JHH f United States
Environmental Protection
R mAgency
EP A-822-P-23 -003
PUBLIC REVIEW DRAFT
Framework for Estimating Noncancer Health Risks
Associated with Mixtures of Per- and Poly fluoroalkyl
Substances (PFAS)
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Framework for Estimating Noncancer Health Risks Associated with Mixtures
of Per- and Polyfluoroalkyl Substances (PFAS)
Prepared by:
U.S. Environmental Protection Agency
Office of Water & Office of Research and Development
Washington, DC
EPA Document Number: EPA-822-P-23-003
March 2023
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Notices
This document has been reviewed in accordance with U.S. Environmental Protection Agency
(EPA) policy and approved for publication.
This document provides a framework for estimating the likelihood of noncancer human health
risks associated with mixtures of per- and polyfluoroalkyl substances (PFAS), based on
longstanding EPA mixtures guidelines and guidance. This document is not a regulation and does
not impose legally binding requirements on EPA, states, tribes, or the regulated community, and
might not apply to a particular situation based on the circumstances. The extent of the utility of
this document for a particular programmatic application will need to be assessed on a case-by-
case basis within each specific decision context under applicable statutory and regulatory
authority. The framework included in this document does not supersede previously published
EPA guidance on mixtures (e.g., EPA, 1986, 2000b) or EPA approaches used to assess
cumulative risks of contaminants including chemical mixtures under various environmental
statutes (e.g., Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA); Food Quality
Protection Act (FQPA); Comprehensive Environmental Response, Compensation, and Liability
Act (CERCLA)). Based upon evolving availability of human health risk assessment relevant
information and increasing confidence in New Approach Methods, EPA may change certain
aspects of this document in the future.
Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.
Dedication
This document is dedicated to the memory of Dr. Jane Ellen Simmons and Mr. Jeffrey Swartout.
Jane Ellen and Jeff were both dedicated civil servants in EPA's Office of Research and
Development for more than 30 years where they conducted rigorous chemical mixtures research
and championed cumulative risk assessment approaches for exposure to multiple stressors. Their
contributions to the field live on in this framework document.
i
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Authors, Contributors, and Reviewers
Prepared by
Justin Conley, PhD
Colleen Flaherty, MS (co-lead)
Earl Gray, PhD
Brittany Jacobs, PhD
Jason C. Lambert, PhD, DABT (co-lead)
Alex Lan, MPH
Casey Lindberg, PhD
Kathleen Raffaele, PhD
Contributors
Kelly Cunningham, MS
Hannah Hoi singer, MPH
Amanda Jarvis, MS
James R. Justice, MS
Internal Technical Reviewers
Andrew Kraft, PhD
Allison Phillips, PhD
Glenn Rice, ScD
Jane Ellen Simmons, PhD
Executive Direction
Elizabeth Behl
Eric Burneson, P.E.
Santhini Ramasamy, PhD
Jamie Strong, PhD
Russell Thomas, PhD
Tim Watkins, PhD
Formatting By
Tetra Tech, Inc.
11
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Contents
EXECUTIVE SUMMARY 1
1.0 Introduction and Background 3
1.1 Purpose 3
1.2 EPA Science Advisory Board (SAB) Review 4
1.3 Background on PFAS 6
1.4 Occurrence of PFAS Mixtures 10
1.5 Evidence of PFAS Exposure in Humans 12
1.6 Brief Summary of State, National, and International Approaches to Address
PFAS Mixtures in Water 13
1.7 Overview of Proposed Framework for Estimating Health Risks for PFAS
Mixtures 17
2.0 Background on EPA Mixtures Additivity Guidance 20
2.1 Component-Based Mixtures Assessment Methods 21
2.1.1 Application of Dose Addition as EPA's Default Assumption 22
3.0 Dose Additivity for PFAS 23
3.1 Overview of Assessment Approaches for Chemical Mixtures 23
3.2 Examples of Chemical Classes and Toxicological Pathways Utilizing Mixture
Assessment Approaches 24
3.2.1 Dioxin-Like Chemicals and Aryl Hydrocarbon Receptor Pathway Toxic
Equivalence Factors (TEFs) 24
3.2.2 Pyrethroids/Pyrethrins - Central Nervous System and Behavior 25
3.2.3 Organophosphates - Lethality, Central Nervous System and Behavior 25
3.2.4 Estrogen agonists - Mixture Effects on the Female Reproductive Tract 26
3.2.5 Phthalates in utero - Mixture Effects on the Female Reproductive Tract 26
3.2.6 Antiandrogens - Male Reproductive Tract Development 27
3.3 Systematic Reviews of Mixtures Toxicity: Quantification of Deviations from
Dose Additivity 29
3.3.1 Deviation from Additivity 30
3.4 PFAS Dose Additivity 30
4.0 Introduction to Estimating Noncancer PFAS Mixture Hazard or Risk 35
4.1 Whole Mixtures Approach 35
4.2 Data-Driven Component-Based Mixtures Approaches for PFAS 35
4.2.1 Conceptual Framework of the Approach 36
4.2.2 Introduction to a Hypothetical Example with Five PFAS 45
5.0 Hazard Index (HI) Approach 55
5.1 Background on the HI Approach 55
in
-------
PUBLIC REVIEW DRAFT MARCH 2023
5.2 Illustrative Example Application of the General HI to a Hypothetical Mixture of
Five PFAS 59
5.2.1 General HI Step 1: Assemble/derive component health effects endpoints
(Chronic oral RfDs) 59
5.2.2 General HI Step 2: Assemble/derive health-based media concentrations
(HBWC) 63
5.2.3 General HI Step 3: Select exposure estimates (measured water
concentrations) 66
5.2.4 General HI Step 4: Calculate PFAS mixture potency (component HQs
and overall HI) 66
5.2.5 General HI Step 5: Compare PFAS mixture potency (HI) to existing
health-based value (1.0) 67
5.3 Illustrative Example Application of the Target-Organ Specific HI (TOSHI) to a
Hypothetical Mixture of Five PFAS 67
5.3.1 TOSHI Step 1: Assemble/derive component health effects endpoints
(RfDs or target-organ toxicity doses) 67
5.3.2 TOSHI Step 2: Assemble/derive health-based media concentrations
(HBWC) 68
5.3.3 TOSHI Step 3: Select exposure estimates (measured water
concentrations) 69
5.3.4 TOSHI Step 4: Calculate PFAS mixture potency (component HQs and
overall TOSHI) 69
5.3.5 TOSHI Step 5: Compare PFAS mixture potency (HI) to existing health
benchmark (1.0) 69
5.4 Advantages and Challenges of the General HI Approach 70
5.5 Advantages and Challenges of the Target Organ Specific Hazard Index 71
6.0 Relative Potency Factor (RPF) Approach 71
6.1 Background on RPF Approach 71
6.2 Illustrative Example Application of RPF to a Hypothetical Mixture of Five PFAS .. 74
6.2.1 RPF Step 1: Assemble/Derive component health effects endpoints (select
ICs, PODheds) 75
6.2.2 RPF Step 2: Assemble/derive health-based media concentrations
(HBWCs for the Index Chemicals) 75
6.2.3 RPF Step 3: Select exposure estimates (measured water concentrations) 76
6.2.4 RPF Step 4: Calculate PFAS mixture potency (RPFs and ICECs for each
effect domain) 76
6.2.5 RPF Step 5: Compare PFAS mixture potency (Total ICECmix) to existing
health-based value (HBWC) 82
6.3 Advantages and Challenges of the Relative Potency Factor Approach 83
7.0 Mixture-BMD Approach 85
7.1 Background on the Mixture-BMD Approach 85
IV
-------
PUBLIC REVIEW DRAFT
MARCH 2023
7.2 Illustrative Example Application of the Mixture Benchmark Dose Approach to a
Hypothetical Mixture of Five PFAS 89
7.2.1 Mixture BMD Step 1: Assemble/derive component health effects
endpoints (BMDx) 89
7.2.2 Mixture BMD Step 2: Assemble/derive health-based media
concentrations (HBWC) 90
7.2.3 Mixture BMD Step 3: Select exposure estimates (measured water
concentrations) 90
7.2.4 Mixture BMD Step 4: Calculate PFAS mixture potency (Mixture BMD
HBWC) 90
7.2.5 Mixture BMD Step 5: Compare PFAS mixture potency (total PFAS
mixture concentration) to health-based value (Mixture BMD HBWC) 92
7.3 Advantages and Challenges of the Mixture BMD Approach 92
8.0 Comparison of Component-Based Approaches 94
References 96
Figures
Figure 2-1. Flow chart for evaluating chemical mixtures using component-based additive
methods 22
Figure 3-1. AOP network for chemicals that disrupt AR-mediated cellular signaling leading
to adverse effects on the development of male reproductive tract resulting from
in utero exposure 28
Figure 4-1. Framework for data-driven application of component-based assessment
approaches for mixtures of PFAS based on an assumption of dose additivity 37
Figure 4-2. Example literature inventory heatmap for traditional epidemiological or
experimental animal studies for five PFAS currently under development/review
in EPA/ORD's IRIS program (heat map circa 2018). Health effects are based
on groupings from the IRIS website
(https://cfpub.epa.gov/ncea/iris/search/index.cfm) 39
Figure 4-3. Example plot illustrating in vitro cell bioactivity expressed in AEDs which are
an estimated oral exposure dose that results in an internal steady-state
concentration consistent with the in vitro concentration associated with a
biological perturbation or activity. The example shows the distribution of
AEDs; the vertical orange dashed line indicates the 5th percentile on the
bioactivity landscape. The green dashed line corresponds to an upper bound on
the estimated general population-based median exposure (generated from
EPA's Systematic Empirical Evaluation of Models (SEEM3);
https://www.epa.gov/chemical-research/computational-toxicology-
communities-practice-systematic-empirical-evaluation) 42
Figure 4-4. Example PFAS-specific literature search string applied to toxicity information
databases such as the four listed (e.g., PubMed, WoS, Toxline, and TSCATS) 47
v
-------
PUBLIC REVIEW DRAFT MARCH 2023
Figure 4-5. Example PECO criteria and considerations used to determine study relevance in
the systematic review and evaluation of a literature inventory for chemicals
such as PFAS 48
Figure 4-6. Example literature screening logic flow for hypothetical PFAS using an EPA
systematic review approach. The figure depicts example PECO-dependent
development of evidence bases to support human health assessment
application(s). Note: only 4 of 5 hypothetical PFAS (i.e., PFAS 1-4) are
represented as PFAS 5 is data-poor 49
Figure 4-7. (A). Example exposure-response arrays for the PFAS 1-3 identified as having
existing human health risk assessment values for one or more exposure
durations; (B). Empirical clearance or plasma half-life data for PFAS 1-5.
ND = no data available. *Clearance = elimination rate constant (ke) x volume of
distribution (Vd); **Data needed to calculate clearance or Vd were not available
for rodents or humans as such only plasma half-life is presented where
available; ''Male only 51
Figure 4-8. Evidence integration across three target health effect domains for a mixture of
five hypothetical PFAS. The heat map indicates strength of evidence supporting
an effect of the PFAS in a domain. (+++) indicates likely effect; (++) evidence
suggests an effect in the domain; (—) evidence is inadequate to determine an
effect in the domain; *Although PFAS 5 has no traditional human or
experimental animal assay data available, in vitro cell bioactivity data are
available from assays performed predominately in hepatocyte cell lines 52
Figure 5-1. General steps to derive bioactivity-based reference value (RfV) using
bioactivity data in human or animal tissue/cells 62
Figure 6-1. General cell signaling pathways associated with PFAS-induced liver injury 78
Figure 6-2. Proposed process for integrating NAM-based RPFs and ICECs into mixtures
assessment 79
Figure 7-1. Example comparison of observed data with model predictions using two dose
addition-based mixture models and a response addition model for a binary
mixture study (adapted from Gray et al., 2022). The two chemicals displayed
individual dose response curves with widely disparate slopes for the endpoint
(reduced organ weight). The two dose addition models either assume
component chemicals have similar dose response slopes (red solid line) for the
effect or have non-congruent dose response curve slopes (black dashed line).
For these chemicals with disparate slopes the dose addition model that does not
assume equal slopes provided a better fit of the observed data (see table below) 87
Tables
Table 1-1. Two Primary Categories of PFASa 8
Table 1-2. Groups, Structural Traits and Examples of Perfluoroalkyl Acids (PFAA),
including Perfluoroalkylether Acids (PFEAAs)a 9
vi
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Table 1-3. Characterization System of Short-Chain and Long-Chain PFAAa 9
Table 1-4. Summary of U.S. and International Approaches to Addressing the Combined
Toxicity of Multiple PFAS in Drinking Water or Groundwater51'13 (only
combined PFAS approaches are presented) 15
Table 4-1. Drinking water concentrations for five hypothetical PFAS; the values represent
the median of a distribution of sampling data collected across a community
over time 46
Table 4-2. Analytical quantitation limits for drinking water for five hypothetical PFAS 46
Table 4-3. Data array to inform decisions in Steps 2 and 3 of the framework approach for
component-based mixtures assessment of PFAS 54
Table 5-1. EPA and ATSDR Peer-Reviewed Human Health Assessments Containing
Noncancer Toxicity Values (RfDs or MRLs) for PFAS that Are Either Final or
Under Development 57
Table 5-2. Calculation of estimated clearance values for PFAS 4 in female rats and humans 61
Table 5-3. Summary of PODheds and RfDs for hypothetical PFAS in a mixture 63
Table 5-4. EPA Exposure Factors for Drinking Water Intake 64
Table 5-5. Calculation of HBWCs for hypothetical PFAS in a mixture 66
Table 5-6. Calculation of individual component hazard quotients (HQs) for the hypothetical
PFAS mixture 66
Table 5-7. Target Organ Toxicity Doses (TTDs) for the hypothetical mixture PFAS; the
bolded numbers represent the overall RfD for each respective PFAS 68
Table 5-8. Calculation of Developmental Effect-Specific HBWCs for hypothetical PFAS in
a mixture using TTDs 68
Table 5-9. Calculation of individual component hazard quotients (HQs) specifically for
developmental effects associated with the hypothetical PFAS mixture. The
HBWCs in this TOSHI application are derived from TTDs for the
developmental effect domain 69
Table 6-1. Summary of PODheds for three selected health effect domains for a mixture of
five hypothetical PFAS 75
Table 6-2. Example Liver Effect RPFs and ICECs for a Hypothetical Mixture of Five
PFAS 80
Table 6-3. Example Thyroid Effect RPFs and ICECs for a Hypothetical Mixture of Five
PFAS 81
Table 6-4. Example Developmental Effect RPFs and ICECs for a Hypothetical Mixture of
Five PFAS 81
Table 7-1. "Best Model" based upon AIC values 87
Table 7-2. Summary of PODheds for three selected health effect domains for a mixture of
five hypothetical PFAS 90
vii
-------
PUBLIC REVIEW DRAFT MARCH 2023
Table 7-3. M-BMD Approach: Hypothetical Water Sample 91
Vlll
-------
PUBLIC REVIEW DRAFT
MARCH 2023
List of Abbreviations and Acronyms
6:2 FTS 6:2 fluorotelomer sulfonic
acid
AED
Administered Equivalent
Dose
AFFF
aqueous film forming
foam
AhR
aryl hydrocarbon receptor
AIC
Akaike Information
Criteria
AOF
adsorbable
organofluorine
AOP
adverse outcome pathway
AR
androgen receptor
AT SDR
Agency for Toxic
Substances and Disease
Registry
AUC
area under the
concentration vs. time
curve
BBP
butyl benzyl phthalate
BMD
benchmark dose
BMDL
lower statistical bound on
a BMD
BMR
benchmark response
C-F
carbon-fluoride bond
CAA
Clean Air Act
CAR
constitutive androstane
receptor
CCL
Contaminant Candidate
List
CERCLA
Comprehensive
Environmental Response,
Compensation, and
Liability Act
CPSC CHAP
Consumer Product Safety
Commission Chronic
Hazard Advisory Panel
DA
dose addition
DAF
dosimetric adjustment
factor
DBP
di-n-butyl phthalate
DEHP
di(2-ehtylhexyl)phthalate
DIBP
diisobutyl phthalate
DLC
dioxin-like chemical
DWI
drinking water intake
DWI-BW
body weight-based
drinking water intake
E
duration-relevant
exposure
ECx
effect concentration
EDx
effective dose in x
percent of test animals
EFSA
European Food Safety
Authority
EOF
extractable
organofluorine
EPA
U.S. Environmental
Protection Agency
EU
European Union
FT4
free serum thyroxine
FQPA
Food Quality Protection
Act
FIFRA
Federal Insecticide,
Fungicide, and
Rodenticide Act
GenX chemicals
hexafluoropropy 1 ene
oxide (HFPO) dimer acid
and HFPO dimer acid
ammonium salt
GD
gestation day
HBWC
Health-Based Water
Concentration
HED
human equivalent dose
HepaRG
epoxide hydrolase
endpoint in liver
HFPO
hexafluoropropy 1 ene
oxide
HI
hazard index
HQ
hazard quotient
IA
integrated addition
IC
index chemical
IX
-------
PUBLIC REVIEW DRAFT
ICEC
Index Chemical
Equivalent Concentration
IC EC mix
total mixture Index
Chemical Equivalent
Concentration
ICECnam
New Approach
Methodology (NAM)-
based ICEC
ICED
index chemical equivalent
dose
IRIS
Integrated Risk
Information System
IVIVE
in vitro to in vivo
extrapolation
KE
key event
LOAEL
lowest observed adverse
effect level
M-BMD
mixture benchmark dose
MAC
maximum acceptable
concentration
MCL
Maximum Contaminant
Level
MCLG
Maximum Contaminant
Level Goal
mg/kg/day
milligrams per kilogram
per day
MIE
molecular initiating event
MOA
mode of action
MRL
minimal risk level
NAM
New Approach
Methodology(ies)
NAS
National Academy of
Sciences
NBP2
Nafion byproduct 2
ng/g
nanograms per gram
ng/L
nanograms per liter
NHANES
National Health and
Nutrition Examination
Survey
NIEHS
National Institute of
Environmental Health
Sciences
MARCH 2023
NOAEL
no observed adverse
effect level
NRC
National Research
Council
NTP
National Toxicology
Program
OECD
Organisation for
Economic Co-operation
and Development
OP
organophosphate
ORD
Office of Research and
Development
osRfV
organ-specific reference
value
PCB
polychlorinated biphenyl
PCDD
polychlorinated dibenzo-
p-dioxins
PCDF
polychlorinated
dibenzofuran
PECO
Population, Exposure,
Comparator, and
Outcome
PFAA
perfluoroalkyl acids
PFAS
per- and polyfluoroalkyl
substances
PFBA
perfluorobutanoic acid
PFBS
perfluorobutanesulfonic
acid
PFCA
perfluoroalkyl carboxylic
acid
PFDA
perfluorodecanoic acid
PFDoDA
perfluorododecanoic acid
PFDS
perfluorodecanesulfonate
PFECHS
perfluoroethylcyclohexane
sulfonate
PFHpA
perfluoroheptanoic acid
PFHpS
perfluoroheptanesulfonic
acid
PFHxA
perfluorohexanoic acid
PFHxS
perfluorohexanesulfonic
acid
PFNA
perfluorononanoic acid
x
-------
PUBLIC REVIEW DRAFT
MARCH 2023
PFNS
perfluorononanesulfonic
acid
RSC
PFOA
perfluorooctanoic acid
rTK
PFOS
perfluorooctanesulfonic
SAB
acid
SARA
PFOSA
perfluorooctane
sulfonamide
SDWA
PFPA
perfluoropropanoic acid
T3
PFPeA
perfluoropentanoic acid
T4
PFPeS
perfluoropentanesulfonic
acid
TCDD
PFPIA
perfluoroalkyl phosphinic
acid
TD
PFPS
perfluoropropane sulfonic
acid
TEC
PFSA
perfluoroalkane sulfonic
TEF
acid
TEQ
PFSIA
perfluoroalkane sulfinic
TK
acid
TOSHI
PFTA
perfluorotetradecanoic
acid
TOSHIi
PFTrDA
perfluorotridecanoic acid
PFUnA
perfluoroundecanoic acid
PND
post-natal day
TSCA
POD
point-of-departure
PODhed
human-equivalent point-
TTD
of-departure
UCMR
PPAR.a
peroxisome proliferator
activated receptor alpha
UF
PPAR.y
peroxisome proliferator
activated receptor gamma
UFa
PPRTV
Provisional Peer-
Reviewed Toxicity Value
UFd
ppt
parts per trillion
UFh
PWS
public water system
RA
response addition
RfD
reference dose
UFl
RfV
reference value
RPF
relative potency factor
UFs
RPFnam
New Approach
Methodology (NAM)-
based RPF
relative source
contribution
reverse toxicokinetic
Science Advisory Board
Superfund Amendments
and Reauthorization Act
Safe Drinking Water Act
triiodothyronine
serum thyroxine
2,3,7,8-
tetrachl orodib enzo-p-
dioxin
toxicodynamic
toxic equivalent
concentrations
toxic equivalence factor
toxic equivalents
toxicokinetic
target organ specific
hazard index
target organ specific
hazard index for
developmental effects
Toxic Substances Control
Act
target-organ toxicity dose
Unregulated Contaminant
Monitoring Rule
uncertainty factor
interspecies uncertainty
factor
database uncertainty
factor
human interindividual
variability uncertainty
factor
LOAEL-to-NOAEL
uncertainty factor
extrapolation from
sub chronic to a chronic
exposure duration
uncertainty factor
XI
-------
PUBLIC REVIEW DRAFT
MARCH 2023
EXECUTIVE SUMMARY
The U.S. Environmental Protection Agency (EPA) is releasing for public review the Framework
for Estimating Noncancer Health Risks Associated with Mixtures of Per- and Polyfluoroalkyl
Substances (PFAS) ("PFAS Mixtures Framework"). This document is designed to communicate
and illustrate the practical application of existing EPA chemical mixtures approaches and
methods to two or more PFAS co-occurring in environmental media, using drinking water
examples. In November 2021, EPA released a draft version of this document for Science
Advisory Board (SAB) review. EPA has considered the SAB comments and revised the
document accordingly.
In a mixtures risk assessment context, while it would be optimal to leverage whole mixture
hazard and dose-response data, such data are extremely rare, particularly at component-chemical
proportions and concentrations consistent with environmentally occurring mixtures. As such,
mixtures risk assessment commonly relies upon integration of available toxicity information for
the individual component chemicals that co-occur in environmental media. This PFAS Mixtures
Framework document describes flexible, data-driven approaches that facilitate practical
component chemical-based mixtures evaluation of two or more PFAS, based on an assumption
of dose additivity. Studies with PFAS and other classes of chemicals support the health
protective assumption that a mixture of chemicals with similar apical effects should be assumed
to act in a dose additive manner unless data demonstrate otherwise. Descriptions of dose
additivity-based approaches such as the Hazard Index (HI), Relative Potency Factor (RPF), and
Mixture Benchmark Dose (M-BMD) are presented to demonstrate application to PFAS mixtures,
but they are not intended to provide a comprehensive treatise on the methods themselves; EPA
mixtures guidelines and guidance (EPA, 1986, 2000b) exist for such a purpose. EPA's mixture
assessment concepts and associated illustrative examples presented in this framework may
inform PFAS evaluation(s) by federal, state, and tribal partners, as well as public health experts,
drinking water utility personnel, and other stakeholders interested in assessing the potential
noncancer human health hazards and risks associated with PFAS mixtures.
It is not the intent of the framework to ignore potential carcinogenic effects associated with
PFAS exposure. Rather, at present, few PFAS have information available to evaluate potential
carcinogenic effects via any route of exposure. Should such information become available for an
increasing number of PFAS in the future, EPA would consider approaches for addressing joint
carcinogenic effects. EPA's National PFAS Testing Strategy (EPA, 2021f) is underway to
develop and issue test orders that may help inform human health hazard data needs, including
potential for carcinogenicity.
PFAS are a large and diverse structural family of compounds used in myriad commercial
applications due to their unique physicochemical properties. Although PFAS have been
manufactured and used broadly in commerce since the 1940s, particular concern over potential
adverse effects on human health grew in the early 2000s with the discovery of perfluorooctanoic
acid (PFOA) and perfluorooctanesulfonic acid (PFOS) in human blood. Since that time,
hundreds of PFAS have been identified in water, soil, and air. Many PFAS, or their precursors or
degradants, are environmentally persistent, bioaccumulative, and have long half-lives in humans,
particularly the longer-chain perfluorocarboxylic acid (PFCA) and perfluorosulfonic acid
(PFSA) species such as PFOA and PFOS, respectively. PFCAs/PFSAs with shorter carbon chain
1
-------
PUBLIC REVIEW DRAFT
MARCH 2023
length, such as perfluorobutanesulfonic acid (PFBS) and hexafluoropropylene oxide (HFPO)
dimer acid and HFPO dimer acid ammonium salt (also known as GenX Chemicals1), were
subsequently developed and integrated into various consumer products and industrial
applications because they have the desired industrial properties and characteristics associated
with this class of compounds but are more quickly eliminated from the human body than PFOA
and PFOS. The range of PFAS encountered in environmental media is often a diverse milieu of
linear, branched, cyclic, and/or aromatic parent species, metabolites, and/or abiotic degradants,
leading to significant potential for PFAS mixture exposures in aquatic, terrestrial, and human
populations.
As of February 2023, final EPA human health assessments are available for PFBS (EPA, 2021a),
HFPO-DA (EPA, 2021b), and perfluorobutanoic acid (PFBA; EPA 2022e). EPA is in the process
of updating the 2016 health assessments for PFOA and PFOS (EPA, 2023b,c,d,e), and they will
be publicly reviewed as part of the development of a National Primary Drinking Water
Regulation for PFAS. In addition, EPA's Integrated Risk Information System (IRIS) program is
developing four additional PFAS human health assessments for perfluorohexanoic acid
(PFHxA), perfluorohexanesulfonic acid (PFHxS), perfluorononanoic acid (PFNA), and
perfluorodecanoic acid (PFDA), which are expected to be completed by 2024. In May 2021, the
Agency for Toxic Substances and Disease Registry (ATSDR) published a "Toxicological Profile
for Perfluoroalkyls" that included an additional nine PFAS that EPA has not yet formally
assessed. However, beyond PFOA and PFOS, ATSDR derived quantitative minimal risk levels
(MRLs) for PFHxS and PFNA.
A significant challenge in evaluating PFAS is the lack of hazard and dose-response data suitable
for human health risk assessment for the large majority of individual PFAS. In response to the
critical need, EPA and the National Institute of Environmental Health Sciences (NIEHS) are
actively engaged in research and testing to help address data gaps for a broad landscape of PFAS
(approximately 150 structures at the time of the drafting of this document). Until results from
ongoing research and testing efforts are available, the evaluation of potential toxicity/risk
associated with mixtures of PFAS is primarily limited to existing hazard and dose-response data
under the purview of human health assessments by federal, state, and/or international entities.
To facilitate the use of potentially disparate sources of PFAS information in a mixture context,
the application of the component-based methods presented in this framework document is
demonstrated using a hypothetical mixture of five PFAS as follows:
(1) PFAS 1 = comprehensively studied, most potent for effect(s), and has formal noncancer
human health assessment value(s) (i.e., reference dose or RfD) and a health-based water
concentration (HBWC) available; (2) PFAS 2 = well-studied, second most potent for effect(s)
among PFAS 1-3, and has formal noncancer human health assessment value(s) and HBWC
available; (3) PFAS 3 = studied, least potent for effect(s) among PFAS 1-3, and has formal
noncancer human health assessment value(s) and HBWC available; (4) PFAS 4 = experimental
animal toxicity data available but no formal human health assessment and no HBWC; and
(5) PFAS 5 = data-poor. The hypothetical PFAS mixture is purposefully designed to
demonstrate how this framework allows for flexible integration of information derived from
:EPA notes that the chemical HFPO-DA is used in a processing aid technology developed by DuPont to make fluoropolymers
without using PFOA. The chemicals associated with this process are commonly known as GenX Chemicals and the term is often
used interchangeably for HFPO-DA along with its ammonium salt.
2
-------
PUBLIC REVIEW DRAFT
MARCH 2023
health assessment data sources (e.g., federal, state, international), available human
epidemiological and/or experimental animal hazard and dose-response data (that have not yet
been formally evaluated in an assessment product), and information from New Approach
Methodologies (NAMs). Opportunities for integrating additional PFAS into the context of a
mixture assessment is expected to evolve over time and will depend on the decision context and
availability of hazard and dose-response data from traditional and/or NAM-based assays and in
silico platforms.
1.0 Introduction and Background
1.1 Purpose
PFAS are an urgent public health and environmental issue facing communities across the United
States. In April 2021, Administrator Michael Regan established EPA's Council on PFAS, and
charged the Council to develop a bold, strategic, whole-of-EPA strategy to protect public health
and the environment from the impacts of PFAS. In October 2021, EPA released the PFAS
Strategic Roadmap2 ('The Roadmap') which lays out EPA's approach to tackling PFAS and sets
timelines by which the agency plans to take concrete actions to deliver results for the American
people. The Roadmap is built on a number of key principles, including considering the lifecycle
of PFAS, getting upstream of the problem, holding polluters accountable, ensuring science-based
decision-making, and prioritizing protection of disadvantaged communities.
Recognizing that PFAS tend to occur in mixtures in environmental media (see Section 1.4), EPA
has developed this data-driven framework for estimating the noncancer human health risks
associated with oral exposures to mixtures of PFAS. The approaches presented in this document
are based on longstanding EPA guidelines and guidance related to human health risk assessment
for mixtures (EPA, 1986, 2000b). Although the framework and illustrative examples provided in
this document include examples for PFAS in water, the framework itself is not limited to specific
media and may be useful for understanding the potential noncancer health effects of PFAS
mixtures under various authorities or decision contexts. The approach presented here is not
intended to be used to assign groups or subclasses or otherwise classify PFAS (instead see the
EPA National PFAS Testing Strategy for categorization efforts; EPA, 2021f). Rather, the
framework is designed for practical application of EPA chemical mixtures approaches and
methods to gain insight on the potential joint toxicity associated with exposure to mixtures of
PFAS. The mixture assessment concepts and associated illustrative examples presented in this
framework may inform PFAS evaluation(s) by federal, state, and tribal partners, as well as public
health experts, drinking water utility personnel, and other stakeholders interested in assessing the
potential human health risks associated with PFAS mixtures.
The framework and example calculations presented here are intended to demonstrate data-driven
application of EPA component-based mixture methods based on gradations of data availability
and completeness, anticipated to occur in real-world scenarios for PFAS. While the examples
provided are focused on drinking water, the approaches described in this framework could also
2 https://www.epa.gov/pfas/pfiis-strategic-roadmap-epas-commitments-action-2021-2024
3
-------
PUBLIC REVIEW DRAFT
MARCH 2023
be applied to other environmental media with oral3 exposure routes (e.g., soil, fish/shellfish,
food). Due to the constantly evolving science related to PFAS, the approaches presented herein
have the flexibility to consider information as it becomes available, including forthcoming EPA
human health assessments, assessments from other sources (e.g., federal, state, international),
available hazard and dose-response data in the public domain, information from high(er)
throughput bioassays and other NAMs including data submitted to the Agency through the Toxic
Substances Control Act (TSCA) Section 4 authority for developing and issuing PFAS test orders.
Experimental evidence supports dose additive effects from combined exposure to multiple
PFAS. Dose additivity, described in detail in Section 3.0, means that each of the component
chemicals in the mixture behaves as a concentration or dilution of every other chemical in the
mixture differing only in relative potency for toxicity. Several alkyl acid species (PFAAs) of
PFAS tested to date have been shown to elicit common adverse effects on several biological
systems including thyroid hormone levels, lipid synthesis and metabolism, as well as on
development, and immune and liver function (ATSDR, 2021; EFSA, 2018, 2020; USEPA,
2022c).
The document is not a regulation and does not impose legally binding requirements on EPA,
states, tribes, or the regulated community, and might not apply to a particular situation based on
the circumstances. Based upon peer-review and/or evolving availability of information, including
public comment, EPA may change certain aspects of this document in the future.
1.2 EPA Science Advisory Board (SAB) Review
In November 2021, EPA released the Draft Framework for Estimating Noncancer Health Risks
Associated with Mixtures of PFAS ("Draft PFAS Mixtures Framework"; EPA, 202 le) for EPA
SAB review. The SAB held public meetings on December 16, 2021; January 4, 6, and 7, 2022;
and July 20, 2022, to discuss the Draft PFAS Mixtures Framework and three other technical
documents supporting EPA's development of a National Primary Drinking Water Regulation for
PFAS under the Safe Drinking Water Act (SDWA). EPA sought SAB comment on whether the
framework and illustrative examples provided in the draft document were scientifically
supported, clearly described, and informative for assessing potential health risk(s) associated
with exposure to mixtures of PFAS. EPA asked specific charge questions on the assumption of
dose additivity and three component-based approaches: HI, RPF, and M-BMD. A draft of the
written SAB recommendations was published on April 1, 2022, and EPA received the final
report from the SAB on August 22, 2022 (EPA SAB, 2022).
EPA received a generally favorable review from SAB (EPA SAB, 2022) for its development of
component-based mixture assessment approaches that rely on a health protective assumption of
dose additivity based on a common health outcome, instead of a common mode of action
(MOA), to evaluate risk from PFAS mixtures in drinking water and other environmental media.
EPA has responded to the SAB's consensus advice in the development of this final Framework
for Estimating Noncancer Health Risks Associated with Mixtures of Per- and Polyfluoroalkyl
3 In general, the component-based approaches presented in this document may also be applicable in assessing health
risks associated with inhalation exposures of PFAS mixtures. However, the dosimetry differences across categories
of (volatile/semi-volatile) PFAS gases/vapors would need to be considered in such an assessment. Data regarding
the volatilization and toxicity of inhaled PFAS are generally limited.
4
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Substances (PFAS). The SAB's overarching consensus recommendations and EPA's responses
are summarized below. To view EPA's complete responses to SAB comments on the Draft
PFAS Mixtures Framework please see EPA (2023a).
• "The SAB supports dose additivity based on a common outcome, instead of a common
mode of action as a health protective default assumption and does not propose another
default approach. However, EPA should more thoroughly and clearly present the
uncertainties associated with this approach along with information supporting this
approach." (EPA SAB, 2022)
o EPA has added text in the Section, "Dose Additivity for PFAS" to address the
SAB's comments related to uncertainties associated with assuming dose addition
as the default assumption for assessment of PFAS mixtures. EPA has added
further discussion on deviations from dose additivity, such as synergy or
antagonism, but available evidence suggests that dose addition should be
considered as the default model.
• "The SAB expressed concern regarding the requirement for "external peer review" of
toxicity values developed by states and recommends that this phrase in the draft
framework be broadened to recommend the need for scientific input and review in
general." (EPA SAB, 2022)
o In response to this point of clarification, EPA has removed the text related to
external peer review. The text now reads, "If de novo derivation of toxicity values
is necessary, it is recommended that experts in hazard identification and dose
response assessment be consulted for scientific input and review, and the
associated uncertainties (e.g., data gaps) be transparently characterized."
