
0.049
0.229
1.025
1.654
2.230
4.724
8.055
F = C/D
0.012
0.074
0.388
0.710
1.091
2.362
4.027
Abbreviations: MGD - million gallons per day.
Notes:
a Median system populations are from USEPA Safe Drinking Water Information System Federal (SDWIS/Fed) fourth quarter
2021 "frozen" dataset that contains information reported through January 14,2022.
b Flow estimates are based on regression equations that relate population and design or average flows, derived in U.S. EPA
(2000b).
c Entry point data from 2006 Community Water System Survey (U.S. EPA, 2009), Table 13, values rounded to nearest whole
number.
Proposed PFAS Rule Economic Analysis
N-4
March 2023
-------
DRAFT FOR PUBLIC COMMENT
Table N-4: Model System Characteristics for Surface Water Systems
MARCH 2023
System Size Category (Population Served)
Parameter
25-500
501-
3,301-
10,001-
50,001-
100,001-
>500,000
3,300
10,000
50,000
100,000
500,000
System Population3
A
110
1,140
5,476
18,660
66,549
163,947
826,664
System Design
Flowb (MGD)
System Average
Flowb (MGD)
B
0.050
0.459
2.026
6.459
21.500
50.438
232.945
C
0.015
0.156
0.748
2.538
9.017
22.155
111.175
Entry Points0
D
1
1
1
1
2
2
4
Entry Point Design
Flow (MGD)
Entry Point Average
Flow (MGD)
E = B/D
0.050
0.459
2.026
6.459
10.750
25.219
58.236
F = C/D
0.015
0.156
0.748
2.538
4.509
11.077
27.794
Abbreviations: MGD - million gallons per day.
Notes:
^Median system populations are from USEPA Safe Drinking Water Information System Federal (SDWIS/Fed) fourth quarter
2021 "frozen" dataset that contains information reported through January 14,2022.
bFlow estimates are based on regression equations that relate population and design or average flows, derived in U.S. EPA
(2000b).
cEntry point data from 2006 Community Water System Survey (U.S. EPA, 2009), Table 13, values rounded to nearest whole
number.
Given these characteristics, EPA considered three types of systems:
• Baseline System: this model system has occurrence of all three potentially regulated
PFAS included in the national analysis (PFOA, PFOS, and PFHxS). It reflects the costs
that are covered in the national analysis and provides a basis for comparison.
• System Type 1: this model system has no detections of PFOA, PFOS, or PFHxS.
However, it has occurrence of all the other PFAS included in the HI (HFPO-DA, PFBS,
and PFNA). EPA considered two scenarios for this system type: high occurrence of the
three PFAS and medium occurrence of the three PFAS. This system type represents
additional systems that are not currently captured in the national costs but would incur
treatment costs under the HI.
• System Type 2: this model system has occurrence of PFOA, PFOS, and PFHxS identical
to the baseline system. It also has occurrence of all the other PFAS included in the HI
(HFPO-DA, PFBS, and PFNA). Like System Type 1, EPA considered two scenarios:
high occurrence of the three other PFAS and medium occurrence of those PFAS. This
system type illustrates a range of potential incremental treatment costs for systems that
are already treating to remove PFOA, PFOS, and/or PFHxS in the national cost analysis.
Table N-5 shows the occurrence assumptions for each system type. Concentrations for PFOA,
PFOS, and PFHxS correspond to the median for each contaminant from the UCMR3 data,
considering detected values only. Concentrations for the other PFAS are 95th percentile and
median values based on EPA's analysis of state-level occurrence data.
Proposed PFAS Rule Economic Analysis
N-5
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Table N-5: PFAS Occurrence Assumptions for Model System Analysis (ppt)
PFAS
Compound
Baseline
System3
System Type la
System Type 2a
High
Medium
High
Medium
PFOA
30
0
0
30
30
PFOS
60
0
0
60
60
PFHxS
73
0
0
73
73
HFPO-DA
0
29.65
12.6
29.65
12.6
PFBS
0
25.285
5.2
25.285
5.2
PFNA
0
56.4
8.5
56.4
8.5
Abbreviations: HFPO-DA - Hexafluoropropylene oxide dimer acid; PFAS - Per- and polyfluoroalkyl substances; PFBS -
Perfluorobutanesulfonic acid; PFHxS - Perfluorohexanesulfonic acid; PFNA - Perfluorononanoic acid; PFOA - Perfluorooctanoic
Acid PFOS - Perfluorooctanesulfonic Acid.
Notes:
aValues of zero indicate no detection of that PFAS.
Given these occurrence assumptions and basic characteristics, EPA estimated a range of costs for
model systems in each size category for each of the three treatment technologies (GAC, IX, and
RO/NF). The range of costs reflects all combinations of two source waters (ground and surface)
and two cost levels (low and high). For GAC and IX, the range of costs also incorporates two
bed life scenarios corresponding to a range of influent water quality, as discussed below.
For GAC, the lower end of the cost range reflects a bed life corresponding to an influent TOC of
0.5 mg/L, which is a typical detection limit for TOC. The upper end of the range corresponds to
an influent TOC of 2 mg/L, which is approximately the median for surface water systems and the
85th percentile for ground water systems. Beyond 2 mg/L influent TOC, GAC bed life may
make GAC usage less practical, based on the linear equations from U.S. EPA (2023d). However,
the maximum influent TOC value of 2 mg/L used in this analysis should not be regarded as a
strict limit on the practicality of GAC. The bed life equations are based on pooled data from a
limited number of studies and reflect central tendency results under varying water quality
conditions. They should not be used in lieu of site-specific engineering analyses or pilot studies
to estimate bed life or treatment costs for specific individual treatment systems. Individual
systems might achieve longer GAC bed lives and lower treatment costs at higher influent TOC
concentrations, particularly if their influent concentrations of PFAS are lower than the values
assumed in this analysis.
For IX, the lower end of the cost range reflects a bed life corresponding to a total influent PFAS
concentration that is the sum of the initial influent concentrations of the regulated PFAS shown
in Table N-5 (i.e., the lower bound assumes that no other PFAS compounds are present). The
upper end of the range assumes additional PFAS compounds are present such that total influent
PFAS is approximately 7,000 ppt. Data are not available to estimate bed life for higher influent
concentrations using the linear equations from U.S. EPA (2023d). IX costs are uncertain beyond
this value, but it should not be regarded as a strict limit on the feasibility of the technology.
Sources of uncertainty in this analysis include the following:
Proposed PFAS Rule Economic Analysis
N-6
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• EPA does not have sufficient quantitative data to include PFNA in the linear equations
used to estimate bed life for GAC and IX. For GAC, EPA assumes the bed life for these
PFAS is the same as PFOS. For IX, EPA assumes the bed life is the same as PFOA.
Given the chain length of PFNA, these assumptions likely underestimate the actual bed
life and err on the side of higher costs for GAC and IX for model systems type 1 and 2.
• EPA does not have sufficient quantitative data to include HFPO-DA in the bed life
estimates for IX. The IX costs for model system types 1 and 2 do not account for
occurrence of HFPO-DA and, therefore, are underestimates.
Table N-6 shows the results for baseline model systems. The costs shown are in $l,000's
per year ($2020) and reflect total capital costs annualized over the useful life of the
technology utilizing a 7 percent cost of capital rate.57, plus annual O&M costs. The cost
estimates for baseline systems form the basis of comparison for the other model systems.
Given the PFAS concentrations shown in Table N-5, the baseline GAC costs are
controlled by the removal requirements for PFHxS, IX costs are controlled by removal of
PFOA, and RO/NF costs are controlled by removal of PFOS.
Table N-6: Annualized Costs for Baseline Systems ($l,000's per year)
System Size
(Population Served)
GAC
IX
RO/NF
25-500
$17 to $31
$16 to $28
$161 to $203
501-3,300
$62 to $142
$58 to $103
$273 to $526
3,301-10,000
$293 to $680
$254 to $457
$705 to $999
10,001-50,000
$656 to $1,828
$665 to $1,285
$1,667 to $2,816
50,001-100,000
$2,026 to $5,773
$2,188 to $4,175
$4,855 to $9,070
100,001-500,000
$3,902 to $12,266
$4,874 to $9,134
$9,555 to $17,873
>500,000
$16,070 to $62,515
$22,817 to $46,556
$40,655 to $82,519
Abbreviations: GAC - granular activated carbon; IX - ion exchange; RO/NF - reverse osmosis/nanofiltration.
Table N-7 and Table N-8 show results for the type 1 model systems with high and medium
occurrence, respectively. These results reflect potential costs at additional systems triggered into
treatment and not captured in the national analysis of treatment costs. Overall, type 1 systems
have estimated costs ranging from 0.70 to 1.77 times baseline system costs. These results vary by
occurrence scenario and by technology, as discussed below.
Type 1 systems with high occurrence have estimated costs slightly lower to somewhat higher
than systems captured in the national analysis (0.92 to 1.77 times baseline). These results vary by
technology as follows:
57 The 7 percent cost of capital represents the cost to systems for debt service associated with expenditures on capital equipment
for treatment and is used in the annualization of engineering capital cost over the useful life of the technology. Note the use of 7
percent in this case is distinct from the 7 and 3 percent values used to adjust all national level costs and benefits to account for the
differential timing of compliance costs and resultant benefit impacts.
Proposed PFAS Rule Economic Analysis
N-7
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
• For GAC, removal of the other HI compounds results in a shorter bed life than required
to remove PFHxS in the baseline scenario, with correspondingly higher costs. The impact
of the shorter bed life increases with increasing system size, resulting in costs from 1.00
to 1.77 times baseline.
• For IX, removal of the other HI compounds results in a bed life that is not substantially
different from that required to remove PFOA in the baseline scenario. As discussed in the
body of this document, the WBS cost curves for IX differentiate at 20,000 bed volume
increments. The difference in estimated bed life between baseline and type 1 systems
with high occurrence is less than this increment, so cost results are coincidentally
identical (1.00 times baseline). As discussed above, however, the IX costs in Table N-7
do not account for occurrence of HFPO-DA and, therefore, are underestimates.
• For RO/NF, the removal efficiency required for the other HI compounds is slightly lower
than that required to remove PFOS in the baseline scenario. Correspondingly, costs are
also slightly lower (0.92 to 1.00 times baseline).
Type 1 systems with medium occurrence have estimated costs that are the same as or somewhat
lower than systems captured in the national analysis (0.70 to 1.00 times baseline). GAC costs are
very similar (0.99 to 1.00 times baseline). Estimated costs for IX and RO/NF are lower than
baseline (0.77 to 0.96 for IX, 0.70 to 0.98 for RO/NF). As discussed above, however, the IX
costs in Table N-7 do not account for occurrence of HFPO-DA and, therefore, are
underestimates.
