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


-------