• "EPA should consider using a menu-based framework to support selection of fit-for-
purpose approaches, rather than a tiered approach as described in the draft Mixtures
document. Tiered approaches that require increasingly complex information before
reaching a final decision point can be extremely challenging for data-poor chemicals such
as PFAS.
o Based on this and other SAB comments, EPA has eliminated the tiered approach
and restructured the framework as a data-driven, flexible approach to facilitate
PFAS mixture assessment in various decision contexts (e.g., at a contaminated
site, water system, etc.) (see Section 4.2 and Figure 4-1). With "fit-for-purpose
assessment" in mind, EPA has included discussion of key steps in the framework,
including problem formulation and scoping, assembling information, evaluating
data objectives, considering the data landscape to select component-based
approach(es), and performing component-based mixture assessment approach(es)
(see Section 4.2.1).
• "EPA should provide clarification regarding the conceptual similarities and differences
between the target-organ-specific hazard index (TOSHI) approach, the relative potency
factor (RPF) approach, and the mixture benchmark dose (BMD) approach, since all are
based on health effect-specific values (i.e., Reference Values (RfVs) or RPFs) for the
individual PFAS in the PFAS mixture. More discussion and comparison of approaches,
as well as when they converge, is needed. For instance, given the mathematical
correspondence between the RPF and mixture BMD approaches, EPA should consider
revising the discussion of these two approaches to present them as essentially the same
5
-------
PUBLIC REVIEW DRAFT
MARCH 2023
(or highlighting any essential differences), and perhaps also merging them into a single
section." (EPA SAB, 2022)
o EPA has added a new section (Section 8.0) that describes similarities and
differences among the different component-based mixtures assessment
approaches. In addition, EPA has made the revision to use the same hypothetical
example mixture of five PFAS (ranging from data poor to well-studied) for all of
the illustrative examples so that the user can better understand
similarities/differences among the approaches.
• "For both the RPF and mixture BMD approach, EPA's approach would be strengthened
by using PODs from animal studies that are based on human equivalent doses (HEDs)
rather than administered doses. The SAB found it difficult to envision situations in which
the mixture BMD was advantageous; therefore, EPA should provide additional
information on how the proposed Mixtures BMD approach will be applied in practice."
(EPA SAB, 2022)
o Text has been added in several places to indicate that it is optimal to calculate and
use HEDs rather than oral administered dose in test animals where and when
possible. This includes additional text that walks the reader through EPA's logic
flow for cross-species scaling (see new Subsection 5.2.1). Regarding the Mixture
BMD (M-BMD), text has been added to better articulate when this specific
approach is more appropriate (e.g., component chemical data that indicate
common health outcome but with non-congruent dose-response functions).
Further, Subsection 7.3 has been revised to reiterate the conditions that warrant
consideration of this specific component-based mixtures approach (as opposed to
the RPF method).
1.3 Background on PFAS
PFAS are a large group of structurally diverse anthropogenic chemicals that include PFOA,
PFOS, and thousands of other fully or partially fluorinated chemicals. Based on three related
structural definitions associated with EPA's identification of PFAS to be included in the fifth
Contaminant Candidate List (CCL; see below), the universe of environmentally relevant PFAS,
including parent chemicals, metabolites, and degradants, is approximately 12,000 compounds.4
The Organisation for Economic Co-operation and Development (OECD) New Comprehensive
Global Database of Per- and Polyfluoroalkyl Substances (PFASs) includes over 4,700 PFAS
(OECD, 2018). Comparatively, Buck et al. (2021) proposed that the number of PFAS currently
used in commercial products at the time of the drafting of this document is approximately 250
substances.
PFAS have been manufactured and used in a wide variety of industries around the world,
including in the United States since the 1940s. In general, PFAAs studied to date have strong,
stable carbon-fluorine (C-F) bonds, making them resistant to hydrolysis, photolysis, microbial
degradation, and metabolism (Ahrens, 2011; Beach et al., 2006; Buck et al., 2011; Evich et al.,
2022). Conversely, the larger PFAS universe is more structurally and physicochemically diverse
and includes categories of substances that may be more or less stable, persistent, and/or
bioaccumulative compared to PFAAs studied thus far (see National PFAS Testing Strategy;
4 See EPA List of PFAS Substances (Version 4) available at: fattps://eomptox.epa. gov/dashboard/chemica 1-
lists/PF ASM ASTER
6
-------
PUBLIC REVIEW DRAFT
MARCH 2023
EPA, 2021f). The chemical structures and physical-chemical properties of some PFAS make
them repel water and oil, remain chemically and thermally stable, and exhibit surfactant
properties; these properties are what make PFAS useful for commercial and industrial
applications and purposes, but these are also what make some PFAS persistent in the human
body and the environment (Calafat et al., 2007, 2019). Due to their widespread use,
physicochemical properties, persistence, and bioaccumulation potential, many PFAS co-occur in
exposure media (e.g., air, water, ice, sediment), and in tissues and blood of aquatic and terrestrial
organisms, and humans.
There are many families or subclasses of PFAS, and each contains many individual structural
homologues and can exist as either branched-chain or straight-chain isomers (Buck et al., 2011;
EPA, 2021c). These PFAS families can be divided into two primary categories: non-polymers
and polymers. The non-polymer PFAS include perfluoroalkyl and polyfluoroalkyl substances.
Polymer PFAS include fluoropolymers, perfluoropolyethers, and side-chain fluorinated polymers
(Table 1-1). Several U.S. federal, state, and industry stakeholders as well as European entities
have posited various definitions of what constitutes a PFAS. OECD, an international
organization comprised of 38 countries, recently published a practical guidance regarding the
terminology of PFAS (OECD, 2021). The OECD-led "Reconciling Terminology of the Universe
of Per- and Polyfluoroalkyl Substances: Recommendations and Practical Guidance" workgroup
provided an updated definition of PFAS, originally posited in part by Buck et al. (2011), as
follows: "PFASs are defined as fluorinated substances that contain at least one fully fluorinated
methyl or methylene carbon atom (without any H/Cl/Br/I atom attached to it), i.e. with a few
noted exceptions, any chemical with at least a perfluorinated methyl group (-CF3) or a
perfluorinated methylene group (-CF2-) is a PFAS". It is not within the scope of this framework
to compare and contrast the various definitions, or the nuances associated with defining or
scoping PFAS; rather the reader of this document is referred to OECD (2021) for review.
However, for the purpose of development of EPA's CCL 5, the structural definition of PFAS
includes chemicals that contain at least one of the following three structures:
• R-(CF2)-CF(R')R", where both the CF2 and CF moieties are saturated carbons, and none
of the R groups can be hydrogen (TSCA draft definition);
• R-CF2OCF2-R', where both the CF2 and CF moieties are saturated carbons, and none of
the R groups can be hydrogen; and
• CF3C(CF3)R'R ", where both the CF2 and CF moieties are saturated carbons, and none of
the R groups can be hydrogen.
It should also be noted that what defines or constitutes a PFAS may change or evolve over time
and under different purviews (e.g., federal, state, international).
7
-------
PUBLIC REVIEW DRAFT
Table 1-1. Two Primary Categories of PFASa
MARCH 2023
PFAS Non-polymers
Structural Elements
Example PFAS Families
Perfluoroalkyl acids
Compounds in which all carbon-
hydrogen bonds, except those on
the functional group, are replaced
with carbon-fluorine bonds
Perfluoroalkyl carboxylic and
sulfonic acids (e.g., PFOA,
PFOS), perfluoroalkyl phosphonic
and phosphinic acids,
perfluoroalkylether carboxylic and
sulfonic acids
Polyfluoroalkyl acids
Compounds in which all carbon-
hydrogen bonds on at least one
carbon (but not all) are replaced
with carbon-fluorine bonds
polyfluoroalkyl carboxylic acids,
polyfluoroalkylether carboxylic
and sulfonic acids
PFAS Polymers
Structural Elements
Example PFAS Families
Fluoropolymers
Carbon-only polymer backbone
with fluorines directly attached
polytetrafluoroethylene,
polyvinylidene fluoride,
fluorinated ethylene propylene,
perfluoroalkoxyl polymer
Polymeric
perfluoropolyethers
Carbon and oxygen polymer
backbone with fluorines directly
attached to carbon
F-(CmF2mO-)nCF3, where the
CmF2mO represents -CF20, -
CF2CF20, and/or -CF(CF3)CF20
distributed randomly along
polymer backbone
Side-chain fluorinated
polymers
Non-fluorinated polymer
backbone with fluorinated side
chains with variable composition
n: 1 or n:2 fluorotelomer-based
acrylates, urethanes, oxetanes, or
silicones; perfluoroalkanoyl
fluorides; perfluoroalkane sulfonyl
fluorides
Notes:
a Amalgamation of information from Figure 9 (OECD, 2021) and Buck et al. (2011).
PFOA and PFOS belong to the perfluoroalkyl acids (PFAA) of the non-polymer perfluoroalkyl
substances category of PFAS and are among the most researched PFAS in terms of human health
toxicity and biomonitoring (for review see Podder et al., 2021). The PFAA family includes
perfluoroalkyl carboxylic, phosphonic, and phosphinic acids and perfluoroalkane sulfonic and
sulfinic acids (Table 1-2). PFAA are highly persistent and are frequently found in the
environment (Ahrens, 2011; Brendel et al., 2018; Wang et al., 2017). Although EPA defines,
specifically for purposes under the purview of TSCA, long-chain perfluoroalkyl carboxylate
substances as having perfluorinated carbon chain lengths equal to or greater than seven carbons
and less than or equal to 20 carbons (85 FR 45109, July 27, 2020), a more comprehensive
delineation of what constitutes short-chain vs. long-chain PFAAs is provided by the OECD
(OECD, 2021). Specifically, the OECD established long-chain perfluoroalkyl carboxylic acids
(PFCAs) as those species with eight or more carbons (seven or more carbons are perfluorinated),
and short-chain PFCAs are identified as those with seven or fewer carbons (six or fewer carbons
are perfluorinated). Conversely, long-chain perfluoroalkane sulfonic acids (PFSAs) are identified
8
-------
PUBLIC REVIEW DRAFT
MARCH 2023
as those species with six or more carbons (six or more carbons are perfluorinated), and short-
chain PFSAs are identified as those with five or fewer carbons (five or fewer carbons are
perfluorinated) (see Table 1-3).
Table 1-2. Groups, Structural Traits and Examples of Perfluoroalkyl Acids (PFAA),
including Perfluoroalkylether Acids (PFEAAs)a
Group Functional Group
Examples
Perfluoroalkyl carboxylic acids -COOH
(PFCAs)
Perfluorooctanoic acid (PFOA),
C7F15COOHb
Perfluoroalkane sulfonic acids -S03H
(PFSAs)
Perfluorooctane sulfonic acid
(PFOS), C8F17S03H
Perfluoroalkyl phosphonic acids -P03H2
(PFPAs)
Perfluorooctyl phosphonic acid
(C8-PFPA)
Perfluoroalkyl phosphinic acids -P02H
(PFPIAs)
Bis(perfluorooctyl) phosphinic acid
(C8/C8-PFPIA)
Perfluoroalkylether carboxylates -0C2F40CF2C00H
(PFECAs)
Perfluoro-2-methyl-3-oxahexanoic
acid (GenX), 4,8-Dioxa-3H-
perfluorononanoic acid (ADONA)
Perfluoroalkylether sulfonic acids -0CF2CF2S03H
(PFESAs)
Nafion byproduct 2
Perfluoroalkyl dicarboxylic acids HOOC-CnF2n-COOH
(PFdiCAs)
Perfluoro-1,10-decanedicarboxylic
acid, Perfluorosebacic acid
Perfluoroalkane disulfonic acids H03S-CnF2n-S03H
(PFdiSAs)
Perfluoroalkane sulfinic acids -S02H
(PFSIAs)
Perfluorooctane sulfinic acid
Notes:
a Modified from Figure 9 in OECD, 2021
b The anionic form is most prevalent in water matrices.
Table 1-3. Characterization System of Short-Chain and Long-Chain PFAAa
Total # of carbons
3
4
5
6
7
8
9
10
# of fluorinated
carbons
2
3
4
5
6
«)
PFCAs
Shorl-chain PI-'CAs
Long-chain PI-'C'As
PI PA
PI liA
PI IVA
PI 1l\A
PI 1 IpA
PI'OA
PI \.\
PI DA
# of fluorinated
carbons
3
4
5
6
7
8
9
10
PFSAs
PI PS
PI liS
PI IVS
PI ll.\S
PI 1 IpS
PI OS
PI AS
PI l)S
9
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Total # of carbons
3
4
5
6
7
8
9
10
# of fluorinated
carbons
2
3
4
5
6
«)
PFCAs
Shorl-chain PI-'CAs
Long-chain PI-'C'As
PI PA
PIliA
PI IVA
PI ll\A
PI 1 IpA
PI-OA
Pl \ \
PI-DA
Shorl-chain PI-'SAs
l.ong-chain PI-'SAs
Notes: PFPA = perfluoropropanoic acid; PFBA = perfluorobutanoic acid; PFPeA = perfluoropentanoic acid;
PFHxA = perfluorohexanoic acid; PFHpA = perfluoroheptanoic acid; PFOA = perfluorooctanoic acid;
PFNA = perfluorononanoic acid; PFDA = perfluorodecanoic acid; PFPS = perfluoropropane sulfonic acid;
PFBS = perfluorobutanesulfonic acid; PFPeS = perfluoropentanesulfonic acid; PFEtxS = perfluorohexanesulfonic acid;
PFHpS = perfluoroheptanesulfonic acid; PFOS = perfluorooctanesulfonic acid; PFNS = perfluorononanesulfonic acid;
PFDS = perfluorodecanesulfonate.
For brevity, Table 1-3 only includes PFAAs of 3-10 carbons; the long-chain class of PFCAs and PFSAs can be expanded
considerably.
a Modification of Table 2-2 (ITRC, 2022)
Although many PFAS are manufactured in various salt forms (e.g., potassium (K+) PFBS), they
typically fully dissociate to their protonated acid and/or anionic forms depending on their acid
strength (pKa value) in aqueous environmental media, soils, or sediments, and the human body.
Importantly, the protonated and anionic forms may have different physicochemical and
environmental fate and transport properties. It should also be noted that the structural diversity of
PFAS is far greater than the PFCAs and PFSAs indicated in Table 1-3. There are branched,
cyclic, aromatic, and multi-component (e.g., polymers) structures that have been or are currently
classified as PFAS. However, in general, the linear PFCAs and PFSAs have been the most well-
studied PFAS to date and indeed have been the primary focus of formal human health risk
assessment activities in the federal and state sectors.
1.4 Occurrence of PFAS Mixtures
Improved analytical monitoring and detection methods have enabled detection of the co-
occurrence of multiple PFAS in drinking water, ambient surface waters, aquatic organisms,
biosolids (sewage sludge), and other environmental media.5 PFOA and PFOS have historically
been target analytes, but recent monitoring studies have begun to focus on additional PFAS via
advanced analytical instruments/methods and non-targeted analysis (De Silva et al., 2020;
McCord and Strynar, 2019; McCord et al., 2020). The proposed framework for estimating the
likelihood of human health risks associated with oral exposures to mixtures of PFAS (described
in Section 4) is flexible to accommodate information for any PFAS mixture of interest, provided
sufficient hazard and dose-response information is available.
EPA uses the Unregulated Contaminant Monitoring Rule (UCMR) to collect data for
contaminants that are suspected to be present in drinking water and do not have health-based
standards set under the SDWA. Between 2013 and 2015, EPA's third UCMR (i.e., UCMR 3)
required all large public water systems (PWSs) (serving more than 10,000 people) and a
statistically representative national sample of 800 small PWSs (serving 10,000 people or fewer)
5 For a more detailed discussion of the occurrence of PFOA, PFOS, and other PFAS in potential human exposure
sources see the relative source contribution (RSC) sections in EPA (2022a,b,c,d).
10
-------
PUBLIC REVIEW DRAFT
MARCH 2023
to monitor for 30 unregulated contaminants in drinking water, including six PFAS: PFOS,
PFOA, PFNA, PFHxS, perfluoroheptanoic acid (PFHpA), and PFBS. UCMR 3 data
demonstrated that two or more of those six PFAS co-occurred in 48% (285/598) of sampling
events with PFAS detected, and PFOA and PFOS co-occurred in 27% (164/598) of sampling
events with two or more PFAS detected (EPA, 2019b; Guelfo and Adamson, 2018). EPA found
that 4% of PWSs reported results for which one or more of the six UCMR 3 PFAS were
measured at or above their respective minimum reporting levels.6 Outside of the UCMR 3 data
collection, many states have undertaken individual efforts to monitor for PFAS in both source
and finished drinking water. These results show occurrence in multiple geographic locations
consistent with what was observed during UCMR 3 monitoring (EPA, 202Id). Additionally,
these results show that PFAS are very likely to co-occur as mixtures in the environment. For
water systems, these data suggest that systems with high concentrations of one PFAS compound
are more likely to have higher concentrations of other PFAS and that there is notable co-
occurrence at elevated concentrations (Cadwallader et al., 2022).
PFAS mixtures have also been reported in U.S. ambient surface waters and in aquatic biota
(Ahrens, 2011; Benskin et al., 2012; Burkhard, 2021; Nakayama et al., 2007; Remucal, 2019;
Zareitalabad et al., 2013; McCord and Strynar, 2019). Most environmental monitoring of PFAS
in surface waters has focused on sites of historical manufacturing and known contamination (3M
Company, 2000; Boulanger et al., 2004; Cochran, 2015; Hansen et al., 2002; Jarvis et al., 2021;
Konwick et al., 2008; Nakayama et al., 2007). Simcik and Dorweiler (2005) consistently
detected both PFOA and PFHpA in all 12 surface waters sampled across the U.S. Midwest, and
PFOS in all but two locations. Sinclair and Kannan (2006) detected PFOA and PFOS in all
effluent-dominated samples collected across New York State. In addition to PFOA and PFOS,
Sinclair and Kannan (2006) also detected PFHxS; however, PFBS and perfluorooctane
sulfonamide (PFOSA) were below detection limits in all samples. De Silva et al. (2011) detected
PFOS and additional short chain PFAS (i.e., perfluoropentanoic acid (PFPeA) (C5), PFHxA
(C6), PFHpA (C7), and PFOA (C8)) co-occurring as mixtures in all surface water samples
(n = 32) collected across the five Laurentian Great Lakes. Relatively longer chain PFAS,
including PFNA (C9), PFDA (C10), perfluoroundecanoic acid (PFUnA) (CI 1), PFBS, PFHxS,
perfluoroethylcyclohexane sulfonate (PFECHS), and perfluoromethylcyclohexane sulfonate,
were also quantified in at least 20 of the 32 samples collected from the Great Lakes.
PFAS mixtures in the environment can be linked to direct application of manufactured products
that contain a specific mixture of PFAS. For example, aqueous film forming foam (AFFF) used
in firefighting and training activities can contain hundreds of polyfluoroalkyl precursors (Ruyle
et al., 2021). Anderson et al. (2016) quantified PFAS in ambient surface waters across 10 U.S.
Air Force bases where there were known historic uses of AFFF. PFOA and PFOS largely co-
occurred with one another and were detected in 88% and 96% of samples, respectively.
Anderson et al. (2016) also detected PFBA, PFBS, PFPeA, PFHxA, PFHxS, and PFHpA
in > 80% of samples.
Environmental monitoring of PFAS in aquatic biota has primarily focused on fish. Generally,
PFCAs are less bioaccumulative than PFSAs in aquatic systems, with longer chain PFAS being
6 The 4% figure is based on 198 PWSs reporting measurable PFAS results for one or more sampling events from
one or more of their sampling locations. Those 198 PWSs serve an estimated total population of approximately 16
million (EPA, 2019b,c).
11
-------
PUBLIC REVIEW DRAFT
MARCH 2023
more bioaccumulative than short chain PFAS (Burkhard, 2021; Conder et al., 2008; Kannan et
al., 2005). Within the United States, PFAS in aquatic biota have been measured in several
estuaries, in the Laurentian Great Lakes region, and targeted studies of impacted (e.g., industrial)
sites. Sedlak et al. (2017) measured PFAS in composite samples containing yellowfin gobies
(Acanthogobius flavimanus), chameleon/cheekspot gobies (Tridentiger trigonocephalus/Ilypnus
gilberti), northern anchovy (Engraulis mordax), shiner surfperch (Cymatogaster aggregata), and
staghorn sculpin (Leptocottus armatus) that were collected from the San Francisco Bay estuary.
PFOS and PFOSA were detected in nearly all composite samples and at relatively high
concentrations (geometric mean PFOS = 3.9 nanograms (ng) per gram (g); geometric mean
PFOSA = 3.2 ng/g). Other longer chain PFAS, including PFNA, PFDA, PFUnA, and
perfluorododecanoic acid (PFDoDA), were also frequently detected in the fish composite
samples, but at relatively low concentrations (geometric mean concentrations < 2.4 ng/g).
Shorter chain PFAS, including PFBS, PFBA, PFHxA, and PFHpA, were not detected in any of
the fish composite samples. Houde et al. (2006) measured whole body PFAS in six fish species
in Charleston Harbor, South Carolina, and in five fish species in Sarasota Bay, Florida. Out of
the six species from Charleston Harbor, PFOA, PFOS, PFNA, PFDA, PFUnA, PFDoDA,
PFHxS, and PFOSA were all commonly detected in fish tissues. Charleston Harbor was the more
developed of the two sites and had higher overall PFAS concentrations. PFOS and PFDoDA
were the only two PFAS that were detected at elevated concentrations in the fish species residing
in Sarasota Bay (Houde et al., 2006). De Silva et al. (2011) measured PFAS from lake trout
(Sa/ve/inus namaycush) samples collected in 2001 from each of the Great Lakes. Eight different
PFAS (i.e., PFNA, PFDA, PFUnA, PFDoDA, perfluorotridecanoic acid (PFTrDA),
perfluorotetradecanoic acid (PFTA), PFHxS, and PFOS) were detected in lake trout tissues
across all of the Great Lakes, with PFOA, PFECHS, and perfluorodecanesulfonic acid (PFDS)
also being detected in Lake Ontario (De Silva et al., 2011). A study in New Jersey found co-
occurrence of PFAS in ambient water, sediment, and fish at sites with historic and current
industrial activities (Goodrow, et al., 2020). Fish tissue concentrations of PFOS were generally
higher than other PFAS and high enough in nearly all fish species to trigger fish consumption
advisories.
Within the United States, PFAS occurrence in invertebrate tissues, such as shellfish, has not been
as extensively monitored as PFAS occurrence in fish. Kannan et al. (2005) measured PFAS in
several species, including zebra mussels, from two rivers in southern Michigan (Raisin River, St.
Claire River), and one in northern Indiana (Calumet River). Overall, PFAS concentrations in
zebra mussels were lower than in fish. Nevertheless, PFOS and PFOSA were both detected in
zebra mussels in the Raisin River (PFOS concentration = 3.1 ng/g wet weight; PFOSA
concentration = 2.7 ng/g wet weight). Interestingly, PFOA was not detected in zebra mussel
tissues even though it was detected in elevated concentrations in the Raisin River water column
(PFOA water concentration = 17.7 ng/liter (L)), suggesting that chemical-specific considerations
(e.g., carbon chain length, functional group differences) affect bioaccumulation dynamics in
aquatic organisms and resultant human exposures to PFAS mixtures via ingestion of fish and
shellfish (Kannan et al., 2005).
1.5 Evidence of PFAS Exposure in Humans
Humans can be exposed to PFAS through a variety of sources, including food that is packaged in
PFAS-containing materials, processed with equipment that use PFAS, or grown or raised in
12
-------
PUBLIC REVIEW DRAFT
MARCH 2023
PF AS-contaminated soil or water (including livestock and seafood); commercial household
products, including stain- and water-repellent fabrics, nonstick products, polishes, waxes, paints,
and cleaning products; the fire suppressant, AFFF; production facilities or industries that use
PFAS; and drinking water, where these chemicals have contaminated water supplies. Although
humans may be exposed to PFAS via dermal and inhalation routes, the primary focus of this
document is the oral route of exposure, including drinking water, food, fish/shellfish, and
incidental soil/dust ingestion (Egeghy and Lorber, 2010; Lorber and Egeghy, 2011; Poothong et
al., 2020).
The Centers for Disease Control and Prevention's National Health and Nutrition Examination
Survey (NHANES) has measured blood serum concentrations of several PFAS in the general
U.S. population since 1999. Results, from a nationally representative, biomonitoring study in
which data were gathered from 1999-2000 through 2015-2016, documented measurable serum
levels of PFOS, PFOA, PFHxS and PFNA in greater than 95% of participants, indicating
widespread exposure to these PFAS in the U.S. population. PFOA and PFOS have been detected
in up to 98% of serum samples collected in biomonitoring studies that are representative of the
U.S. general population; however, blood levels of PFOA and PFOS dropped 60% to 80%
between 1999 and 2014, presumably due to restrictions on their commercial use in the United
States. Under EPA's PFOA Stewardship Program, the eight major companies of the
perfluoropolymer/fluorotelomer industry agreed to voluntarily reduce facility emissions and
product content of PFOA, precursor chemicals that can break down to PFOA, and related higher
homologue chemicals, including PFNA and longer-chain PFCAs, by 95% on a global basis by no
later than 2010 and eliminate these substances in products by 2015 (EPA, 2021c). However,
since the voluntary phase out of these longer-chain PFAS compounds in the United States,
manufacturers are shifting to shorter-chain and alternative forms of PFAS compounds such as
HFPO-DA. Additionally, other PFAS compounds were found in blood samples from recent
(2011-2016) NHANES surveys, for example, PFDA, PFDoDA, PFHpA, PFHxS, PFNA, and 2-
(N-Methyl-perfluorooctane sulfonamido) acetic acid. Studies of residents in locations of
suspected PFAS contamination show higher serum levels of PFAS compared to the general U.S.
population reported by NHANES (Kotlarz et al., 2020; Yu et al., 2020; Table 17-6 in ITRC
2022; ATSDR, 2022). There is less publicly available information on the occurrence and health
effects of these replacements for PFOA and PFOS and other members of the carboxylic acid and
sulfonate PFAS families.
1.6 Brief Summary of State, National, and International Approaches to Address
PFAS Mixtures in Water
In 2016, EPA finalized drinking water Health Advisories of 70 parts per trillion (ppt) for PFOA
and PFOS, for the individual chemicals and when present as a mixture (EPA, 2016a,b) because
the RfDs were based on developmental effects and numerically identical. Since then, some states
have developed state-specific cleanup levels, drinking water or groundwater guidelines,
advisories or standards for PFOS and PFOA. In some cases, the state values are the same as
EPA's 2016 drinking water Health Advisory (70 ppt for the individual and/or combined
concentration of PFOA and PFOS); in other cases, states have developed different values. As of
October 2022, Alaska, Colorado, Florida, Montana, Ohio, and Rhode Island have followed an
approach similar to EPA and have adopted or otherwise applied a value of 70 ppt (e.g., as a
13
-------
PUBLIC REVIEW DRAFT
MARCH 2023
guideline, advisory, or enforceable standard for water resources) to account for the combined
toxicity of PFOA and PFOS (Table 1-4).
Several states have developed values below 70 ppt and/or included additional PFAS (beyond
PFOA and PFOS) in their combined toxicity approach based on similarity in chemical structure
and/or toxicity (Table 1-4).Wisconsin has established a maximum concentration of 20 ppt for
combined PFOA and PFOS (WIDHS, 2019a,b), while Massachusetts and Maine derived a
maximum concentration of 20 ppt for any combination of the following six PFAS: PFOA, PFOS,
PFNA, PFHxS, PFDA and PFHpA based on "close similarities in chemical structure and similar
toxicities for this subgroup of PFAS" (Maine DEP, 2021; Mass DEP, 2019). Similarly, Vermont
established a limit of 20 ppt for the same PFAS as Massachusetts and Maine with the exclusion
of PFDA based on several criteria, including that "PFHxS, PFHpA and PFNA ... are considered
sufficiently similar to PFOA and PFOS" (VT DEC, 2021).
In June 2022, EPA issued draft interim updated health advisories for PFOA and PFOS and final
health advisories for HFPO-DA and PFBS (EPA, 2022a,b,c,d). EPA's interim updated health
advisories for PFOA and PFOS are 0.004 ng/L and 0.02 ng/L, respectively, and the final health
advisories for HFPO-DA and PFBS are 10 ng/L and 2,000 ng/L, respectively. Each of the health
advisory documents provides an example of how to use the HI approach to assess the potential
noncancer risk of a mixture of PFOA, PFOS, HFPO-DA, and PFBS (EPA, 2022a, b,c,d)
consistent with the approach presented in this framework document.
International approaches to addressing multiple PFAS in drinking water have resulted in a range
of proposed and promulgated standards, guidance values, and a variety of grouping methods
(Table 1-4). Canada has adopted a method similar to the HI to estimate mixture toxicity by
adding the ratio of the PFOA concentration to its maximum acceptable concentration (MAC) to
the ratio of the PFOS concentration to its MAC. If the sum of the ratios is equal to or lower than
one, the drinking water is considered safe to drink. Australia has established a combined level of
70 ppt for PFOS and PFHxS, as a precaution, based on the assumption that PFHxS is similar in
toxicity to PFOS (i.e., PFOS tolerable daily intake also applies to PFHxS). Several countries
have expanded the combined toxicity approach to include a variety of other PFAS chemicals. For
instance, Denmark has set a limit of 100 ppt to account for any combination of the following:
C4-C10 PFCAs, PFBS, PFHxS, PFOS, PFOSA, and 6:2 fluorotelomer sulfonic acid (6:2 FTS).
Sweden has adopted the same approach, not including PFOSA, and set a maximum limit of
90 ppt. In both Denmark and Sweden, it is assumed that these PFAS are similar in toxicity to
PFOS. Most recently, the European Union (EU) adopted a level of 100 ppt for the sum of 20
PFAS including C4-C13 PFSAs and C4-C13 PFCAs and a level of 500 ppt for all PFAS, as
measured by extractable or adsorbable organofluorine (EOF/AOF) (Cousins et al., 2020; EU,
2020). Further, Sweden and the Netherlands have evaluated the potential human health risk(s)
associated with mixtures of PFAS using component-based methods consistent with the HI or
RPF approaches presented in EPA's framework (Borg et al., 2013; RIVM, 2018). Although not
specifically related to drinking water, the European Food Safety Authority (EFSA) has also taken
PFAS mixture toxicity into consideration in their development of a Tolerable Weekly Intake for
the sum of PFOA, PFNA, PFHxS, and PFOS (4.4 ng/kg/week) (EFSA, 2020).
14
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Table 1-4. Summary of U.S. and International Approaches to Addressing the Combined
Toxicity of Multiple PFAS in Drinking Water or Groundwatera'b (only combined PFAS
approaches are presented)
Entity
Date
Cone (ng/L)
Sum of PFAS
Background
EPA (EPA,
2022
0.004, PFOA
HI example for
Interim Updated Drinking
2022a,b,c,d;
(supersedes
(Interim)
PFOA, PFOS,
Water Health Advisory for
EPA, 2016a,b;)
2016
0.02, PFOS
HFPO-DA and
PFOA and PFOS based on
advisory)
(Interim)
PFBS
human epidemiological data. HI
10, HFPO-
example assumes dose additive
DA
toxicity of PFOA, PFOS,
2000, PFBS
HFPO-DA, and PFBS.
Drinking Water Health
2016
70
PFOA and PFOS
Advisory. Assumes dose
additive toxicity of PFOA and
PFOS.
Alaska (USA)
2019
70
PFOA and PFOS
Application of EPA 2016
(Alaska DEC,
Health Advisory.
2019)
Colorado (USA)
2020
70
PFOA and PFOS
Application of EPA 2016
(CDPHE, 2020)
Health Advisory.
Delaware (USA) 2018
17°
PFOA and PFOS
Based on the sum of
(DE DHHS,
approximately 50% of each
2021)
individual MCLd
Florida (USA)
2019
70
PFOA and PFOS
Application of EPA 2016
(Florida Health,
Health Advisory.
2020)
Maine (USA)
2021
20
PFOA, PFOS,
Based on similarities in
(Maine DEP,
PFNA, PFHxS,
chemical structure and
2021)
PFHpA, and
toxicities of six PFAS to PFOS
PFDA
and PFOA. Same approach as
EPA 2016 Health Advisory but
includes an additional
uncertainty factor.
Massachusetts
2019
20
PFOA, PFOS,
Based on similarities in
(USA) (Mass
PFNA, PFHxS,
chemical structure and
DEP, 2019)
PFHpA, and
toxicities of six PFAS to PFOS
PFDA
and PFOA. Same approach as
EPA 2016 Health Advisory but
includes an additional
uncertainty factor.
15
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Entity
Date
Cone (ng/L)
Sum of PFAS
Background
Montana (USA)
(MT DEQ,
2020)
2019
70
PFOA and PFOS
Application of EPA 2016
Health Advisory
Ohio (USA)
(Ohio EPA,
2019)
2019
70
PFOA and PFOS
Application of EPA 2016
Health Advisory
Rhode Island
(USA) (RIDEM,
2017)
2019
70
PFOA and PFOS
Application of EPA 2016
Health Advisory
Vermont (USA) 2019 20 PFOA, PFOS PFHxS, PFHpA and PFNA are
(VT DEC, 2021) PFNA, PFHxS considered sufficiently similar
and PFHpA to PFOA and PFOS. Difference
to EPA Health Advisory is due
to Vermont's calculation being
based on infant consumption
rates.
Based on ATSDR's2021
intermediate MRL, with
additional modifying factor of
10 for immunotoxicity; HI
approach.
"PFAS Total" proposed to be
enforced through measurement
of EOF/AOF once validated or
100 ppt for the sum of 20 PFAS
considered to be a concern for
drinking water (implementation
January 12, 2023).
C4-C10 PFCAs, Assumes all 12 PFAS are
PFBS, PFHxS, similarly toxic as PFOS.
PFOS, PFOSA, Rationale: PFOS is the most
and 6:2 FTS toxic and toxicity data are
limited on PFAS other than
PFOS and PFOA.
Sweden 2014
90
C4-C10 PFCAs,
Assumes all 11 PFAS are
(Swedish Food
PFBS, PFHxS,
similarly toxic as PFOS.
Agency, 2021)
PFOS and 6:2
Rationale: PFOS is the most
FTS
toxic and toxicity data are
limited on PFAS other than
PFOS and PFOA.
Wisconsin 2019 20 PFOA and PFOS
(USA) (WI
DHS, 2019a,b)
European Union 2020 100 100 ng/L for sum
(EU, 2020) 500 of 20 PFAS (C4-
C13 PFSAs and
C4-C13 PFCAs)
500 ng/L for
"PFAS Total" -
the total of all
PFAS
Denmark 2015 100
(Danish
Environmental
Protection
Agency, 2015)
16
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Entity
Date
Cone (ng/L)
Sum of PFAS
Background
Australia
(Australian
Government
Department of
Health, 2019)
2017
70
PFOS and
PFHxS
combined, if
both present
Assumes PFHxS is similarly
toxic as PFOS. Rationale:
PFOS is the most toxic and
toxicity data are limited on
PFAS other than PFOS and
PFOA.