Proposed PFAS Rule Economic Analysis
N-8
March 2023
-------
DRAFT FOR PUBLIC COMMENT
Table N-7: Results for Type 1 Systems for High PFAS Occurrence
MARCH 2023
System Size
(Population Served)
GAC
IX
RO/NF
Annualized Cost ($l,000's per year)
25-500
$17 to $38
$16 to $28
$161 to $202
501-3,300
$63 to $201
$58 to $103
$269 to $521
3,301-10,000
$296 to $902
$254 to $457
$682 to $920
10,001-50,000
$666 to $2,703
$665 to $1,285
$1,622 to $2,740
50,001-100,000
$2,063 to $9,195
$2,188 to $4,175
$4,714 to $8,819
100,001-500,000
$3,995 to $21,084
$4,874 to $9,134
$9,229 to $17,396
>500,000
$16,539 to $110,936
$22,817 to $46,556
$39,203 to $80,018
Ratio to Corresponding Baseline Cost
25-500
1.00 to 1.23
1.00
0.99 to 1.00
501-3,300
1.02 to 1.41
1.00
0.98 to 0.99
3,301-10,000
1.01 to 1.33
1.00
0.92 to 0.97
10,001-50,000
1.01 to 1.48
1.00
0.97
50,001-100,000
1.02 to 1.59
1.00
0.97
100,001-500,000
1.02 to 1.72
1.00
0.97
>500,000
1.03 to 1.77
1.00
0.96 to 0.97
Abbreviations: GAC - granular activated carbon; IX - ion exchange; RO/NF - reverse osmosis/nanofiltration.
Proposed PFAS Rule Economic Analysis
N-9
March 2023
-------
DRAFT FOR PUBLIC COMMENT
Table N-8: Results for Type 1 Systems for Medium PFAS Occurrence
MARCH 2023
System Size
(Population Served)
GAC
IX
RO/NF
Annualized Cost ($l,000's per year)
25-500
$17 to $31
$16 to $27
$158 to $197
501-3,300
$62 to $142
$51 to $93
$235 to $478
3,301-10,000
$291 to $680
$219 to $403
$508 to $764
10,001-50,000
$653 to $1,828
$545 to $1,095
$1,260 to $2,172
50,001-100,000
$2,012 to $5,773
$1,763 to $3,457
$3,634 to $6,904
100,001-500,000
$3,865 to $12,266
$3,829 to $7,275
$6,837 to $13,638
>500,000
$15,880 to $62,515
$17,567 to $36,085
$28,432 to $62,378
Ratio to Corresponding Baseline Cost
25-500
1.00
0.96
0.97 to 0.98
501-3,300
0.99 to 1.00
0.87 to 0.91
0.86 to 0.91
3,301-10,000
0.99 to 1.00
0.86 to 0.88
0.72 to 0.76
10,001-50,000
0.99 to 1.00
0.82 to 0.85
0.76 to 0.77
50,001-100,000
0.99 to 1.00
0.81 to 0.83
0.75 to 0.76
100,001-500,000
0.99 to 1.00
0.79 to 0.80
0.72 to 0.76
>500,000
0.99 to 1.00
0.77 to 0.78
0.70 to 0.76
Abbreviations: GAC - granular activated carbon; IX - ion exchange; RO/NF - reverse osmosis/nanofiltration.
Table N-9 and Table N-10 show results for type 2 model systems with high and medium
occurrence, respectively. Comparing these costs to those in the baseline system costs in Table N-
6 shows the potential additional costs at systems incurring treatment costs under the national
analysis if additional PFAS occurrence data were available. Overall, the need to remove the other
HI compounds could increase treatment costs by 0 to 77 percent on a per-system basis. These
results vary by occurrence scenario and by technology, as discussed below.
For type 2 systems with high occurrence using GAC, treatment costs could increase by 0 to 77
percent. At the upper bound of the cost range, the high TOC influent combined with the need to
remove the other HI compounds (particularly HFPO-DA) results in a shorter bed life. The impact
of the shorter bed life on costs increases with increasing system size (23 to 77 percent increase in
costs). At the lower bound of the cost range, the change in bed life is less significant with respect
to other capital and operating costs, so the percent increase in cost is small (0 to 3 percent). For
type 2 systems with medium occurrence, the change in GAC bed life is small, resulting in a
relatively small increase in cost (0 to 9 percent).
For type 2 systems using IX in both occurrence scenarios, there is no increase in treatment cost.
IX performance and cost remain controlled by removal of PFOA. As discussed above, however,
the IX costs in Table N-9 and Table N-10 do not account for occurrence of HFPO-DA and,
therefore, are underestimates. HFPO-DA removal could result in increased cost over baseline for
type 2 systems.
Proposed PFAS Rule Economic Analysis
N-10
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
For type 2 systems with high occurrence using RO/NF, the increase in removal efficiency needed
to remove the other HI compounds is very small. The resulting increase in cost rounds to 0
percent (i.e., it is less than 0.5 percent). For type 2 systems with medium occurrence, RO/NF
performance and cost remain controlled by PFOS, so there is no increase in treatment cost.
Table N-9: Results for Type 2 Systems for High PFAS Occurrence
System Size
(Population Served)
GAC
IX
RO/NF
Annualized Cost ($l,000's per year)
25-500
$17 to $38
$16 to $28
$161 to $203
501-3,300
$63 to $201
$58 to $103
$274 to $527
3,301-10,000
$296 to $902
$254 to $457
$707 to $1,001
10,001-50,000
$666 to $2,703
$665 to $1,285
$1,671 to $2,822
50,001-100,000
$2,063 to $9,195
$2,188 to $4,175
$4,867 to $9,091
100,001-500,000
$3,995 to $21,084
$4,874 to $9,134 !
$9,582 to $17,913
>500,000
$16,539 to $110,936
$22,817 to $46,556 $40,777 to $82,733
Percent Increase from Baseline Cost
25-500
0% to 23%
0%
0%
501-3,300
2% to 41%
0%
0%
3,301-10,000
1% to 33%
0%
0%
10,001-50,000
1% to 48%
0%
0%
50,001-100,000
2% to 59%
0%
0%
100,001-500,000
2% to 72%
0%
0%
>500,000
3% to 77%
0%
0%
Abbreviations: GAC - granular activated carbon; IX - ion exchange; RO/NF - reverse osmosis/nanofiltration.
Proposed PFAS Rule Economic Analysis
N-ll
March 2023
-------
DRAFT FOR PUBLIC COMMENT
Table N-10: Results for Type 2 Systems for Medium PFAS Occurrence
MARCH 2023
System Size
(Population Served)
GAC
IX
RO/NF
Annualized Cost ($l,000's per year)
25-500
$17 to $32
$16 to $28
$161 to $203
501-3,300
$63 to $149
$58 to $103
$273 to $526
3,301-10,000
$294 to $711
$254 to $457
$705 to $999
10,001-50,000
$661 to $1,921
$665 to $1,285
$1,667 to $2,816
50,001-100,000
$2,043 to $6,114
$2,188 to $4,175
$4,855 to $9,070
100,001-500,000
$3,945 to $13,213
$4,874 to $9,134 !
89,555 to $17,873
>500,000
$16,288 to $67,984
$22,817 to $46,556 $40,655 to $82,519
Percent Increase from Baseline Cost
25-500
0%to 3%
0%
0%
501-3,300
l%to 5%
0%
0%
3,301-10,000
l%to 5%
0%
0%
10,001-50,000
l%to 5%
0%
0%
50,001-100,000
1% to 6%
0%
0%
100,001-500,000
l%to 8%
0%
0%
>500,000
l%to 9%
0%
0%
Abbreviations: GAC - granular activated carbon; IX - ion exchange; RO/NF - reverse osmosis/nanofiltration.
Proposed PFAS Rule Economic Analysis
N-12
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Appendix O. Appendix References
Agency for Healthcare Research and Quality. (2011). MEPSHC-128: 2009 Medical Conditions
File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H€~ 128
Agency for Healthcare Research and Quality. (2012a). MEPSHC-137: 2010 Medical Conditions
File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H€~ 13 7
Agency for Healthcare Research and Quality. (2012b). MEPS HC-139: Panel 14 Longitudinal
Data File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H€~ 13 9
Agency for Healthcare Research and Quality. (2013a). MEPSHC-146: 2011 Medical Conditions
File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H€~ 146
Agency for Healthcare Research and Quality. (2013b). MEPS HC-148: Panel 15 Longitudinal
Data File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H€~ 148
Agency for Healthcare Research and Quality. (2014a). MEPSHC-154: 2012 Medical Conditions
File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H€~ 154
Agency for Healthcare Research and Quality. (2014b). MEPS HC-156: Panel 16 Longitudinal
Data File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H€~ 156
Agency for Healthcare Research and Quality. (2015a). MEPS HC-162: 2013 Medical Conditions
File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail. isp?cboPuf
Number=H€~ 162
Agency for Healthcare Research and Quality. (2015b). MEPS HC-164: Panel 17 Longitudinal
Data File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail. isp?cboPuf
Number=H(
Agency for Healthcare Research and Quality. (2016a). MEPS HC-170: 2014 Medical Conditions
File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail. isp?cboPuf
Number=H€~ 170
Agency for Healthcare Research and Quality. (2016b). MEPSHC-172: Panel 18 Longitudinal
Data File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail. isp?cboPuf
Number=HC-172
Proposed PFAS Rule Economic Analysis
0-13
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Agency for Healthcare Research and Quality. (2017a). Healthcare Cost and Utilization Project.
Retrieved from: https://hcupnet.ahrq.gov
Agency for Healthcare Research and Quality. (2017b). MEPSHC-180: 2015 Medical Conditions
File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail. isp?cboPuf
Number=H€~ 180
Agency for Healthcare Research and Quality. (2017c). MEPSHC-183: Panel 19 Longitudinal
Data File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail. isp?cboPuf
Number=H€~ 183
Agency for Healthcare Research and Quality. (2018). MEPSHC-193: Panel 20 Longitudinal
Data File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H(
Agency for Healthcare Research and Quality. (2019a). MEPSHC-190: 2016Medical Conditions
File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H€~ 190
Agency for Healthcare Research and Quality. (2019b). MEPSHC-199: 2017Medical Conditions
File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H€~ 199
Agency for Healthcare Research and Quality. (2019c). MEPSHC-202: Panel 21 Longitudinal
Data File. Medical Expenditure Panel Survey. Retrieved from:
https://www.meps.ahrq.eov/mepsweb/data stats/download data files detail.isp?cboPuf
Number=H€~202
Agency for Toxic Substances and Disease Registry. (2018). Toxicological Profile for
Perfluoroalkyls.
Ahrens, K. A., Rossen, L. M., Thoma, M. E., Warner, M., & Simon, A. E. (2017). Birth order
and injury-related infant mortality in the U.S. American Journal of Preventive Medicine,
53(4), 412-420.
Alio, A. P., Mbah, A. K., Kornosky, J. L., Wathington, D., Marty, P. J., & Salihu, H. M. (2011).
Assessing the impact of paternal involvement on racial/ethnic disparities in infant
mortality rates. Journal of Community Health, 36, 63-68. doi:10.1007/sl0900-010-9280-
3
Almond, D., Chay, K. Y., & Lee, D. S. (2005). The costs of low birth weight. The Quarterly
journal of economics, 720(3), 1031-1083. doi: 10.1093/qje/120.3.1031
Ambavane, A., Yang, S., Atkins, M. B., Rao, S., Shah, A., Regan, M. M., . . . Michaelson, M. D.
(2020). Clinical and economic outcomes of treatment sequences for intermediate- to
poor-risk advanced renal cell carcinoma. Immunotherapy, 72(1), 37-51. doi:10.2217/imt-
2019-0199
Anderson, R. N. (1999). Methodfor constructing complete annual US life tables (Vol. 129):
National Ctr for Health Statistics.
Apelberg, B. J., Witter, F. R., Herbstman, J. B., Calafat, A. M., Halden, R. U., Needham, L. L.,
& Goldman, L. R. (2007). Cord serum concentrations of perfluorooctane sulfonate
Proposed PFAS Rule Economic Analysis
0-14
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
(PFOS) and perfluorooctanoate (PFOA) in relation to weight and size at birth. Children's
Health, 775(11), 1670-1676.
Arias, E., Heron, M., & Tejada-Vera, B. (2013). United States life tables eliminating certain
causes of death, 1999-2001.
Arias, E., & Xu, J. (2019). United States Life Tables, 2017.
Arizona Department of Environmental Quality. (2021). Public Water System PFAS Data.
Retrieved from https://static.azdeq.eov/wqd/pfas/lit isults.pdf
Ashley-Martin, J., Dodds, L., Arbuckle, T. E., Bouchard, M. F., Fisher, M., Morriset, A. S., . . .
Piatt, R. W. (2017). Maternal concentrations of perfluoroalkyl substances and fetal
markers of metabolic function and birth weight: the Maternal-Infant Research on
Environmental Chemicals (MIREC) study. American journal of epidemiology, 185(3),
185-193. doi: 10.1093/aje/kww213
Bach, C. C., Bech, B. H., Nohr, E. A., Olsen, J., Matthiesen, N. B., Bonefeld-Jorgensen E.C., . . .
Henriksen, T. B. (2016). Perfluoroalkyl acids in maternal serum and indices of fetal
growth: the Aarhus Birth Cohort. Environmental health perspectives, 124(6), 848-854.
doi:10.1289/ehp,1510046.
Backes, C. H., Markham, K., Moorehead, P., Cordero, L., Nankervis, C. A., & Giannone, P. J.
(2011). Maternal preeclampsia and neonatal outcomes. Journal of pregnancy.
Bang, H., & Tsiatis, A. A. (2000). Estimating medical costs with censored data. Biometrika,
87(2), 329-343.
Bartell, S. M. (2003). Statistical Methods for Nonsteady State Exposure Inference using
Biomarkers. University of California-Davis, Davis, CA.
Bartell, S. M. (2017). Online serum PFOA calculator for adults. Environmental health
perspectives, 725(10), 104502. doi:10.1289/EHP2820.
Bartell, S. M., Calafat, A. M., Lyu, C., Kato, K., Ryan, P. B., & Steenland, K. (2010). Rate of
decline in serum PFOA concentrations after granular activated carbon filtration at two
public water systems in Ohio and West Virginia. Environ Health Perspect, 118(2), 222-
228. doi: 10.1289/ehp.0901252
Beatty, A. L., Ku, I. A., Bibbins-Domingo, K., Christenson, R. H., DeFillippi, C. R., Ganz, P., . .
. Whooley, M. A. (2015). Traditional risk factors versus biomarkers for prediction of
secondary events in patients with stable Coronary Heart Disease: from the Heart and Soul
Study. Journal of the American Heart Association, 4. doi: 10.1161/JAHA. 114.001646
Begg, C. B., & Mazumdar, M. (1994). Operating characteristics of a rank correlation test for
publication bias. Biometrics, 50(4), 1088-1101. doi:https://doi.org/10.
Bell, E. M., Yeung, E. H., Ma, W., Kannan, K., Sundaram, R., Smarr, M. M., & Louis, G. M. B.
(2018). Concentrations of endocrine disrupting chemicals in newborn blood spots and
infant outcomes in the upstate KIDS study. Environment international, 121, 232-239.
Bhattachaijee, S., Gharaibeh, M., Kamal, M. U., & Riaz, I. B. (2017). Influence of Health and
Functional Status and Co-occurring Chronic Conditions on Healthcare Expenditures
Among Community-dwelling Adults With Kidney Cancer in the United States: A
Propensity-score-matched Analysis. Clinical Genitourinary Cancer, 75(3), e357-e368.
doi:https://doi.ore/.10 101 >/i.clec.J0 i>- IJ 01
Bureau of Labor Statistics. (2022). Employment Cost Index (NAICS). Total compensation for All
Civilian workers in All industries and occupations, Index. Series ID:
CIS10100000000001. Retrieved from: https://data.bls.gov/pdq/SurveyOutputServlet
Proposed PFAS Rule Economic Analysis
0-15
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Cadwallader, A., A. Greene, H. Holsinger, A. Lan, M. Messner, M. Simic, & Albert, R. (2022).
A Bayesian hierarchical model for estimating national PFAS drinking water occurrence.
A WW A Water Science, el284.
California Division of Drinking Water. (2020). Perfluorooctanoic acid (PFOA) and
Perjluorooctanesulfonic acid (PFOS). Retrieved from
https://www.waterboards.ca.gov/drinking water/certlic/drinkinewatei' l.html
Callaghan, W. M., MacDorman, M. F., Shapiro-Mendoza, C. K., & Barfield, W. D. (2017).
Explaining the recent decrease in US infant mortality rate, 2007-2013. American journal
of obstetrics and gynecology, 216(1), 73. e71-73. e78.
Callan, A. C., Rotander, A., Thompson, K., Heyworth, J., Meueller, J. F., Odland, J. O., &
Hinwood, A. L. (2016). Maternal exposure to perfluoroalkyl acids measured in whole
blood and birth outcomes in offspring. Science of the total environment, 569-570, 1107-
1113. doi:10.1016/j.scitotenv.2016.06.177
Cao, W., Liu, X., Liu, X., Zhou, Y., Zhang, X., Tian, H., . . . Bhatti, P. (2018). Perfluoroalkyl
substances in umbilical cord serum and gestational and postnatal growth in a Chinese
birth cohort. Environment international, 116, 197-205.
Centers for Disease Control and Prevention. (2013a). 2011-2012 Data Documentation,
Codebook, and Frequencies. Cholesterol - HDL (HDL G). National Health and
Nutrition Examination Survey. Retrieved from:
https://wwwn. cdc. gov/Nch s/Nh an es/^
Centers for Disease Control and Prevention. (2013b). 2011-2012 Data Documentation,
Codebook, and Frequencies. Cholesterol - Total (TCHOL G). National Health and
Nutrition Examination Survey. Retrieved from:
https://wwwn. cdc. gov/Nch s/Nh an es/^
Centers for Disease Control and Prevention. (2013c). 2011-2012 Data Documentation,
Codebook, and Frequencies. Diabetes (DIQ G). National Health and Nutrition
Examination Survey.
Centers for Disease Control and Prevention. (2013d). 2011-2012 Data Documentation,
Codebook, and Frequencies. Smoking - Cigarette Use (SMQ G). National Health and
Nutrition Examination Survey. Retrieved from:
https://wwwn.cdc.eov/Nchs/Nhanes/2011 -2012/SM
Centers for Disease Control and Prevention. (2015a). 2011-2012 Data Documentation,
Codebook, and Frequencies. Demographic Variables & Sample Weights (DEMO G).
National Health and Nutrition Examination Survey. Retrieved from:
https://wwwn.cdc.eov/Nchs/Nhanes/2011 -2012/DE.N ^ N ^ I
Centers for Disease Control and Prevention. (2015b). 2013-2014 Data Documentation,
Codebook, and Frequencies. Demographic Variables & Sample Weights (DEMO H).
National Health and Nutrition Examination Survey. Retrieved from:
https://wwwn.cdc.eov/Nchs/Nhanes/201 '« .014/DEN ^ N I
Centers for Disease Control and Prevention. (2015c). 2013-2014 Data Documentation,
Codebook, and Frequencies. Diabetes (DIQ H). National Health and Nutrition
Examination Survey. Retrieved from: https://wwwn.cdc.eov/Nchs/Nhanes/2011 -
2
Centers for Disease Control and Prevention. (2015d). 2013-2014 Data Documentation,
Codebook, and Frequencies. Smoking - Cigarette Use (SMQ H). National Health and
Proposed PFAS Rule Economic Analysis
0-16
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Nutrition Examination Survey. Retrieved from:
https://wwwn.cdc.eov/Nchs/Nhanes/201 I JO IJ 'SM1-s U ^ I
Centers for Disease Control and Prevention. (2016a). 2013-2014 Data Documentation,
Codebook, and Frequencies. Cholesterol - HDL (HDL H). National Health and
Nutrition Examination Survey. Retrieved from:
https://wwwn.cdc.gov/Nchs/Nhanes/2013-2Q14/HDL H.XPT
Centers for Disease Control and Prevention. (2016b). 2013-2014 Data Documentation,
Codebook, and Frequencies. Cholesterol - Total (TCHOL H). National Health and
Nutrition Examination Survey. Retrieved from:
https://wwwn.cdc.gov/Nchs/Nhanes/^ 'TCHOL H.XPT
Centers for Disease Control and Prevention. (2016c, February 11). STDs During Pregnancy -
CDC Detailed Fact Sheet. Retrieved from https://www.cdc. gov/std/pregnancv/stdfact-
pregnancv-detailed.htm#r5
Centers for Disease Control and Prevention. (2017a). 2015-2016 Data Documentation,
Codebook, and Frequencies. Cholesterol - HDL (HDL I). National Health and Nutrition
Examination Survey. Retrieved from: https://wwwn.cdc.gov/Nchs/Nhanes/2Q 15-
2016/HDL I.XPT
Centers for Disease Control and Prevention. (2017b). 2015-2016 Data Documentation,
Codebook, and Frequencies. Cholesterol - Total (TCHOL I). National Health and
Nutrition Examination Survey. Retrieved from:
https://wwwn.cdc.eov/Nchs/Nhanes/2015-2016/TCHQL I.XPT
Centers for Disease Control and Prevention. (2017c). 2015-2016 Data Documentation,
Codebook, and Frequencies. Demographic Variables & Sample Weights (DEMO I).