Canada (Health
Canada, 2018)
2018
200
600
PFOA
PFOS
When PFOS and PFOA are
found together in drinking
water, HI approach is applied.
Notes:
a Modified from Cousins et al. (2020).
b As of July 2021, several states have passed or proposed compound-specific Maximum Contaminant Levels (MCLs) or Health
Advisories, e.g., California, Illinois, Michigan, Minnesota, New Jersey, New York, Pennsylvania, Texas, and Washington.
Some states have applied EPA's Health Advisory to interpret narrative water quality standards under the Clean Water Act, e.g.,
Colorado, Montana. Only approaches using the sum of PFAS parameters are presented in this table.
c Proposed level based on the Delaware PFOA and PFOS MCL Implementation Plan
d Based on a PFOA Maximum Contaminant Level (MCL) of 21 ppt and PFOS MCL of 14 ppt.
1.7 Overview of Proposed Framework for Estimating Health Risks for PFAS
Mixtures
This document describes a framework of options with different levels of data requirements and
objectives for estimating the noncancer human health risks associated with mixtures of PFAS,
based on longstanding EPA chemical mixtures guidance. To address concerns over health risks
from multichemical exposures, EPA issued the Guidelines for the Health Risk Assessment of
Chemical Mixtures in 1986 (EPA, 1986). The 1986 guidelines were followed in 2000 by the
Supplementary Guidance for Conducting Health Risk Assessment of Chemical Mixtures (EPA,
2000b). These documents define a chemical mixture as "any combination of two or more
chemical substances, regardless of source or of spatial or temporal proximity, that can influence
the risk of chemical toxicity in the target population" (EPA, 1986, 2000b); this definition is used
in this framework document.
Several laws direct EPA to address health risks posed by exposures to chemical mixtures,
including the Comprehensive Environmental Response, Compensation, and Liability Act
(CERCLA) of 1980, the Superfund Amendments and Reauthorization Act (SARA) of 1986, and
amendments in 2002 (CERCLA, 2002; SARA, 2002) (commonly referred to as Superfund); the
Clean Air Act (CAA) Amendments of 1990 (CAA, 1990); the Safe Drinking Water Act (SDWA)
Amendments of 1996 (SDWA, 1996); and the Food Quality Protection Act (FQPA) of 1996
(FQPA, 1996). Both the 1986 Chemical Mixtures Guidelines (EPA, 1986) and the 2000
Supplementary Chemical Mixtures Guidance (EPA, 2000b) were developed, in part, to be
responsive to these laws. When developing assessment information for exposures to chemical
mixtures, risk assessors and risk managers in EPA's programs currently implement
environmental laws through regulations that rely on the guidance and methods articulated in the
1986 Chemical Mixtures Guidelines and the 2000 Supplementary Chemical Mixtures Guidance.
This framework does not supersede previously published EPA guidance on mixtures or
longstanding EPA approaches used to assess health risks of contaminants including chemical
mixtures under various environmental statutes (e.g., Federal Insecticide, Fungicide, and
17
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Rodenticide Act (FIFRA); FQPA; Comprehensive Environmental Response, Compensation, and
Liability Act (CERCLA)).
The objective of this document is to provide a flexible, data-driven framework that facilitates
practical component-based mixtures evaluation of two or more PFAS under an assumption of
dose additivity. All approaches presented involve integrating dose-response metrics that have
been scaled based on the potency of each PFAS in the mixture. Three approaches are presented:
1) The HI approach is likely the most health protective approach because it is based on the
most sensitive health outcome (lowest RfD) for each chemical and provides a risk
indicator for exposure to a PFAS mixture of concern (Section 5),
2) The RPF approach provides a mixture toxicity estimate by scaling the potency of
component chemicals, for a common health effect, relative to a well-characterized
member of the mixture, referred to as the index chemical (IC) (Section 6);
3) The M-BMD approach uses a DA model-based equation (similar to the Berenbaum
equation; Section 4.2.6 in EPA, 2000b) to calculate a BMD (e.g., BMDiohed) for the
mixture (Section 7).
The HI facilitates estimation of potential joint toxicity associated with co-occurrence of
chemicals in environmental media (e.g., water, soil) (EPA, 2000b). The RPF method is more
data intensive than the HI approach in that the mixture component chemicals typically must meet
two requirements: (1) there are data to demonstrate or suggest that component chemicals share
either a similar toxicological MO A7 or have a conserved toxicological target (i.e., share a
common apical endpoint/effect); and (2) the dose-response functions for the effect of concern are
congruent (similar shape and slope) over the exposure ranges most relevant to the decision
context (EPA, 2000b). The RPF method is illustrated in Section 6 using common target
organs/pathways including liver, thyroid, and developmental effects. A MOA for a given toxic
effect is a detailed description of the source to outcome pathway, including the key
molecular/cellular or organellar events, leading to a defined health effect or syndrome of effects
(e.g., "developmental" can be a collection of related outcomes). In general, a health effect or
outcome is the terminus of one or more operant MOA(s). In addition to the HI and RPF methods,
the assumption of similarity in MOA or toxicological target is also inherent when applying the
M-BMD approach; however, in contrast to the RPF method there is no necessity or assumption
of congruent dose response functions (i.e., same/similar shape or slope) across chemicals. This
approach provides more accurate predictions of a mixture effect even if the slopes of the dose
response curves differ among the chemicals (Section 7). Considering that PFAS are an emerging
chemical class of note for toxicological evaluations and human health risk assessment, MOA
data may be limited or not available for many PFAS. As such, this framework focuses the
biological level of organization for evaluation of potential dose additivity on similarity of
toxicological endpoint/effect/adverse outcome rather than similarity in MOA, which is
consistent with EPA mixtures guidance (EPA, 2000b) and expert opinion from the National
Academy of Sciences, National Research Council (NRC, 2008).
7 Mode of action is a sequence of key events and processes, starting with interaction of an agent and a cell,
proceeding through operational and anatomical changes, and resulting in a noncancer effect or cancer formation.
18
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Recognizing the evolving and dynamic nature of PFAS science, the component-based mixtures
assessment approaches described herein are flexible to allow for consideration of new or
evolving dose-response data and toxicity assessments as they become available. Additionally,
because publicly available traditional (e.g., in vivo mammalian) toxicity studies are limited to
only a small fraction of the -12,000 PFAS, this framework also provides suggestions for
practical integration of validated NAMs such as toxicogenomics (e.g., in vitro cell bioactivity)
and in silico platforms (e.g., structure-activity, read-across) into the HI, RPF, and M-BMD
approaches. The illustrative examples in Sections 5, 6, and 7 are intended to demonstrate the
application of dose-additive-based component-chemical mixture approaches using hypothetical
human health relevant toxicity and exposure information.
19
-------
PUBLIC REVIEW DRAFT
MARCH 2023
2.0 Background on EPA Mixtures Additivity Guidance
Exposure to mixtures of environmental chemicals occurs in human populations through
ingestion, inhalation, and/or dermal contact with contaminated media (e.g., water, air, food). It
should be noted that a "mixture" of chemicals may be a function of both co-occurrence in
exposure media and/or internal bioaccumulation and persistence in biological matrices. In
recognition of the need for methods and approaches that inform evaluation of potential health
risks associated with chemical mixtures, EPA developed the 1986 Chemical Mixtures Guidelines
and subsequently the 2000 Supplementary Chemical Mixtures Guidance (EPA, 1986, 2000b). In
those guidance documents, EPA proposed a hierarchy of mixtures approaches where the
preferred approach is to evaluate toxicity using hazard and dose-response data for a specific
whole mixture of concern, or alternatively a sufficiently similar mixture. However, whole
mixture data are rare; there are often too many chemical combinations and proportions in the
environment (e.g., parent chemicals, metabolites, and/or abiotic degradants) introducing a level
of complexity that is difficult to evaluate and characterize. Further, most controlled experimental
toxicity data derive from single chemical exposures, or at best, small mixtures (i.e., limited
number of component chemicals at fixed proportions/ratios). As such, EPA also developed
multiple component-chemical based mixtures assessment approaches. Component based
methods are used more frequently than whole-mixture methods. These component methods are
based on assumptions of how the chemicals behave when co-occurring. Although observed
toxicity could be related to direct chemical-to-chemical interaction(s), the manner in which co-
occurring chemicals induce toxicity in a coordinated or independent way is the basis for the
concept of "additivity." Basic tenets of EPA mixtures additivity theory and practice are as
follows:
• Additivity based methods are used to estimate the probability or magnitude of a given
health outcome (e.g., incidence and/or severity, or change in magnitude, of a noncancer
target organ effect) associated with exposure to mixtures of two or more component
chemicals. In the 1986 and 2000 EPA mixtures guidelines and guidance documents,
development of component-based mixture approaches were informed by two main
concepts, simple similar action and simple independent action, as described by Bliss
(1939) and Finney (1971).
• Simple similar action applies to mixtures of chemicals that cause a common health effect
via toxicologically similar pathway(s). Under simple similar action (i.e., DA), the
evidence associated with toxic responses to mixture component chemicals demonstrate or
suggest coordinated (i.e., same/similar) pathway events. DA is generally applied when
mixture chemicals are assumed to act through simple similar action.
• Simple independent action applies to mixtures of chemicals that cause a common health
effect via toxicologically independent pathways. Under simple independent action (i.e.,
response addition; RA), the evidence associated with toxic responses to different mixture
component chemicals demonstrate or suggest independent pathway events. RA is
generally applied when mixture chemicals are assumed to act through simple independent
action.
20
-------
PUBLIC REVIEW DRAFT
MARCH 2023
2.1 Component-Based Mixtures Assessment Methods
Component-based methods that EPA has developed for evaluating potential additivity of dose,
response, or both are shown in Figure 2-1. Based primarily on similarity in toxicity
endpoint/health effect of PFAS, this framework document focuses on the use of dose-additive,
component-based methods (left side of Figure 2-1; shaded box), specifically the HI (Section 5),
RPF (Section 6), and M-BMD (Section 7) approaches. The methods involve different
assumptions and data requirements and objectives for evaluating the joint toxicity of component
chemicals in a "mixture." Each of the methods are introduced and detailed in Sections 5-7 and
include demonstration of application using a hypothetical five component PFAS mixture.
Specifically, to facilitate the use of potentially disparate sources and types of PFAS information
in a mixture context, the data-driven application of the component-based methods presented in
this framework document is demonstrated using a hypothetical mixture of five PFAS as follows:
PFAS 1 = comprehensively studied, most potent for effect(s), and has formal noncancer
human health assessment value(s) and an HBWC available;
PFAS 2 = well-studied, second most potent for effect(s) among PFAS 1-3, and has
formal noncancer human health assessment value(s) and HBWC available;
PFAS 3 = studied, least potent for effect(s) among PFAS 1-3, and has formal noncancer
human health assessment value(s) and HBWC available;
PFAS 4 = experimental animal toxicity data available but no formal human health
assessment and no HBWC; and
PFAS 5 = data-poor.
This hypothetical PFAS mixture is purposefully designed to demonstrate how the framework
allows for flexible integration of information derived from diverse data types and sources.
Opportunities for integrating PFAS into a mixture assessment is expected to evolve over time
and will depend on the decision context and availability of hazard and dose-response data from
traditional and/or NAM-based assays and/or in silico platforms.
An important property of DA-based methods is that they can aid in the indication or estimation
of effects of a mixture even when all of the individual component chemical exposures are at or
below their individual no-observed-adverse-effect levels (NOAELs; i.e., 'something from
nothing'). In dose additivity models such as the RPF approach, the sum of the scaled IC8
equivalent doses/concentrations for each component can exceed the equivalent threshold dose of
the mixture and result in a detectable response, which has been supported experimentally (Jonker
et al., 1996; Silva et al., 2002).
8 An IC is that mixture component that is typically the most toxicologically well-studied member. The qualitative
and quantitative hazard and dose-response data for an index chemical serve as an index or anchor against which all
other components are compared. IC equivalent doses/concentrations represent scaled dose(s) of mixture
components, based on potency for a given toxicity endpoint/health effect, in a corresponding dose of the index
chemical.
21
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Component data available for Exposure and Dose-Response
Components Grouped per Evidence of Toxicological Similarity
Toxicologically
Similar
Components
Mix of Toxicologically
Similar and Independent
Components
Toxicologically
Independent
, Components
Hazard Index (HI)
• Screening-level
• TOSH I
• Interactions-based HI
'Relative Potency Factors (RPF) ^
RPF-based Hazard/Risk
Estimate
• RPF-based MOE
* RPF-based HQs
Integrated Addition
Mixture-based BMD
Hazard Estimate
Notes:
Modification of Figure 4-3b (EPA, 2007a)
Component-based methods selection is based on the relevant evidence supporting toxicological similarity (DA) or toxicological
independence (RA or effect summation). Integrated addition methods are reserved for mixtures of component chemicals that
demonstrate a profile of both toxicological similarity and independence.
BMD = benchmark dose; HI = hazard index; HQ = hazard quotient; MOE = margin of exposure; RPF = relative potency factor;
TOSHI = target-organ specific hazard index.
Figure 2-1. Flow chart for evaluating chemical mixtures using component-based additive
methods.
2.1.1 Application of Dose Addition as EPA's Default Assumption
Several in vivo studies have examined predicted mixture responses based on dose-addition
models for specific groups of chemicals (e.g., Altenburger et al., 2000; Crofton et al., 2005;
EPA, 2007a; Gennings et al., 2004; Hass et al., 2017; Howdeshell et al., 2015; Kortenkamp and
Haas, 2009; Moser et al., 2005, 2012; Mwanza et al., 2012; Rider et al., 2008, 2009, 2010;
Walker et al., 2005), focusing primarily on whether experimentally observed toxicity is
consistent with modeled predictions of dose-additivity. Many of these studies examined groups
of chemicals that are thought to target the same biological signal transduction pathways (Moser
et al., 2012; Mwanza et al., 2012; Walker et al., 2005), while others have examined chemicals
thought to target disparate pathways that lead to the same health outcome (Van Der Ven et al.,
2022; NRC, 2008; Rider et al., 2009). In general, the results of such studies listed here, and many
others, support the continued application of DA as EPA's default component-based mixture
assessment approach. Further discussion and examples of the basis for use of dose additivity for
component-based evaluation of PFAS mixtures is provided in Section 3.
22
-------
PUBLIC REVIEW DRAFT
MARCH 2023
3.0 Dose Additivity for PFAS
This section presents a review of in vivo chemical mixture studies for different biological
pathways that provide information on how mixtures of chemicals with similar and dissimilar
molecular initiating events (MIEs) and/or MO As interact. Section 3.2 discusses the evidence
demonstrating that mixtures of chemicals disrupting common pathways typically produce dose
additive alterations. In in vivo studies that rigorously tested accuracy for DA, Integrated Addition
(IA) and RA model predictions for mixtures with components that disrupted common pathways,
DA models provided predictions that were better than or equal to IA and RA predictions of the
observed mixture effects (Section 3.2). Consistent with the conclusions of the National Academy
of Sciences (NAS) (NRC, 2008), Boobis et al. (2011) and Martin et al. (2021) found that
published studies in the literature (Section 3.2) support the assumption of DA as the default
model for estimating mixture effects, even when the mixtures included chemicals with diverse
MO As (but common targets of toxic action). Further, these two large systematic reviews of the
literature on chemical mixtures found little evidence for deviations from dose additivity, such as
synergy or antagonism (Boobis et al. 2011; Martin et al., 2021). For example, Martin et al.
(2021), following a review of more than 1,200 mixture studies (selected from > 10,000 reports),
concluded that there was little evidence for synergy or antagonism among chemicals in mixtures
and that DA should be considered as the default model. Taken together, this supports the health
protective assumption that a mixture of chemicals with similar apical effects should be assumed
to also act in a DA manner unless data demonstrate otherwise. Further, experimental data
demonstrate that PFOS, PFOA, and other PFAS disrupt signaling of multiple biological
pathways resulting in common adverse effects on several biological systems and functions
including thyroid hormone functioning, lipid synthesis and metabolism, developmental toxicity,
and immune and liver function, and are reviewed in Section 3.4. Finally, in Section 3.4, a
summary is provided for two ongoing EPA Office of Research and Development (ORD) PFAS
developmental toxicity mixture studies (of which one study uses a mixture of PFOA and PFOS)
that provide robust evidence that PFAS behave in a DA manner.
3.1 Overview of Assessment Approaches for Chemical Mixtures
Over 30 years ago, scientists developed quantitative dose metrics and methods to assess the
combined toxicity of mixtures of large classes of chemicals that disrupt a common pathway
(NATO, 1988). Toxicity equivalence factors (TEFs) were initially developed in the mid-1980s
for hundreds of dioxin-like polychlorinated biphenyls (PCBs), polychlorinated dibenzofurans
(PCDFs), and polychlorinated dibenzo-p-dioxins (PCDDs) based upon their potency relative to
2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). Many of the lessons learned about assessing the
effects of mixtures of dioxin-like chemicals (DLCs) also are applicable to assessing the effects of
PFAS mixtures. Since that time, TEF-like approaches have been used to evaluate mixtures of
other chemical classes. The emerging picture is that some chemicals, regardless of MIE or MO A,
produce mixture effects on common apical endpoints that generally are well predicted using DA
models.
The general applicability of DA models is based on reviews of studies specifically designed to
evaluate how well different mixture models predict the way chemicals in a mixture interact to
produce effects. Studies evaluating mixture models and effects typically include an evaluation of
individual chemical dose response curves and apply this information to different statistical
models. In general, systematic reviews have noted that many mixture studies do not include
23
-------
PUBLIC REVIEW DRAFT
MARCH 2023
information conducive for evaluating the utility of different mixture models. The data from
several studies (reviewed in Section 3.2) indicate that chemicals that produce common adverse
effects will typically interact in a DA manner when they occur together in a mixture. Thus, the
effects of any combination of co-occurring chemicals can be predicted when sufficient chemical
dose-response data are available for all of the individual components within an environmentally
relevant mixture. For example, the Consumer Product Safety Commission Chronic Hazard
Advisory Panel (CPSC CHAP) on Phthalates used DA models to predict the hazard posed by
mixtures of phthalates to pregnant women and children. In their assessment, phthalate mixture
exposures from NHANES data were used to predict individual hazard scores for each person and
then determine the percentage of people who exceeded a point-of-departure (POD) (CHAP,
2014).
In the absence of an adequate in vivo database to evaluate mixture models, it should be assumed
that any mixture acts in a DA manner if the individual chemicals produce common effects. This
approach was fully endorsed by NRC (2008) and Martin et al. (2021).
3.2 Examples of Chemical Classes and Toxicological Pathways Utilizing Mixture
Assessment Approaches
3.2.1 Dioxin-Like Chemicals and AryS Hydrocarbon Receptor Pathway Toxic
Equivalence Factors (TEFs)
In 2010, EPA published guidance for the use of TEFs for human health risk assessments of
DLCs, which produce many of their adverse effects by acting as aryl hydrocarbon receptor
(AhR) agonists (EPA, 2010). It should be noted that the TEF approach is a specialized
application under the RPF umbrella but is only applicable when all mixture components induce
an effect via an identical MIE/MOA (e.g., AhR agonism). DLCs such as PCBs, PCDFs, and
PCDDs have been identified as AhR agonists. As such, for DLC mixtures, EPA recommended
use of the TEF methodology and the World Health Organization's TEFs to evaluate the risks
associated with exposure to mixtures of 2,3,7,8-TCDD and DLCs for human health (EPA, 1987,
1989, 2003) and ecological risk assessments (EPA, 2008). TEFs can be calculated for each DLC
based on dietary dose or internal whole body toxic equivalent concentrations (TECs).
The joint toxicity of a DLC mixture is based on toxic equivalents (TEQs) which are toxicity-
weighted masses of mixtures of PCDDs, PCDFs, and PCBs. The TEQ for each chemical in the
mixture is calculated by multiplying each TEF by the corresponding chemical concentration in
the mixture. The individual TEQs are then summed to calculate the TEQ of the mixture. The
reported TEQ provides toxicity information about the mixture of chemicals and is more
meaningful than reporting the total mass of DLCs in grams.
This approach assumes:
• Chemicals interact in a DA manner;
• They all affect a common pathway via the AhR, among other pathways;
• Synergistic and antagonistic interactions are uncommon within the group (Safe, 1994);
and
• TEFs and TEQs based on AhR agonism are not necessarily predictive of chemical
potency for effects mediated by other receptors or pathways.
24
-------
PUBLIC REVIEW DRAFT
MARCH 2023
EPA's TEFs have undergone several revisions (Van den Berg et al., 2006). In 2010, EPA
published recommended TEFs for human health risk assessment of DLCs (EPA, 2010).
Although the AhR is present in all classes of vertebrates, vertebrate species vary greatly in their
sensitivity to environmental TEQ levels. Sensitive species include terns and cormorants (bill
deformities), herons (embryo mortality), and mink (lethality and reproductive failure) (Beckett et
al., 2008; Restum et al., 1998), for example. Adverse effects also occur in frogs (amphibians)
(Gutleb et al., 2000), fish (Monosson, 2000), and snapping turtles (reptiles) (Bishop et al., 1998;
Gale et al., 2002). EPA (2008) stated that the TEQ methodology was appropriate for evaluating
risks to fish, birds, and mammals associated with AhR agonists.
Studies of AhR agonists in various species indicate:
• Species and tissues differ in sensitivity to the effects of the mixture; and
• Even though the AhR pathway is conserved, the adverse outcomes can vary greatly from
species to species.
One common effect of DLCs is a reduction in serum thyroxine (T4). Crofton et al. (2005)
conducted a mixture study of 18 thyroid-disrupting DLCs consisting of 12 PCBs, 4 PCDFs, and
2 PCDDs at 6 dilutions of the highest dose, which contained effective dose (ED30) concentrations
of each chemical in the high dose. This mixture reduced serum T4 in a dose-related manner. The
reduction in T4 was dose additive in the low dose range of interest, but the observed reduction in
T4 in the high dose (46% reduced) exceeded DA predictions (28% reduced) by about 18%. In a
review of the literature on the effects of mixtures on the thyroid axis, Crofton (2008) concluded
"To date, the limited data from thyroid disrupting chemical mixture studies suggest that DA is
reasonably accurate in predicting the effects on serum T4 concentrations."
3.2.2 Pyrethroids/Pyrethrins - Centra! Nervous System and Behavior
Pyrethrins and pyrethroids share the ability to interact with voltage-gated sodium channels
ultimately leading to neurotoxicity. Wolansky et al. (2009) administered a mixture of
11 pyrethroid pesticides to adult male rats acutely by oral gavage using a fixed-ratio dilution
design at eight dose levels and measured locomotor activity on the day of dosing. The reduction
in exploratory activity by the mixture was accurately predicted by DA modeling.
EPA has determined that naturally occurring pyrethrins and synthetic pyrethroid pesticides form
a common mechanism group. This common mechanism grouping is based on 1) shared
structural characteristics, 2) shared ability to interact with voltage-gated sodium channels
(VGSC), resulting in disruption of membrane excitability in the nervous system, and 3)
ultimately neurotoxicity characterized by two different toxicity syndromes. In 2011, after
establishing a common mechanism grouping for the pyrethroids and pyrethrins, EPA conducted
a cumulative risk assessment using a RPF approach (EPA, 201 la).
3.2.3 Organophosphates - Lethality, Centra! Nervous System and Behavior
In the late 1950s Murphy and Dubois (1957) reported that O-ethyl O-p-nitrophenyl
phenylphosphonothioate potentiated the lethality of malathion when the two chemicals were
given simultaneously. Subsequently, all organophosphate (OP) pesticides in use were evaluated
in binary mixture studies to determine if non-additivity was a common outcome among this class
of insecticides (reviewed by Moser et al., 2005; Padilla, 2006). An examination of the
25
-------
PUBLIC REVIEW DRAFT
MARCH 2023
interactions of 43 pairs of OP insecticides revealed that 4 pairs showed greater-than-additive
effects on lethality (Dubois, 1961). Moser et al. (2005, 2006) reported a range of responses with
mixtures of 4 or 5 OPs. The ratios of the predicted-to-observed ED20S and ED50S of the mixtures
indicated that several effects displayed greater-than-additive effects (ratios = 1.2 to 2.6), a few
were less than additive (ratio = 0.5 to 0.9), and most were dose additive (ratio = 1).
In 1999, EPA determined that the OPs form a common mechanism group based on their shared
ability to bind to and phosphorylate the enzyme acetylcholinesterase, leading to accumulation of
acetylcholine and ultimately cholinergic neurotoxicity (EPA, 1999). As such, cumulative risk to
OPs has been assessed using AChE inhibition as the source of dose response data. The most
recent OP cumulative assessment was conducted in 2006 employing a RPF approach (EPA,
2006).
Further, in 2018 ATSDR concluded that the "default assumption of dose-additive joint action at
shared targets of toxicity (i.e., effects on neurological endpoints) be used for screening level
assessments of the potential adverse health outcome from concurrent oral exposure to mixtures
of pyrethroids, organophosphorus, and carbamate insecticides." (ATSDR, 2018).
3.2.4 Estrogen agonists - Mixture Effects on the Female Reproductive Tract
Scientists have examined the effects of mixtures of estrogenic chemicals in the female rat using a
uterotropic assay, an EPA Endocrine Disruptor Screening Program Test Guideline that is a
sensitive in vivo test for estrogenicity (EPA, 2009). In this assay, immature or adult
ovariectomized female rats are typically exposed to test chemicals for 3-4 days, after which
uterine weights are taken. Exposures can be administered orally or through subcutaneous
injections. Tinwell and Ashby (2004) exposed immature female rats for 3 days to several known
xenoestrogens, either individually or as mixtures. In a reanalysis of the data, predictions of a DA
model for a binary mixture of bisphenol A and genistein were consistent with the observed
effects of the mixture with an average deviation of observed results vs. the DA model of 4%.
Similarly, Conley et al. (2016) found that the effects of mixtures of bisphenol S + methoxychlor,
bisphenol AF + methoxychlor, and bisphenol F + bisphenol S + methoxychlor + bisphenol
C + ethinyl estradiol, administered orally to female rats, produced effects that were comparable
to predictions using DA models. Because the chemicals all stimulate uterine growth via a
common estrogen receptor alpha pathway and produce a common effect, Conley et al. (2016)
determined that DA was the most appropriate model for mixtures of these estrogenic compounds.
3.2.5 Phthaiates in utero - Mixture Effects on the Female Reproductive Tract
Hannas et al. (2013) reported that administration of a mixture of five phthaiates (> 520 mg total
phthalate) to pregnant rats from gestational days (GDs) 8 to 13 induced reproductive tract
malformations in female rat offspring. These malformations included complete to partial uterine
agenesis and agenesis of the lower vagina, an effect similar to a human congenital condition
known as the Mayer-Rokitansky-Kiister-Hauser syndrome that occurs in about 1 in 4,500 female
newborns. The phthalate mixture was a fixed-ratio dilution and contained five phthaiates that do
not produce malformations in either female or male offspring when administered individually at
the doses used in the mixture. These malformations have been seen in dibutyl-(500 milligrams
per kilogram per day (mg/kg/day)) and diethylhexyl (750 mg/kg/day) phthalate studies at a low
incidence and at high doses but were not seen in similar studies with the other three phthaiates.
Although there was not enough individual phthalate data to compare DA and RA prediction
26
-------
PUBLIC REVIEW DRAFT
MARCH 2023
models, it is clear these effects exceed RA (i.e., 0 + 0 + 0 + 0 + 0 = 75% for uterine agenesis)
and is an example of "something from nothing" (Silva et al., 2002).
3.2.6 Antiandrogens - Male Reproductive Tract Development
Historically, it has been hypothesized that mixtures of chemicals with dissimilar MIEs would
interact in a RA or IA manner. However, this conclusion is not currently supported by a large
body of literature on the effects of chemical mixtures and was rejected by NRC (2008). Studies
on the effects of mixtures on male reproductive development provide one of the larger databases
supporting the use of DA models as the default model. These studies include chemical mixtures
with common MIEs and those with multiple MIEs that converge on a common KE in multiple
AOPs in an AOP network. These studies focus on chemicals that disrupt androgen signaling in
utero during the critical period of mammalian sexual differentiation. For over 20 years, scientists
have examined the in utero effects of mixtures of chemicals that disrupt androgen signaling on
the male reproductive tract (e.g., Gray et al, 2001; reviewed by Haas et al., 2007; Howdeshell et
al., 2017; Metzdorff et al., 2007). These studies include defined binary or multi-chemical fixed-
ratio dilution mixtures and were designed to compare the observed effects to DA, RA, and IA
model predictions. The numbers of chemicals used in these studies range from 2 to 18,
administered at a range of doses enabling one to discriminate additive from antagonistic or
synergistic interactions. In all of these studies, the DA model predicted the effects of the mixture
on the male reproductive tract more accurately than IA or RA. Likewise, Metzdorff et al. (2007)
concluded that the "Effects of a mixture of similarly acting anti-androgens can be predicted fairly
accurately based on the potency of the individual mixture components by using the DA concept.
Exposure to anti-androgens, which individually appear to exert only small effects, may induce
marked responses in concert with, possibly unrecognized, similarly acting chemicals."
In addition, two recent studies were designed to specifically address a gap in the literature
identified by the CPSC CHAP (Lioy et al., 2015). At the time of their review there were no
published studies that addressed whether or not phthalate mixtures exhibited mixture effects
when administered at levels below the lowest observed adverse effect levels (LOAELs) of each
individual chemical. In the first study, a mixture of 18 administered chemicals induced effects at
dose levels about 80-fold below each chemical's individual LOAEL (Conley et al., 2018). These
18 chemicals disrupt androgen signaling via five different MIEs (Figure 3-1) and multiple AOPs
that converge on common KEs resulting in common adverse reproductive effects in male rat
offspring.
27
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Fina^rde —»• 5a-reduciase
inlvbition
p'p'-DDE Androgen
PynNuquinazon y rf?teplni
Procymidone antagonism
Vinck)7olin
Flutamsde
Piochlaraz
Linu ron
Simvastatin
Phthalates
/
Reduced DHT
synthesis
fa
Reduced AR
dependent
mRNA/protein
synthesis
Reproductive
malformations
T
Inhibition of
CYP
steroidogenic
cn/ynms
Inhibition of
HMG-CoA
reductase
Unknown Mil
\ Reduced
4 testosterone
synthesis
t
Reduced
cholesterol
synthesis
Abnormal cell
apoptosis/
proliferation
\
Reduced
androgen
dependent
tissue weights
*
Reduced
sperm
production
Infertility i
Suppressed
development of
gubernacular
cords
Toxicol Sci, Volume 164, issue 1, July 2018, Pages 166-178. https://doi.ora/10.1093/toxsci/kfv069
The content of this slide may be subject to copyright: please see the slide notes for details.
Undescended
testes
Reproductive
tract cancers
OXFORD
UNIVERSITY PRESS
Notes:
Adapted from Conley et al., 2018
The bold outlined KE indicates the critical node that links the various MIEs to the downstream adverse outcomes.
DHT = dihydrotestosterone; AR = androgen receptor; CYP = cytochrome P450; HMG-CoA = 3-hydroxy-3-methyl-glytaryl
coenzyme A.
Figure 3-1. AOP network for chemicals that disrupt AR-mediated cellular signaling leading
to adverse effects on the development of male reproductive tract resulting from in utero
exposure.
In the second study (Conley et al., 2021a), 15 chemicals (acting via at least 3 MIEs)
demasculinized male rat offspring at dose levels 2- to 4-fold lower than the individual no
observed effect levels for each chemical, and the DA models were always as good or better than
RA or IA models. For example, 60% of male offspring were found to have penile malformations
that resulted in infertility and this effect was accurately predicted by DA, whereas IA and RA
predicted that none of the males would be malformed. This is not a unique observation; rather, it
is a typical finding with male reproductive tract malformations.
Recently, Gray et al. (2022) demonstrated that the in utero effects of a PFAS pesticide,
pyrifluquinazon (contains a heptafluoroisopropyl side chain; see
https://comptox.epa.gov/dashboard/chemical/details/DTXSID6058057) combined with the di-
ortho phthalate ester dibutyl phthalate produced dose-additive effects that were more accurately
predicted with DA models that did not require parallel slopes than with RA models for multiple
male reproductive abnormalities.
28
-------
PUBLIC REVIEW DRAFT
MARCH 2023
All of these endocrine active chemicals act via AOPs that converge on a common KE in an AOP
network (Figure 3-1) that regulates the sequence of molecular events in cells that are involved in
the development of androgen-dependent tissues. Each of the identified chemicals/classes reduce
the number of androgen receptor (AR) dimers, AR/AR, activated by an androgen agonist. AR
antagonists, like vinclozolin or procymidone, accomplish abrogation of androgen-dependent
signaling by blocking androgens from binding to ARs, and the PFAS pesticide pyrifluquinazon
has been hypothesized to act by enhancing AR degradation (Gray et al., 2019; Yasunaga et al.,
2013). Chemicals like the phthalates di-n-butyl phthalate (DBP), di(2-ethylhexyl)phthalate
(DEHP), dipentyl phthalate (DPeP), butyl benzyl phthalate (BBP), and diisobutyl phthalate
(DIBP) reduce the levels of androgens available to respondent cell populations (Hannas et al.,
2011; Howdeshell et al., 2008; Furr et al., 2014). In contrast, chemicals like finasteride inhibit
the enzyme in tissues that converts testosterone to dihydrotestosterone (a more active androgen
that has higher affinity for the AR) (Clark et al., 1990). The result of these related androgen
disrupting AOPs is that fewer activated AR/AR heterodimers bind the promoter region on the
DNA of androgen-regulated genes, androgen-dependent mRNA and protein synthesis levels are
reduced, and growth and differentiation of androgen-dependent tissues in the fetus is inhibited.
As a result, male offspring display agenesis or hypoplasia or malformations in androgen-
dependent tissues. In summary, an examination of KEs disrupted in androgen signaling pathways
by chemicals such as those identified in Figure 3-1, at the cellular-molecular level, explains why
one should expect the mixtures to behave in a DA manner.
In summary, an examination of the literature on the effects of mixtures on male reproductive
tract development demonstrates that common effects can be adequately modeled by DA, the
chemicals acted in a DA manner even when including chemicals with different MIEs, and IA and
RA models can underestimate the hazard of a mixture of chemicals acting on a common KE and
with a common apical effect.