National Health and Nutrition Examination Survey. Retrieved from:
https://wwwn. cdc. eov/Nchs/Nhanes/2015 -2016/DE
Centers for Disease Control and Prevention. (2017d). 2015-2016 Data Documentation,
Codebook, and Frequencies. Diabetes (DIQ I). National Health and Nutrition
Examination Survey. Retrieved from: https://wwwn.cdc.eov/Nchs/Nhanes/2015-
2
Centers for Disease Control and Prevention. (2017e). 2015-2016 Data Documentation,
Codebook, and Frequencies. Smoking - Cigarette Use (SMQ I). National Health and
Nutrition Examination Survey. Retrieved from:
https://wwwn.cdc.eov/Nchs/Nhanes/2015-2016/SMi M \< I
Centers for Disease Control and Prevention. (2017f). User Guide to the 2017 Period/2016
Cohort Linked Birth/Infant Death Public Use File.
Centers for Disease Control and Prevention. (2018). User Guide to the 2018 Period/2017 Cohort
Linked Birth/Infant Death Public Use File.
Centers for Disease Control and Prevention. (2020a). Commit to Healthy Choices to Prevent
Birth Defects. Retrieved from https://www.cdc.eov/ncbddd/birthdefects/prevention.html
Centers for Disease Control and Prevention. (2020b). Pregnancy Complications.
Centers for Disease Control and Prevention. (2020c). Underlying Cause of Death, 1999-2020 on
CDC WONDER Online Database, released in 2020. Data are from the Multiple Cause of
Death Files, 1999-2019, as compiledfrom data provided by the 57 vital statistics
jurisdictions through the Vital Statistics Cooperative Program. Retrieved from:
http://wonder.cdc.eov/ucd-icdlO.html
Proposed PFAS Rule Economic Analysis
0-17
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Centers for Disease Control and Prevention. (2022). National Report on Human Exposure to
Environmental Chemicals. Updated March 2022. Retrieved from
https://www.cdc.eov/exposurereport/
Chang, H. Y., Nuyten, D. S. A., Sneddon, J. B., Hastie, T., Tibshirani, R., S0rlie, T., . . . van de
Vijver, M. J. (2005). Robustness, scalability, and integration of a wound-response gene
expression signature in predicting breast cancer survival. Proc. Natl. Acad. Sci. U. S. A.,
702(10), 3738-3743. doi:10.1073/pnas.0409462102
Chateau-Degat, M.-L., Pereg, D., Dallaire, R., Ayotte, P., Dery, S., & Dewailly, E. (2010).
Effects of perfluorooctanesulfonate exposure on plasma lipid levels in the Inuit
population of Nunavik (Northern Quebec). Environmental research, 110(7), 710-717.
Chen, M. H., Ha, E. H., Wen, T. W., Su, Y. N., Lien, G. W., Chen, C. Y., . . . Hsieh, W. S.
(2012). Perfluorinated compounds in umbilical cord blood and adverse birth outcomes.
PLoS One, 7(8), e42474.
Chen, M. H., Ng, S., Hsieh, C. J., Lin, C. C., Hsieh, W. S., & Chen, P. C. (2017). The impact of
prenatal perfluoroalkyl substances exposure on neonatal and child growth. Science of the
total environment, 607-608, 669-675. doi:10.1016/j.scitotenv.2017.06.273
Chu, C., Zhou, Y., Li, Q.-Q., Bloom, M. S., Lin, S., Yu, Y.-J., . . . Yang, B.-Y. (2020). Are
perfluorooctane sulfonate alternatives safer? New insights from a birth cohort study.
Environment international, 135, 105365.
Cochran, W. G. (1954). The combination of estimates from different experiments. Biometrics,
70(1), 101-129. doi:htfps://dot.org/10.2307/3001666
Cole, T. J., Hey, E., & Richmond, S. (2010). The PREM score: A graphical tool for predicting
survival in very preterm births. Archives of Disease in Childhood-Fetal and Neonatal
Edition, 95(1), F14-F19.
Collins Jr, J. W., & David, R. J. (2009). Racial disparity in low birth weight and infant mortality.
Clinics in perinatology, 36(1), 63-73.
Colorado Department of Public Health and Environment. (2020). 2020 PFAS Sampling Project.
Convertino, M., Church, T. R., Olsen, G. W., Liu, Y., Doyle, E., Elcombe, C. R., . . . Evans, T.
R. J. (2018). Stochastic Pharmacokinetic-Pharmacodynamic Modeling for Assessing the
Systemic Health Risk of Perfluorooctanoate (PFOA). Toxicological Sciences, 163(1),
293 -3 06. doi: 10.1093/toxsci/kfy03 5
Cooke, R. W. I. (2010). Does Neonatal and Infant Neurodevelopmental Morbidity of Multiples
and Singletons Differ? Paper presented at the Seminars in Fetal and Neonatal Medicine.
Crainiceanu, C. M., Stedinger, J. R., Ruppert, D., & Behr, C. T. (2003). Modeling the U.S.
national distribution of waterborne pathogen concentrations with application to
Cryptosporidium parvum. Water Resources Research, 39(9).
doi :htfps://dot .org/10.1029/2002WR001664
D'Agostino, R. B., Grundy, S., Sullivan, L. M., Wilson, P., & Group, C. H. D. R. P. (2001).
Validation of the Framingham coronary heart disease prediction scores: results of a
multiple ethnic groups investigation. Jama, 286(2), 180-187. doi:10.1001/jama.286.2.180
D'Agostino, R. B., Russell, M. W., Huse, D. M., Ellison, R. C., Silbershatz, H, Wilson, P. W., &
Hartz, S. C. (2000). Primary and subsequent coronary risk appraisal: new results from the
Framingham study. Am. Heart J., 139(2 Pt 1), 272-281. doi: 10.1067/mhj.2000.96469
D'Agostino, R. B., Vasan, R. S., Pencina, M. J., Wolf, P. A., Cobain, M., Massaro, J. M., &
Kannel, W. B. (2008). General cardiovascular risk profile for use in primary care.
Circulation, 777(6), 743-753.
Proposed PFAS Rule Economic Analysis
0-18
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
D'Agostino, R. B., Wolf, P. A., Belanger, A. J., & Kannel, W. B. (1994). Stroke risk profile:
adjustment for antihypertensive medication. The Framingham Study. Stroke, 25(1), 40-
43. doi:10.1161/01 str.25.1.40
Darrow, L. A., Stein, C. R., & Steenland, K. (2013). Serum Perfluorooctanoic Acid and
Perfluorooctane Sulfonate Concentrations in Relation to Birth Outcomes in the Mid-Ohio
Valley, 2005-2010. Environ. Health Persped., 727(10), 1207-1213.
doi: 10.1289/ehp. 1206372
Deeks, J. J. (2002). Issues in the selection of a summary statistic for meta-analysis of clinical
trials with binary outcomes. Statistics in medicine, 27(11), 1575-1600.
Delaware Office of Drinking Water. (2021). Delaware Drinking Water Watch Data. Retrieved
from: https://drinkingwater.dhss.delaware.gov/
DerSimonian, R., & Laird, N. (1986). Meta-analysis in clinical trials. Controlled clinical trials,
7(3), 177-188.
Donahue, S. M., Kleinman, K. P., Gillman, M. W., & Oken, E. (2010). Trends in birth weight
and gestational length among singleton term births in the United States: 1990-2005.
Obstetrics and gynecology, 115(2 Pt 1), 357.
Dong, Z., Wang, H., Yu, Y. Y., Li, Y. B., Naidu, R., & Liu, Y. (2019). Using 2003-2014 US
NHANES data to determine the associations between per-and polyfluoroalkyl substances
and cholesterol: Trend and implications. Ecotoxicology and environmental safety, 173,
461-468.
Duval, S., & Tweedie, R. (2000a). A nonparametric "trim and fill" method of accounting for
publication bias in meta-analysis. J Am Stat Assoc, 95(449), 89-98.
doi:10.1080/01621459.2000.10473905
Duval, S., & Tweedie, R. (2000b). Trim and fill: A simple funnel-plot-based method of testing
and adjusting for publication bias in meta-analysis. Biometrics, 56(2), 455-463.
doi: 10.1111/j .0006-341X.2000.00455.X
Dzierlenga, M., Allen, B., Clewell, H., & Longnecker, M. (2020). Pharmacokinetic bias analysis
of an association between clinical thyroid disease and two perfluoroalkyl substances.
Environ Int, 141: 105784.
Dzierlenga, M., Crawford, L., & Longnecker, M. (2020). Birth weight and perfluorooctane
sulfonic acid: a random-effects meta-regression analysis. Environmental Epidemiology,
4(3).
Egger, M., Smith, G., Schneider, M., & Minder, C. (Eds.). (2008). Systematic reviews in health
care: Meta-analysis in context. London, UK: BMJ Publishing Group.
Egger, M., Smith, G. D., Schneider, M., & Minder, C. (1997). Bias in meta-analysis detected by
a simple, graphical test. BMJ, 375(7109), 629-634.
doi:https://doi.ore/10.1136/btiit '< I ^ 10° .9
Ely, D. M., & Driscoll, A. K. (2020). Infant Mortality in the United States, 2018: Data From the
Period Linked Birth/Infant Death File. Retrieved from
https://www.cdc.gov/nchs/data/nvsr/nvsr69/NVSR-69-7-508.pdf
Engels, E. A., Schmid, C. H., Terrin, N., Olkin, I., & Lau, J. (2000). Heterogeneity and statistical
significance in meta-analysis: an empirical study of 125 meta-analyses. Statistics in
medicine, 79(13), 1707-1728.
Eriksen, K. T., Raaschou-Nielsen, O., McLaughlin, J. K., Lipworth, L., Tj0nneland, A., Overvad,
K., & S0rensen, M. (2013). Association between plasma PFOA and PFOS levels and
total cholesterol in a middle-aged Danish population. PLoS One, 8(2), e56969.
Proposed PFAS Rule Economic Analysis
0-19
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Fan, Y., Li, X., Xu, Q., Zhang, Y., Yang, X., Han, X., . . . Lu, C. (2020). Serum albumin
mediates the effect of multiple per-and polyfluoroalkyl substances on serum lipid levels.
Environmental pollution, 266, 115138.
Fei, C., McLaughlin, J. K., Tarone, R. E., & Olsen, J. (2007). Perfluorinated Chemicals and Fetal
Growth: A Study within the DanishNational Birth Cohort. Environmental health
perspectives, 775(11), 1677-1682.
Fisher, M., Arbuckle, T. E., Wade, M., & Haines, D. A. (2013). Do perfluoroalkyl substances
affect metabolic function and plasma lipids?—Analysis of the 2007-2009, Canadian
Health Measures Survey (CHMS) Cycle 1. Environmental research, 121, 95-103.
Fitz-Simon, N., Fletcher, T., Luster, M. I., Steenland, K., Calafat, A. M., Kato, K., & Armstrong,
B. (2013). Reductions in serum lipids with a 4-year decline in serum perfluorooctanoic
acid and perfluorooctanesulfonic acid. Epidemiology (Cambridge, Mass.), 24(4), 569.
Fragki, S., Dirven, H., Fletcher, T., Grasl-Kraupp, B., Bjerve Giitzkow, K., Hoogenboom, R., . . .
Luijten, M. (2021). Systemic PFOS and PFOA exposure and disturbed lipid homeostasis
in humans: what do we know and what not? CritRev Toxicol, 141-164.
Fried, L. P., Borhani, N. O., Enright, P., Furberg, C. D., Gardin, J. M., Kronmal, R. A., . . .
Newman, A. (1991). The cardiovascular health study: design and rationale. Annals of
epidemiology, 7(3), 263-276.
Friedman, G. D., Cutter, G. R., Donahue, R. P., Hughes, G. H., Hulley, S. B., Jacobs Jr, D. R., . .
. Savage, P. J. (1988). CARDIA: study design, recruitment, and some characteristics of
the examined subjects. Journal of clinical epidemiology, ¥7(11), 1105-1116.
Fromme, H., Mosch, C., Morovitz, M., Alba-Alejandre, I., Boehmer, S., Kiranoglu, M., . . .
Koletzko, B. (2010). Pre-and postnatal exposure to perfluorinated compounds (PFCs).
Environmental Science & Technology, 44(18), 7123-7129.
Fu, Y., Wang, T., Fu, Q., Wang, P., & Lu, Y. (2014). Associations between serum concentrations
of perfluoroalkyl acids and serum lipid levels in a Chinese population. Ecotoxicology and
environmental safety, 106, 246-252.
Gabry, J., & Mahr, T. (2020). bayesplot: Plotting for Bayesian Models. Retrieved from
https://mc-stan.ore/bayesplot/
Gabry, J., Veen, D., Stan Development Team, Andreae, M., Betancourt, M., Carpenter, B., . . .
Trangucci, R. (2018). shinystan: Interactive Visual and Numerical Diagnostics and
Posterior Analysis for Bayesian Models. Retrieved from: https://cran.r-
proiect.org/web/packages/shinystan/index.html
Gao, K., Zhuang, T., Liu, X., Fu, J., Zhang, J., Fu, J., . . . Song, M. (2019). Prenatal exposure to
per-and polyfluoroalkyl substances (PFASs) and association between the placental
transfer efficiencies and dissociation constant of serum proteins-PFAS complexes.
Environmental Science & Technology, 53(11), 6529-6538.
Georgia Environmnetal Protection Division. (2020). EPA Sampling Report for Summerville.
Retrieved from https://epd.eeoreia.eov/pfoa-and-pfos-information
Goff, D. C., Lloyd-Jones, D. M., Bennett, G., Coady, S., D'Agostino, R. B., Gibbons, R., . . .
Wilson, P. W. F. (2014). 2013 ACC/AHA guideline on the assessment of cardiovascular
risk: A report of the American College of Cardiology/American Heart Association task
force on practice guidelines. Circulation, 129(25 suppl 2), 49-73.
Govarts, E., Remy, S., Bruckners, L., Hond, E. D., Sioen, I., Nelen, V., . . . Schoeters, G. (2016).
Combined Effects of Prenatal Exposures to Environmental Chemicals on Birth Weight.
Proposed PFAS Rule Economic Analysis
O-20
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
International journal of environmental research and public health, 73(495), 1-19.
doi:10.3390/ijerphl3050495
Hamm, M. P., Cherry, N. M., Chan, E., Martin, J. W., & Burstyn, I. (2010). Maternal exposure
to perfluorinated acids and fetal growth. Journal of Exposure Science and Environmental
Epidemiology, 20, 589-597. doi: 1559-0631/10
He, X., Liu, Y., Xu, B., Gu, L., & Tang, W. (2018). PFOA is associated with diabetes and
metabolic alteration in US men: National Health and Nutrition Examination Survey
2003-2012. Science of the total environment, 625, 566-574.
Higgins, J., Chandler, J., Cumpston, M., Page, M., & Welch, V. E. (2021). Cochrane Handbook
for Systematic Reviews of Interventions version 6.2 (updated February 2021). Cochrane.
Higgins, J., Thomas, J., Chandler, J., Cumpston, M., Li, T., Page, M. J., & Welch, V. A. (2019).
Cochrane handbook for systematic reviews of interventions: John Wiley & Sons.
Higgins, J., Thompson, S., Deeks, J., & Altman, D. (2003). Measuring inconsistency in meta-
analyses. BMJ, 327(7414), 557-560. doi:https://doi.ore/10.1 136/bi^i . Ill
Hjermitslev, M. H., Long, M., Wiels0e, M., & Bonefeld-J0rgensen, E. C. (2020). Persistent
organic pollutants in Greenlandic pregnant women and indices of foetal growth: The
ACCEPT study. Science of the total environment, 698, 134118.
Hoffman, K., Webster, T. F., Bartell, S. M., Weisskopf, M. G., Fletcher, T., & Vieira, V. M.
(2011). Private drinking water wells as a source of exposure to perfluorooctanoic acid
(PFOA) in communities surrounding a fluoropolymer production facility. Environmental
health perspectives, 779(1), 92-97.
Hollenbeak, C. S., Nikkei, L. E., Schaefer, E. W., Alemao, E., Ghahramani, N., & Raman, J. D.
(2011). Determinants of Medicare All-Cause Costs Among Elderly Patients with Renal
Cell Carcinoma. Journal of Managed Care Pharmacy, 77(8), 610-620.
doi:10.18553/jmcp.2011.17.8.610
Horon, I., Munson, K.M. (2020). Discussion of the Merits of Using Singleton Data from the
CDC to Develop a Relationship Between Birthweight and Infant Mortality. Email
Correspondence 11/10/2020.
Huang, M., Jiao, J., Zhuang, P., Chen, X., Wang, J., & Zhang, Y. (2018). Serum polyfluoroalkyl
chemicals are associated with risk of cardiovascular diseases in national US population.
Environment international, 119, 37-46.
ICF. (2020). Memorandum re: Trends in Infant Mortality by Birth Weight Categories, EPA
Contract No. 68HE0C18D0001, TO 68HERC20F0107. Submitted to U.S. Environmental
Protection Agency Office of Groundwater and Drinking Water July 30, 2020. Retrieved
from https://www.reeiilations.eov/docket/EPA-HQ-OW-2022-
ICF. (2021). Memorandum re: Review ofPFAS Lipids Study Quality. Submitted to U.S.
Environmental Protection Agency Office of Groundwater and Drinking Water March 23,
2021. Retrieved from https://www.reeulations.gov/docket/* " \ i iQ-QW-202 J 01 I I
ICF. (2022). Memorandum re: Literature Review of Population Attributable Fraction Estimates
and Next Steps for Renal Cell Carcinoma PFAS Benefits Modeling. Submitted to U.S.
Environmental Protection Agency Office of Groundwater and Drinking Water July 26,
2022. Retrieved from
Illinois Environmental Protection Agency. (2021). Illinois Drinking Water Watch. Retrieved
from: http://water.epa.state.il.us/dww/index.isp
Institute of Medicine. (1985). Preventing Low Birthweight. Washington, DC: The National
Academies Press.
Proposed PFAS Rule Economic Analysis
0-21
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Jacob, J. A. (2016). U.S. infant mortality rate declines but still exceeds other developed
countries. Journal of the American Medical Association, 315(5), 451-452.
Jain, R. B., & Ducatman, A. (2019). Roles of gender and obesity in defining correlations
between perfluoroalkyl substances and lipid/lipoproteins. Science of the total
environment, 653, 74-81.
Johnson, P. I., Sutton, P., Atchley, D. S., Koustas, E., Lam, J., Sen, S., . . . Woodruff, T. J.
(2014). The Navigation Guide—Evidence-Based Medicine Meets Environmental Health:
Systematic Review of Human Evidence for PFOA Effects on Fetal Growth.
Environmental health perspectives, 722(10), 1028-1039. doi:10.1289/ehp.1307893
Joyce, C., Goodman-Bryan, M., & Hardin, A. (2012). Preterm Birth and Low Birth Weight.
Retrieved from The Urban Child Institute:
http://www.iirbanchildinstitute.ore/sites/all/files ~and~LBW.pdf
Kannel, W. B., D'Agostino, R. B., Silbershatz, H., Belanger, A. J., Wilson, P. W., & Levy, D.
(1999). Profile for estimating risk of heart failure. Arch. Intern. Med., 759(11), 1197-
1204. doi:10.1001/archinte,159.11.1197
Kashino, I., Sasaki, S., Okada, E., Matsuura, H., Goudarzi, H., Miyashita, C., . . . Kishi, R.
(2020). Prenatal exposure to 11 perfluoroalkyl substances and fetal growth: A large-scale,
prospective birth cohort study. Environment international, 136, 105355.
Kassambara, A. (2020). ggpubr: 'ggplot2' Based Publication Ready Plots. Retrieved from:
https://cran.r-proiect.org/web/packages/ggpubr/index.html
Kentucky Department for Environmental Protection. (2019). Evaluation of Kentucky Community
Drinking Water for Per- & Poly-Fluoroalkyl Substances. Retrieved from
https://eec.ky. eov/Docum ents%20for%20IJRLs/PF AS%20Drinkin g%20W ater%20Repor
t%20Final.pdf
Kiely, J. L., Brett, K. M., Yu, S., & Rowley, D. L. (1994). Low birth weight and intrauterine
growth retardation. Birth Outcomes, 185-202.
Kim, S., Choi, K., Ji, K., Seo, J., Kho, Y., Park, J., . . . Giesy, J. P. (2011). Trans-Placental
Transfer of Thirteen Perfluorinated Compounds and Relations with Fetal Thyroid
Hormones. Environmental Science & Technology, 45(11), 7465-7472.
doi:10.1021/es202408a
Kochanek, K. D., Murphy, S. L., Xu, J., & Arias, E. (2019). Deaths: Final data for 2017.
National vital statistics reports : from the Centers for Disease Control and Prevention,
National Center for Health Statistics, National Vital Statistics System, 68(9), 1-77.
Kwon, E. J., Shin, J. S., Kim, B. M., Shah-Kulkarni, S., Park, H., Kho, Y. L., . . . Ha, E. H.
(2016). Prenatal Exposure to Perfluorinated Compounds Affects Birth Weight Through
GSTM1 Polymorphism. JOEM, 58(6), el98-e205.