3.3 Systematic Reviews of Mixtures Toxicity: Quantification of Deviations from
Dose Additivity
Boobis et al. (2011) examined the literature from 1990 to 2008 that discussed synergy in
mammalian test systems with an emphasis on "low dose" studies. Of the 90 papers identified, 43
papers had original data from which synergy could be examined, and only 11 studies reported the
magnitude of the difference between the dose additive estimates of toxicity with the observed
results. Of these 11 studies, 6 reported magnitudes of synergy that were generally less than 2-
fold with a maximum value of 3.5-fold. As a result, the authors concluded that deviations from
DA at low doses were not common.
The issue of the occurrence of greater-than-DA (sometimes referred to as synergistic) vs. DA or
less-than-DA (sometimes referred to as antagonistic) interactions was recently reassessed by
Martin et al. (2021). The authors conducted a systematic literature review and quantitative
reappraisal of 10 years of a broad range of mixture studies from 2007 to 2017. Martin et al.
(2021) identified 1,220 mixture studies, -65% of which did not incorporate more than 2
components. They reported that "relatively few claims of synergistic or antagonist effects stood
up to scrutiny in terms of deviations from expected additivity that exceed the boundaries of
acceptable between-study variability," and that the observed effects were not more than 2-fold
greater than the predicted effects of the mixture predicted by DA.
29
-------
PUBLIC REVIEW DRAFT
MARCH 2023
3.3.1 Deviation from Additivity
Although the literature indicates that significant deviations from dose additivity are not common
among mixtures containing chemicals that disrupt common targets via common AOPs or AOP
networks, greater-than-additive (i.e., synergism) and less-than-additive (i.e., antagonism)
interactions may occur with co-exposure to chemicals that affect different target organs or
different, unrelated AOPs. There are several examples of chemical interactions that deviate from
DA in which one chemical has the capability to alter the metabolism of the other chemical(s).
For example, twenty years of research has identified at least 85 drugs whose metabolism is
inhibited by a chemical in grapefruit, potentially resulting in serious side effects (Bailey, 2013).
Furanocoumarins in grapefruit bind to the active site on the CYP3 A4 enzyme causing
irreversible inactivation that prolongs the half-life and AUC (the area under the concentration vs.
time curve) of some drugs, like some statins for example.
The effects of metabolic alterations of chemical toxicity are not limited to drug-drug interactions.
Hodgson (2012) published a comprehensive review of the effects of metabolism on the toxicity
of a large number of pesticides and also described the metabolic mechanisms of chemical
activation and/or inactivation.
In addition to metabolic activity leading to synergistic or antagonistic interactions among
chemical mixtures, there are other examples of deviations from DA that do not include chemicals
that disrupt common KEs, AOPs, AOP networks, or target organs. For example, although
melamine and its derivatives, including cyanuric acid, individually present low toxicity, together
the compounds can lead to the formation of cyanurate crystals in nephrons, causing kidney
effects and kidney failure in mammals. Such impacts have been observed in cats and dogs from
adulteration of pet food (Jacob et al., 2011), as well as infants and young children in China from
contaminated infant formula and related dairy products (WHO, 2008).
3.4 PFAS Dose Additivity
Per- and polyfluoroalkyl acids (PFAA), such as PFOA and PFOS, with linear or branched alkyl
or alkyl ether chains and sulfonic or carboxylic acid functional groups, as well as PFAA
precursors, while not necessarily toxicologically identical, do share common toxicological
impacts of exposure on multiple cellular receptors, tissues, life stages, and species (ATSDR,
2021; EFSA et al., 2018, 2020). As described above (Section 3.2), precedents of prior research
conducted on mixtures of various chemical classes with common KEs and adverse outcomes,
support the use of dose additive models for estimating mixture-based risks, even in instances
where chemicals with disparate MIEs were included. Thus, in the absence of detailed
characterization of molecular mechanisms for most PFAS, it is considered a reasonable health-
protective assumption that PFAS which can be demonstrated to share one or more KEs or
adverse outcomes will produce dose-additive effects from co-exposure. PFOA and PFOS have
historically been the most studied and well-characterized PFAS, but recent work has also
provided supportive evidence of similar effects of other PFAAs including ether-linked structures.
Below is a brief overview of similarities and differences in MIEs, KEs, and adverse outcomes
that have been reported for those PFAS studied to-date and experimental evidence which
supports dose additive effects from combined exposure to multiple PFAS. This overview
highlights study results from, among others, the NIEHS National Toxicology Program (NTP) 28-
day repeat dose guideline toxicity studies of perfluoroalkyl carboxylates (PFHxA, PFOA, PFNA,
30
-------
PUBLIC REVIEW DRAFT
MARCH 2023
and PFDA) (NTP, 2019a) and perfluoroalkyl sulfonates (PFBS, PFHxS, and PFOS) (NTP,
2019b). The NTP studies provide high quality side-by-side comparisons of multiple PFAS from
experiments conducted by a single lab with rigorous exposure characterization and multiple
endpoints spanning MIEs, KEs, and AOPs. More comprehensive reviews of PFAS toxicity
endpoints in experimental animal studies and observational human studies can be found
elsewhere (ATSDR, 2021; EFSA et al., 2018, 2020).
Mechanistically, in vitro and in vivo studies have demonstrated the activation of multiple nuclear
receptors from exposure to a range of PFAS indicating several potential MIEs for PFAS-relevant
AOPs. The most commonly reported MIE associated with many PFAS is activation of
peroxisome proliferator activated receptor alpha (PPARa) based on in vitro binding and
transcriptional activation assays (Behr et al., 2020; Evans et al., 2022; Ishibashi et al., 2019;
Nielsen et al., 2022; Takacs and Abbott, 2007; Vanden Heuvel et al., 2006; Wolf et al., 2012), in
vitro upregulation of PPARa target genes (Bjork et al., 2011), and in vivo tissue-specific
upregulation of PPARa target genes (Bjork et al., 2008; Rosen et al., 2007). All PFAA
carboxylates and sulfonates included in the NTP 28-day studies displayed upregulation of the
PPARa target genes Acoxl and Cyp4al in male and female rat livers. PPARa is a highly
conserved transcription factor that regulates pleiotropic effects on mammalian energy
homeostasis and lipid metabolism, among others. Similarly, multiple PFAS have been shown to
activate peroxisome proliferator activated receptor gamma (PPARy) in vitro (Evans et al., 2022;
Houck et al., 2021; Vanden Heuvel et al., 2006) and upregulate PPARy target genes in vitro
(Marques et al., 2022) and in vivo (Rosen et al. 2017). PPARy is also a highly conserved
transcription factor that regulates multiple physiological processes including adipogenesis and
glucose metabolism. Further, in vivo studies of tissue-specific gene expression patterns have also
demonstrated activation of constitutive androstane receptor (CAR.) for both PFOA and PFOS,
among other PFAS, due to upregulation of CAR-dependent genes (Rosen et al., 2017). Similar to
the PPARa target genes above, all PFAA studied by NTP displayed upregulation of the CAR-
inducible genes Cyp2bl and Cyp2b2 in adult male and female rat livers (NTP, 2019a,b).
Additional in vitro data indicate potential involvement from several other nuclear receptors
including estrogen receptor alpha (Evans et al., 2022; Houck et al., 2021, Kjeldsen and Bonefeld-
Jorgensen, 2013), pregnane X receptor (Bjork et al., 2011; Houck et al. 2021), farnesoid X
receptor (Bjork et al., 2011), and liver X receptor (Bjork et al. 2011, Houck et al. 2021). Multiple
PFAAs, including PFOA and PFOS, activate multiple nuclear receptors and gene transcription
pathways, which is a primary basis for positing shared or overlapping AOPs across PFAAs.
In addition to the cell and/or tissue specific gene expression changes described above, multiple
KEs downstream of the above potential MIEs are also shared between PFOA, PFOS, and other
PFAAs. In both rodent and non-human primate studies, serum lipids (cholesterol, triglycerides)
are consistently reduced and markers of liver injury or dysfunction (ALT, AST, and/or ALP) are
consistently elevated in a dose-responsive manner (ATSDR, 2021; EFSA et al., 2018, 2020).
Specifically, the NTP 28-day studies reported reduced serum cholesterol, triglycerides, and
globulin and elevated serum ALT, AST (males only), ALP, and bile acids from exposure to
PFHxA, PFOA, PFNA, PFDA, PFBS, and PFOS (NTP, 2019a,b). Further, circulating thyroid
hormone concentrations are reduced from oral PFAA exposure with all compounds tested in the
NTP 28-day studies being associated with decreased serum total and free thyroxine (T4) (NTP,
2019a,b). Ether linked PFAAs have also been shown to reduce circulating thyroid hormone
concentrations (Conley et al. 2019, 2022). In combination with the nuclear receptor activity and
31
-------
PUBLIC REVIEW DRAFT
MARCH 2023
gene expression profiles, there is a pronounced similarity in the serum clinical chemistry and
thyroid hormone-based KEs for PFOA, PFOS, and several other studied PFAS. Characterization
of PFAS relevant KEs is currently an area of high research activity as additional pathways
associated with the reported MIEs across various life stages and species continue to be
investigated.
Similar adverse outcomes at the organ and whole animal levels have been described for PFAAs
including PFOA and PFOS. Developmental exposure studies with PFOA, PFOS, PFNA, HFPO-
DA and Nafion byproduct 2 (NBP2) in rats and/or mice have reported consistent effects on pups
including reduced offspring survival/viability and reduced offspring body weight (Abbott et al.,
2007, 2009; Blake et al., 2020; Butenhoff et al., 2004; Conley et al., 2021b, 2022a; Das et al.,
2015; Lau et al., 2003, 2006; Luebker et al., 2005a,b; Thibodeaux et al., 2003). PFAS studied by
NTP (2019a,b) have been observed to increase rat liver weights and produce hepatocyte
hypertrophy. PFOA and PFOS, and potentially other PFAS, have also been shown to produce
functional immunotoxicity (i.e., reduced antibody response) in animal studies (NTP, 2016).
Taken together, there is a broad spectrum of adverse effects in laboratory animals that are
conserved across PFAS such as PFOA, PFOS and other PFAAs, and plausibly associated with
the common molecular mechanisms and KEs. However, it is important to recognize that while
there are same/similar qualitative effect profiles across many PFAS (e.g., liver injury; decreased
thyroid hormones) there are quantitative oral potency differences in reported effects, and not all
effects appear to be shared across all PFAS, or even across all PFAAs, at the dose levels reported
in published studies. For example, PFHxS did not affect serum ALT in male or female rats at any
dose in the NTP 28-day studies, while all other PFAAs tested increased serum ALT levels. The
specific molecular mechanisms or precise modes of action for a given adverse outcome may be
disparate across some PFAS. Studies utilizing transgenic mice with PPARa deletion have
demonstrated that some effects, such as survival of neonates following in utero PFOA exposure,
were dependent on PPARa involvement, while this effect appeared independent of PPARa for
PFOS (Abbott et al. 2007, 2009). It is important to note in those studies that fetal mortality (as
opposed to neonatal mortality) was independent of PPARa genotype for PFOA. The relevance of
rodent PPARa-based effects has been debated in the toxicology literature for decades, largely as
it relates to hepatocarcinogenesis, yet the pharmacological utility of PPARa modulation is
widely accepted and exploited in the development of therapeutics. Further, studies of PPARa
knockout mice have also demonstrated that many liver effects are PPARa independent for PFOA
and PFOS (Abbott et al., 2007, 2009). Although there is potential for disparate MIEs in PFAS
related AOPs and there is a lack of mechanistic characterization for most PF AS-mediated effects,
it is a reasonable health-protective assumption that health effects shared across PFAS in a given
mixture will be dose additive.
Limited work has been conducted on combined exposure to PFAS in experimental systems either
in vitro or in vivo. A few in vitro studies have directly assessed the mixture-based effects of
combined PFAS exposures by comparing observed experimental data with model-based
predictions. For example, Wolf et al. (2014) evaluated in vitro PPARa activation and observed
joint toxicity of combined exposure to binary combinations of PFOA and PFOS, PFNA, PFHxA,
and PFHxS that were consistent with dose additivity in the lower dose ranges, but the authors
reported slightly greater than additive effects at higher mixture doses. In contrast, Carr et al.
(2013) reported slightly less than additive responses for in vitro PPARa activation of binary
mixtures of PFAAs including PFOA, PFNA, PFOS, and PFHxS. Further, Ojo et al. (2020)
32
-------
PUBLIC REVIEW DRAFT
MARCH 2023
reported both synergistic and antagonistic effects compared to a concentration addition model for
binary, ternary, and multi-component mixtures of PFAAs for cytotoxicity in HepG2 cells. Most
recently, Nielsen et al. (2022) demonstrated the utility of generalized concentration addition for
PPARa activation of PFAA mixtures in an in vitro system with variable efficacy across
compounds; however, there are no studies indicating this model variation on simple
concentration addition is applicable for prediction of in vivo mixture effects. Importantly, as
described above, systematic reviews of chemical mixture studies across various compound
classes indicate that departures from dose additivity are uncommon and rarely exceed minor
deviations (~2-fold) from predictions based on additivity (Martin et al. 2021). Similarly, recent
PFAS mixture studies in zebrafish reported interactions for combinations of PFOA and PFOS,
but departures from additive models were also minor (Ding et al., 2013). Menger et al. (2020)
reported zebrafish behavioral effects from a PFAS mixture that were less than individual PFAS,
however evaluation of chemical dose response and comparison to mixture models was not
conducted. Regarding zebrafish PFAS effects, it is notable that fish PPARy has relatively low
sequence homology to that of mammalian PPARy (Zhao et al., 2015) and the potent
PPARy agonist rosiglitazone activates this rat, mouse and man receptor in vitro but not in three
species of fish or the clawed frog (Medvedev et al., 2020). The interactions described in the
literature thus far for combined in vitro exposure to PFAAs demonstrate results that are either
consistent with or have relatively minor deviations from predictions based on concentration
additive models.
Mammalian in vivo toxicity studies evaluating exposure to multiple PFAS are more limited but
recent studies indicate that exposure to a mixture of PFOA, PFOS, and PFHxS in mice (Marques
et al., 2021), a mixture of PFOA, PFOS, PFNA, PFHxS, and HFPO-DA in mice (Roth et al.,
2021), and a mixture of PFBA, PFPeA, PFHxA, PFHpA, PFOA, PFNA, PFDA, PFBS, PFHxS,
and PFOS in rabbits produced numerous significant health effects compared to control animals,
which were consistent with the spectrum of individual PFAS effects described above (e.g., liver
injury; thyroid hormone alterations). However, these studies did not include individual PFAS
dose response data or conduct any mixture model-based analyses, so it is not possible to
ascertain if the mixtures behaved in a DA or RA manner, or if interactions occurred.
Currently, developmental toxicity studies of PFAS mixtures in rats are ongoing at EPA ORD.
One study recently investigated in vivo effects in maternal rats and offspring from combined
exposure to PFOA and PFOS during gestation and early lactation (Conley et al., 2022b). The
study included a series of experiments designed to characterize dose response curves across
multiple endpoints for PFOA and PFOS individually, followed by a mixture study of the two
chemicals combined. The mixture experiment was designed to test for shifts in the PFOA dose
response curves from combined exposure to a fixed dose of PFOS, compare DA and RA model
predictions, and conduct an RPF analysis as clear demonstrations of mixture effects. Exposure to
binary combinations of PFOA and PFOS significantly shifted the PFOA dose response curves
left towards elicitation of effects at lower doses compared to PFOA only exposure. This clearly
indicated mixture effects for a range of endpoints including decreased pup survival, maternal and
pup bodyweight, pup serum T3 and glucose, and increased maternal kidney weight, maternal and
pup liver weight, and pup bile acids, BUN, and bilirubin. Maternal kidneys and maternal and pup
livers in the mixture study also displayed a range of treatment related histopathological lesions.
For nearly all endpoints amenable to mixture model analyses, the DA equation produced
equivalent or better estimates of observed data than RA. Similarly, for nearly all maternal and
33
-------
PUBLIC REVIEW DRAFT
MARCH 2023
neonatal endpoints modeled, the RPF approach produced accurate estimates of dose additive
mixture effects. Only maternal bodyweight at term and gestational weight gain demonstrated
departures from additivity and these effects were less than additive. This work is ongoing with
multiple KE analyses still to be conducted on samples collected during the studies. However,
results thus far support the hypothesis of joint toxicity on shared endpoints from PFOA and
PFOS co-exposure, and dose additivity as a reasonable assumption for predicting mixture effects
of co-occurring PFAS. Further, Conley et al. (2021a) presented preliminary data on an
unpublished mixture study of PFOS, HFPO dimer acid, and NBP2 (an emerging
polyfluoroethersulfonic acid compound recently detected in human serum (Kotlarz et al. 2020)),
which produced neonatal mortality that was accurately predicted by DA modeling, among other
mixture effects. The results discussed above provide robust evidence of combined toxicity of
PFOA, PFOS, and other PFAS on multiple maternal and developmental endpoints and the
greater accuracy of DA for predicting mixture effects in vivo than RA.
In summary, systematic identification and assembly of PFAS data reported in the literature
support an assumption of similarity in toxicity profiles for several health effect domains (for
review see Carlson et al., 2022). Importantly, study results reported in this section across
multiple chemical classes, biological effects, and study designs clearly support a dose-additive
mixture assessment approach. For example, recent efforts to characterize in vivo mixture effects
from combined exposure to multiple PFAS provide key supportive evidence that co-exposure
produces dose additive effects on several endpoints within the range of "same/similar" endpoints
that are shared across the spectrum of PFAS effects. Further, the National Academies of
Sciences, Engineering, and Medicine (NASEM, 2022) recently recommended clinicians apply an
additive approach for evaluating patient levels of PFAS currently measured in NHANES. EPA
will continue to review how mixtures of PFAS and other chemicals interact. Dose additivity is
proposed as the "default" model and other models will be evaluated when data empirically
support or demonstrate significant deviations from dose additivity.
34
-------
PUBLIC REVIEW DRAFT
MARCH 2023
4.0 Introduction to Estimating Noncancer PFAS Mixture Hazard or
Risk
4.1 Whole Mixtures Approach
The preferred hazard and dose-response knowledge base for any mixture of environmental
chemicals would be derived from exposure to a whole mixture of concern. However, the
exponential diversity of chemicals such as PFAS co-occurring in different component
associations and proportions makes whole mixture evaluations difficult and complex. That is, in
the environment, due to differing fate and transport properties of chemicals, biotic (metabolism)
and abiotic (degradation) processes, pH, ultraviolet radiation, media temperature, and so on,
components commonly co-occur in an array of parent species, metabolites, and/or abiotic
degradants making characterization of any given mixture complicated. In controlled
experimental study designs, whole mixtures can be assembled with defined component
membership and proportions. However, the relevance of toxicity associated with exposure to a
defined mixture in a laboratory setting may not be translatable to environmental mixtures of
different component associations and proportions in the field. In the context of PFAS, increasing
environmental evidence (e.g., environmental water, air, and soil sampling results) suggests that
the complexities briefly summarized above with regard to the diversity of chemicals co-
occurring in different component associations and proportions make evaluating each unique
whole mixture of PFAS intractable, which is why component-based mixture approaches are
considered particularly useful and appropriate for addressing human exposure to mixtures of
PFAS (see Sections 5-7).
EPA's Supplemental Guidance for Conducting Health Risk Assessment of Chemical Mixtures
(EPA, 2000b) indicates that there may be opportunities to infer hazard and dose-response for a
mixture of concern from a 'sufficiently similar mixture.' A mixture is considered sufficiently
similar to a mixture of concern when the components and respective proportions exist in
approximately the same pattern. There are clearly gradations of expert "judgment" involved in
what constitutes a sufficiently similar mixture, but determinations should be based on a
comparison of similarities or differences in the components' chemical fate and transport in the
environment, persistence, bioaccumulative potential, kinetics, and toxicity profile. If no
significant qualitative differences are identified in a systematic comparison of mixtures of
chemicals, the hazard and dose-response information associated with the sufficiently similar
chemical could be used as a surrogate for the mixture of concern. However, as with a whole
mixture of concern, information pertaining to a sufficiently similar mixture is rare. The whole
mixture options should be investigated prior to moving to a component-based mixtures approach.
4.2 Data-Driven Component-Based Mixtures Approaches for PFAS
As a result of both the complexities associated with characterization and evaluation of whole
mixtures (see Section 4.1 above) and the reality that most toxicological information derives from
exposure-response studies of individual chemicals, component-based mixtures risk assessment is
particularly relevant (Figure 4-1). In addition, although the methodological approaches and
associated illustrative examples in this framework are targeted at application to water, the
concepts may facilitate evaluation of PFAS mixtures in other exposure media as well (e.g., soil,
air). As outlined in earlier sections of this framework, while EPA component-based methods and
approaches are available for evaluation of mixtures of chemicals under different assumptions of
35
-------
PUBLIC REVIEW DRAFT
MARCH 2023
additivity (EPA, 2000b), the currently available evidence on PFAS, and several other classes of
environmental chemicals, supports an assumption of dose additivity (see Section 3). The HI and
RPF are two component-based mixture approaches based on dose additivity, that are well
validated, supported by peer-reviewed guidance, and actively used by EPA. These two
approaches are discussed below and include illustrative examples that are based on a
hypothetical five component mixture of PFAS (see Sections 5 and 6). An alternative "M-BMD"
approach, generally based on the Berenbaum equation (see Section 4.2.6 in EPA mixtures
guidance (2000b)), is also a dose additive approach that is described and illustrated (see Section
7). The primary difference in the RPF and M-BMD approaches is that RPF assumes component
chemical dose response slopes are congruent, while the M-BMD approach is more applicable for
mixture component chemicals with dissimilar slopes. It should be noted that others have recently
demonstrated the application of the HI and RPF approaches in the evaluation of PFAS (Bil et al.,
2021, 2022; Mumtaz et al., 2021), lending confidence to the direction of this framework
document in guiding formal component-based assessment of PFAS mixtures. The M-BMD
approach was described and supported in both EPA's mixtures guidance (2000) and by the
National Research Council (NRC) (NRC, 2008), and laboratory studies have provided empirical
evidence (Gray et al., 2022 and example in Section 7).
4.2.1 Conceptual Framework of the Approach
A pragmatic data-driven approach to the application of component-based evaluation of mixtures
of PFAS with variable hazard and dose-response databases is presented in Figure 4-1.
36
-------
PUBLIC REVIEW DRAFT MARCH 2023
Problem formulation and scoping
5^-pp ^ Assemble PFAS mixture component chemical toxicity information
Stop 2
Step 3
Step 4
Do components have
existent oral toxicity
values (e.g., RfD)?
Yoc
CD
Are component
toxicity values based
on same effect?
No"
Hazard Index
(seeSection 5.2)
No
Yes
Do components have
available in vivo
hazard and dose-
response data?
Does component
data support de novo
derivation of toxicity
values?
Target-Organ Specific
Hazard Index
(see Section 5.3}
No
Do components have
available NAM data
to inform dose-
response evaluation?
[\jo f"'a§ mixture
component(s) for
data needs
Yes (a) Yes (b)
Are component dose-
f\JO response functions
(e.g., shape, slope) for
similar effect
congruent ?
Yes
Relative Potency Factors
and ICECs (see Section 6)
No
Are dose-response
No data amenable to
benchmark dose
modeling (BMD)?
V
Yes
Mixture BMD
i (see Section 7)
Figure 4-1. Framework for data-driven application of component-based assessment approaches for mixtures of PFAS based
on an assumption of dose additivity.
37
-------
PUBLIC REVIEW DRAFT
MARCH 2023
The general steps of the component-based approach as shown in Figure 4-1 are as follows:
Problem formulation and scoping. Problem formulation is the part of the risk assessment
framework that articulates the purpose of the assessment and defines the problem (e.g., PFAS
source and occurrence, fate and transport, populations/subpopulations potentially at risk, health
endpoints). Problem formulation also typically includes development of a conceptual model and
analysis plan and includes engagement with potentially affected stakeholders (e.g., co-regulators
like states and tribes, risk managers, affected community) to discuss foreseeable science and
implementation issues.
Step 1: Assemble information.
Step 1 of the data-driven mixture assessment approach is to identify available hazard and dose-
response information for: (1) a whole mixture of the PFAS of potential concern at component
proportions consistent with the environmental sampling data; (2) a sufficiently similar mixture;
and/or (3) data for the individual component PFAS. If whole toxicity data for the mixture itself
or a sufficiently similar mixture are not available or are insufficient, then a structured search,
collection, and assembly of all available toxicity data for mixture component PFAS of potential
concern is conducted. Although the optimal approach would be to utilize formal systematic
literature search and review principles as set forth by EPA (please see the systematic review
protocol for PFB A, PFHxA, PFHxS, PFNA, and PFDA as an example9), the user of this
framework may employ a structured literature search approach of their choosing so long as the
underpinning decisions resulting in the literature inventory and data landscape used in steps 2-4
of the framework approach are transparent. It should be noted that while this step is primarily
intended for identification and harvesting of traditional human epidemiological and/or
experimental animal toxicity data, it is ideal to also assemble information such as toxicokinetic
(TK)-relevant parameters (e.g., clearance, plasma/serum half-life, volume of distribution),
mechanistic pathway data, empirical (or predicted) physicochemical properties, and if available,
validated NAM data such as cell bioactivity, high(er)-throughput transcriptomics, and/or
structure-activity/read-across.
Step 2: Evaluate data objectives.
Once harvested, curated, and arrayed, the user should be able to evaluate the following primary
data objectives: (a) Existence of human health risk assessment values (e.g., EPA RfD; ATSDR
MRL); duration-specific values (e.g., subchronic (note: ATSDR refers to this duration as
"intermediate"), or chronic RfVs) may be available from various sources and should be
assembled and incorporated into the data-driven mixture assessment approach(es) as deemed
appropriate by the user; and (b) Development of health effect domain profiles (see example
literature inventory in Figure 4-2 excerpted from a PFAS systematic review protocol10), and
associated dose-response information sorted based on exposure duration (e.g., short[er]-term
(i.e., acute, short-term, dev/repro), longer-term (i.e., subchronic, chronic)), across mixture PFAS
supported by the assembled traditional (i.e., human epidemiological and/or experimental animal)
toxicity data from Step 1.
9 https://cfpnb.epa.gov/ncea/iris drafts/recordisplav.cfm?deid=345065
10 https://cfpnb.epa.gov/ncea/iris drafts/reeordisptav.cfm?deid=345065
38
-------
PUBLIC REVIEW DRAFT
MARCH 2023
n in u
¦! \ i Hi
3FH>
inci s-d I ts
1u.,
Figure 4-2. Example literature inventory heatmap for traditional epidemiological or experimental animal studies for five
PFAS currently under development/review in EPA/ORD's IRIS program (heat map circa 2018). Health effects are based on
groupings from the IRIS website (https://cfpub.epa.gov/ncea/iris/search/index.cfm).
39
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Users of this framework may find that many PFAS of interest are data-poor (i.e., no traditional
human health assessment relevant epidemiological or experimental animal study data are
available). In such cases, NAM platforms or assays might provide opportunities to inform
potential hazard and dose-response for PFAS mixture components. For example, read-across is a
NAM approach that could potentially be leveraged to identify surrogate dose-response metrics
(e.g., POD, ECx, ICx) for integration into the component-based mixtures assessment approaches
presented in the subsections below. Analog-based read-across, in general, is a process in which
chemicals (i.e., analogs) with relatively replete toxicity databases are compared to a data-poor
target chemical across similarity domains including structural, physicochemical, TK, and/or
toxicodynamic (TD) similarity (Wang et al., 2012; Wu et al., 2010). Based on weight-of-
evidence for similarity between a data-poor target chemical and candidate analogs, hazard and
dose-response data (e.g., POD) are then adopted from a selected (single-best) analog as surrogate
for the target chemical. This read-across approach might facilitate incorporation of data-poor
PFAS into the component-based methods presented in this framework, as surrogate PFAS data
that inform similarity of toxic endpoint/health effect and dose-response could potentially: (a) be
used in the derivation of a noncancer RfD (using uncertainty factors appropriate for the data-poor
target chemical) and subsequent calculation of a hazard quotient (HQ); (b) be used in the
calculation of RPF(s); or (c) the surrogate health-effect dose-response data could be BMD
modeled and included in the calculation of an overall M-BMD. EPA's ORD has been employing
expert-driven analog-based read-across in the evaluation of data-poor chemicals for over a
decade; for an illustrative example human health assessment application please see Appendix A
of the Provisional Peer-Reviewed Toxicity Value (PPRTV) document for 2,3-toluenediamine at
https://cfpub.epa.gOv/ncea/pprt.v/recordisplav.cfm?deid=352932).
Another opportunity for integration of NAMs into the proposed mixture approaches involves cell
bioactivity or transcriptomic data from experimental animals and/or in vitro cell cultures. For
over a decade, EPA and NIEHS have invested significant resources into high(er)-throughput
assay development and application to hundreds of chemicals (Richard et al., 2016; Kavlock et
al., 2012; Dix et al., 2007). The assays are primarily targeted at nuclear receptor activity but also
include several other assays that inform cell viability, enzyme activity, DNA reactivity, cell
transport, and macromolecular/cellular dysfunction. The bioactivity information is quality
assured, assembled, curated, and presented in a manner that is intended to facilitate incorporation
into risk-based decision contexts such as human health assessment (see example bioactivity plot
in Figure 4-3). Although there are inherent complexities and challenges associated with study
designs and associated data interpretation using NAM assays/platforms, such as in vitro cell
culture, recent investigation has demonstrated that quantitative data (e.g., PODs) from in vitro
bioactivity is a reasonable estimate of in v/'vo-based PODs (Paul-Friedman et al., 2020).
Likewise, over the past decade, systematic comparisons of pathway-based (e.g., Gene Ontology
or 'GO'u) transcriptomic PODs to phenotypic health outcome PODs has illustrated that for most
chemicals evaluated to date, the dose-response relationship between genotype and phenotype for
toxic effects is typically within an order of magnitude (Johnson et al., 2020; Thomas et al., 2013;
2011). For many chemicals, bioactivity assays can also provide information on the potential to
disrupt specific MIEs and KEs of known or postulated MO As or AOPs and may inform the
relevance of specific pathways to humans. While in vitro assays are critical they are not without
limitations. For example, in vitro dose metrics in a component-based mixture context (e.g., RPFs
11 http://geneontology.org/docs/ontology-documentation/
40
-------
PUBLIC REVIEW DRAFT
MARCH 2023
or BMDs) cannot be directly extrapolated to in vivo RPFs or BMDs and used for the prediction
of PFAS mixture effects in vivo due to the lack of chemical-specific ADME processes. Rather,
converting in vitro bioactivity concentrations to estimated human in vivo doses (i.e.,
Administered Equivalent Doses; AEDs) requires application of in vitro-to-in vivo extrapolation
(IVIVE) and reverse toxicokinetics (rTK) which may not be possible for many data-poor PFAS.
In addition, several in vitro cell-based assays to date employ truncated nuclear receptors with
only the ligand binding domain and, as a consequence, the transcriptional events that follow
binding may not be fully representative of quantitative chemical potencies compared to that seen
with full length receptors in native in vivo systems. Further, for such approaches to gain
widespread regulatory acceptance it will be important to demonstrate that the NAMs under
consideration are reproducible, robust, and can be transferred to other laboratories and produce
results that are relevant to in vivo adverse effects.
41
-------
PUBLIC REVIEW DRAFT
MARCH 2023
:i
i
i
-•T-
rn
ij.:
-
i
1 i
¦ n
L . L;
- iX>
-"Ti
-CD
—.fT
H_l
-rp-
:u
--LU-
_TU
J'
it:
Administered Equivalent Dose (ioglO-mg/kg/day)
Additional Values
I AED Sth %te
, SEEMS US5
Figure 4-3. Example plot illustrating in vitro cell bioactivity expressed in AEDs which are
an estimated oral exposure dose that results in an internal steady-state concentration
consistent with the in vitro concentration associated with a biological perturbation or
activity. The example shows the distribution of AEDs; the vertical orange dashed line
indicates the 5th percentile on the bioactivity landscape. The green dashed line corresponds
to an upper bound on the estimated general population-based median exposure (generated
from EPA's Systematic Empirical Evaluation of Models (SEEM3);
https://www.cpa.gov/cheniical-iesearch/conmiitational-toxicoloi;v-coimminities-nractice-
svstematic-einpirical-evaliiation).
42
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Step 3: Consider data landscape to select component-based approach(es).
Consideration of the data landscape across mixture component PFAS in Step 2 will help the user
ascertain which option(s) in Step 4 is the most supportable. For example, dose-response data for
PFAS mixture components may be applicable to more than one option in step 4, however
characteristics of the data (e.g., shape or slope of dose-response between components) may not
be congruent potentially leading the user to select a M-BMD approach over the RPF approach.
Another example entails dose-response data for components that do not indicate "like" health
outcome or MO A; in this case it may be more practicable to use the available data in a general
HI approach (assuming human health assessment values already exist, or dose-response data
support de novo derivation of assessment values). When using any of the approaches, it is
optimal to calculate and use HEDs rather than oral administered doses in test animals where and
when possible. The user is not precluded from applying available hazard and dose-response data
across the entirety of the options in Step 4, however study data types, design (e.g., exposure
duration, route), and confidence will likely dictate optimal selection of component-based
approach for PFAS on a case-by-case basis. In most cases, it is likely that a user would select
only one of the approaches, based on the data available for the PFAS components in a mixture of
concern (e.g., if toxicity values and HBWCs are available or can be calculated, then the HI
approach is appropriate). If instead toxicity values are not available or cannot be readily derived,
then the RPF or M-BMD approach may be more suitable.
Step 4: Perform component-based approach(es).
a. HI and TOSHI. Although limited in number, for those mixture PFAS that have human
health risk assessment value(s), the user in Step 3 should have determined if the critical
effect(s) on which duration-specific assessment values were derived, for any two or more
PFAS, fit into the same health effect domain (e.g., liver, thyroid, developmental). If not,
then those PFAS are entered into a general HI approach (Section 5). For many PFAS, no
formal health assessment exists, however human epidemiological and/or experimental
animal hazard and dose-response data may be available in the public domain. Should
such data be available, the user has the option of performing de novo derivation of
duration-specific noncancer toxicity values using assessment guidance and practices
accepted under their specific purview. Again, if the de novo derived values are not based
on same/similar effect then the general HI approach is recommended. In brief, the HI
approach entails use of duration-relevant exposure (E) and toxicity values (e.g., RfV), for
each component PFAS, in a simple ratio (E/RfV) to calculate an HQ. The general HI
involves the use of RfVs for each PFAS mixture component irrespective of health
outcome domain. Because each mixture component HQ is calculated using a
corresponding RfV, the mixture HI may represent a conservative indicator of potential
mixture risk. The component PFAS HQs are then summed to generate a mixture HI (see
Equation 5-1). A mixture HI approaching or exceeding 1.0 indicates potential concern for
health risk(s) associated with a given environmental media or site. The HI provides an
indication of: (1) risk associated with the overall PFAS mixture; and (2) potential driver
PFAS (i.e., those PFAS with high(er) HQs). Conversely, those PFAS with low(er) HQs
(e.g., < 0.0X) might be deprioritized as they may not have significant impact(s) on overall
mixture risk at the specific media concentrations identified. It should be noted that a user
of this approach should consider the potential exposure (e.g., water concentration), the
potency for toxic effect (e.g., low(er) or high(er) RfVs, PODs), the duration associated
43
-------
PUBLIC REVIEW DRAFT
MARCH 2023
with exposure and toxicity, and qualitative and quantitative uncertainty (i.e., totality of
UF application) for each PFAS mixture component.