Lau, C., Ambalavanan, N., Chakraborty, H, Wingate, M. S., & Carlo, W. A. (2013). Extremely
low birth weight and infant mortality rates in the United States. Pediatrics, 131(5), 855-
860.
Lauritzen, H. B., Larose, T. L., 0ien, T., Sandanger, T. M., Odland, J. O., van der Bor, M., &
Jacobsen, G. W. (2017). Maternal serum levels of perfluoroalkyl substances and
organochlorines and indices of fetal growth: a Scandinavian case-cohort study. Pediatric
research, 57(1), 33-42. doi:10.1038/pr.2016.187
Lee, E. S., Han, S., & Oh, J. E. (2016). Association between perfluorinated compound
concentrations in cord serum and birth weight using multiple regression model.
Reproductive Toxicology, 59, 53-59. doi:10.1016/j.reprotox.2015.10.020
Proposed PFAS Rule Economic Analysis
0-22
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Lenters, V., Portengen, L., Rignell-Hydbom, A., Jonsson, B. A. G., Lindh, C. H., Piersma, A. H.,
. . . Vermeulen, R. (2016). Prenatal Phthalate, Perfluoroalkyl Acid, and Organochlorine
Exposures and Term Birth Weight in Three Birth Cohorts: Multi-Pollutant Models Based
on Elastic Net Regression. Environmental health perspectives, 124, 365-372.
doi: 10.1289/ehp. 1408933
Li, M., Zeng, Z. W., Qian, Z. M., Vaughn, M. G., Suauve, S., Paul, G., . . . Dong, G. H. (2017).
Isomers of perfluorooctanesulfonate (PFOS) in cord serum and birth outcomes in China:
Guangzhou Birth Cohort Study. Environment international, 102, 1-8.
doi:10.1016/j.envint.2017.03.006
Li, S., Peng, Y., Wang, X., Qian, Y., Xiang, P., Wade, S. W., . . . Handelsman, Y. (2019).
Cardiovascular events and death after myocardial infarction or ischemic stroke in an
older Medicare population. Clinical cardiology, 42(3), 391-399.
Li, Y., Barregard, L., Xu, Y., Scott, K., Pineda, D., Lindh, C. H., . . . Fletcher, T. (2020).
Associations between perfluoroalkyl substances and serum lipids in a Swedish adult
population with contaminated drinking water. Environmental Health, 79(33).
Li, Y., Fletcher, T., Mucs, D., Scott, K., Lindh, C. H., Tallving, P., & Jakobsson, K. (2018).
Half-lives of PFOS, PFHxS and PFOA after end of exposure to contaminated drinking
water. Occupational and environmental medicine, 75(1), 46-51.
Liao, S., Yao, W., Cheang, I., Tang, X., Yin, T., Lu, X., . . . Li, X. (2020). Association between
perfluoroalkyl acids and the prevalence of hypertension among US adults. Ecotoxicology
and environmental safety, 196, 110589.
Lin, C. Y., Lee, H. L., Hwang, Y. T., & Su, T. C. (2020). The association between total serum
isomers of per-and polyfluoroalkyl substances, lipid profiles, and the DNA
oxidative/nitrative stress biomarkers in middle-aged Taiwanese adults. Environmental
research, 182, 109064.
Lin, L., & Chu, H. (2018). Quantifying publication bias in meta-analysis. Biometrics, 74(3), 785-
794.
Lin, P.-I. D., Cardenas, A., Hauser, R., Gold, D. R., Kleinman, K. P., Hivert, M.-F., . . . Horton,
E. S. (2019). Per-and polyfluoroalkyl substances and blood lipid levels in pre-diabetic
adults—longitudinal analysis of the diabetes prevention program outcomes study.
Environment international, 129, 343-353.
Lind, D. V., Priskorn, L., Lassen, T. H., Nielsen, F., Kyhl, H. B., Kristensen, D. M., . . . Jensen,
T. K. (2017). Prenatal exposure to perfluoroalkyl substances and anogenital distance at 3
months of age in a Danish mother-child cohort. Reproductive Toxicology, 68, 200-206.
Liu, H.-S., Wen, L.-L., Chu, P.-L., & Lin, C.-Y. (2018). Association among total serum isomers
of perfluorinated chemicals, glucose homeostasis, lipid profiles, serum protein and
metabolic syndrome in adults: NHANES, 2013-2014. Environmental pollution, 232, 73-
79.
Lockwood, J. R., Schervish, M. J., Gurian, P., & Small, M. J. (2001). Characterization of arsenic
occurrence in source waters of U.S. community water systems. Journal of the American
Statistical Association, 96(456), 1184-1193. doi: 10.1198/016214501753381832
Louis, G. M. B., Zhai, S., Smarr, M. M., Grewal, J., Zhang, C., Grantz, K. L., . . . Honda, M.
(2018). Endocrine disruptors and neonatal anthropometry, NICHD fetal growth studies-
singletons. Environment international, 119, 515-526.
Lu, S., & Bartell, S. M. (2020). Serum PFAS Calculator for Adults. Retrieved from:
www.ics.uci.edu/~sbartell/pfascalc.html
Proposed PFAS Rule Economic Analysis
0-23
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Ma, S., & Finch, B. K. (2010). Birth Outcome Measures and Infant Mortality. Population
Research and Policy Review, 29(6), 865-891.
Mahmood, S. S., Levy, D., Vasan, R. S., & Wang, T. J. (2014). The Framingham Heart Study
and the epidemiology of cardiovascular disease: a historical perspective. The Lancet,
353(9921), 999-1008.
Maine Department of Environmental Protection. (2020). MAINE PFAS DATA (2007 - 2020).
Retrieved from https://wwwl.maine.gov/dep/spills/topics/pfas/PFAS-current-results-
10262020.pdf
Maisonet, M., Terrell, M. L., McGeehin, M. A., Christensen, K. Y., Holmes, A., Calafat, A. M.,
& Marcus, M. (2012). Maternal concentrations of polyfluoroalkyl compounds during
pregnancy and fetal and postnatal growth in British girls. Environmental health
perspectives, 720(10), 1432-1437.
Manzano-Salgado, C. B., Casas, M., Lopez-Espinosa, M.-J., Ballester, F., Iniguez, C., Martinez,
D., . . . Schettgen, T. (2017). Prenatal exposure to perfluoroalkyl substances and birth
outcomes in a Spanish birth cohort. Environment international, 108, 278-284.
Marks, K. J., Cutler, A. J., Jeddy, Z., Northstone, K., Kato, K., & Hartman, T. J. (2019).
Maternal serum concentrations of perfluoroalkyl substances and birth size in British boys.
International journal of hygiene and environmental health, 222(5), 889-895.
Massachusetts Department of Energy and Environmental Affairs. (2021). Energy and
Environmental Affairs Data Portal. In.
Mathews, T. (2013). Infant mortality statistics from the 2010 period linked birth/infant death data
set. National vital statistics reports : from the Centers for Disease Control and
Prevention, National Center for Health Statistics, National Vital Statistics System, 62(8),
1-26.
McCormick, M. C. (1985). The contribution of low birth weight to infant mortality and
childhood morbidity. New England Journal of Medicine, 312(2), 82-90.
Mclntire, D. D., Bloom, S. L., Casey, B. M., & Leveno, K. J. (1999). Birth weight in relation to
morbidity and mortality among newborn infants. New England Journal of Medicine,
340(16), 1234-1238.
Meng, Q., Inoue, K., Ritz, B., Olsen, J., & Liew, Z. (2018). Prenatal exposure to perfluoroalkyl
substances and birth outcomes; an updated analysis from the Danish National Birth
Cohort. International journal of environmental research and public health, 15(9), 1832.
Michigan Environment, G. L., and Energy,. (2021). Michigan Compliance Monitoring Data.
Retrieved from https://www.michigan.gOv/pfasresponse/0.9Q38.7~365~
-~~.00.html
Minatoya, M., Itoh, S., Miyashita, C., Araki, A., Sasaki, S., Miura, R., . . . Kishi, R. (2017).
Association of prenatal exposure to perfluoroalkyl substances with cordblood adipokines
and birth size: The Hokkaido Study on environment and children's health. Environmental
research, 156, 175-182. doi:10.1016/j.envres.2017.03.033
Mishra, S., Ram, B., Singh, A., & Yadav, A. (2018). Birth order, stage of infancy and infant
mortality in India. Journal of biosocial science, 50(5), 604-625.
Mitchell, A. P., Dey, P., Ohn, J. A., Tabatabai, S. M., Curry, M. A., & Bach, P. B. (2020). The
Accuracy and Usefulness of the National Comprehensive Cancer Network Evidence
Blocks Affordability Rating. Pharmacoeconomics, 38(1), 737-745. doi:10.1007/s40273-
020-00901-x
Proposed PFAS Rule Economic Analysis
0-24
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Monroy, R., Morrison, K., Teo, K., Atkinson, S., Kubwabo, C., Stewart, B., & Foster, W. G.
(2008). Serum levels of perfluoroalkyl compounds in human maternal and umbilicalcord
blood samples. Environmental research, 108, 56-62. doi:10.1016/j.envres.2008.06.001
Mora, S. (2016). Nonfasting for Routine Lipid Testing: From Evidence to Action. JAMA Intern
Med, 176, 1005-1006.
National Center for Health Statistics, H., MD. (2019). Health, United States, 2018. Retrieved
from https://www.cdc.eov/nchs/data/hus/hiisl8.pdf
Negri, E., Metruccio, F., Guercio, V., Tosti, L., Benfenati, E., Bonzi, R., . . . Moretto, A. (2017).
Exposure to PFOA and PFOS and fetal growth: a critical merging of toxicological and
epidemiological data. Critical reviews in toxicology, 10.1080/10408444.2016.1271972,
489-515. doi: 10.1080/10408444.2016.1271972
Nelson, J. W., Hatch, E. E., & Webster, T. F. (2010). Exposure to polyfluoroalkyl chemicals and
cholesterol, body weight, and insulin resistance in the general US population.
Environmental health perspectives, 118(2), 197-202.
New Hampshire Department of Environmental Services. (2021). One Stop Data Portal. In.
New Jersey Department of Environmental Protection. (2021). New Jersey Drinking Water
Watch. In.
Ngui, E. M., Cortright, A. L., & Michalski, K. (2015). Relationship of paternity status, welfare
reform period, and racial/ethnic disparities in infant mortality. American Journal of
Men's Health, 9(5), 350-359. doi:10.1177/1557988314543906
North Dakota Department of Environmental Quality. (2020). North Dakota Statewide 2020 Per-
andPolyfluoroalkyl Substances (PFAS) Presence/Absence Survey. Retrieved from
https://deq.nd.eov/Publications/MF/PFAS Report 2020.pdf? v=l
O'Sullivan, A. K., Rubin, J., Nyambose, J., Kuznik, A., Cohen, D. J., & Thompson, D. (2011).