If hazard data across PFAS mixture components indicate same/similar effect, then a
target-organ specific HI (TOSHI) is recommended (see Section 5.3). The TOSHI
approach is exactly as the name suggests, that is, it entails calculating component
chemical HQs and corresponding mixture His for specific target-organ effects/endpoints
using target-organ toxicity doses (TTDs) (note: some TTDs could also be the overall RfD
for a given PFAS). In practice, it is recommended to calculate TTDs for all health
outcomes associated with a given mixture PFAS, where/when evidence support. This
may facilitate calculation of TOSHIs for more mixture component PFAS across more
health outcomes thus enriching the evaluation of PFAS co-occurring in a given
environmental medium (e.g., drinking water).
If data are available that indicate and support deviations from dose additivity (e.g.,
synergy or antagonism), an interactions-based HI may be employed (see EPA, 2000b).
However, in this framework, based on an assumption of dose additivity only the HI and
TOSHI are included. Specifically, data to inform deviations from dose additivity (e.g.,
interactions such as synergism or antagonism) are virtually non-existent for PFAS co-
occurring in mixture, as such, an interactions-based HI is not feasible at present.
b. RPF. In contrast to the HI, the RPF approach provides a PFAS mixture risk estimate (see
Section 6). In the RPF approach, potency for an effect across each mixture PFAS is
scaled to a selected IC for critical health effect domains of concern. In the illustrative
example in Section 6.2, application of the RPF method is demonstrated for liver, thyroid,
and developmental effects associated with the hypothetical five component PFAS
mixture. These three health effect domains were selected primarily because: (1) the
effects are common across several PFAS assessed by EPA (and ATSDR) thus far; and (2)
each hypothetical example PFAS has differing levels of hazard and dose-response data
available across the three health effect domains, to best illustrate demonstration of the
RPF methodology. In practice, for application in a water context, each respective PFAS
RPF is multiplied by its corresponding specific media concentration (e.g., water
concentration), resulting in an Index Chemical Equivalent Concentration (ICEC). The
ICECs across PFAS mixture components are summed to generate an overall mixture
ICEC (see Equation 6-2), which is effectively a total concentration of the IC, for each
health effect domain. In traditional EPA mixtures risk assessment practice, the mixture
ICEC is then mapped to the dose-response function of the IC to arrive at a "mixture
response." In this framework, in the context of water, the mixture ICEC (i.e., total dose of
IC) is compared directly to an HBWC (e.g., Health Advisory, Maximum Contaminant
Level Goal (MCLG)) based on the relevant health effect domain (e.g., liver, thyroid,
developmental) for the IC. If the mixture ICEC for one or more of the selected effect
domains exceeds the corresponding IC HBWC then there may be cause for concern for
the mixture at the reported/measured component water concentrations. Conversely, if the
mixture ICECs for all effect domains are below the corresponding IC HBWC, health risk
is not anticipated. Additionally, individual mixture PFAS with large(r) RPFs and
corresponding ICECs should be flagged regardless of whether the total mixture ICEC is
above or below an IC HBWC.
44
-------
PUBLIC REVIEW DRAFT
MARCH 2023
c. M-BMD. An additional option entails calculation of a M-BMD (see Section 7) and is
applicable even when PFAS in the mixture have dissimilar dose response curves. In
contrast to the RPF approach, there is no need for identification of mixture ICs,
calculation of RPFs or ICECs, or existence of HBWCs as the final determination of risk
is based on comparison of the observed total mixture concentration with each effect-
based M-BMD. Similar to the RPF approach, hazard dose response data across one or
more health effect domains for each PFAS in the mixture are needed to determine the
corresponding benchmark response (BMR) for each PFAS component (i.e., each
component PFAS benchmark dose). Then the individual component chemical BMDs are
scaled based on their proportion in the mixture and added using a simple dose-addition
based equation to arrive at a total M-BMD. The M-BMD approach does not require that
component chemicals meet an assumption of statistically similar dose response functions
(i.e., slopes). The resulting M-BMD (i.e., POD) could be converted into a mixture RfD
using appropriate UF application, and subsequently incorporated into the calculation of a
corresponding mixture-specific HBWC (e.g., Health Advisory or MCLG). However, it is
cautioned that such values would be specific to a given mixture of PFAS at defined
component proportions (e.g., individual PFAS water concentrations). The illustrative
example in Section 7 utilizes data for a five-component hypothetical PFAS mixture with
hazard and dose response data for the three selected effect domains. Similar to the RPF
approach, if the total mixture concentration exceeds the M-BMD for one or more effect
domains, then there may be cause for concern for the mixture at the reported/measured
component water concentrations. If the total mixture concentration is below all M-BMDs
calculated, then health risk is not anticipated.
4.2.2 Introduction to a Hypothetical Example with Five PFAS
In the following sections, the HI (Section 5), RPF (Section 6), and M-BMD (Section 7)
approaches will be detailed and accompanied by demonstration of practical application to a
hypothetical five component mixture of PFAS. As a reminder, PFAS 1-5 are as follows:
PFAS 1 = comprehensively studied, most potent for effect(s), and has formal noncancer
human health assessment value(s) and an HBWC available;
PFAS 2 = well-studied, second most potent for effect(s) among PFAS 1-3, and has
formal noncancer human health assessment value(s) and HBWC available;
PFAS 3 = studied, least potent for effect(s) among PFAS 1-3, and has formal noncancer
human health assessment value(s) and HBWC available;
PFAS 4 = experimental animal toxicity data available but no formal human health
assessment and no HBWC; and
PFAS 5 = data-poor.
To help introduce the illustrative case study, the hypothetical drinking water scenario is as
follows: Periodic targeted and non-targeted analysis of drinking water samples obtained at the
tap across a community revealed the presence of five PFAS, referred to as PFAS 1-5 (median
45
-------
PUBLIC REVIEW DRAFT
MARCH 2023
concentrations shown in Table 4-1), above hypothetical analytical quantitation limits12
(Table 4-2).
Table 4-1. Drinking water concentrations for five hypothetical PFAS; the values represent
the median of a distribution of sampling data collected across a community over time.
PFAS Exposure Estimates (Measured in Drinking Water) (ng/L)
PFAS 1 PFAS 2 PFAS 3 PFAS 4 PFAS 5
Median 4.8 52 172 58 69
Table 4-2. Analytical quantitation limits for drinking water for five hypothetical PFAS.
PFAS Analytical Quantitation Limits (ng/L)
PFAS 1 PFAS 2 PFAS 3 PFAS 4 PFAS 5
Analytical limit 3.0 5.0 5.0 4.0 5.0
Problem formulation and scoping. For simplicity, the problem formulation is scoped to 'What
are the potential (noncancer) public health risks associated with exposure to the mixture of PFAS
1-5 in drinking water for the community?' A formal problem formulation and scoping exercise
might include identification of population/community exposure details (e.g., distribution of
sexes, ages, exposure frequency, exposure duration), seasonal variations in PFAS 1-5 levels in
the drinking water, groundwater/surface water PFAS 1-5 concentrations, density of wells in the
community, and other mitigating circumstances or factors.
Step 1: Assemble information.
The structured literature search for the mixture of PFAS 1-5 included comprehensive Boolean
search strings and were applied across information databases such as PubMed, Web of Science,
Toxline, and TSCATS (Figure 4-4). Please note that the specific PFAS names, synonyms, and
CASRN in Figure 4-4 are for illustrative purposes only. In application, the search string(s) would
need to be scoped and developed to optimize the literature search for specific mixture PFAS on a
case-by-case basis.
12 Analytical quantitation limits for chemicals are generally defined as the lowest detectable concentration of an
analyte where the accuracy achieves the objectives of the intended purpose. For example, in EPA's UCMR Program,
the agency establishes "minimum reporting levels" to ensure consistency in the quality of the information reported
to the agency. Under UCMR 5, an analytical quantitation limit is the minimum quantitation level that, with 95%
confidence, can be achieved by capable analysts at 75% or more of the laboratories using a specified analytical
method.
46
-------
PUBLIC REVIEW DRAFT MARCH 2023
• PubMed [Main nal Library of Medicine)
• Web of Science (Thomson Reuters)
• Toxline f National Library of Medicine!
• TSCATS (Toxic S tifaslatices Control Act Test Submissions!
Search
Search strategy
Dates of search
PubMed
Search
terms
375-22-4{mJ OR "Heptafluoro-l-butanoieacidn[twl OR "Heptafluorobutartoic
acid"[tw} OR "Heptafluorobutyric acid"[tw] OR "Kyseiina
heptafluormaselna*[tw] Oft "Perfluorobutanoic addH{tw] OR
"Perfltiorobutyrfc aci# ftwj OR "Perfluoropropanecarboxylk: acid*ltwj OR
"2,2,3,1,4,4»4-heflaf1uoro-iulaii0ic aetd"ftw] OR "Butanoic acid,
2»2»B3,4J%4-heptafluero-"IlwJ OR "Butanoie acid, heptafluoro-"ItwJ OR
"Perfluoro-n-butanoic aeitf'ltw} OR "Perfluorobutanoate"[twJ OR
*2,2,3,3,4,4,4-Heptafluorobutanotc adef ftwj OR "Butyric acid,
heptafluoro-tt{tw] Oft "Fluorad FC 23"[twJ OR "H 0024"(tw] OR "NSC 82rftwJ
OR (fpf BAItwl OR-FC 23"itwl OR HfBAftwJ) AND (fluorocarbon»{twl OR
fluorotelomer'llw] OR po«yfluoro*ttwl OR perffuoro-'ftwj OR
perfluoroa'Itw] OR perfluorob*{tw] OR peffluoroc','flwl OR perfluorod*[twl
OR perfluoroe*{tw] OR perf!uoroh»|twI OR perfluoron»ltw} OR
perfluoroo*ttwl OR pertluorop*itwl OR perfluoros*[twI OR perfluorou*Etw]
OR pcrftuorinatedltw] OR fluorIr»3t«J|twl OR Pf ASftwf OR PFOSftw} OR
PFOAftwIW
No date
limit
Figure 4-4. Example PFAS-specific literature search string applied to toxicity information
databases such as the four listed (e.g., PubMed, WoS, Toxline, and TSCATS).
The assembled literature inventory was then screened at the level of title and abstract to
determine preliminary relevance to informing human health risk assessment using defined
Population, Exposure, Comparator, and Outcome (PECO) elements as illustrated in Figure 4-5.
Again, specific details provided in Figure 4-5 are for illustrative purposes only; mention of PFAS
other than the hypothetical PFAS 1-5 should not be construed to be the basis of the illustrative
PFAS mixture example in subsequent sections.
47
-------
PUBLIC REVIEW DRAFT
MARCH 2023
PECO
Evidence
Populations
Human: Any population and lifestage (occupational or general population, including children and
other sensitive populations). The following study designs will he included; controlled exposure,
cohort, case-control, and cross-sectional, (Note: Case reports and case series will be tracked as
potential supplemental material!
Animal: Nonhuman mammalian animal species {whole organism,} of any lifestage {including
Other: in vitro, in silica, or nonmammalian models of genotoxiclty. (Note; Other in vitro, in silico,
or nonmammalian models will be tracked as potential supplemental material,}
Exposures
Human: Studies providing quantitative estimates of PFAS exposure based on administered dose
or concentration, biomonitoring data {e.g., urine, Wood, or other specimens,), environmental or
occupational-setting measures {e.g., water levels or air concentrations, residential location
and/or duration, fob title, or work title), (Note; Studies that provide qualitative, but not
quantitative, estimates of exposure will be tracked as supplemental mate-rial,J
Animal; Oral or inhalation studies including quantified exposure to a WAS of interest based on
administered dose, dietary level, or concentration, {Note; Nonoral and noninhalation studies
will be tracked as potential supplemental material.) PFAS mixture studies are included if they
employ an experimental arm that involves exposure to a single PFAS of interest {Note: Other
PFAS mixture studies are tracked as potential supplemental material.)
Studies must address exposure to one or more of the following: PFDA (CASRN 335-76-2), PFDA
ammonia salt (CASRN 3108-42-7}, PFDA sodium salt (CASRN 3SM-45-3J, PFNA (CASRN 375-95-1),
PFN.A ammonium sail (CASRN 4149-60-4), PFNA sodium sail (CASRN 21043-33-B), PFHxA
(CASRN 307-24-4), PFHxA sodium salt (CASRN 2923-26-4), PfHxA ammonium salt (CASRN 21615-
47-4), PFHxS {CASRN 3S5-46-4), PffixS potassium salt (CASRN 3871-99-6), PFBA
{CASRN 375-22-4J, or PFBA ammonium salt (CASRN 10495-86-0). I Note: although while these
PFAS are not metabolized or transformed in the body, there are precursor compounds known to
be biotransfornied to a PFAS of interest; for example, 6:2. Ruoratebmer alcohol is metabolized
to PFHxA and PFBA {Russell et al . 2015). Thus, studies of precursor WAS that identify and
quantify a PFAS of Interest will be tracked as potential supplemental material {e.g., for ADME
analyses or interpretations)J
Comparators
Human: A comparison or reference population exposed to lower levels (or no
exposure/exposure below detection levels} or for shorter periods of lime.
Animal: Includes connparlsofis to histories 1 controls or 3 concurrent control group that is
unexposed, exposed to vehide-only or air-only exposures, (Mote: Experiments including
sxpffisufs to PFAS scross different durations or exposure levels without incluclifig oris of these
control groups will be tracked as potential supplemental material fe.g.t for evaluating key
science Issues; Section 2.4).)
Outcomes
All cancer and noncancer health outcomes. (Note; Other than genotoxitity studies, studies
including only molecular endpoints [e.g., gene or protein changes; receptor binding or
activation) or other nonphenotypie endpoints addressing the potential biological or chemical
Figure 4-5. Example PECO criteria and considerations used to determine study relevance
in the systematic review and evaluation of a literature inventory for chemicals such as
PFAS.
48
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Following removal of duplicate references and systematic screening of the initial inventory using
the defined PECO, non-relevant studies/reports were excluded, and the remaining references
were full text screened. The full-text screening resulted in three buckets of references: (1) studies
or reports meeting PECO; (2) studies or reports tagged as supplemental (i.e., useful for risk
assessment but not a toxicity study); and (3) studies or reports that upon further review were
excluded as not PECO-relevant (bottom of Figure 4-6).
fes'ijiii
Literature Searches {through 2018) [ Other
I AT5DR assessment (n
^.,Miiirt*i! t ^ Fr'A Jfi
WOS | f ToxLine J I TSCATS | I Non-English or non-f
Title & Abstract Sc
2 records after dupti
(710
Fill I TEXT I Titl# & Abttract Sen
| {SIS r«cor)
AtkMtums* Stralegips
Fuil-Taxt
(n =
TITLE AND ABSTRACT
FULL TEXT SCREE
|S0? records sfft#r t
Exdudud (rt- )
FULL TEXT SCREENING
Figure 4-6. Example literature screening logic flow for hypothetical PFAS using an EPA
systematic review approach. The figure depicts example PECO-dependent development of
evidence bases to support human health assessment application(s). Note: only 4 of 5
hypothetical PFAS (i.e., PFAS 1-4) are represented as PFAS 5 is data-poor.
49
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Step 2: Evaluate data objectives.
For the hypothetical example, the systematic literature search and screen resulted in PECO
relevant studies for PFAS 1-4 only. PFAS 5 was identified as data-poor and further interrogated
in EPA's Computational Toxicology Dashboard (https://comptox.epa. gov/dashboard/) for
presence of alternative toxicity testing data (e.g., in vitro cell bioactivity assay data). At the
conclusion of the data gathering exercise for the five PFAS in Step 1 of the framework approach,
and once all of the hazard and dose-response studies and data were assembled and evaluated, it
was determined under Step 2 that: (1) there are no whole mixture/sufficiently similar mixture
studies available for the combination of PFAS of concern; (2) PFAS 1-3 have existing human
health risk assessment values for the oral route of exposure (Figure 4-7A); (3) PFAS 4 has
existing hazard and dose-response data but no known health assessment value(s); and (4) PFAS
5, although data-poor, has predicted physicochemical property and empirical in vitro cell
bioactivity data. In addition, across the five hypothetical PFAS, different levels/types of
toxicokinetic data were identified. Specifically, clearance13 values for experimental rats and
humans were located for example PFAS 1-3. No clearance or volume of distribution (Vd)14
values were identified for PFAS 4; only serum half-life data were harvested where available.
Lastly, only rat serum half-life data were located for PFAS 5 (Figure 4-7B).
13 Clearance represents the combined intrinsic ability of organs and tissues to remove chemical(s) from the plasma
and is commonly expressed in units of volume/time-body mass (e.g., L/day-kg body weight); it is typically
calculated as the product of the elimination rate constant (ke) x volume of distribution (Vd). An elimination rate
constant represents the faction of chemical eliminated from the body per unit of time, commonly expressed in units
of hour(s) or day(s).
14 The volume of distribution represents the degree to which a chemical is distributed in body tissues. For example,
chemicals that are highly bound to plasma proteins and not broadly distributed in tissues have a low Vd; conversely,
chemicals that have low affinity for plasma proteins typically have a high Vd and distribute broadly across
tissues/compartments. Vd is commonly expressed in units of volume/body mass (e.g., L/kg).
50
-------
PUBLIC REVIEW DRAFT
MARCH 2023
(A)
Oral Reference Values
PFAS 1
cute I Short Term
Subchronic
Chronic
1-
+IT ' Q R®
I
PFAS 2
Oral Reference Vaiues
cut® < Short Term
Subchrontc
PFAS 3
Oral Reference Values - 1
Short Term ] Subchronic
Chronic
*MOH M)
I ©MMDESf?®
I BlTCEOmffi'
10 100 1,000
Duration (Days)
10,000 100,000
(B)
PFAS 1
PFAS 2
PFAS 3
PFAS 4
PFAS 5
IBBl
Clearance (L/day-kg)
Plasma half-life (hours)"
1
Female
Male
Female Male
Female
Male
Female Male
Female Male
Rat
0.0031
0.0023
0.42 0.16
0.0024
0.00064
33.6 734.4
1406 958
Mouse
0.00057
0.0003S
0.17 0.13
0.00017
0.00015
1128 1236
ND
Monkey
0.000034
ND
0.00005
0.00003
ND
ND
Human
0.9 F.-5
0.00002"
24,52.8
ND
Figure 4-7. (A). Example exposure-response arrays for the PFAS 1-3 identified as having
existing human health risk assessment values for one or more exposure durations; (B).
Empirical clearance or plasma half-life data for PFAS 1-5. ND = no data available.
"Clearance = elimination rate constant (ke) x volume of distribution (Vd); "Data needed to
calculate clearance or Vd were not available for rodents or humans as such only plasma
half-life is presented where available; fMale only
51
-------
PUBLIC REVIEW DRAFT
MARCH 2023
For brevity, in the evaluation of the hypothetical five component PFAS mixture, the health effect
domains are truncated to three targets: Liver, Thyroid, and Developmental. Across these three
target health effect domains, the PECO relevant studies and data were subjected to systematic
review principles and practice (e.g., risk of bias analysis; evidence integration) across the
available data sources/streams (Figure 4-8).
Step 1
Strength of the Evidence
Judgment of the evidence for
an effect in human studies
Judgment of the evidence for
an effect in animal studies
Step 2
Inference Aemsi Lines of Evident*
• information on the tiuro#r» relevance of
the animal and mechanistic evidence
• Coherence across lines of evidence or
with related health effects, information
on susceptible populations, other
read-across)
cVIQence InWgrailOfl
Condusfon
Overall conclusion
across lines of
for a human
health effect
dm_ n^, 3tu. ie,r.h * ¦iirasl Mas ¦>
& itlls I dll'-ilT L I s c.ni ^:ii" ¦ ' o-jv .o Tir*- kg
fey U^itu^ nc dStJ " e el sU*.peiT
I-ucmi |rui«ui i|j, *
RMe of Ofgrif of wmitivitv. $*ww% $wse*ptiWlity
evidence, little
Moderns 5 or £u"nt
Stipht
inconsistent ci little-
Srtderenriui,i.c^ tuiMCus- |: Evidence,
Cornpi'llm^
i'vicR'nc^ol
rtn
^ncsjid-llly
fvUteJTOtfkelY
Evidcnct» suggest*
htidertce
Strong pv'dfnct' o*
no effect
Effect domain
Liver
Thyroid
Developmental
PFAS 1
¦Bw||
¦¦Mi
PFAS 2
^^M
++
PFAS 3
+++
PFAS 4
¦Mil
PFAS 5
*
Figure 4-8. Evidence integration across three target health effect domains for a mixture of
five hypothetical PFAS. The heat map indicates strength of evidence supporting an effect of
the PFAS in a domain. (+++) indicates likely effect; (++) evidence suggests an effect in the
domain; (—) evidence is inadequate to determine an effect in the domain; * Although PFAS
5 has no traditional human or experimental animal assay data available, in vitro cell
bioactivity data are available from assays performed predominately in hepatocyte cell lines.
Step 3: Consider data landscape to select component-based approach(es).
As illustrated in the evidence heat map in Figure 4-8, PFAS 1, 2 and 4 have traditional
experimental animal assay hazard and dose-response data available for component-based mixture
assessment of liver effect(s); for thyroid effect(s), PFAS 1-3; and for developmental effect(s),
PFAS 1-4. The landscape of duration relevant dose-response data (e.g., chronic) is then
interrogated for identification of single best study/dataset for a critical effect to represent the
PFAS in a given effect domain. "Single best" study/dataset will be subjective and user-
52
-------
PUBLIC REVIEW DRAFT
MARCH 2023
dependent, however considerations such as robustness of the study(ies) (e.g., power of
study/high N per treatment group, comprehensiveness and transparency of toxicity evaluation;
statistics), effect level identification (e.g., are both a LOAEL and NOAEL identifiable?), and are
data amenable to benchmark dose modeling? are just a few factors for which a study might be
selected. For those PFAS that have an existing health assessment (e.g., in the hypothetical
example, PFAS 1-3), such decisions have already been made by the publishing authors. In
practice, if the user of this framework deviates from use of existing health assessment dose-
response metrics, clear rationale must be provided in the mixtures assessment. The dose-response
data/metrics (e.g., PODs, dose-response curves) selected across mixture PFAS should be clearly
presented, as in Table 4-3 for the hypothetical five component PFAS mixture.
53
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Table 4-3. Data array to inform decisions in Steps 2 and 3 of the framework approach for
component-based mixtures assessment of PFAS.
PFAS 1
PFAS 2
PFAS 3
PFAS 4
PFAS 5
RfD:3 E-8
POD . : 0.00001
(BMDL0.1ER)
U Ff: 300
Critical effect: Delayed
growth and development in
offspring (ABC et a!. 2022)
RfD: 1 E-5
POD. . 0.0013 (BMDL10)
U F;: 100
Critical effect: Liver necrosis
(DEF et al. 2022)
RfD:7 E-4
POD. . : 0.21 (BMDL1SD)
UF,: 300
Critical effect: Decreased
thyroid hormones (T4/T3)
(GHI et al. 2022)
S-D rat; single generation
repro/dev study;
Daily gavage GD 1-20 .
(ABC etal. 2022) :
S-D rat; 2-year bioassay;
DW ad libitum
(DEF et al. 2022)
Ci>7BL6 mouse;
90-day gavage;
(GHI et al. 2022)
F344 Rat; two generation
repro/dev study;
Multiple dev
outcomes In
offspring; feed ad
libitum (JKL et al. 2022)
Bioactivity profile in ToxCast;
Biological perturbation at 5%
AED (2.8 E-4 mg/kg-day) = 4^
cyt c oxidase in HepaRG cells;
1" ox. Stress
- - Bioactivity profile in ToxCast;
Biological perturbation at 5%
AED (8 E-5 mg/kg-day) =
epoxide hydrolase in HepaRG
cells; 'T' ox. stress
Notes: AED = administered equivalent dose15; GD = gestational day; HHRA = human health risk assessment; PODhed = human
equivalent point of departure; RfD = oral reference dose; LTFc = composite uncertainty factor.
Step 4: Perform component-based mixture assessment approach(es).
At this juncture in the data-driven framework approach, information has been assembled to
facilitate application in Step 4. In practice, the user may choose to select one component-based
mixture approach over others, based on data evaluation/interpretation, or apply data where
appropriate to more than one of the approaches. For the purposes of demonstrating practical
application using the PFAS 1-5 mixture, all approaches (i.e., Step 4/bottom row of Figure 4-1)
will be selected and demonstrated in the following sections.
15 An AED is an estimated oral exposure dose that results in an internal steady-state concentration in humans
consistent with the in vitro concentration associated with a biological perturbation or activity.
54
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Within each component-based approach (i.e., HI/TOSHI, RPF, and M-BMD), these five general
steps apply:
1. Assemble/derive component health effects endpoints (e.g., PODhed, RfD);
2. Assemble/derive health-based media concentrations (e.g., HBWC);
3. Select exposure estimates (e.g., measured water concentrations);
4. Calculate PFAS mixture potency; and
5. Compare PFAS mixture potency to existing health-based value (e.g., HI = 1.0, HBWC).
5.0 Hazard Index (HI) Approach
5.1 Background on the HI Approach
The HI is the most commonly used component-based mixture risk assessment method in EPA.
Because the HI employs a population level exposure and human health assessment value, such as
an oral RfD, this ratio provides an indication of potential health risk(s). That is, the HI is a
decision aid; it is not a mixture risk estimate in that it is not expressed as a probability, nor is it
an estimate of specific toxicity (e.g., embryo toxicity). The HI is based on an assumption of DA
among the mixture components (EPA, 2000b; Svendsgaard and Hertzberg, 1994). In the HI
approach, an HQ is calculated as the ratio of human exposure to a health-based RfV for each
mixture component chemical (i) (EPA, 1986). The HI is unitless, so in the HI formula, E and the
RfV must be in the same units (Equation 5-1). For example, if E is the oral intake rate (mg/kg-
day), then the RfV could be the RfD, which has the same units. Alternatively, the exposure
metric can be a media-specific metric such as water concentration and the toxicity value is best
represented as a duration-specific HBWC, for example, an EPA lifetime drinking water Health
Advisory (e.g., EPA 2022a,b) or MCLG, or a similar value (e.g., developed by a state). In this
case, the HQ is calculated as the ratio of water concentration (in mass/volume) to a HBWC (also
in mass/volume). The component chemical HQs are then summed across the mixture to yield the
HI, as illustrated in Equation 5-1.
n ri tz
Hi=y\ hq. = y—
A ' (Eqn.5-1)
Where:
HI = Hazard Index
HQ; = Hazard Quotient for chemical i
E; = Exposure, i.e., dose (mg/kg-day) or media concentration, such as in drinking
water (ng/L), for chemical i
RfVi = Reference value (e.g., oral RfD or MRL (mg/kg-day), or corresponding health-
based, media-specific value; e.g., such as a HBWC, for example, a drinking water
Health Advisory or MCLG) for chemical i (ng/L)
Because the numerator of each component chemical HQ is the estimated population-level human
exposure, the noncancer health RfVs used in the denominator must be based on human toxicity.
55
-------
PUBLIC REVIEW DRAFT
MARCH 2023
These RfVs are derived either directly from human epidemiological/occupational study-based
PODs (or measured or modeled EDx from exposure-response data in a cohort or population) or
as human-equivalent PODs converted from experimental animal studies (e.g., conversion of a
rodent POD to a human equivalent dose (PODhed) using cross-species TK-based modeling or
allometric body-weight scaling).
The HI approach in practical application may be subdivided into a "general" HI and a "target-
organ specific" HI (TOSHI). In either case, following the logic flow in Figure 4-1 to the general
HI or the TOSHI, they are both applied under an assumption of dose additivity. In the HI, the
RfV for each mixture component chemical is used in the calculation of a HQ, irrespective of the
effect on which each component RfV is based (e.g., RfD for mixture chemical 1 may be based on
liver effect, for chemical 2 thyroid effect, and chemical 3 developmental effect). The resultant HI
is generally a health-protective indicator because often the most sensitive health effects are used
as the basis for each respective chemical HQ. Conversely, the TOSHI entails derivation of HQs
for each mixture component chemical based on a "similar" effect. For example, in the case of a
liver-specific HI, for some mixture components the liver effect(s) may indeed be the basis for the
RfD whereas for other components, the liver might be among the least sensitive of effects. To
use this approach, organ-specific reference values (osRfVs) or TTDs are needed (note: these are
the same type of noncancer values, just with different naming conventions) for each mixture
component of potential concern. For chemicals lacking hazard and dose-response data from
traditional or NAM-based data streams for the selected effect, it may not be possible to
determine their potential contribution to the mixture, which may result in an underestimation of
the overall mixture risk.
A HI greater than 1.0 is generally regarded as an indicator of potential adverse health risks
associated with exposure to the mixture. A HI less than or equal to 1.0 is generally regarded as
having no appreciable risk (recall that an RfV, such as an oral RfD, represents an estimate at
which no appreciable risk of deleterious effects exists), typically requiring no further analysis
(EPA, 1986, 1991, 2000b). However, in some circumstances the user may want to consider a HI
less than 1.0, for example, for screening when multiple contaminants of concern are present at a
site or one or more are present in multiple exposure media. In the case of PFAS, final peer-
reviewed toxicity assessments are only available for a small proportion of the approximately
11,000 environmentally relevant PFAS (e.g., see summary of EPA and ATSDR PFAS
assessments in Table 4-4). EPA's primary source of peer-reviewed toxicity assessments is its
IRIS program, but in some cases (e.g., when no IRIS assessment exists or there is a more current
assessment from another authoritative source), the agency relies on assessments from other EPA
program offices, and other state, national, and international programs. U.S. federal human health
assessments, such as EPA's IRIS16, PPRTV17, EPA Office of Water toxicity assessments18,
TSCA risk evaluations19, and ATSDR's ToxProfiles20, undergo rigorous peer and public review
processes; as a result, they are considered to be of high scientific quality. The chronic RfDs for
16 https://www.epa.gov/iris
17 https://www.epa.gov/pprtv
18 e. g.. https://www.epa.gov/svstem/files/doctiments/2 zenx-chemicals-toxicitv-assessment tech-edited oct-
-I ''('8.pdf
19 https://www.epa.gov/assessing-and-managing-chemicals-niider-tsca/risk-evalnations-existlng-cheniicals-niider-
tsca
20 https://www.atsdr.cdc.gov/toxprofiledocs/index.htinl
56
-------
PUBLIC REVIEW DRAFT
MARCH 2023
PFBS (EPA, 2021a), HFPO-DA (EPA, 2021b), and PFBA (EPA, 2022e) represent the only
currently available/final EPA toxicity values, although the agency is updating its PFOA and
PFOS assessments (EPA, 2023b,c,d,e) and several more PFAS assessments are under
development in EPA/ORD (e.g., PFHxA, PFHxS, PFNA, and PFDA; see Table 5-1 below) that
can be considered in the future. Additionally, use of this approach could consider other PFAS
toxicity values (e.g., ATSDR MRLs) for which EPA has not yet developed values.
Table 5-1. EPA and ATSDR Peer-Reviewed Human Health Assessments Containing
Noncancer Toxicity Values (RfDs or MRLs) for PFAS that Are Either Final or Under
Development
Chemical
EPA Chronic Oral RfD
ATSDR Intermediate MRLa
PFOA
Draft 2023 RfD = 3 x 10"8 mg/kg/day
2021 MRL = 3 x 10"6 mg/kg/day
PFOS
Draft 2023 RfD = 1 x 10"7 mg/kg/day
2021 MRL = 2 x 10"6 mg/kg/day
PFNA
Under development in the EPA IRIS
program
2021 MRL = 3 xlO"6 mg/kg/day
PFDA
Under development in the EPA IRIS
program
N/A
PFBA
2022 RfD = 1 x 10"3 mg/kg/day
N/A
PFBS
2021 RfD = 3 x 10"4 mg/kg/day
N/A
PFHxA
Under development in the EPA IRIS
program
N/A
PFHxS
Under development in the EPA IRIS
program
2021 MRL = 2 x 10"5 mg/kg/day
HFPO-DA
2021 RfD = 3 x 10"6 mg/kg/day
N/A
Notes: N/A = Not available.
aNote that MRLs and RfDs are not necessarily equivalent (e.g., intermediate duration MRL vs. chronic duration RfD; EPA and
ATSDR may apply different uncertainty /modifying factors) and are developed for different purposes.
Some state health agencies publish toxicological assessments for PFAS that could potentially be
used in HI calculations. For example, the Minnesota Department of Health publishes
Toxicological Summaries that include the assessment of available toxicological information and
subsequent development of oral toxicity values if adequate data are available (MN DOH, 2021).
It should be noted that state or other (e.g., international) assessments may have varying levels of
peer and public review and may reflect different risk assessment practices or policy choices as
compared to EPA or ATSDR assessments.
There may be scenarios where a final peer-reviewed toxicity assessment for one or more
component chemicals is not available for a mixture. In these cases, an evaluation of available
hazard and dose-response information for PFAS in the mixture may be necessary under a HI
approach. For instance, there may be a need to develop toxicity value(s) to estimate potential
risks associated with site-specific/localized contamination from a PFAS mixture with a
component(s) that may not be relevant to other areas, sites, or exposure sources, and/or has not
been prioritized for assessment at the federal level. In such cases, the user of this framework
57
-------
PUBLIC REVIEW DRAFT
MARCH 2023
might have a need to develop a targeted, fit-for-purpose assessment, if possible (i.e., based on
availability of hazard and dose-response data, resources, and expertise). Excluding component
PFAS that lack off-the-shelf toxicity values from further analysis could result in underestimation
of the potential risk of the mixture. If de novo derivation of toxicity values is necessary, it is
recommended that experts in hazard identification and dose response assessment be consulted for
scientific input and review, and the associated uncertainties (e.g., data gaps) be transparently
characterized. EPA has published several peer-reviewed guidance documents that may assist in
efforts to derive chronic (or subchronic) oral RfDs for chemicals with no available peer-reviewed
toxicological assessment (for more information see EPA's Human Health Risk Assessment
website at https://www.epa.eov/risk/human-health-risk-assessment).
To date, the majority of environmental chemicals including PFAS are data-poor, having no
known or available information to inform hazard or dose-response in a screening/prioritization or
assessment context. Considering that the number of legacy and new(er) chemicals present in
commerce and the environment is in the tens of thousands, the generation of traditional animal
toxicity data to support hazard identification and dose-response assessment would take decades
and extraordinary numbers of animals and fiscal resources to complete. As human populations
and biota are currently exposed to mixtures of chemicals such as PFAS, it is critical to identify
methods, approaches, and platforms that can provide some reasonable context for potential
human health hazard(s) and associated dose-response/potency for effects associated with
exposure to multiples of PFAS (i.e., two or more co-occurring PFAS). A diverse set of resources
has been developed over the past 15 + years that entails, in general, high(er)-throughput assays in
cell culture (or cell free) systems, in silico computational prediction models, alternative animal
species (e.g., zebrafish), and refined short-term laboratory rodent assays and databases and
platforms to collate and deliver such data to end-users. These methods, assays, and platforms are
collectively referred to as NAMs. In the absence of traditional animal bioassay and human
epidemiological information, validated NAMs could potentially play a pivotal and
transformational role in human health (and ecological) risk assessment, particularly in evaluating
hazard and dose-response of PFAS that co-occur in mixtures.
Individually or in concert, NAMs such as in vitro cell bioactivity and in silico platforms (e.g.,
read-across) might inform identification or prediction of data that can be used in PFAS-specific
hazard and dose-response assessment. For example, in vitro concentration-bioactivity data from
resources such as ToxCast and Tox21 can be transformed into estimated in vivo exposure-
response using IVIVE and rTK (Rotroff et al., 2010; Wambaugh et al., 2015; Wetmore et al.,
2012, 2014). These administered human-equivalent dose datasets could potentially then be used
to identify PODs (e.g., BMDs, NOAELs, LOAELs), and with expert-driven application of
appropriate UFs, be leveraged into the derivation of corresponding noncancer toxicity values.
These NAM-based toxicity values could then be converted into corresponding HBWCs and used,
with exposure data, to calculate HQs for data-poor PFAS.
A critical consideration in using NAM-based hazard and concentration/dose-response data is
recognizing that for some high(er) throughput platforms or bioassays, perturbations of
underlying biological pathways may not be readily identifiable as being directly related to
specific apical toxic effects. That is, chemical exposures may elicit a myriad of perturbations or
responses at the molecular, macromolecular, or cellular level, with some alterations being critical
or key to eliciting an apical toxic effect level response, whereas many other alterations may
58
-------
PUBLIC REVIEW DRAFT
MARCH 2023
seemingly have no relationship to toxic effect(s) (e.g., general stress, housekeeping). However,
the dose-response relationship associated with non-apical perturbations or effects (e.g., cell-
based bioactivity) may be considered in an in vivo effect agnostic context. Specifically, although
there may not be clear qualitative linkages between non-apical biological perturbations and a
specific, apical tissue- or organ-level effect, corresponding dose-response relationships for
biological perturbations have been shown to provide a reasonable quantitative approximation for
dose-response (e.g., POD) associated with traditional apical effects (Paul-Friedman et al., 2020;
Johnson et al., 2020; Thomas et al., 2011, 2013). The implication for use of NAM data such as in
vivo or in vitro cell-based bioactivity or transcriptomics, for example, is that pathway- or cell
function-based response levels (e.g., effect concentration 50 (ECso), inhibitory concentration 50
(IC50), or other biologically supported response levels of interest), could potentially be leveraged
and applied in the mixture component approaches proposed in this chapter (e.g., HI, RPF, M-
BMD), irrespective of direct linkage(s) to a phenotypic apical effect.
In summary, considering the lengthy and resource-intensive processes and study protocols (e.g.,
OECD Test Guidelines-type studies) typically involved in generating traditional repeat-dose
bioassay data for human health assessment of chemicals, incorporating NAMs could potentially
serve an important role for PFAS screening and assessment, including in a mixture context. It is
recognized that practical application of NAMs in an assessment, whether for a single chemical or
mixtures of chemicals, would be dependent on whether the results provide information that fits a
decision context or purpose, and this may not be intuitive. It is recommended that experts in
NAM data interpretation be consulted for potential integration into mixtures
screening/assessment to appropriately contextualize the applicability of results, and that they
transparently communicate uncertainties associated with a given platform or assay output(s) in
human health assessment.
5.2 Illustrative Example Application of the General HI to a Hypothetical Mixture
of Five PFAS
As mentioned previously, final human health assessments with chronic oral RfDs exist for
hypothetical PFAS 1-3. Based on the RfDs for PFAS 1-3 (Table 4-3), PFAS 1 is a
comprehensively studied chemical that is most potent for effect(s); PFAS 2 is also well-studied,
but is less potent than PFAS 1 for effect(s); and PFAS 3 has been studied and is even less potent
than PFAS 1 or 2. PFAS 4 has experimental animal toxicity data available but no formal human
health assessment. Finally, PFAS 5 is data-poor, and was identified as having only bioactivity
data available under step 2 of the framework approach to inform hazard and dose-response
(Table 4-3). As PFAS 1-3 have existing human health assessment values, integration into the HI
approach is simplified. However, for both PFAS 4 and 5, integration would necessitate de novo
calculation of values in order to develop component HQs and an overall PFAS mixture HI
(Equation 5-1). For the purposes of the defined illustrative example for the hypothetical five
component PFAS mixture, this process is as follows:
5.2.1 General HI Step 1: Assemble/derive component health effects endpoints
(Chronic oral RfDs)
PFAS 1-3: Upon review of the available information harvested in the literature search in Step 1
of the framework approach, formal human health assessments containing oral RfDs were
identified (Table 4-3). However, the critical effect on which each corresponding RfD was
59
-------
PUBLIC REVIEW DRAFT
MARCH 2023
derived are in different effect domains; PFAS 1 critical effect = developmental effect in
offspring; PFAS 2 critical effect = liver effect in adults; and PFAS 3 critical effect = thyroid
hormone effect in adult females (in a repro/developmental life stage). As such, application of the
general HI is optimal in this scenario and will entail use of the overall RfD (or MRL), regardless
of underlying critical health effect. If a subchronic RfD or an MRL is only available for an
intermediate duration (akin to subchronic for EPA purposes), additional uncertainty (e.g.,
subchronic-to-chronic duration) may be considered for extrapolation to a corresponding chronic
duration value, unless subchronic/intermediate duration is the target.
PFAS 4: No federal, state, or other assessments with a RfV are available, but traditional hazard
and dose-response (e.g., experimental animal study) data were judged adequate to support
derivation. Systematic review and evaluation of the animal study data led to identification of a
single best study (e.g., hypothetical 2-Gen repro/dev rat study; Table 4-3) and multiple
developmental health outcomes as candidate critical effects such as delayed growth and
development at post-natal day 1 (PND 1) and decreased neonatal viability and thyroid hormone
levels at PND 4. Thus, the user may choose to calculate a RfV using appropriate dose-response
metrics (i.e., POD) and application of UFs. Appropriate characterization of hazard conclusions
and qualitative and quantitative confidence and uncertainty(ies) in de novo derivation of RfVs
for PFAS in this category is imperative. For the specific hypothetical PFAS example, the dose-
response data associated with delayed growth and development in PND 1 offspring provided the
most robust endpoint and confidence in dose-response for PFAS 4; following benchmark dose
modeling (as per EPA BMD guidance (EPA, 2012)), a rat lower statistical bound on a BMD
(BMDLisd) of 1.06 mg/kg-day was calculated.
As shown in Figure 4-7B, TK data exists for PFAS 4 in relevant animal species (i.e., rats) and
humans, such that a data-informed adjustment approach for estimating the dosimetric adjustment
factor (DAF) can be used. In Recommended Use of Body Weight314 as the Default Method in
Derivation of the Oral Reference Dose (EPA, 201 lb), EPA endorses a hierarchy of approaches
to derive human equivalent oral exposures using data from laboratory animal species, with the
preferred approach being physiologically based TK modeling. Other approaches might include
using chemical-specific information, without a complete physiologically based TK model. In the
absence of chemical-specific models or data to inform the derivation of human equivalent oral
exposures, EPA endorses BW3/4 as a default to extrapolate toxicologically equivalent doses of
orally administered agents from laboratory animals to humans for the purpose of deriving an RfD
under certain exposure conditions. In this illustrative hypothetical mixture example, it was
determined that: (1) Clearance values were included in the dosimetric adjustment of PODs used
in the derivation of non-cancer human health assessment values for PFAS 1-3; and (2) kinetic
data for PFAS 4 are sufficient to support a data-informed dosimetric adjustment of the rat POD.
Briefly, while specific TK data needed to estimate clearance or volume of distribution in rodents
or humans for PFAS 4 were not available, clearance values for humans and rats could be
estimated, under the assumption that the volume of distribution in human females is equal to
female adult rats (i.e., the PFAS-exposed unit leading to effects in PND1 offspring), as follows:
Clearance = elimination rate constant (ke) x volume of distribution (Vd)
Where ke= (ln2/plasma half-life) = (0.693/plasma half-life), and Vd is assumed
equivalent between female rats and humans
60
-------
PUBLIC REVIEW DRAFT MARCH 2023
Table 5-2. Calculation of estimated clearance values for PFAS 4 in female rats and humans.
PFAS4
Plasma half-life (hr)
Elimination rate
constant (hr1)
Volume of Distribution
Estimated Clearance
(L/day)
{L/day-kg)
Female rats
33.6
0.021
1.0
0.021
Humans
24,528
0.000028
1.0
0.000028
*The value of 1.0 was used for volume of distribution (Vd) strictly for the purpose of calculation
of an estimated clearance value; the Vd of 1.0 is not based on empirical evidence for PFAS 4.
Having made this estimate, the ratio of clearance values (Table 5-2) in human females to that in
female rats, CLh:CLa, can be used to calculate the DAF, and the resulting human equivalent
dose (HED) can be calculated using equation 5-2 as follows:
POD = the rat BMDLisd of 1.06 mg/kg-day
DAF = CLh/CLa
CLa = 0.021 L/day-kg (female adult rat; the effect in offspring is a function of maternal
intake)
CLh = 0.000028 L/day-kg
The application of the DAF of 0.001 to the rat POD results in a PODhed of 0.0011 mg/kg-day.
This PODhed was then divided by a composite UF of 100 which included a human
interindividual variability UF (UFh) of 10 for human interindividual variability as there were no
data to inform this parameter, an interspecies UF (UFa) of 3 for extrapolation from rats to
humans, as the kinetic differences between species were accounted for in part by the dosimetric
adjustment above, a LOAEL-to-NOAEL UF (UFl) of 1 because the POD is a BMDL, an
extrapolation from subchronic to a chronic exposure duration UF (UFs) of 1 for sub chronic-to-
chronic extrapolation because the effect was observed in a developmental population following
gestational exposure (developmental exposures are considered duration independent in EPA),
and a database UF (UFd) of 3 as the totality of the hazard and dose-response database included
repeat-dose studies in rats and mice of longer-term duration (i.e., subchronic), as well as single-
and two-generation reproductive and developmental studies in rats. The resulting RfD for PFAS
4 = PODhed/UF = 0.0011 mg/kg-day/100 = 1 E-5 mg/kg-day.
PFAS 5: Because no final federal, state, or other RfD or MRL or traditional hazard and dose-
response data are available, NAM data streams could be surveyed and leveraged for PFAS
information that might facilitate development of a POD, and potentially, derivation of a NAM-
based RfV using application of UFs consistent with the data scenario (Judson et al., 2011; Parish
et al., 2020). It is recommended that data be systematically evaluated for suitability in supporting
the derivation of RfVs using accepted approaches and practice. Unfortunately, no formal EPA
technical guidance or guidelines currently exist to guide the approach for use of NAM-based
PODs in quantitative human health risk assessment applications. However, for the purposes of
demonstrating potential application of NAM data (e.g., in vitro cell bioactivity) in the
HED =POD X ^
cla
(Eqn. 5-2)
Where:
61
-------
PUBLIC REVIEW DRAFT
MARCH 2023
hypothetical PFAS mixture evaluation, the general process within the context of this framework
approach is as shown in Figure 5-1.
Bioactivity
data in human
or animal
tissue/cells
Convert in vitro
concentrations to
in vivo
'administered
equivalent doses
(AEDs)' in human
using IVIVE/rTK
Perform dose-
response analyses
on AED data (e.g.,
benchmark dose)
to identify POD
Apply uncertainty
factors to human
PODs to derive a
bioactivity-based
RfV
Figure 5-1. General steps to derive bioactivity-based reference value (RfV) using bioactivity
data in human or animal tissue/cells.
The detailed steps and mechanics of the bioactivity > IVIVE/rTK > AED process above are
beyond the scope of this framework document; the reader is referred instead to (Paul-Friedman et
al., 2020; Wambaugh et al., 2018; Wetmore et al., 2012, 2014, 2015) for better context for the
conversion of in vitro cell-based exposure concentrations to approximately equivalent human
external exposure doses using IVIVE and rTK. The AED identified as a NAM-based human
POD for PFAS 5 is a BMD, modeled off of the AED-based dose-response data for the decreased
epoxide hydrolase endpoint in liver (HepaRG) cells; the BMDaedisd = 0.004 mg/kg-day. This
human POD was then divided by a composite UF of 100 which included a UFh of 1 for human
interindividual variability as the rTK model used in the conversion of in vitro cell-based
concentrations to corresponding human AEDs takes into account population level human
variability, a UFa of 1 for extrapolation from animals to humans since the resulting BMDaedisd
is an estimated human POD, a UFl of 1 since the POD is a BMD, a UFs of 10 for subchronic-to-
chronic extrapolation since the effect was observed in a short-term exposure duration and it is
unclear how longer exposure durations might impact underlying biology in a liver health
outcome, and a UFd of 10 as the database is limited to NAM data; no traditional toxicity studies
were identified. The resulting RfD for PFAS 5 = 0.004 mg/kg-day/100 = 4 E-5 mg/kg-day.
Appropriate characterization and denoting of confidence and qualitative and quantitative
uncertainty(ies) in RfVs derived for PFAS in this category is imperative. Consultation with
experts in the field of NAM data interpretation and risk assessment application is recommended.
Summary of RfDs
In summary, as shown in Table 5-3, RfDs for PFAS 1-5 range from 10"5 to 10"8 mg/kg-day, with
PFAS 1 being the most potent overall. Note that PFAS 4 and 5 have similar RfDs despite their
data limitations.
62
-------
PUBLIC REVIEW DRAFT MARCH 2023
Table 5-3. Summary of PODheds and RfDs for hypothetical PFAS in a mixture.
Liver Thyroid Developmental
PODhed PODhed PODhed RfD
(mg/kg-day) (mg/kg-day) (mg/kg-day) (mg/kg-day) Confidence
PFAS 1
0.044
0.24
0.00001
(BMDLo. ierhed)
3 E-8
High (formal
toxicity assessment)
PFAS 2
0.0013
(BMDLiohed)
0.23
0.0051
1 E-5
High (formal
toxicity assessment)
PFAS 3
N/A
0.21
(BMDLisdhed)
2.1
7 E-4
High (formal
toxicity assessment)
PFAS 4
50
N/A
0.0011
(BMDLisdhed)
1 E-5
Medium (high
quality in vivo data)
PFAS 5
0.004
(BMDisdaed)
N/A
N/A
4 E-5
Low (bioactivity-
based)
Note:
Bold indicates lowest (most sensitive) health outcome selected for RfD derivation.
5.2.2 General HI Step 2: Assemble/derive health-based media concentrations (HBWC)
Dependent upon the problem formulation, the user has the option to either use the oral RfV
calculated for mixture components, or to leverage such values in the calculation of media-
specific values, such as HBWCs for drinking water. Care should be taken to ensure that all
HBWCs are applicable to the same exposure duration. In the following examples, the HBWCs
are derived using chronic oral RfDs and thus are considered health protective values over a
lifetime of exposure.
How to Calculate a HBWC for Drinking Water
The following equation is used to derive a noncancer HBWC. A noncancer HBWC, such as a
lifetime HA or MCLG, is designed to be protective of noncancer effects over a lifetime of
exposure, including sensitive populations and life stages, and is typically based on data from
chronic experimental animal toxicity and/or human epidemiological studies. The calculation of a
HBWC includes an oral RfV such as a RfD (or chronic MRL or duration relevant user-provided
value), body weight-based drinking water intake (DWI-BW), and a relative source contribution
(RSC) factor as presented in equation 5-3.
Noncancer HBWC = (RfD/(DWI-BW)) * RSC (Eqn. 5-3)
Where:
RfD = chronic reference dose—an estimate (with uncertainty spanning perhaps an order
of magnitude) of a daily oral exposure of the human population to a substance that is
likely to be without an appreciable risk of deleterious effects during a lifetime, (see HI
step 1 above).
DWI-BW = the 90th percentile drinking water intake (DWI) for the selected population
or life stage, adjusted for body weight (BW), in units of liters of water consumed per
kilogram body weight per day (L/kg bw-day). The DWI-BW considers both direct and
63
-------
PUBLIC REVIEW DRAFT
MARCH 2023
indirect consumption of drinking water (indirect water consumption encompasses water
added in the preparation of foods or beverages, such as tea or coffee).
RSC = relative source contribution—the percentage of the total oral exposure attributed
to drinking water sources (EPA, 2000b), with the remainder of the exposure allocated to
all other routes or sources.
When developing HBWCs, the goal is to protect all ages of the general population including
potentially sensitive populations or life stages such as children. The approach to select the DWI-
BW and RSC for the HBWC includes a step to identify sensitive population(s) or life stage(s)
(i.e., populations or life stages that may be more susceptible or sensitive to a chemical exposure)
by considering the available data for the contaminant. Although data gaps can make it difficult to
identify the most sensitive population (e.g., not all windows or life stages of exposure or health
outcomes may have been assessed in available studies), the critical effect and POD that form the
basis for the RfD can provide some information about sensitive populations because the critical
effect is typically observed at the lowest tested dose among the available data. Evaluation of the
critical study, including the exposure interval, may identify a particularly sensitive population or
life stage (e.g., pregnant women, formula-fed infants, lactating women). In such cases, the user
can select the corresponding DWI-BW for that sensitive population or life stage from the
Exposure Factors Handbook (EPA, 2019a) to derive the HBWC. In practice, when multiple
populations or life stages are identified based on the principal study design and critical effect or
other health effects data (from animal or human studies), EPA selects the population or life stage
with the greatest DWI-BW because it is the most health protective. This approach ensures that all
populations and life stages are protected at the HBWC, and in the case of the HI approach, that
each component HQ and the overall HI is protective of all populations and life stages. In the
absence of information indicating a sensitive population or life stage (e.g., non-developmental
critical effect as in PFAS 2 and 3, or NAM-based reference dose as in PFAS 5), the DWI-BW
corresponding to all ages of the general population may be selected (Table 5-4).
Table 5-4 shows EPA exposure factors for DWI for some sensitive populations and life stages.
Other populations or life stages may also be considered depending on the available information
regarding study design and sensitivity to health effects after exposure to a contaminant.
Table 5-4. EPA Exposure Factors for Drinking Water Intake.
Population or
Life Stage
DWI-BW
(L/kg bw-day)
Description of Exposure Metric
Source
General
population (all
ages)
0.0338
90th percentile direct and indirect
consumption of community
water, consumer-only two-day
average, all ages.
2019 Exposure Factors
Handbook Chapter 3,
Table 3-21, NHANES
2005-2010 (EPA, 2019a)
Children
0.143
90th percentile direct and indirect
consumption of community
water, consumer-only two-day
average, birth to < 1 year.
2019 Exposure Factors
Handbook Chapter 3,
Table 3-21, NHANES
2005-2010 (EPA, 2019a)
64
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Population or DWI-BW
Life Stage (L/kg bw-day) Description of Exposure Metric
Source
Formula-fed
infants
0.249
90th percentile direct and indirect
consumption of community
water, formula-consumers only, 1
to < 3 months. Includes water
used to reconstitute formula plus
all other community water
ingested.
Kahn et al. (2013),
Estimates of Water
Ingestion in Formula by
Infants and Children
Based on CSFII 1994-
1996 and 1998a'b
Pregnant
women
0.0333
90th percentile direct and indirect
consumption of community
water, consumer-only two-day
average.
2019 Exposure Factors
Handbook Chapter 3,
Table 3-63, NHANES
2005-2010 (EPA, 2019a)
Women of
childbearing
age
0.0354
90th percentile direct and indirect
consumption of community
water, consumer-only two-day
average, 13 to < 50 years.
2019 Exposure Factors
Handbook Chapter 3,
Table 3-63, NHANES
2005-2010 (EPA, 2019a)
Lactating
women
0.0469
90th percentile direct and indirect
consumption of community
water, consumer-only two-day
average.
2019 Exposure Factors
Handbook Chapter 3,
Table 3-63, NHANES
2005-2010c (EPA, 2019a)
Notes: CSFII = continuing survey of food intake by individuals; L/kg bw-day = liter per kilogram body weight per day.
a The sample size does not meet the minimum reporting requirements as described in the Third Report on Nutrition Monitoring in
the United States (LSRO, 1995).
b Chapter 3.2.3 in EPA (2019a) cites Kahn et al. (2013) as the source of drinking water ingestion rates for formula-fed infants.
While EPA (2019a) provides the 95th percentile total direct and indirect water intake values, Office of Water/Office of Science
and Technology (OW/OST) policy is to utilize the 90th percentile DWI-BW.
c Estimates are less statistically reliable based on guidance published in the Joint Policy on Variance Estimation and Statistical
Reporting Standards on NHANES III and CSFII Reports: Human Nutrition Information Service (HNIS)/National Center for
Health Statistics (NCHS) Analytical Working Group Recommendations (NCHS, 1993).
To account for potential aggregate risk from exposures and exposure pathways other than oral
ingestion of drinking water, EPA applies an RSC when calculating HBWCs to ensure that total
human exposure to a contaminant does not exceed the daily exposure associated with the RfD.
When data are available for multiple sensitive populations or life stages, the most health-
protective RSC is selected. The RSC represents the proportion of an individual's total exposure
to a contaminant that is attributed to drinking water ingestion (directly or indirectly in beverages
like coffee, tea, or soup, as well as from transfer to dietary items prepared with drinking water)
relative to other exposure pathways. The remainder of the exposure equal to the RfD is allocated
to other potential exposure sources (EPA, 2000a). The purpose of the RSC is to ensure that the
level of a contaminant (e.g., HBWC value), when combined with other identified potential
sources of exposure for the population of concern, will not result in exposures that exceed the
RfD (EPA, 2000a).
To determine the RSC, EPA follows the Exposure Decision Tree for Defining Proposed RfD (or
POD/UF) Apportionment in EPA's guidance, Methodology for Deriving Ambient Water Quality
Criteria for the Protection of Human Health (EPA, 2000a). EPA considers whether there are
65
-------
PUBLIC REVIEW DRAFT
MARCH 2023
significant known or potential uses/sources other than drinking water, the adequacy of data and
strength of evidence available for each relevant exposure medium and pathway, and whether
adequate information on each source is available to quantitatively characterize the exposure
profile. The RSC is developed to reflect the exposure to the general population or a sensitive
population within the general population exposure.
Per EPA's guidance, in the absence of adequate data to quantitatively characterize exposure to a
contaminant, EPA typically recommends an RSC of 20%. When scientific data demonstrating
that sources and routes of exposure other than drinking water are not anticipated for a specific
pollutant, the RSC can be raised as high as 80% based on the available data, thereby allocating
the remaining 20% to other potential exposure sources (EPA, 2000a). For the illustrative
hypothetical PFAS mixture, an RSC of 0.2 (i.e., 20%) is selected as no information was
identified to suggest a higher value. The calculation of HBWCs for PFAS 1-5 are presented in
Table 5-5.
Table 5-5. Calculation of HBWCs for hypothetical PFAS in a mixture.
Chemical
Oral Reference Dose
(mg/kg-day)
DWI-BW (L/kg-day)
RSC
HBWC (ng/L)
PFAS 1
3 E-8
0.0354
0.2
0.2
PFAS 2
1 E-5
0.0338
0.2
60
PFAS 3
7 E-4
0.0338
0.2
4,000
PFAS 4
1 E-5
0.0469
0.2
43
PFAS 5
4 E-5
0.0338
0.2
200
5.2.3 General HI Step 3: Select exposure estimates (measured water concentrations)
Select appropriate exposure estimates consistent with the problem formulation. Specifically, the
user may choose to calculate or use exposure estimates that are for the oral route in general (i.e.,
total intake in mg/kg-day) or media-specific concentrations. In the hypothetical PFAS mixture
example, 'exposure' is represented by the drinking water monitoring data in Table 4-1.
5.2.4 General HI Step 4: Calculate PFAS mixture potency (component HQs and overall
HI)
Using the median of the drinking water monitoring data (Table 4-1) and the calculated HBWCs
for PFAS 1-4 (Table 5-5), individual component HQs are derived as shown in Table 5-6.
Table 5-6. Calculation of individual component hazard quotients (HQs) for the
hypothetical PFAS mixture.
Hypothetical Drinking Water
Exposure Estimate (ng/L)
HBWC (ng/L)
General HQ
PFAS 1
4.8
0.2
24.0
PFAS 2
52
60
0.9
66
-------
PUBLIC REVIEW DRAFT
MARCH 2023
Hypothetical Drinking Water
Exposure Estimate (ng/L)
HBWC (ng/L)
General HQ
PFAS 3 172
4,000
0.04
PFAS 4 58
43
1.3
PFAS 5 69
200
0.3
MIXTURE GENERAL HAZARD INDEX (HI)
27
Notes: HQ = DW Exposure Estimate/HBWC; HI = the sum of individual HQs.
5.2.5 General HI Step 5: Compare PFAS mixture potency (HI) to existing health-based
value (1.0)
The HI (27) is significantly greater than 1.0, indicating health risks to the mixture of PFAS at the
measured drinking water concentrations. Further, as illustrated by the individual component
HQs, PFAS 1 and 4 are risk drivers of the mixture HI with individual HQs greater than 1.0;
PFAS 2 and 5 also appear to be contributors with HQs of 0.9 and 0.3, respectively. Assessment
of PFAS 2 and 5 in isolation (individually) would indicate no/low risk (i.e., HQs < 1), but
assessment of the binary mixture of PFAS 2 and 5 would indicate appreciable risk (HI = 1.2).
Conversely, with a HQ of 0.04, PFAS 3 is less influential compared to the other mixture
components. In this hypothetical scenario, clearly PFAS 1 and 4, and potentially PFAS 2 and 5,
might be prioritized for remediation activity(ies).
It should be noted that in the hypothetical PFAS mixture, the HBWC for PFAS 1 (0.2 ng/L) is
lower than its corresponding hypothetical drinking water analytical quantitation limit of 4 ng/L
by over an order of magnitude. In such cases, any detectable level (i.e., of PFAS 1) will result in
an HI greater than 1.0 for the whole mixture.
5.3 Illustrative Example Application of the Target-Organ Specific HI (TOSHI) to a
Hypothetical Mixture of Five PFAS
5.3.1 TOSHI Step 1: Assemble/derive component health effects endpoints (RfDs or
target-organ toxicity doses)
Application of the TOSHI is essentially identical to the steps for the general HI. The critical
nuance is that use of human health/toxicity values across mixture components are effect/endpoint
specific. For some PFAS, this might be the overall RfD or MRL; for other PFAS, this will
involve TTDs (i.e., an RfD for a specific health effect). In the TOSHI, there is a greater
likelihood that TTDs have not been derived for effects other than the critical effect that
underpins the derivation of an overall RfD for a given PFAS, although in some federal and state
purviews this practice is changing. In those instances where only an overall RfD (or MRL) has
been derived, TTDs could potentially be derived for other health effect domains but should be
accomplished with transparent characterization of qualitative and quantitative uncertainties
associated with hazard and dose-response data on a case-by-case basis. TTDs are derived
identically to RfDs; however, there may be differing circumstances to consider such as type of
POD (e.g., BMD vs NOAEL or LOAEL), cross-species TK dosimetric adjustment (e.g., RfD
may have been derived from a POD based on an adjustment of rat kinetics to human kinetics,
whereas a TTD for the same chemical might be mouse to human resulting in different PODhed),
67
-------
PUBLIC REVIEW DRAFT
MARCH 2023
and/or different qualitative and quantitative uncertainties. In practice, human health assessment
applications including mixtures assessment may be more robust if TTDs are derived across all
health outcome domains that are supported by evidence. For the purposes of the illustrative
hypothetical PFAS mixture example, the calculation of TTDs is limited to the three selected
health effect domains listed in Table 5-7. Several more TTDs could potentially be derived based
on availability of data and confidence in the evidence conclusions.
Table 5-7. Target Organ Toxicity Doses (TTDs) for the hypothetical mixture PFAS; the
bolded numbers represent the overall RfD for each respective PFAS.
Target Organ Toxicity Doses (mg/kg-day)
Effect domain PFAS 1 PFAS 2 PFAS 3 PFAS 4 PFAS 5
~ 'flHHHT 4f~5* v*
. * -• ••••
9E-3 2E-3 ~
Note:
* TTDnam based on in vitro perturbation indicative of oxidative stress in liver cells.
5.3.2 TOSHI Step 2: Assemble/derive health-based media concentrations (HBWC)
To calculate HBWCs for the TOSHI, the TTDs for a specific effect domain across mixture
components are used in the calculation of HQs and a TOSHI. For example, a TOSHI for
developmental effects (TOSHIdev) for the hypothetical PFAS mixture can be calculated using
the developmental TTDs, appropriate DWI-BWs and RSCs (Table 5-8). For this developmental
example, the DWI-BW is the 90th percentile direct and indirect consumption of community
water, consumer-only two-day average, for women of childbearing age (13 to < 50 years).
Table 5-8. Calculation of Developmental Effect-Specific HBWCs for hypothetical PFAS in
a mixture using TTDs.
Chemical
Target Organ Toxicity
Dose (mg/kg-day)
DWI-BW
(L/kg-day)
RSC
TOSHIdev
HBWC (ng/L)
PFAS 1
3 E-8
0.0354
0.2
0.2
PFAS 2
9 E-3
0.0354
0.2
50,000
PFAS 3
2 E-3
0.0354
0.2
10,000
PFAS 4
1 E-5
0.0469
0.2
43
PFAS 5
—
N/A
N/A
ND
Notes: N/A = not applicable; ND = not determined.
Bolded numbers indicate that the TTD for developmental effects is the overall RfD for that PFAS.
68
-------
PUBLIC REVIEW DRAFT
MARCH 2023
5.3.3 TOSH! Step 3: Select exposure estimates (measured water concentrations)
Select appropriate exposure estimates consistent with the problem formulation. Specifically, the
user may choose to calculate or use exposure estimates that are for the oral route in general (i.e.,
total intake in mg/kg-day) or media-specific concentrations. In the hypothetical PFAS mixture
example, 'exposure' is represented by the drinking water monitoring data in Table 4-1.
5.3.4 TOSHI Step 4: Calculate PFAS mixture potency (component HQs and overall
TOSHI)
Using the median of the drinking water monitoring data (Table 4-1) and the calculated HBWCs
for PFAS 1-4 derived from TTDs for the developmental effect domain (Table 5-8), individual
component HQs are derived as shown in Table 5-9.
Table 5-9. Calculation of individual component hazard quotients (HQs) specifically for
developmental effects associated with the hypothetical PFAS mixture. The HBWCs in this
TOSHI application are derived from TTDs for the developmental effect domain.
Chemical
Hypothetical Drinking
Water Exposure
Estimate (ng/L)
TOSHIdev HBWC
(ng/L)
TOSHIdev HQ
PFAS 1
4.8
0.2
24
PFAS 2
52
50,000
0.001
PFAS 3
172
10,000
0.01
PFAS 4
58
43
1.3
PFAS 5
69
ND
ND
MIXTURE TOSHIdev
25
Notes: HQ = DW Exposure Estimate/HBWC; HI = the sum of individual HQs. ND = not determined.
5.3.5 TOSHI Step 5: Compare PFAS mixture potency (HI) to existing health benchmark
(1.0)
The TOSHIdev of 25 indicates concern for developmental effects associated with exposure to the
hypothetical PFAS mixture at the measured drinking water concentrations (Table 4-1). While
this example application shows that use of TTDs did not meaningfully diminish indication of
health risk associated with the mixture (compared to a general HI approach), the individual HQs
clearly demonstrate drivers (PFAS 1 and 4) and relative inerts (PFAS 2 and 3) for developmental
health outcomes. The converse is possible dependent on the TTDs for different health outcomes,
and differing PFAS concentrations in environmental media. Please note that the magnitude of an
HI, TOSHI, or an individual component HQ, should not be directly interpreted as a quantitative
estimate of increased level of concern. For example, a mixture HI of 20 is not necessarily of 10-
fold greater concern than a mixture HI of 2.0. The practical interpretation is that both mixture
approaches would indicate appreciable risk of health effects in exposed populations.
69
-------
PUBLIC REVIEW DRAFT
MARCH 2023
5.4 Advantages and Challenges of the General HI Approach
The HI approach provides an indication of the joint toxicity associated with co-occurrence of
PFAS in environmental media, such as drinking water. One advantage of the HI formula in risk
communication is that interpretation of the results is relatively straightforward. The simplicity of
the method is in taking a ratio of the exposure to hazard to indicate potential concern for a
mixture of PFAS and providing an alert to specific PFAS that may be potential drivers in risk to
human health (i.e., those PFAS where the HQs have greater contribution to a HI > 1.0, relative to
other PFAS members of the mixture).
Another advantage is that the "hazard" does not necessarily have to be the same for general HI
(e.g., all liver or all kidney effects). Specifically, the HI approach can be used where the
individual HQ calculated for each mixture PFAS is based on the most well-characterized, and
oftentimes most sensitive, toxic effect and corresponding noncancer RfV (e.g., oral RfD). As
such, a HI will typically represent the most health-protective indicator of mixture risk, as each
component HQ is based on its overall RfV.
The HI is an indication of appreciable risk, not an estimate of the concentration of the mixture in
water that may result in adverse health outcomes after a specific period of exposure.
Comparisons of HI estimates across different exposure scenarios can be misleading. Because the
HI is based on DA, it implies that if two exposure scenarios involve the same chemicals and their
HI values are the same, then with other factors being equal (e.g., exposure frequency and
duration, similar endpoints, and similar receptor (exposed population) age), the two exposure
scenarios could be judged to have the same potential for causing toxic effects. This interpretation
has the strongest scientific foundation when there are only minor differences in the component
exposures (thus, same exposure route, similar exposure duration for specific receptors, and
roughly similar estimates of the individual HQs) between the two scenarios. Interpretation is
more difficult when the underlying information is poor. For example, if the dominant chemical
(highest HQ) has a highly uncertain exposure estimate, or its RfV was derived using a large
composite UF, then the associated HI is also highly uncertain.
Another challenge of the application of this HI approach to specific media such as water is that it
requires derivation of a health-based, media-specific concentration like a drinking water Health
Advisory or MCLG, in addition to the underlying oral RfV (e.g., RfD). Development of these
values typically requires significant expertise and resources often on a longer timeframe (i.e.,
years). In addition, while a formal hierarchy of preferred human health reference/toxicity values
is not being proposed in this framework, there is a recognized gradation of confidence across the
range of possible PFAS values. Specifically, it would clearly be preferable to use RfVs obtained
from assessment sources that use comprehensive and transparent systematic approaches and
standardized protocols. The level of confidence or certainty in such values would be greater than
RfVs deriving from questionable toxicity data sources, entails non-transparent decision-making,
and/or is associated with higher levels of qualitative and quantitative uncertainty.