Cost estimation of cardiovascular disease events in the US. Pharmacoeconomics, 29(8),
693-704.
Occupational Safety and Health Administration. (2010). Occupational Exposure to Respirable
Crystalline Silica — Review of Health Effects Literature and Preliminary Quantitative
Risk Assessment.
Ohio Department of Health. (2020). Per- and Polyfluoroalkyl Substances (PFAS). Retrieved
from: https://data-oepa.opendata.arceis.com/datasets/pfas-sampline-results/explore
Osterman, M. J., Kochanek, K. D., MacDorman, M. F., Strobino, D. M., & Guyer, B. (2015).
Annual summary of vital statistics: 2012-2013. Pediatrics, 135(6), 1115-1125.
Ott, W. R. (1995). Environmental Statistics and Data Analysis (1st ed.). New York: CRC Press,
Lewis Publishers.
Palmer, T. M., & Sterne, J. A. C. (Eds.). (2016). Meta-analysis in Stata: An updated collection
from the Stata Journal (2nd ed.). College Station, TX: Stata Press.
PEACE Trial Investigators. (2004). Angiotensin-converting-enzyme inhibition in stable
coronary artery disease. New England Journal of Medicine, 351(2.0), 2058-2068.
Pencina, M. J., D'Agostino, R. B., Sr., Larson, M. G., Massaro, J. M., & Vasan, R. S. (2009).
Predicting the 30-year risk of cardiovascular disease: the framingham heart study.
Circulation, 119(24), 3078-3084. doi:10.1161/CIRCULATIONAHA.108.816694
Pencina, M. J., D'Agostino, R. B., Sr., & Steyerberg, E. W. (2011). Extensions of net
reclassification improvement calculations to measure usefulness of new biomarkers. Stat.
Med., 30(1), 11-21. doi:10.1002/sim.4085
Proposed PFAS Rule Economic Analysis
0-25
March 2023
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Pennsylvania Department of Environmental Protection. (2020). PFAS Phase 1 Results Summary.
Retrieved from
http://files.dep.state.pa.us/Water/DrinkingWater/Perfluorinated%20Chemicals/Sampling
Results/PF ASPhaselResultsSummary.pdf
Perrin, A., Sherman, S., Pal, S., Chua, A., Gorritz, M., Liu, Z., . . . Garrison, L. P. (2015).
Lifetime cost of everolimus vs axitinib in patients with advanced renal cell carcinoma
who failed prior sunitinib therapy in the US. Journal of Medical Economics, 18(3), 200-
209.
Powers, D. A., Frisbie, W. P., Hummer, R. A., Pullum, S. G., & Solis, P. (2006). Race/ethnic
differences and age-variation in the effects of birth outcomes on infant mortality in the
US. Demographic Research, 14, 179-216.
Quah, S. R. (2016). International Encyclopedia of Public Health. Waltham, MA: Academic
Press.
R Core Team. (2021). R: A language and environment for statistical computing. Vienna, Austria:
R Foundation for Statistical Computing.
Racsa, P. N., Whisman, T. R., & Worley, K. (2015). Comparing two tyrosine kinase inhibitors
for treatment of advanced renal cell carcinoma in Medicare and commercially insured
patients. Current Medical Research and Opinion, 37(10), 1933-1940.
Rappazzo, K. M., Coffman, E., & Hines, E. P. (2017). Exposure to perfluorinated alkyl
substances and health outcomes in children: a systematic review of the epidemiologic
literature. International journal of environmental research and public health, 14(1), 691.
Ratnasiri, A. W., Parry, S. S., Arief, V. N., DeLacy, I. H., Halliday, L. A., DiLibero, R. J., &
Basford, K. E. (2018). Recent trends, risk factors, and disparities in low birth weight in
California, 2005-2014: A retrospective study. Maternal health, neonatology and
perinatology, 4(1), 1-13.
Raudenbush, S. W. (2009). Analyzing effect sizes: Random-effects models. The handbook of
research synthesis and meta-analysis, 2, 295-316.
Ray, J. G., Park, A. L., & Fell, D. B. (2017). Mortality in infants affected by preterm birth and
severe small-for-gestational-age birth weight. Pediatrics, 140(6).
Regli, S., Chen, J., Messner, M., Elovitz, M. S., Letkiewicz, F. J., Pegram, R. A., . . . Wright, J.
M. (2015). Estimating potential increased bladder cancer risk due to increased bromide
concentrations in sources of disinfected drinking waters. Environmental Science &
Technology, 49(22), 13094-13102.
Reinhorn, D., Sarfaty, M., Leshno, M., Moore, A., Neiman, V., Rosenbaum, E., & Goldstein, D.
A. (2019). A cost-effectiveness analysis of nivolumab and ipilimumab versus sunitinib in
first-line intermediate-to poor-risk advanced renal cell carcinoma. The oncologist, 24(3),
366-371.
Rice, W. S., Goldfarb, S. S., Brisendine, A. E., Burrows, S., & Wingate, M. S. (2017).
Disparities in infant mortality by race among Hispanic and non-Hispanic infants.
Maternal and child health journal, 21(1), 1581-1588.
Robledo, C. A., Yeung, E., Mendola, P., Sundaram, R., Maisog, J., Sweeney, A. M., . . . Louis,
G. M. B. (2015). Preconception maternal and paternal exposure to persistent organic
pollutants and birth size: the LIFE study. Environmental health perspectives, 123(1), 88-
94.
Proposed PFAS Rule Economic Analysis
0-26
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Rodriguez-Barranco, M., Tobias, A., Redondo, D., Molina-Portillo, E., & Sanchez, M. J. (2017).
Standardizing effect size from linear regression models with log-transformed variables
for meta-analysis. BMC medical research methodology, 77(1), 1-9.
Roger, V. L., Go, A. S., Lloyd-Jones, D. M., Benjamin, E. J., Berry, J. D., Borden, W. B., . . .
Ford, E. S. (2012). Heart disease and stroke statistics—2012 update: a report from the
American Heart Association. Circulation, 725(1), e2.
Rowley, D. L., & Hogan, V. (2012). Disparities in infant mortality and effective, equitable care:
Are infants suffering from benign neglect? Annual Review of Public Health, 33, 75-87.
RStudio Team. (2020). RStudio: Integrated Development Environment for R. Retrieved from
https://www.rstudio.com/
Riicker, G., Schwarzer, G., Carpenter, J., & Olkin, I. (2009). Why add anything to nothing? The
arcsine difference as a measure of treatment effect in meta-analysis with zero cells.
Statistics in medicine, 25(5), 721-738.
Russell, R. B., Petrini, J. R., Damus, K., Mattison, D. R., & Schwarz, R. H. (2003). The changing
epidemiology of multiple births in the United States. Obstetrics & Gynecology, 101(1),
129-135.
Sagiv, S. K., Rifas-Shiman, S. L., Fleisch, A. F., Webster, T. F., Calafat, A. M., Ye, X., . . .
Oken, E. (2018). Early-pregnancy plasma concentrations of perfluoroalkyl substances
and birth outcomes in Project Viva: confounded by pregnancy hemodynamics? American
journal of epidemiology, 757(4), 793-802.
Savitz, D. A., Stein, C. R., Elston, B., Wellenius, G. A., Bartell, S. M., Shin, H.-M., . . . Fletcher,
T. (2012). Relationship of perfluorooctanoic acid exposure to pregnancy outcome based
on birth records in the mid-Ohio Valley. Environ. Health Perspect., 720(8), 1201-1207.
doi: 10.1289/ehp. 1104752
Seo, S.-H., Son, M.-H., Choi, S.-D., Lee, D.-H., & Chang, Y.-S. (2018). Influence of exposure to
perfluoroalkyl substances (PFASs) on the Korean general population: 10-year trend and
health effects. Environment international, 113, 149-161.
Shearer, J. J., Callahan, C. L., Calafat, A. M., Huang, W. Y., Jones, R. R., Sabbisetti, V. S., . . .
Hofmann, J. N. (2021). Serum concentrations of per- and polyfluoroalkyl substances and
risk of renal cell carcinoma. Journal of the National Cancer Institute, 113(5), 580-587.
doi:10.1093/jnci/djaal43
Shi, Y., Yang, L., Li, J., Lai, J., Wang, Y., Zhao, Y., & Wu, Y. (2017). Occurrence of
perfluoroalkyl substances in cord serum and association with growth indicators in
newborns from Beijing. Chemosphere, 169, 396-402.
doi:10.1016/j.chemosphere.2016.11.050
Sidik, K., & Jonkman, J. N. (2005). A note on variance estimation in random effects meta-
regression. Journal of Biopharmaceutical Statistics, 75(5), 823-838.
Singh, G. K., & Stella, M. Y. (2019). Infant mortality in the United States, 1915-2017: Large
social inequalities have persisted for over a century. International Journal of Maternal
and Child Health and AIDS, 5(1), 19.
Solis, P., Pullum, S. G., & Frisbie, W. P. (2000). Demographic models of birth outcomes and
infant mortality: An alternative measurement approach. Demography, 37(4), 489-498.
South Carolina Department of Health and Environmental Control. (2020). PFAS Sampling
Results: Surface Water-Sourced Community Drinking Water Systems. Retrieved from
https://scdhec.eov/BOW/pfas-sampline-results
Stan Development Team. (2020). Stan. Retrieved from http://mc-stan.ore/
Proposed PFAS Rule Economic Analysis
0-27
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
Stan Development Team. (2021). Stan Modeling Language Users Guide and Reference Manual,
Version 2.26. Retrieved from https://mc-stan.ore/
Starling, A. P., Adgate, J. L., Hamman, R. F., Kechris, K., Calafat, A. M., Ye, X., & Dabelea, D.
(2017). Perfluoroalkyl substances during pregnancy and offspring weight and adiposity at
birth: examining mediation by maternal fasting glucose in the Healthy Start Study.
Environ Health Perspect, 125, 067016-067011. doi:10.1289/EHP641
StataCorp. (2013a). Stata 13 logit Command Reference Manual. Retrieved from College Station,
TX: https://www.stata.com/manualsl3/rloeit.pdf
StataCorp. (2013b). Stata 13 Margins Command Reference Manual. Retrieved from College
Station, TX: https://www.stata.com/manualsl3/rmareins.pdf
StataCorp. (2019). Stata Statistical Software: Release 16. College Station, TX: StataCorp LP.
Steenland, K., Barry, V., & Savitz, D. (2018). An Updated Meta-Analysis of the Association of
Serum PFOA and Birthw eight, with an Evaluation of Potential Biases. Paper presented at
the ISEE Conference Abstracts.