What might be perceived as a challenge for PFAS human health assessment in general could be
an opportunity to advance risk assessment science and practice. Specifically, in the case of
NAM, dose-response metrics obtained from bioactivity-based assays/platforms (or read-across)
may be assigned some level of a priori uncertainty simply because of lack of confidence by end-
users in interpretation and risk assessment application of such data and outputs. As mentioned
70
-------
PUBLIC REVIEW DRAFT
MARCH 2023
previously in this framework, NAMs may represent the only opportunity to integrate a data-poor
PFAS into mixtures assessment. Further, while the integration of NAMs into applications such as
mixtures risk assessment was demonstrated in the hypothetical example using a POD from a
specific assay-type {in vitro cell bioactivity), available NAM data could be leveraged from a
diverse assay portfolio. For example, transcriptomic data from whole animals or cells in vitro
using platforms such as BioSpyder (i.e., TempO-Seq; see https://www.biospyder.com/).
microarrays, and/or RT-PCR may represent additional opportunities to integrate validated
methods and data into assessment application. A potential future improvement using NAMs such
as cell bioactivity (including transcriptomics) may be the categorical integration of qualitative
and quantitative information from across platforms to develop more comprehensive NAM-based
hazard determinations and identification of candidate PODs (i.e., consensus lower bound BMD;
cross-NAM platform mean; etc.). The end-user of this framework, in consultation with
experts/practitioners in NAM development and application, would be advised to leverage NAM
when and where possible, but always characterizing and transparently communicating qualitative
and quantitative uncertainty(ies) along the continuum from data generation and fit-for-purpose
application (Parish et al., 2020) to RfV and subsequent HQ and HI calculations. The
disadvantage to not using NAM data and approaches when applicable to a given PFAS mixture
is that data-poor PFAS would not be accounted for in the HI, thus potentially underestimating
mixture hazard.
In summary, in scenarios where a diverse amalgamation of different types of RfVs (i.e., deriving
from different assessment sources and/or data types) are used in the calculation of HQs and His,
the respective confidence and qualitative uncertainty characterizations for each PFAS need to be
transparently communicated in overall mixture hazard interpretations.
5.5 Advantages and Challenges of the Target Organ Specific Hazard Index
In a TOSHI, toxicity values are aggregated by the "same" target organ endpoint/effect, and HQ
(and HI) values are developed for each effect domain independently (e.g., liver-specific HI,
thyroid-specific HI). The disadvantage of a TOSHI is that it can only be performed for those
PFAS for which a health effect specific RfD (e.g., TTD) is calculated. For example, for some
PFAS a given health effect might be poorly characterized or not studied at all, or, as a function of
dose may be one of the less(er) potent effects in the profile of toxicity for that particular PFAS.
Another limitation is that so many PFAS species lack human epidemiological or experimental
animal hazard and dose-response information across a broad(er) effect range thus limiting
derivation of TTD values. As with the HI, a TOSHI approach might benefit from consideration
of NAM data and approaches that can inform organ/tissue-specific dose-response.
6.0 Relative Potency Factor (RPF) Approach
6.1 Background on RPF Approach
RPF approaches comprise another basic dose-addition method used most commonly by EPA in
mixtures assessment. There are two key types of the RPF approach: (1) the general RPF
approach that has been applied to pesticides, disinfection by-products (Simmons et al., 2004),
and a few other chemical groups and (2) the TEF approach that was originally developed for
mixtures of dioxins and DLCs. The TEF approach is considered a special case of the RPF
71
-------
PUBLIC REVIEW DRAFT
MARCH 2023
approach wherein mixture components are known to act via an identical MOA (e.g., dioxins and
DLCs and AhR activation).
For chemicals demonstrated to act via a similar MOA, or in the case of this framework, those
shown to induce the same/similar health effect (see Section 3 for discussion and justification), an
RPF represents the relative difference in potency between a mixture IC and other members of the
mixture. The IC does not necessarily have to be the most potent member of a given mixture.
Rather, an IC is typically selected because it has the highest quality and most robust
toxicological database and is considered to be most representative of the type of toxicity caused
by the mixture components (EPA, 1986, 2000b). The role of the IC in the RPF approach is to
serve as the point of reference for standardizing the common toxicity (i.e., scaling the potencies)
of all component chemicals in the analysis. The most important consideration in selecting one
mixture component over another as an IC is that high-quality dose-response data are available
(e.g., for the common toxic effect/species/sex) for the exposure route, duration, and pathways of
interest. Further, the IC must have dose-response data for the dose range of interest; chemicals
with steep slopes that cause an effect and/or induce significant toxicity at all doses tested are not
ideal for IC selection. In most cases, identification of a single best mixture component IC will be
evident. However, in the event that two or more mixture components are identified as candidate
ICs, the user must judge which candidate is most representative of the mixture, or subgroupings
within a mixture, and has the most robust toxicity database. It should be noted that selection of
an IC can be duration-, exposure route-, and/or health outcome-specific. That is, in practical
application, it is possible that different mixture components may be optimal ICs under different
scenarios; for example, mixture components A and B may both be identified as candidate ICs in
general, however, candidate A may be selected as the IC if it has a more robust evidence base for
a specific application of interest (e.g., oral/sub chronic duration). In the RPF approach, the
assumption under dose additivity is that the toxicity of each mixture component chemical
induces effects via a similar pathway of biological perturbation and can operationally be
considered a fixed concentration or dilution of the IC (EPA, 2000b). Mathematically, when using
response-specific doses, the RPF is the ratio of the IC to that of each individual mixture
component chemical (j) at a common point on the corresponding dose-response curves (e.g.,
human equivalent LOAELs, BMDs, or EDx). Ideally, the dose-response functions used to
calculate RPFs across mixture components would be approximately the same in exposure
duration and study design (e.g., sex, species, life stage). Further, considering the known
differences in TK characteristics across PFAS (e.g., internal plasma half-life) between rodents,
non-human primates, and humans, it is advisable to convert experimental animal dose-response
data to human equivalents where possible before calculating RPFs. Lastly, of the options for
dose-response metrics to use in the calculation of RPFs across mixture PFAS, BMDs (e.g., the
central tendency estimate) would be optimal. BMDs incorporate the totality of a given dose-
response and facilitate identification of a dose at a pre-defined BMR level (e.g., 0.5 SD or 1 SD
over control; 10% change in some effect/endpoint). BMD modeling would optimize comparison
of "same" as a function of dose across mixture PFAS for a given health effect or endpoint. It is
recognized that dose-response data for chemicals are sometimes not amenable to BMD
modeling. Human equivalent LOAELs or EDx values are suitable alternatives. No matter which
dose-response metric is used, the RPF for the IC is always one. The potency ratio can be
calculated for each mixture component chemical (j) as the ratio of the effect doses as shown in
Equation 6-1:
72
-------
PUBLIC REVIEW DRAFT
MARCH 2023
RPF. = /L
1 ED10,
(Eqn. 6-1)
where IC refers to the index chemical.
For example, if mixture component chemical 2 is twice as potent as the IC, its LOAEL, BMDx,
or EDx will be half as large and the calculated RPF would be a 2. Conversely, if mixture
component chemical 2 is half as potent as the IC, its LOAEL, BMDx, or EDx will be twice as
large and the RPF would be 0.5. In practice, EPA determines a single RPF for the response range
or dose range of interest. When data are available, RPFs can potentially be determined for more
than one health effect domain and/or exposure scenario (e.g., developmental versus thyroid
toxicity, shorter-term vs. chronic exposure, oral vs. inhalation exposure). As illustrated in the
RPF examples in the next section, that flexibility or scenario specificity is an advantage of the
general RPF approach. Once RPFs are calculated for each mixture component chemical using a
common metric in Equation 6-1, ICECs are then calculated by multiplying each respective RPF,
by the corresponding component chemical's concentration (d7), as shown in Equation 6-2:
The total mixture ICEC (ICECmix) is then obtained by taking the sum of the component
chemical ICECs (including that of the IC) (Equation 6-3). A numerical estimate of risk for
noncancer health effects associated with exposure to the mixture of concern is then obtained by
mapping the ICECmix onto the dose-response function for the IC. For example, if the IC's
dose-response model is denoted f(d), then the RPF-based response to the mixture is estimated as:
where the ICEC is derived from Equation 6-3. In the context of this PFAS mixture framework,
there are important modifications or adaptations of this approach to note that include: (1) use of
ICECs, which are water-specific, correlates to index chemical equivalent doses (ICEDs) (EPA,
2000b) and (2) using effect-specific HBWCs for the IC (e.g., 70 ppt for PFOS-induced
developmental effects (decreased body weight in offspring)) as a benchmark point to compare a
mixture ICEC to rather than directly mapping the mixture ICEC onto the IC dose-response. This
serves the purpose of providing the end-user a basic indication of "yes," there is potential effect-
specific risk associated with the mixture (e.g., ICECmix > IC HBWC), or "no," there is no
anticipated effect-specific risk (e.g., ICECmix < IC HBWC), as well as magnitude of health effect
concern and identification of potential component chemical drivers of an ICEC.
EPA's supplementary guidance (EPA, 2000b) states: "The common mode-of-action assumption
can be met using a surrogate of toxicological similarity, but for specific conditions (endpoint,
route, duration)." This suggests that although the common MOA metric for application of RPFs
is optimal, there is flexibility in the level of biological organization at which "similarity" can be
determined among mixture components. To date, EPA has developed RPFs for only a few
chemical groups, largely pesticides (organophosphorus pesticides, triazines, N-methyl
carbamates, chloroacetanilides, and pyrethrins/pyrethroids), which in each case were based on
MOA-level information (EPA, 2018). However, MOA data are limited or not available for many
PFAS. As such, in the interim, when using the RPF approach, it is advisable to focus the
ICECmix = 2"=i dj* RPFj
(Eqn 6-2)
Ymix — f (JCECmix)
(Eqn 6-3)
73
-------
PUBLIC REVIEW DRAFT
MARCH 2023
biological level of organization for component-based evaluation of potential mixtures additivity
for PFAS on similarity in toxicity endpoint/effect. Further, as empirically demonstrated by
Conley et al. (2022b), due to potential variability of potency for health effects across PFAS,
RPFs can vary by more than an order of magnitude. Thus, where possible, it is preferable for a
given PFAS mixture to evaluate multiple common effect domains or endpoints, where and when
dose-response data are available, to identify the most sensitive endpoint for evaluation of risk.
Using the most sensitive endpoint(s) for the RPF analysis helps to ensure that risks are not
underestimated, and, providing a landscape of candidate RPFs across PFAS and health effects
ensures transparent communication of mixtures risk assessment for decision-making. This is the
approach taken in the illustrative RPF examples below and is consistent with previous NAS
recommendations pertaining to the evaluation of chemicals that cause common adverse health
outcomes presumably through diverse biological pathways (NRC, 2008).
6.2 Illustrative Example Application of RPF to a Hypothetical Mixture of Five
PFAS
The example application of the RPF approach incorporates hazard and dose-response
information for the hypothetical five PFAS mixture presented in the HI sections above.
However, in this context only dose-response data for like/similar health effect(s) are needed.
Recall that PFAS 1-3 have existing hazard and dose-response data that have been formally
evaluated for human health risk assessment purposes; these three PFAS also have existing
HBWCs. PFAS 4 has not undergone risk assessment but has existing experimental animal assay
data. Lastly, PFAS 5 is data-poor with only physicochemical, TK, and in vitro cell-based
bioactivity data. This example focuses on development of RPFs for liver, thyroid, and
developmental effects only (Figure 4-8), which have been reported as toxicity targets of several
compounds within the broader class of PFAS (EPA, 2021a,b; ATSDR, 2021; EFSA et al., 2020;
Section 7.1 in ITRC, 2022). The approach here is to use a construct that allows for combination
of PFAS with shared, common health outcome (e.g., such as delayed growth and development in
offspring), as opposed to a stringent requirement of same MO A, to calculate RPFs across one or
more health effect domains. Inclusion of multiple effects/domains among the constellation of
PFAS effects allows for evaluation of the potential impact of differences in RPFs across PFAS in
the mixture for those effects (e.g., the potency of PFAS 1 relative to PFAS 2 may be different for
effects on the liver as compared to effects on the thyroid) (Mumtaz et al., 2021).
The intention is not necessarily to seek the most sensitive effects/domains; rather, it is to
optimize identification of those that are shared among the PFAS in the mixture being assessed.
However, for purposes of evaluating mixture risk using the RPF approach in a specific
environmental medium (e.g., drinking water), it is critical to have an IC effect-specific value or
metric (e.g., an HBWC) so that the mixture ICEC can be compared to a benchmark point. For
PFAS, given the limited availability of hazard effect and dose-response data, if one seeks to
include several PFAS (i.e., beyond those few congeners with robust toxicity databases) the
approach may be limited to a single effect domain, or only those endpoints for which reasonable
estimation of dose-response metrics (e.g., PODs, EDx) for "same/similar" is possible. However,
leveraging available NAM data, such as in vitro cell bioactivity, may provide opportunities to
integrate those PFAS with poor(er) hazard and dose-response databases.
74
-------
PUBLIC REVIEW DRAFT
MARCH 2023
6.2.1 RPF Step 1: Assemble/Derive component health effects endpoints (select ICs,
PODheds)
As PFAS 1-3 are toxicologically well-characterized and have existing HBWCs, all three are
identified as candidate ICs for the mixture. PFAS 4 is also reasonably well characterized
toxicologically and might be considered as a candidate IC in some RPF contexts, however in a
drinking water specific application another key consideration for IC selection is the existence of
a quantitative benchmark such as a HBWC. This is necessary such that the index chemical
equivalent concentration for the mixture (ICECmix) can be compared to the IC's corresponding
HBWC to determine potential for health risk(s). As such, PFAS 1-3 are the only candidate ICs
identified for the hypothetical five component mixture. Based on the strength of toxicological
evidence (see figure below; note, this is a repeat of Table 4-8), not necessarily the quantitative
potency for effect, ICs were selected as follows: Liver IC = PFAS 2; Thyroid IC = PFAS 3; and
Developmental IC = PFAS 1.
Effect domain PFAS 1 PFAS 2
Developmental .j_f
The dose-response metrics for this RPF example are the same as those used above in the HI
example (Table 5-3). The PODheds for three effect domains used in the calculation of the effect-
specific RPFs and corresponding ICECs are presented below (Table 6-1).
Table 6-1. Summary of PODheds for three selected health effect domains for a mixture of
five hypothetical PFAS.
Liver PODhed
(mg/kg-day)
Thyroid PODhed
(mg/kg-day)
Developmental PODhed
(mg/kg-day)
PFAS 1
0.044
0.24
0.00001 (BMDLo.ierhed)
PFAS 2
0.0013 (BMDLiohed)
0.23
0.0051
PFAS 3
N/A
0.21 (BMDLisdhed)
2.1
PFAS 4
50
N/A
0.0011 (BMDLisdhed)
PFAS 5
0.004 (BMDisdaed)
N/A
N/A
Note:
Bolded numbers represent those PODs used in the derivation of corresponding oral RfVs.
6.2.2 RPF Step 2: Assemble/derive health-based media concentrations (HBWCs for the
Index Chemicals)
For this illustrative RPF example, the HBWCs are the same as those used above in the General
HI example (Table 5-5). Specifically, the PFAS 1 HBWC is 0.2 ng/L (IC for developmental
PFAS 3
PFAS 4
¦BBHi
PFAS 5
*
75
-------
PUBLIC REVIEW DRAFT
MARCH 2023
effects), PFAS 2 HBWC is 60 ng/L (IC for liver effects), and PFAS 3 HBWC is 4,000 ng/L (IC
for thyroid effects).
6.2.3 RPF Step 3: Select exposure estimates (measured water concentrations)
Select appropriate exposure estimates consistent with the problem formulation. Specifically, the
user may choose to calculate or use exposure estimates that are for the oral route in general (i.e.,
total intake in mg/kg-day) or media-specific concentrations. In the hypothetical PFAS mixture
example, 'exposure' is represented by the drinking water monitoring data in Table 4-1.
6.2.4 RPF Step 4: Calculate PFAS mixture potency (RPFs and ICECsfor each effect
domainJ
Liver: Available traditional animal assay data indicate liver effects for PFAS 1, 2, and 4. PFAS 5
has only bioactivity data however the molecular and cellular perturbations were observed
primarily in hepatocyte cell cultures (e.g., HepG2; HepaRG). As such, there is an opportunity to
integrate NAM-based information into the RPF approach specifically for the liver effect domain.
Across the landscape of experimental rodent studies that inform liver toxicity for hypothetical
PFAS 1, 2 and 4, several effects were noted after oral exposures such as increased absolute and
relative organ weights, increased incidence of macro- and microvesicular steatosis (i.e., lipid
accumulation in hepatocytes), histopathological evidence of focal hepatocellular necrosis, and
increased serum ALT, AST, and ALP, indicative of hepatocyte or biliary epithelium injury,
respectively. In addition, in vitro cell bioactivity data for PFAS 2 and 5 indicate increased pro-
oxidation/oxidative stress, mitochondrial stress, and altered lipid homeostasis in the lower tested
concentration range. Many of these observed cellular effects are considered KEs in signal
transduction pathways leading to liver tissue alteration and injury (Figure 6-1). Of the effects
observed in experimental rodents across PFAS 1, 2, and 4, histopathological evidence of
significantly increased incidence of hepatocellular death was common across studies. Further,
this effect was the basis for the derivation of an oral RfD and corresponding HBWC for PFAS 2.
As such, increased incidence of hepatocellular death is identified as the common effect for the
liver domain for PFAS 1, 2 and 4. The liver effect-specific RPFs are calculated by dividing the
selected liver effect PODhed for the IC PFAS 2 by the PODhed for PFAS 1 and 4 for the same
effect (Table 6-2). Each RPF is multiplied by the corresponding chemical-specific measured
water concentration to derive a PFAS 2 ICEC (Table 6-2). The example Mixture Total PFAS 2
IC EC mix is then compared to the HBWC for PFAS 2, which is based on the effect of increased
incidence of hepatocellular death.
For PFAS 5, the dose-response used in this specific example RPF application, obtained from
IVIVE/rTK of the in vitro cell bioactivity data, is based on the lowest bioactivity event21
associated with the IC; that is, noncancer bioactivity at the lower end of the distribution for the
IC is the driver for identification of "like" effect for the data-poor mixture component PFAS. In
this hypothetical example, the bioactivity for the IC (PFAS 2) in HepaRG cells was decreased
mitochondrial cytochrome c oxidase activity (Table 4-3). Importantly, cytochrome c oxidase
activity is a key component in proper mitochondrial respiration and function; disruption of this
mitochondrial enzyme has been associated with increased oxidative stress, decreased ATP
production, and cell death. Surveying the landscape of available cell bioactivity data for PFAS 5
21 The "lowest" bioactivity for noncancer application purposes should not be a potential carcinogenic event (e.g.,
mutagenicity or clastogenicity).
76
-------
PUBLIC REVIEW DRAFT
MARCH 2023
revealed that the same effect occurred in hepatocytes, although it was not the most sensitive
perturbation for PFAS 5. The objective of this NAM-based approach is to scale the potency of
the selected bioactive event for the data-poor chemical(s) to the same/similar bioactive event for
the IC, where or when available data allow. The NAM-based RPF (RPFnam) is calculated by
taking the ratio of the BMDaedso for the selected bioactivity event of the IC to the BMDaedso for
the same event associated with the data-poor mixture component chemical. The rationale for
using the BMDaedso is that the quantitative relationship between a KE and an adverse health
outcome is typically unknown. As such, in NAM practice it is common to default to a
50th percentile for comparative biology purposes. The resulting RPFnam represents the relative
potency between the data poor PFAS (PFAS 5) and the IC (PFAS 2) for the selected bioactive
event. This RPFnam is then multiplied by the data-poor chemical (e.g., PFAS 5) exposure metric
(e.g., measured water concentration) to obtain a NAM-based ICEC (ICECnam); to convert the
ICECnam to a mixture ICEC that comports with the other traditional assay-based component
PFAS ICECs, the ICECnam is multiplied by the ratio of BMDx-hed for the critical effect of the
IC (in this example, the BMDiohed for increased incidence of hepatocellular death) to the
BMDaedso for the bioactive event of the IC. The resulting ICEC represents the estimated
contribution of PFAS 5 to the overall risk of the liver-specific effect, but represented as a dose
scaled for potency, relative to the IC, across different levels of biological organization (i.e.,
PFAS 5 in vitro to PFAS 2 in vivo). This process is illustrated in Figure 6-2.
77
-------
PUBLIC REVIEW DRAFT
'!¦=! ¦¦ ii . .1: • ¦ " " >'!i' '
••"I11 •
fesgs 4 Woyf>^ I®? Eft
Figure 6-1. General cell signaling pathways associated with PFAS-induced liver injury.
78
-------
PUBLIC REVIEW DRAFT FEBRUARY 2023
Bioactivity for Index Chemical
Bioactivity for 'data-poor' chemical
:: ¦ ¦- V
Data-poormember Modeled BMDa;-j5l -
RPFnam = Selected Bioactivity BMDflmmfor 1C
Selected Bioactivity BMDAED50for data-poor component
1CECnam = RPFnamx Data-poor component exposure level {e.g., DW cone.)
1CEC = ICECnam x BMDx.HEDfor IC critical effect / Selected Bioactivity BMDAED50for 1C
Figure 6-2. Proposed process for integrating NAM-based RPFs and ICECs into mixtures
assessment.
79
-------
PUBLIC REVIEW DRAFT FEBRUARY 2023
Table 6-2. Example Liver Effect RPFs and ICECs for a Hypothetical Mixture of Five PFAS
PODhed (mg/kg-day);
Hypothetical
Mixture Increased incidence of
Exposure Estimate
PFAS 2ICEC
Component hepatocellular death
Example RPF
(ng/L)
(ng/L)
PFAS 1 0.044 (BMDLiohed)
0.03
4.8
0.1
PFAS 2 (IC) 0.0013 (BMDLiohed)
1
52
52
PFAS 3 N/A
N/A
172
—
PFAS 4 50
0.00003
58
0.002
PFAS 5 0.037 (BMDaedso)"
1.4 (RPFnam)15
69
24°
Mixture Total PFAS 2 ICEC (ppt)
76
Notes:
a The BMD was modeled from the AED-based dose-response for the selected bioactivity event (e.g., decreased mitochondrial
cytochrome c oxidase activity) at a BMR of 50% over control.
b RPFnam for PFAS 5 was calculated as the ratio of the BMDaedso for PFAS 2 (IC) / BMDaedso for PFAS 5; in this example
application, 0.0052 mg/kg-day / 0.0037 mg/kg-day = 1.4.
c The ICEC for PFAS 5 was calculated by first deriving the ICECnam as follows: RPFnam x Exposure estimate for PFAS
5 = 1.4 x 69 (ng/L) = 97 ng/L; the ICECnam was then multiplied by the ratio of the BMDLiohed for PFAS 2 / BMDaedso for
PFAS 2 = 97 ng/L x (0.0013/0.0052) = 24 ng/L.
Thyroid: Available traditional animal assay data indicate thyroid effects for PFAS 1, 2, and 3.
PFAS 4 and 5 have no data available to support inclusion in the RPF analysis for this health
effect domain. Across the landscape of experimental rodent studies that inform thyroid toxicity
for hypothetical PFAS 1, 2 and 3, two effects were noted after oral exposures including
decreased levels of total and free thyroxine (T4) and triiodothyronine (T3) and increased
absolute and/or relative thyroid weight. Importantly, effects on thyroid hormone levels were
observed across sexes and life stages (e.g., newborn and adult rats and mice) and was a common
observation across PFAS 1-3. Further, decreased thyroid hormone levels, particularly total and
free T4 (TT4 and FT4, respectively) in females, was the basis for derivation of the oral RfD and
the corresponding HBWC for PFAS 3. As such, decreased TT4 andFT4 is identified as the
common effect for the thyroid domain for PFAS 1-3. The thyroid effect-specific RPFs are
calculated by dividing the selected thyroid effect PODhed for the IC PFAS 3 by the PODhed for
PFAS 1 and 2 for the same effect (Table 6-3). Each RPF is multiplied by the corresponding
chemical-specific measured water concentration to derive a PFAS 3 ICEC (Table 6-3). The
example Mixture Total PFAS 3 ICEC is then compared to the HBWC for PFAS 3, which is
based on the effect of decreased TT4 and FT4. The calculation of the thyroid-specific RPFs and
corresponding ICECs are presented in Table 6-3.
80
-------
PUBLIC REVIEW DRAFT FEBRUARY 2023
Table 6-3. Example Thyroid Effect RPFs and ICECs for a Hypothetical Mixture of Five
PFAS
Mixture
Component
PODhed (mg/kg-day);
Decreased TT4 and FT4
Example
RPF
Hypothetical
Exposure Estimate
(ng/L)
PFAS 3ICEC
(ng/L)
PFAS 1
0.24 (BMDLisdhed)
0.9
4.8
4
PFAS 2
0.23 (BMDLisdhed)
0.9
52
47
PFAS 3 (IC)
0.21 (BMDLisdhed)
1
172
172
PFAS 4
N/A
N/A
58
—
PFAS 5
N/A
N/A
69
—
Mixture Total PFAS 3 ICEC (ppt)
223
Developmental: Developmental effects associated with oral exposures to PFAS 1, 2, 3, or 4
were observed in rats and mice; the studies available were predominately single-generation
reproductive-developmental design however PFAS 4 also had a two-generation study in rats.
PFAS 5 had no studies/data to suggest effects in the developmental domain. The landscape of
developmental effects in newborns or neonates for PFAS 1-4 was broad and included decreased
body weight at PND 0, delayed growth and development at PND 14 (e.g., reduced body weight,
delayed eye opening and vaginal patency), decreased thyroid hormones at PND 0, and reduced
ossification of phalanges. However, decreased body weight in offspring at birth was common
across PFAS 1-4; further this effect was used as the basis for the derivation of the oral RfD and
corresponding HBWC for the IC, PFAS 1. As such, decreased body weight in offspring was
selected as the common developmental effect for the purposes of this RPF illustrative example.
The developmental effect-specific RPFs are calculated by dividing the PODhed for the selected
effect associated with the IC PFAS 1 by the PODhed for PFAS 2, 3, and 4 for the same effect
(Table 6-4). Each RPF is multiplied by the corresponding chemical-specific measured water
concentration to derive a PFAS 1 ICEC (Table 6-4).
Table 6-4. Example Developmental Effect RPFs and ICECs for a Hypothetical Mixture of
Five PFAS
PODhed (mg/kg-day);
Hypothetical
Mixture
Decreased Body Weight
Exposure Estimate
PFAS1ICEC
Component
in Offspring
Example RPF
(ng/L)
(ng/L)
PFAS 1 (IC)
0.00001 (BMDLisdhed)
1
4.8
5
PFAS 2
0.0051 (BMDLisdhed)
0.002
52
0.1
PFAS 3
2.1 (BMDLisdhed)
5 E-6
172
0.0009
PFAS 4
0.0011 (BMDLisdhed)
0.009
58
0.5
PFAS 5
N/A
N/A
69
—
Mixture Total PFAS 1 ICEC (ppt)
6
81
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
6.2.5 RPF Step 5: Compare PFAS mixture potency (Total ICECmix) to existing health-
based value (HBWC)
In the liver-specific RPF application (Table 6-2), the health risk(s) associated with the mixture is
represented by comparing the PFAS 2 ICECmix to the IC HBWC which is based on the specified
effect for that hazard domain (e.g., for this example, increased incidence of hepatocellular death).
In this hypothetical example, the PFAS 2 ICECmix of 76 ppt exceeds the PFAS 2 HBWC of
60 ppt, indicating potential for risk of liver effects in individuals or populations exposed to a
mixture of the five PFAS at the hypothetical water exposure estimates provided. Importantly,
PFAS 2 and 5 appear to be drivers for the liver health risk associated with the hypothetical
mixture.
In the thyroid-specific RPF application (Table 6-3), the health risk(s) associated with the mixture
is represented by comparing the mixture total PFAS 3 ICECmix to the IC HBWC which is based
on the specified effect for hazard domain (e.g., for this example, decreased TT4 and FT4). In this
hypothetical example, the PFAS 3 ICECmix of 223 ppt is far below the PFAS 3 HBWC of
4000 ppt, indicating no apparent risk of thyroid effects in exposed individuals or populations to a
mixture of the five PFAS at the hypothetical water exposure estimates provided.
In the developmental effect-specific RPF application (Table 6-4), the health risk(s) associated
with the mixture is represented by comparing the mixture total PFAS 1 ICECmix to the IC
HBWC which is based on the specified effect for hazard domain (e.g., for this example,
decreased body weight in offspring). In this hypothetical example, the PFAS 1 ICECmix of 6 ppt
exceeds the PFAS 1 HBWC of 0.2 ppt by over an order of magnitude, indicating significant
potential for health risks in developmental populations exposed to a mixture of the five PFAS at
the hypothetical water exposure estimates provided.
As illustrated in the RPF examples above, PFAS can have different potencies across health effect
domains. Due to differences in both TK and TD, PFAS may exhibit complex gradations of
potency for different effects, and this will be reflected in the corresponding RPFs. Some PFAS
may be exquisitely potent for some effects and yet virtually inactive in others, however
expanding the number of PFAS and the toxicity endpoint profiles across the structural landscape
will be key to illustrating such a diversity in relative potency(ies). Thus, calculating RPFs for as
many endpoints/effects as possible helps to ensure that subsequent PFAS risk management
strategies are health protective. In the example above, risk would have been underestimated if the
RPF analysis was limited to liver and thyroid effects: developmental effects are the risk driver in
this scenario. Further, another critical consideration illustrated in the RPF examples is the impact
of component chemical concentrations. That is, in practical field application, PFAS
concentrations in water, soil, or air may be drastically different dependent on a number of factors
(e.g., different physicochemical and environmental fate and transport properties; proximity to
PFAS manufacturing or use locales; water sources (well water vs. finished drinking water);
waste handling; temporal and spatial variability). In application, transparent presentation and
communication of hazard and dose-response data sources, RPFs, media concentrations, ICECs,
and any associated uncertainties, across as many health effect domains as is practicable is ideal
for RPF-based evaluation of PFAS mixtures. As mentioned previously, a limitation for PFAS is
the availability of human health assessment grade toxicity data; Section 7 offers an alternative to
the RPF approach in such a scenario.
82
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
6.3 Advantages and Challenges of the Relative Potency Factor Approach
An advantage of the RPF approach is that formal toxicity or RfV derivation is not necessary for
the component chemicals. Rather, only effects/endpoints and associated dose-response metrics
(e.g., NOAEL, BMDx, EDx) are needed to perform the exercise. While it would be ideal to
conduct potency comparisons between mixture components for same effect/endpoint using same
dose metrics from same study design/durations, calculation of RPFs across PFAS may in
practical application entail or necessitate use of effect data deriving from diverse study designs
and exposure durations. As such, in some cases, there may be a need to normalize or adjust
available quantitative metrics such that potency comparisons are at comparable points on a given
dose-response. For example, there might be a need to selectively apply UFs in the RPF method,
in particular, the LOAEL-to-NOAEL (UFl) and/or subchronic-to-chronic duration (UFs) factors
to convert quantitative metrics to NOAELs from an estimated chronic-duration exposure. This
flexibility is needed as in some cases, effect data for mixture component PFAS may come from a
variety of study designs such as reproductive/developmental in mice (e.g., GDs 1-20), less than
lifetime repeat-dose (e.g., 28- or 90-days) in rats, and/or 2-year bioassays in rats. For the
expressed purpose of deriving RPFs, applying a UFs of 10 to convert a subchronic NOAELhed
to a corresponding chronic NOAELhed, or, converting a LOAELhed to a NOAELhed, provides
the opportunity for a more 1:1 comparison of potency for a given effect (e.g., developmental
body weight, increase in liver weight) among component PFAS. A critical facet to this is to be
transparent about such POD adjustments (i.e., purpose/rationale) when applied.
RPFs were generally intended for use when mixture components are demonstrated to have
similar/same MOA. This presents a problem as it pertains to practical application of RPF
methodology in that a vast majority of environmental chemicals, including PFAS, have limited-
to-no MOA data available. EPA mixtures guidance does provide flexibility in use of data from
different levels of biological organization in dose additive approaches such as RPF. As
demonstrated in this framework document, this flexibility is an advantage in that there is greater
probability of identifying effect/endpoint and associated dose-response data (e.g., effect-specific
PODs) for mixture components than there is for MOA type data. However, as the data for PFAS
evolve over time, the toxicity profiles including number of effect types and granularity of
biological perturbations (e.g., potential KE data that inform proposed MOA(s)) may eventually
support MOA-based evaluations.
Another advantage is that the RPF method facilitates calculation of an actual mixture toxicity
dose or concentration estimate, as opposed to the HI which is considered an indicator of potential
hazard/toxicity. Although a given mixture ICEC is traditionally mapped to the IC's effect-
specific dose-response function to arrive at a corresponding "mixture response," an advantage of
the RPF approach is that the mixture ICEC may alternatively be used to inform mixture risk in
the context of the relationship to a media-specific health-based value (such as a HBWC).
A clear challenge, not uniquely associated with the RPF approach, is the use of potentially
disparate hazard and dose-response data across mixture components. The implicit assumption for
dose-response data selection in the calculation of RPFs in this framework, is that the same dose-
response data that underpinned the derivation of corresponding RfVs (overall RfDs or organ-
specific TTDs) for use as input(s) for HQs and His would also be leveraged in RPF and/or M-
BMD approaches (see Sections 6 and 7)). However, although ideal, this is not an expressed
requirement of the framework. The user should be afforded the flexibility to make decisions
83
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
regarding suitable dose-response selection for RPF calculations on a case-by-case basis. Key to
this flexibility is transparent characterization and communication of literature searching strategy
and review results, hazard data selection, dose-response evaluation (e.g., BMD preferred, effect
levels such as NOAELs are acceptable), and qualitative and quantitative uncertainties or
confidence in what could potentially be a diverse assembly of data/metrics to support RPF
application(s).
Another challenge is that depending on data availability for the component PFAS, the effect
domains used for the RPF analysis may not be the overall most sensitive out of the total
constellation of common PFAS effects. In the RPF examples shown above, risk is indicated
based on the liver and developmental RPFs, but not for the thyroid effect domain. To effectively
use the RPF approach, the user needs effect data for at least one common endpoint among the
effects for all component PFAS in the mixture. Ideally this would include the most sensitive
effect across PFAS in the mixture of interest in order to provide a conservative (health
protective) risk-based scenario.