Steenland, K., Tinker, S., Frisbee, S., Ducatman, A., & Vaccarino, V. (2009). Association of
perfluorooctanoic acid and perfluorooctane sulfonate with serum lipids among adults
living near a chemical plant. American journal of epidemiology, 770(10), 1268-1278.
Steenland, K., & Woskie, S. (2012). Cohort mortality study of workers exposed to
perfluorooctanoic acid. American journal of epidemiology, 176, 909-917.
Surveillance Research Program - National Cancer Institute. (2020a). SEER*Stat software
Incidence-Based Mortality - SEER Research Data, 18 Registries. Retrieved from:
seer, cancer. gov/ seerstat
Surveillance Research Program - National Cancer Institute. (2020b). SEER*Stat software
Incidence - SEER Research Limited-Field Data, 21 Registries. Retrieved from:
seer, cancer. gov/ seerstat
Thom, T., Kannel, W., Silbershatz, H., & D'Agostino, R. (2001). Cardiovascular diseases in the
United States and prevention approaches. Hurst's the heart, 7, 3-17.
U.S. Bureau of Economic Analysis. (2021). Disposable personal income: Per capita
[A229RC0A052NBEA]. Retrieved from:
https://fred.stlouisfed.ore/serit ?R€ 0 AO 5 2NBE A
U.S. Bureau of Economic Analysis. (2022). USBEA National Income and Product Accounts.
Table 1.1.9. Implicit Price Deflators for Gross Domestic Product, flndex numbers,
2012=100],
U.S. Bureau of Labor Statistics. (2020). CPIfor All Urban Consumers (CPI-U) All items in U.S.
city average, all urban consumers, not seasonally adjusted. Base period 1982-84-100.
Series ID: CUUR0000SA0,CUUS0000SA0. Retrieved from:
https://beta.bls.eov/dataViewer/view/timeseries/CUUR0000S AO.
U.S. Census Bureau. (2017). American Community Survey Single-Year Estimates. American
Community Survey Demographic and Housing Estimates. Retrieved from:
https://data.census.eov/cedsci/
U.S. Census Bureau. (2020). Annual County Resident Population Estimates by Age, Sex, Race,
and Hispanic Origin: April 1, 2010 to July 1, 2019, downloaded at
https://www2.census.gov/programs-surveys/popest/ on May 3, 2022.
U.S. Energy Information Administration. (2021). Annual Energy Outlook 2021. Retrieved
August 10, 2021 https://www, eia. gov/outlooks/aeo/
Proposed PFAS Rule Economic Analysis
0-28
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
U.S. Environmental Protection Agency. (2000a). Arsenic Occurrence in Public Drinking Water
Supplies. EPA-815-D-00-001. Washington, D.C.
U.S. Environmental Protection Agency. (2000b). Geometries and Characteristics of Public
Water Systems. (EPA-815-R-00-24).
U.S. Environmental Protection Agency. (2001). Water Treatment Plant Model Version 2.0.
Retrieved from https://www.epa.eov/sites/default/files/
03/documents/wtp model inual508.pdf
U.S. Environmental Protection Agency. (2005). Occurrence and Exposure Assessment for the
Final Long Term 2 Enhanced Surface Water Treatment Rule, EPA 815-R-06-002. (EPA
815-R-06-002).
U.S. Environmental Protection Agency. (2008). Final Ozone NAAQSRegulatory Impact
Analysis. (EPA 452/R-08-003). Research Triangle Park, North Carolina
U.S. Environmental Protection Agency. (2009). Community Water System Survey, Volume II:
Detailed Tables and Survey Methodology. EPA 815-R-09-002. Retrieved from
http://nepis.epa.gOv/Exe/ZyPDF.c ;key=P 1009U S A. txt
U.S. Environmental Protection Agency. (2010). Guidelines for Preparing Economic Analyses.
U.S. Environmental Protection Agency. (2011a). The Benefits and Costs of the Clean Air Act
from 1990 to 2020. Final Report - Rev. A. Retrieved from
https://www.epa.eov/sites/defaiilt/files/2015-07/dociiments/fullreport rev a.pdf
U.S. Environmental Protection Agency. (201 lb). Exposure Factors Handbook, 2011 Edition
(Final). (EPA-600-R-09-025F). U.S. Environmental Protection Agency, Washington, DC
U.S. Environmental Protection Agency. (2012). Revisions to the Unregulated Contaminant
Monitoring Regulation (UCMR 3) for Public Water Systems; Final Rule. Federal Register
Retrieved from https://www.federalreeister.eov/documents/2012/05/02/2
9978/revisions-to-the-unregulated-contaminant-monitorine-regulation-ucmr-3-for-public-
water-systems
U.S. Environmental Protection Agency. (2015). Benefit and Cost Analysis for the Effluent
Limitations Guidelines and Standards for the Steam Electric Power Generating Point
Source Category. (EPA-821-R-15-005). US EPA, Office of Water, Washington DC
20460
U.S. Environmental Protection Agency. (2016a). Health effects support document for
Perfluorooctane Sulfonate (PFOS). (EPA 822-R-l6-002). Retrieved from
https://www.epa.eov/sites/production/files/201 0 ^
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
815-D-21-001). Retrieved from
https://sab.epa.gov/ords/sab/f?n KV r 968 1 > > 1 -> ! 1 ! 1)1
U.S. Environmental Protection Agency. (2021b). Safe Drinking Water Information System 2021
Quarter 4 Database. Retrieved from https://www.epa.gov/ground-water-and-drinking-
water/safe-drinking-water-information-svstem-sdwis-federal-reporting
U.S. Environmental Protection Agency. (2022). Review of EPA's Analyses to Support EPA's
National Primary Drinking Water Rulemaking for PFAS. EPA-SAB-22-008. Washington,
D.C. Retrieved from
https://sab.epa.gov/ords/sab/f?v- 100 i s I i90947S s • * s • M * _601#report
U.S. Environmental Protection Agency. (2023a). Public Comment Draft - Toxicity Assessment
and Proposed Maximum Contaminant Level Goal for Perfluorooctane Sulfonic Acid
(PFOS) in Drinking Water. EPA-822-P-23-007. Washington, D.C.
U.S. Environmental Protection Agency. (2023b). Public Comment Draft - Toxicity Assessment
and Proposed Maximum Contaminant Level Goal for Perfluorooctanoic Acid (PFOA) in
Drinking Water. EPA-822-P-23-005. Washington, D.C.
U.S. Environmental Protection Agency. (2023c). Technical Support Document - Per- and
Polyfluoroalkyl Substances (PFAS) Occurrence & Contaminant Background. EPA-822-
P-23-010.
U.S. Environmental Protection Agency. (2023d). Technical Support Document - Technologies
and Cost for Removing Per- and Polyfluoroalkyl Substances (PFAS) from Drinking
Water. EPA-822-P-23-011.
Uno, H., Cai, T., Pencina, M. J., D'Agostino, R. B., & Wei, L. J. (2011). On the C-statistics for
evaluating overall adequacy of risk prediction procedures with censored survival data.
Stat. Med., 30(10), 1105-1117. doi:10.1002/sim.4154
Valvi, D., Oulhote, Y., Weihe, P., Dalgard, C., Bjerve, K. S., Steuerwald, U., & Grandjean, P.
(2017). Gestational diabetes and offspring birth size at elevated environmental pollutant
exposures. Environment international, 107, 205-215.
Vermont Department of Environmental Conservation. (2021). Per and Polyfluoroalkyl
Substances (PFAS) & Drinking Water. Retrieved from
https://dec.vermont.gov/water/drinkine-water/water-qiiality-monitorine/pfas
Verner, M. A., Loccisano, A. E., Morken, N. H., Yoon, M., Wu, H., McDougall, R., . . .
Longnecker, M. P. (2015). Associations of Perfluoroalkyl Substances (PFAS) with Lower
Birth Weight: An Evaluation of Potential Confounding by Glomerular Filtration Rate
Using a Physiologically Based Pharmacokinetic Model (PBPK). Environmental health
perspectives, 723(12), 1317-1324. doi: 10.1289/ehp. 1408837
Vieira, V. M., Hoffman, K., Shin, H., Weinberg, J. M., Webster, T. F., & Fletcher, T. (2013).
Perfluorooctanoic acid exposure and cancer outcomes in a contaminated community: A
geographic analysis. Environmental health perspectives, 121(3), 318-323.
von Hippel, P. T. (2015). The heterogeneity statistic 12 can be biased in small meta-analyses.
BMC medical research methodology, 75(1), 1-8.
Wan, X., Zhang, Y., Tan, C., Zeng, X., & Peng, L. (2019). First-line nivolumab plus ipilimumab
vs sunitinib for metastatic renal cell carcinoma: a cost-effectiveness analysis. JAMA
Oncology, 5(4), 491-496.
Wang, Y., Adgent, M., Su, P.-H., Chen, H.-Y., Chen, P.-C., Hsiung, C. A., & Wang, S.-L.
(2016). Prenatal exposure to perfluorocarboxylic acids (PFCAs) and fetal and postnatal
Proposed PFAS Rule Economic Analysis
O-30
March 2023
-------
DRAFT FOR PUBLIC COMMENT
MARCH 2023
growth in the Taiwan Maternal and Infant Cohort Study. Environmental health
perspectives, 124(11), 1794-1800.
Washino, N., Saijo, Y., Sasaki, S., Kato, S., Ban, S., Konishi, K., . . . Saito, K. (2009).
Correlations between prenatal exposure to perfluorinated chemicals and reduced fetal
growth. Environmental health perspectives, 117(4), 660-667.
Whitworth, K. W., Haug, L. S., Baird, D. D., Becher, G., Hoppin, J. A., Skjaerven, R., . . .
Wilson, R. (2012). Perfluorinated compounds in relation to birth weight in the Norwegian
Mother and Child Cohort Study. American journal of epidemiology, 775(12), 1209-1216.
Whooley, M. A., De Jonge, P., Vittinghoff, E., Otte, C., Moos, R., Carney, R. M., . . . Feldman,
M. D. (2008). Depressive symptoms, health behaviors, and risk of cardiovascular events
in patients with coronary heart disease. Jama, 300(20), 2379-2388.
Wickham, H. (2007). Reshaping Data with the reshape Package. Journal of statistical software,
27(12), 1 -20. doi: 10.18637/jss.v021.il2
Wickham, H. (2016). ggplot2: elegant graphics for data analysis: springer.
Wickham, H., Fran
-------
DRAFT FOR PUBLIC COMMENT MARCH 2023
Zhang, Y., Beesoon, S., Zhu, L., & Martin, J. W. (2013). Biomonitoring of perfluoroalkyl acids
in human urine and estimates of biological half-life. Environmental Science &
Technology, ¥7(18), 10619-10627.
Proposed PFAS Rule Economic Analysis
0-32
March 2023
------- |