An additional potential challenge, that actually may present an opportunity for advancing the
science of mixtures risk assessment is the use of NAM data. The constantly evolving information
coming from alternative toxicity testing assays and platforms may be of paramount importance
to human health assessment of environmental chemicals in general (not just for mixtures
applications), however there are inherent challenges associated with application to hazard
identification and dose-response assessment. In a PFAS mixtures assessment context, for some
mixture component chemicals, NAM data (e.g., read-across or cell-based bioactivity (such as
ToxCast and/or Tox21)) might be the only source(s) of evidence available to inform an RPF
approach. The challenge might then be identifying and assembling "same" or "similar"
effect/endpoint data compared to other PFAS in the mixture that have human epidemiological
and/or experimental animal (i.e., apical (phenotypic) effect level) bioassay data. While the RPF
approach affords flexibility in selection of "effect" data, a key requirement is that the "effect" on
which RPFs are based be the same. For example, one mixture PFAS may have histopathological
evidence of multi-focal liver necrosis from in vivo repeat-dose rat studies, whereas another PFAS
may have evidence of cytochrome c release, mitochondrial damage, and cell death in in vitro rat
hepatocyte cell culture studies only. While in this hypothetical example NAM data clearly
demonstrate hallmarks of cellular demise typically associated with necrotic (and apoptotic) cell
death, pathologically consistent with cell death foci observed in whole rat liver, it may be
difficult to make the case that the in vitro-based concentration-response data (converted to an
AED) is suitable for traditional RPF calculation simply based on the interpretation of "same"
effect. Further investigation is needed to evaluate the qualitative and quantitative merits of
applying hazard and dose-response data from across different levels of biological organization in
a component-based mixtures assessment context. This is particularly true of NAMs where the
possible lumping or splitting of assay/data types to inform an integrated or more individualized
interpretation of hazard and dose-response for data-poor mixture components is in its infancy;
case studies using validated NAM assays and data are needed to help optimize application in
mixtures risk assessment.
84
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
7.0 Mixture-BMD Approach
7.1 Background on the Mixture-BMD Approach
Given the broad range of PFAA congeners and structural diversity across the PFAS class, it is
likely that for some effects used for mixture assessment the dose response functions (i.e., slopes)
will be dissimilar across component chemicals. The use of an IC in the RPF approach assumes
component chemicals have congruent dose response slopes for same health effect (and/or MO A).
In addition, the HI approach requires human health assessment values, such as oral RfDs and
individual HBWCs, because these metrics serve as the denominator in determining if the
exposure exceeds a level estimated to be acceptable for human intake. In some cases, a PFAS
mixture may contain component chemicals that do not have congruent dose response curves or
have available human health assessment values (e.g., RfDs). In these cases, a third approach,
called the M-BMD, can be used to estimate health risk(s) associated with mixture exposure. This
approach is described in EPA's supplementary guidance (2000b) (Section 4.2.6) and NRC
(2008) (Appendix C) and employs a DA model-based calculation of a total M-BMD that
corresponds to a defined BMR (e.g., BMDio) for a PFAS mixture. Similar to the RPF approach,
only effects/endpoints and associated dose-response metrics (e.g., BMDx) are needed to perform
the exercise. Further, the mixture evaluation is based on a similar toxicological effect for
component chemicals and the equation provided can be used to define a single point estimate
(e.g., POD) or derive a full dose response curve for the PFAS mixture of interest.
Because RPFs are special applications of the DA concept, such approaches can be a
straightforward way of making quantitative assessments of the effects of chemicals, including
PFAS. However, application of the RPF concept requires congruent dose-response curves for all
component chemicals for the given effect. When this requirement is not met, RPFs will vary with
the effect levels in the mixture and thus could not be considered a "dilution" of the IC across the
full dose response range. For example, the RPF for a given PFAS may be different in the low
dose range vs. the middle or high dose range depending on slope differences with the IC. In this
regard, published data (Hass et al., 2007; Howdeshell et al., 2008; Metzdorff et al., 2007; and
Rider et al., 2008) reveal that chemical dose-response curves for a common effect can display
very different slopes and shapes even across related structures within a given class. In contrast to
the RPF approach, other DA-based equations can be used for quantitative evaluations of the
effects of chemical mixtures when the slopes for a common effect differ among chemicals in the
mixture. As stated in the NRC (2008) report, "It is a widely held misconception (EPA 2000(b))
that dose addition is applicable only with congruent dose response curves (for a general
discussion, see Gennings et al. 2005 and Kortenkamp et al. 2007)."
The following discussion compares the predictions of two DA models, one assumes that the
individual chemicals in the mixture have congruent slopes, whereas the second DA model
(described by NRC (2008)) does not require similar slopes to yield accurate predictions. As an
example, these two models were recently used to predict the full dose-response curve of a
mixture using the EDsos and slopes of each component chemical in the mixture. The first model,
similar to the RPF method (discussed above), assumes that the chemicals in the mixture have
equal slopes, indicating that the relative potencies are constant across the entire dose response
curve. Equation 7-1 is an example of such a model that uses an average slope value to calculate
the joint toxicity of a mixture with the following equation (Olmstead and LeBlanc, 2005; Rider
and LeBlanc, 2005; Rider et al. 2008):
85
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
R
1
(Eqn 7-1)
( " D Y
v z=i ED50i j
where R is the response to the mixture, A is the dose of chemical i in the mixture, HI)50, is the
dose of chemical i that causes a 50% response, and p is the average power (Hillslope) associated
with the chemicals.
Because the assumption of similar slopes is not always met, DA models, like the M-BMD
method described below, that do not require parallel slopes for the chemicals in the mixture
provide more accurate predictions of the mixture effects. Several of these DA-based models have
been previously described (for example, Altenburger et al., 2000; Kortenkamp et al., 2007;
Metzdorff et al., 2007) and by NRC (NRC 2008). The M-BMD equation below (Equation 7-2) is
an example of such a DA model that calculates a given effect level for the total mixture
(.EDxmixture) where p, is the proportion of chemical i in the mixture and EDxi denotes the dose
producing the given level of response for the z'th chemical in the mixture:
When the slopes of the dose response curves differ among chemicals in the mixture, the two DA
models (i.e., Equation 7-1 and Equation 7-2) can yield different dose response predictions (see
Figure 7-1). Further, there is greater uncertainty in the accuracy of the DA mixture predictions
when the assumption of parallelism among slopes is violated (NRC, 2008). The following
example compares the accuracy of a DA model that assumes parallelism (Equation 7-1) with a
DA model that does not (Equation 7-2), and these model predictions are compared with a RA
model. The data used in this example are from a published binary mixture study, but the
chemicals are not identified (Gray et al., 2022).
In this example two chemicals were mixed using a fixed-ratio design. The top dose of this
mixture contained each chemical at a dose close to their respective EDsos, but they have different
dose response shapes using nonlinear 4 parameter logistic regression (Chemical A slope = 40,
Chemical B slope = 5 using linear Y axis and loglO X axis). The mixture effect described in the
figure below is percent reduction in a reproductive organ weight, ranging from 0% reduction in
the control (0 dose of the mixture) to complete agenesis (100% reduced).
EDXmixture (pi/EDxi + P2/EDX2)'1
(Eqn 7-2)
86
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Reproductive Organ Weight Reduction
s
G
S
£
LU
£
LU
Q.
75-
50-
25-
0-
1 1 1 1 1—
0 25 50 75 100
Percent of Top Dose of the
Two Chemicals in the Mixture
O OBSERVED
^ Dose Addition (Model
assumes equal slopes)
Dose Addition (Model
— does not assume equal
slopes
^ Response Addition
Figure 7-1. Example comparison of observed data with model predictions using two dose
addition-based mixture models and a response addition model for a binary mixture study
(adapted from Gray et al., 2022). The two chemicals displayed individual dose response
curves with widely disparate slopes for the endpoint (reduced organ weight). The two dose
addition models either assume component chemicals have similar dose response slopes (red
solid line) for the effect or have non-congruent dose response curve slopes (black dashed
line). For these chemicals with disparate slopes the dose addition model that does not
assume equal slopes provided a better fit of the observed data (see table below).
The observed data were fit with the model parameters of the two DA and the RA model and
Akaike Information Criteria (AIC) values were calculated to determine the best model of the
observed data (Table 7-1). The lower AIC values indicate a better-fit model, and as a "rule of
thumb" (Burnham and Anderson, 2004) there is little support for two of these models because
the delta-AIC (the difference between the two AIC values being compared) is greater than 7.
Table 7-1. "Best Model" based upon AIC values
"Best Model" based upon AIC values
Model
AIC
DA - does not assume equal slopes
164.1
DA - assumes equal slopes
203.3
RA
223.4
AIC: Akaike Information Criterion
In contrast to the above example of chemicals with disparate slopes, if the dose response curves
of the component chemicals in a mixture have similar slope parameters from nonlinear
regression, then there would be little or no difference between the predictions of the two DA
87
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
models shown here. Hence, if sufficient dose response information is available and the slopes are
not parallel then it is preferable to model the data with the M-BMD equation (Equation 7-2) that
does not assume parallelism, as stated by the NRC (2008).
Estimating PFAS mixture effects using the M-BMD method requires empirical data-driven or
reasonable estimation (e.g., read-across between structures) of effect-equivalent endpoints for all
PFAS in the mixture. Similar to the RPF approach above, ideally the dose-response functions
used to calculate effect endpoints (e.g., BMDx) across mixture components would be
approximately the same in exposure duration and study design (e.g., sex, species, life stage).
Further, considering the known differences in TK characteristics across PFAS (e.g., internal
plasma half-life) between rodents, non-human primates, and humans, it is strongly recommended
to convert experimental animal dose-response data to human equivalents where possible. Lastly,
of the options for dose-response metrics to use across mixture PFAS, BMDs (e.g., the central
tendency estimate) would be optimal. BMDs incorporate the totality of a given dose-response
and facilitate identification of a dose at a pre-defined BMR level (e.g., 0.5SD or 1SD over
control; 10% change in some effect/endpoint). BMD modeling would optimize comparison of
"same" as a function of dose across mixture PFAS for a given health effect or endpoint. It is
recognized that dose-response data for chemicals is sometimes not amenable to BMD modeling.
Human equivalent LOAELs or EDx values are suitable alternatives.
Importantly, the endpoint selected must be the same for all PFAS included in the calculation, for
example BMDios for the same liver effect. The equation will produce an equivalent metric (i.e.,
BMDio) for the total mixture with the given proportions of component PFAS being evaluated. In
the illustrative example below, BMDs associated with a BMR of 10% are estimated for each
chemical in a mixture and are used to determine a "M-BMD". The choice of a 10% BMR is for
illustration only; other values may be selected. If available, it is preferable to calculate the M-
BMDs using HED doses rather than oral mg/kg doses administered to test animals. Effect
equivalent BMDs are more statistically robust, and the equation explanation and example below
will reference BMDx as the model components using Equation 7-3 (similar to Equation 7-2),
where M-BMD is the total mixture dose in mg/kg/day, a* are the fixed proportions of the
component PFAS in the mixture, and BMD; is ith chemical BMD (e.g., a BMDx).
Mixture BMD = (Hf=i'^j~) (Eqn 7-3)
The equation results in a single M-BMD dose for a given BMR, that could then be converted to
an assessment value (e.g., oral RfD) and a corresponding HBWC for the mixture. Then the
original observed PFAS mixture concentration (e.g., 38.6 ng PFAS Mix/L) is compared to the
estimated M-BMD-HBWC from the M-BMD equation, and if the observed concentration is
greater than the mixture-based HBWC there is potential for human health risk. If the observed
concentration is below the mixture based HBWC then risk of health effects is not likely. Further,
the calculation can be repeated at multiple BMR levels to allow for modeling of a full mixture
dose response curve. Finally, similar to RPF, due to the potential for different effect domains to
have variable potencies across PFAS within a given mixture, the DA model should be applied
across more than one effect domain for which data are available for each of the PFAS in the
mixture to identify the lowest mixture-specific endpoint, which indicates the most sensitive
domain for the mixture.
88
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
An example is described in the following sections for a mixture of five hypothetical PFAS from
Table 4-1. The five hypothetical PFAS have existing dose response data for increased incidence
of hepatocellular death (i.e., liver endpoint), reduced serum thyroid hormone concentrations (i.e.,
thyroid endpoint), and decreased body weight in offspring (i.e., developmental endpoint), and in
rodent models for each compound. Dose responses for each chemical and each endpoint are
modeled and BMDx calculated for each compound. These values serve as the denominator
values in Equation 7-3. The numerator values are the proportions of each component PFAS in
the given mixture on a concentration basis. The total M-BMD is the inverse of the sum of the
proportion divided by the BMDx for each PFAS in the mixture. The total M-BMDx represents
an equivalent BMDx as each of the individual chemical BMDs that were used in the calculations
(i.e., if the individual chemical data were BMDio values, the DA calculation derives a BMDio for
the mixture of PFAS with those specific proportions).
The M-BMD, which is in the same units as the component chemical BMDs (e.g., oral dose in a
rodent study such as mg/kg-day), can then be adjusted based on user-defined extrapolation
factors (e.g., application of dosimetric adjustment, RSC, UFs, and life stage-specific drinking
water consumption rates) to derive a unique HBWC for the total PFAS mixture (as opposed to an
IC-specific HBWC as in the RPF approach). The derived M-BMD-HBWC can then be compared
to the actual (measured) mixture concentration and if the actual mixture concentration exceeds
the M-BMD-HBWC there is risk of the specific effect from exposure to that mixture at the
measured concentrations.
In practice, the lowest mixture-specific endpoint indicates the most sensitive effect domain for
the mixture and this endpoint can then be used for the derivation of an equivalent M-BMD-
HBWC and estimation of risk. This M-BMD is then used to derive a M-BMD HBWC for
comparison to the actual total mixture concentration of the sample. If the total mixture
concentration is greater than the M-BMD HBWC, then there is potential risk of developmental
effects in the exposed population.
7.2 Illustrative Example Application of the Mixture Benchmark Dose Approach
to a Hypothetical Mixture of Five PFAS
7.2.1 Mixture BMD Step 1: Assemble/derive component health effects endpoints
(BMDx)
PFAS 1-5 have existing dose response data on increased incidence of hepatocellular death (i.e.,
liver endpoint), reduced serum thyroid hormone concentrations (i.e., thyroid endpoint), and
decreased body weight in offspring (i.e., developmental endpoint), in rodent models for each
compound (Note: Table 7-2 below is a repeat of Table 6-1; these data were also used in the RPF
example). Dose responses for each chemical and each endpoint are modeled and BMDx
calculated for each compound.
89
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Table 7-2. Summary of PODheds for three selected health effect domains for a mixture of
five hypothetical PFAS.
Liver PODhed
(mg/kg-day)
Thyroid PODhed
(mg/kg-day)
Developmental PODhed
(mg/kg-day)
PFAS 1
0.044
0.24
0.00001 (BMDLo.ierhed)
PFAS 2
0.0013 (BMDLiohed)
0.23
0.0051
PFAS 3
N/A
0.21 (BMDLisdhed)
2.1
PFAS 4
50
N/A
0.0011 (BMDLisdhed)
PFAS 5
0.004 (BMDisdaed)
N/A
N/A
7.2.2 Mixture BMD Step 2: Assemble/derive health-based media concentrations
(HBWC)
In the case of the M-BMD approach (unlike the HI/TO SHI and RPF approaches) there is no need
for pre-existing HBWC(s) because the goal of this approach is to develop a unique, mixture-
specific HBWC for comparison to the Mixture Total PFAS concentration. Calculation of the M-
BMD HBWC is shown in Section 7.2.4.
7.2.3 Mixture BMD Step 3: Select exposure estimates (measured water
concentrations)
Select appropriate exposure estimates consistent with the problem formulation. Specifically, the
user may choose to calculate or use exposure estimates that are for the oral route in general (i.e.,
total intake in mg/kg-day) or media-specific concentrations. In the hypothetical PFAS mixture
example, 'exposure' is represented by the drinking water monitoring data in Table 4-1.
7.2.4 Mixture BMD Step 4: Calculate PFAS mixture potency (Mixture BMD HBWC)
In this example, M-BMDs are calculated for three effect domains: Liver, Thyroid, and
Developmental (Table 7-3). Application of Equation 7-3 to the example water sample in Table 4-
1 is used to derive the M-BMD. This example is for the developmental domain as it was the
lowest M-BMD of the three effect domains.
+ —) 1 = 0.00085
n/aJ
Mixture BMD = = (—- + + — + —
V BMDiJ V0.00001 0.0051 2.1 0.001
mg/kg-day
90
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Table 7-3. M-BMD Approach: Hypothetical Water Sample
Median
Measured Water
Concentration
(ng/L)
Mixing
Ratio
(Proportion)
Liver
BMD
(mg/kg/d)
Thyroid
BMD
(mg/kg/d)
Developmental
BMD
(mg/kg/d)
PFAS 1
4.8
0.01
0.044
0.24
0.00001
PFAS 2
52
0.15
0.0013
0.23
0.0051
PFAS 3
172
0.48
N/A
0.21
2.1
PFAS 4
58
0.16
50
N/A
0.0011
PFAS 5
69
0.19
0.004
N/A
N/A
Mixture Total
355.8
1.0
M-BMD
Calculation
0.0061
0.34
0.00085*
Notes: N/A = data not available.
*The lowest M-BMD is converted to a mixture-HBWC using Eqn. 7-3 for comparison to the measured concentration (i.e.,
355.8 ng/L).
The developmental-effect produced the lowest M-BMD (i.e., 0.00085 mg/kg-day), representing
the most sensitive effect domain; this value is selected for calculation of the M-BMD HBWC.
The developmental-based Mixture BMD is first converted to an RfD by applying UFs consistent
with the data being used. Selection of uncertainty factors will likely be different across mixture
component chemicals based on the available hazard and dose-response data. As such there is no
suggested standard application of quantitative uncertainty for a mixture of components although
it is suggested that a user of this approach consider uncertainty across the five areas used in EPA
human health risk assessment practice: (1) Human interindividual variability (UFh); (2)
extrapolation from animal-to-human (UFa); (3) subchronic-to-chronic duration extrapolation
(UFs); (4) LOAEL-to-NOAEL extrapolation (UFl); and (5) database uncertainty (UFd). In the
specific context of application of uncertainty to a M-BMD, a reasonable health protective
approach is to apply factors consistent with the data status of the most data-poor member of the
mixture. For example, in this hypothetical five-component PFAS mixture, a composite UF of
300 is applied based on the uncertainty associated with PFAS 5; this includes a UFh of 10 (no
availability of empirical data to inform human interindividual variability), UFa of 3
(toxicokinetic extrapolation between experimental cells/animals and humans has been accounted
for in the calculation of human equivalent doses), UFs of 1 (developmental effects are considered
duration-independent), UFl of 1 (POD is based on a benchmark dose); and UFd of 10
(significant lack of data across exposure durations and health outcome domains):
RfD = m =
V UF /
'0.00085^/d\
= 0.0000028 mg/kg-day
300
An HBWC can then be derived using Eqn. 7-3. In this example, the DWI-BW is for women of
childbearing age (i.e., 90th percentile direct and indirect consumption of community water,
consumer-only two-day average, 13 to < 50 years), and the RSC is 20% (0.2). The M-BMD
HBWC is calculated as follows:
91
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
/ Rfn \ /0.0000028^i/d\
HBWC = ( * RSC = — * 0.2 = 0.000016 mg/L = 16 ng/L
VDWI-BWV ^ 003544/d / 5 5
7.2.5 Mixture BMD Step 5: Compare PFAS mixture potency (total PFAS mixture
concentration) to health-based value (Mixture BMD HBWC)
In the developmental effect-specific M-BMD application, the health risk(s) associated with the
mixture is represented by comparing the mixture total PFAS concentration (355.8 ng/L) to the
M-BMD HBWC which is based on the specified effect for hazard domain (e.g., for this example,
decreased body weight in offspring). In this hypothetical example, the mixture total PFAS
concentration 355.8 ng/L exceeds the M-BMD HBWC 16 ng/L by over an order of magnitude,
indicating significant potential for health risks in developmental populations exposed to a
mixture of the five PFAS at the hypothetical water exposure estimates provided.
Although not shown in the above example, if the M-BMD was instead based on the liver effect
domain, M-BMD HBWC would be 115 ng/L, well below the measure PFAS concentration
(355.8 ng/L), indicating the potential for liver effects in exposure populations. Alternatively, if
the M-BMD was based on the thyroid effect domain, the resulting M-BMD HBWC would be
6,321 ng/L, well above the measured total PFAS concentration (355.8 ng/L), indicating unlikely
risk for thyroid effects among the exposed population.
7.3 Advantages and Challenges of the Mixture BMD Approach
There are several advantages to the M-BMD approach. First, there is no a priori requirement for
having formal human health assessment values, such as oral RfDs or chemical-specific HBWCs,
for any of the individual PFAS in the mixture. The only data needs are effect endpoints (i.e.,
BMDs) for each of the PFAS in the mixture for the common endpoint(s) being modeled. Another
advantage is that it avoids any potential confusion that could arise from putting the mixture POD
in the units of a single chemical (i.e., the IC from the RPF approach). Rather, the end result is a
POD for the whole mixture that is specific for the assortment and ratios of PFAS in the mixture
being evaluated. It is important to recognize that the DA model calculation of combined mixture
effect (M-BMD) is different for each PFAS mixture depending on: (1) the specific PFAS in the
mixture; (2) the mixing ratio; and (3) the effect or endpoint being modeled. For example, one
could expect that a mixture of PFAS that has a greater concentration of a more potent compound,
and a lower concentration of a less potent compound would have a lower (i.e., more potent) M-
BMD than a similar assortment of compounds that has a lower concentration of the more potent
PFAS and a greater concentration of the less potent PFAS. It is also advantageous that the M-
BMD approach does not actually require or assume that the component PFAS in a given mixture
have congruent dose response curves for each effect being evaluated (reviewed in NRC (2008)).
Finally, it is ideal to have well resolved dose response curves for each component PFAS in a
mixture to estimate equivalent BMDs (e.g., BMDio), and this is necessary if a goal is to model
the entire dose response for the mixture. However, in the absence of such data, M-BMD
modeling is also amenable to simple point estimates such as NOAELs, as long as they are
toxicologically similar across component chemicals (i.e., for same endpoint, such as increased
incidence of hepatocellular death) but use of this type of point data would impede the modeling
of the full mixture dose response curve if desired.
92
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
There are also several challenges with the M-BMD approach. Like the RPF approach, the user
needs effect data for at least one common endpoint from the constellation of PFAS effects for all
components of the mixture. Ideally this would be for one of, if not the most sensitive, effect
across PFAS in the mixture of interest in order to provide a conservative (protective) risk
scenario. For some mixtures that contain less well-studied PFAS there may be limited or no
available dose response data available to derive component chemical BMDs in order to calculate
the M-BMD.
A limitation, that is not unique to this specific approach, is that PFAS mixtures may vary over
time in environmental media. As proportions of component PFAS change in the mixture, the
calculations would need to be recalculated as the composition of the mixture changed from site
to site or over time within the same site. However, the calculation can be readily and easily
repeated for different mixing ratios and mixture concentrations once the component chemical
effect endpoint values have been determined. Finally, for both the RPF and M-BMD approaches,
depending on data availability for the individual compounds, the effect domains modeled may
potentially not be the overall most sensitive out of the total constellation of common PFAS
effects (e.g., in reality developmental effects may be the most sensitive and would produce the
lowest M-BMD, but data are only available for the component PFAS to calculate M-BMDs for
liver and thyroid effects). In this example, the M-BMD HBWC is based on developmental
effects because that is the most sensitive among the three assessed effect domains, and thus is
protective of the other effects (i.e., liver, thyroid). As described in the previous section, if this M-
BMD analysis was instead based on the thyroid effect, the user would conclude that potential
risk is unlikely.
93
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
8.0 Comparison of Component-Based Approaches
This framework document describes the conceptual bases and practical application of data-
driven options for estimating the noncancer health risks associated with human exposure to
mixtures of PFAS. The component-based options described are included in prior EPA mixtures
guidance (EPA, 1986, 2000b) and supported by NRC (2008). Although the approaches and
illustrative examples are provided using drinking water as the exposure route, the technical basis
of each approach could be readily applied or adapted to other sources of oral exposure (e.g., soil,
fish/shellfish). Each of the approaches included require varying levels of data input and have
relatively subtle but substantive differences in assumptions and ultimately produce risk
estimations that may slightly differ based on those exposure assumptions. Importantly,
interpretations of health risks associated with mixtures of PFAS will be highly dependent on the
specific PFAS components within a given mixture and the individual concentrations or
proportions of each of those components. Given the significant lack of toxicity data across the
diverse structural landscape of compounds within the PFAS chemical class, it is likely that many
users of this framework will need to incorporate information NAMs such as in vitro cell
bioactivity, toxicogenomic platforms, and/or structure-activity/read-across in order to facilitate
estimation of health risk for a PFAS mixture of interest; however, in vivo animal or human
toxicity data are strongly recommended where available.
Given the range of data-driven options presented in this framework, an important consideration
is under what circumstances the different options produce similar or dissimilar indications or
estimates of health risk(s). The primary basis for differing risk estimates relates to differences in
data input requirements, model assumptions, and final value derivation. Both the HI and TOSHI
approaches necessitate availability (or de novo derivation) of health assessment values (e.g., oral
RfDs) to calculate mixture component HQs. In contrast, the RPF and M-BMD approaches target
dose-response data for same/similar effect, sans derivation of health assessment values, to inform
mixture risk estimates (e.g., concentrations or doses) for comparison to measured media
concentrations either for an anchor/index chemical (RPF) or across each mixture component (M-
BMD). The general HI approach allows for component PFAS in the mixture to have different
health effects or endpoints as the basis for the component chemical RfVs (see Figure 4-1) and
thus this approach is likely a more health-protective indicator of risk (i.e., produce a component
HQ or mixture HI of > 1) since the representation of toxicity will likely be the most sensitive,
compared to the RPF and M-BMD approaches where similarity in toxicity does not have to a
priori be the most sensitive effect domain. In contrast, the TOSHI approach is more targeted and
assumes the component RfVs are based on the same organ or organ system. This more narrowed
focus is likely to produce a less health-protective indicator of risk than the general HI (i.e., less
likely than general HI to produce HQ > 1) because the range of potential effects has been scoped
to a specific target organ or organ system; for example, for some mixture components the effect
domain identified for TOSHI application may be one of the less potent across a profile of effects.
This important nuance will be dependent on the availability of target organ specific RfVs, and
case-by-case interpretations of "potency" for effect will be a function of both dose-response
(e.g., POD) and uncertainty factor application. The user of the TOSHI approach would be
advised to also perform the general HI for the same mixture and compare the His (and
component HQs) across each approach. It should be noted that any component chemical HQ or
mixture HI >1 indicates potential health risks; magnitude of HI is not an optimal comparator. A
TOSHI risk estimate is likely less conservative than the general HI but may be more
94
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
conservative than the RPF and M-BMD approaches. However, the TOSHI and RPF approach
will give essentially the same answer when the ratio of the POD values used to calculate the
RPFs is equal to the ratio of the endpoint-specific PODs used in the derivation of RfVs used to
calculate the TOSHI. The major difference between the RPF and M-BMD estimates is the RPF
approach assumes congruent slopes whereas the M-BMD does not. If the mixture component
chemicals have congruent dose response curves, for same effect, the RPF and M-BMD
calculations produce nearly identical risk estimates for the same mixture. However, if the
mixture component chemicals display a range of congruent and non-congruent dose response
curves, then the assumptions for application of the RPF method are violated and the M-BMD
approach should be used to produce a more accurate estimate of risk (NRC, 2008).
Another factor in the concordance (or not) of mixture risk estimates across the component-based
approaches presented are the factors used in deriving RfVs or HBWCs from hazard data across
components. The HI and TOSHI approaches are calculated based on an exposure metric divided
by the component chemical RfVs or HBWCs. The input data for calculating RfV and HBWC,
including the critical effect and POD (i.e., BMDLx or NOAEL or LOAEL), correction factors
(i.e., UFs, DAF), and exposure route adjustments (i.e., DWI, RSC) used for each component
PFAS will have impacts on the resulting risk estimates and should be carefully selected based on
available data for each component PFAS. Similarly, when conducting RPF or M-BMD
assessment, the PODs used across component PFAS in a mixture could potentially be derived
from dissimilar response metrics (i.e., LOAEL vs BMDLx) and the resulting M-BMD-HBWC or
Total ICEC comparison will be somewhat dependent on the applicability of the correction and
route adjustment factors used for the M-BMD HBWC or Total ICEC derivation across all
component PFAS in the mixture. For example, the correction factors and route adjustment
factors may not be appropriate for all component PFAS and thus the Total ICEC to IC RfV or
HBWC comparison or M-BMD HBWC to measured concentration may be affected. Thus, it is
strongly encouraged to use comparable PODs across component PFAS where possible and select
adjustment factors given careful consideration of the components for the specific PFAS mixture
being evaluated. It will be key to transparently present and communicate selection of uncertainty
factors or exposure route adjustment factors, and associated rationale(s), such that interpretations
and conclusions of mixture risk are supportable.
A critical consideration for use of the approaches in this framework document is that where data
are available and support, it may be prudent to apply each approach to the same mixture. The
purpose of the comparison is not necessarily to determine which approach provides the most
conservative estimate of mixture risk but rather which approach entails the greatest level of
confidence in the data underlying the components and support for the assumption of dose
additivity in the evaluation of joint toxicity of PFAS.
95
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
References
3M Company. 2000. Voluntary Use and Exposure Information Profile Perfluorooctanoic Acid
and Salts. EPA Administrative Record AR.226-0595. EPA, Washington, DC.
Abbott, B.D., C.J. Wolf J.E. Schmid, K.P. Das, RD. Zehr, L. Helfant, S. Nakayama, A.B.
Lindstrom, M.J. Strynar, and C. Lau. 2007. Perfluorooctanoic acid induced
developmental toxicity in the mouse is dependent on expression of peroxisome
proliferator activated receptor-alpha. Toxicological Sciences 98(2):571-581.
Abbott, B.D., C.J. Wolf, K.P. Das, R.D. Zehr, J.E. Schmid, A.B. Lindstrom, M.J. Strynar, and C.
Lau. 2009. Developmental toxicity of perfluorooctane sulfonate (PFOS) is not dependent
on expression of peroxisome proliferator activated receptor-alpha (PPAR alpha) in the
mouse. Reproductive Toxicology 27(3-4):258-265.
Ahrens, L. 2011. Polyfluoroalkyl compounds in the aquatic environment: a review of their
occurrence and fate. Journal of Environmental Monitoring 13(1):20—31.
Alaska DEC (Alaska Department of Environmental Conservation). 2019. Technical
Memorandum Action Levels for PFAS in Water and Guidance on Sampling Groundwater
and Drinking Water. AK DEC, Division of Spill Prevention and Response Contaminated
Sites Program and Division of Environmental Health Drinking Water Program.
https://dec.alaska.gov/media/15773/pfas-drinking-water-action-levels-technical-
memorandu
Altenburger, R., T. Backhaus, W. Boedeker, M. Faust, M. Scholze, and L.H. Grimme. 2000.
Predictability of the toxicity of multiple chemical mixtures to Vibrio fischerr. mixtures
composed of similarly acting chemicals. Environmental Toxicology and Chemistry: An
International Journal 19(9):2341-2347.
Anderson, R.H., G.C. Long, R.C. Porter, and J.K. Anderson. 2016. Occurrence of select
perfluoroalkyl substances at U.S. Air Force aqueous film-forming foam release sites
other than fire-training areas: field-validation of critical fate and transport properties.
Chemosphere 150:678-685.
ATSDR (Agency for Toxic Substances and Disease Registry). 2018. Interaction Profile for
Mixtures of Insecticides: Pyrethroids, Organophosphorus Compounds, and Carbamates.
U.S. Department of Health and Human Services, Public Health Service, ATSDR, Atlanta,
GA.
ATSDR (Agency for Toxic Substances and Disease Registry). 2021. Toxicological Profile for
Perfluoroalkyls. U.S. Department of Health and Human Services, Public Health Service,
ATSDR, Atlanta, GA.
ATSDR (Agency for Toxic Substances and Disease Registry). 2022. CDC/ATSDR PFAS
Exposure Assessment Community Level Results: Hampden County (MA) near Barnes Air
National Guard Base. ATSDR, Atlanta, GA.
https://www.atsdr.cdc.eov/pfas/commimities/factsheet/Commimity-Level-Resiilts-
Factsheet.html.
96
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Australian Government Department of Health. 2019. Health Based Guidance Values for PFAS.
https://wwwl.health.eov.aii/internet/main/piiblishine.nsf/Content/2200FE086D480353C
*<- 30C900817CDC/$File/HBGV-Factsheei . 01' \v M I mIf
Bailey, D.G. 2013. Grapefruit-medication interactions. CMAJ 185(6):507-508.
Beach, S.A., J.L. Newsted, K. Coady and J.P. Giesy. 2006. Ecotoxicological evaluation of
perfluorooctanesulfonate (PFOS). Reviews of Environmental Contamination and
Toxicology 186:133-174.
Beckett, K.J., B. Yamini., and S.J. Bursian. 2008. The effects of 3,3',4,4',5-pentachlorobiphenyl
(PCB 126) on mink (Mustela vison) reproduction and kit survivability and growth.
Archives of Environmental Contamination and Toxicology 54:123-129.
Behr, A.C., C. Plinsch, A. Braeuning, and T. Buhrke. 2020. Activation of human nuclear
receptors by perfluoroalkylated substances (PFAS). Toxicology In vitro 62:104700.
Benskin, J.P., D.C. Muir, B.F. Scott, C. Spencer, A.O. De Silva, H. Kylin, J.W. Martin, A.
Morris, R. Lohmann, G. Tomy, B. Rosenberg, S. Taniyasu, andN. Yamashita. 2012.
Perfluoroalkyl acids in the Atlantic and Canadian Arctic Oceans. Environmental Science
& Technology 46(11):5815-5823.
Bil, W., M. Zeilmaker, S. Fragki, J. Lijzen, E. Vergruggen, and B. Bokkers. 2021. Risk
assessment of per- and polyfluoroalkyl substance mixtures: a relative potency factor
approach. Environmental Toxicology and Chemistry 40(4): 859-870.
doi:10.1002/etc.4835.
Bil W, Zeilmaker MJ, Bokkers BGH. 2022. Internal Relative Potency Factors for the Risk
Assessment of Mixtures of Per- and Polyfluoroalkyl Substances (PFAS) in Human
Biomonitoring. Environmental Health Perspectives
130(7):77005. doi: 10.1289/EHP10009.
Bishop, C.A., P. Ng, K.E. Petit, S.W. Kennedy, J.J. Stegeman, R.J. Norstrom, and R.J. Brooks.
1998. Environmental contamination and developmental abnormalities in eggs and
hatchlings of the common snapping turtle (Chelydra serpentina serpentina) from the
Great Lakes-St Lawrence River basin (1989-1991). Environmental Pollution 101(1): 143—
156.
Bjork, J.A., C. Lau, S.C. Chang, J.L. Butenhoff, and K.B. Wallace. 2008. Perfluorooctane
sulfonate-induced changes in fetal rat liver gene expression. Toxicology 251(l-3):8-20.
Bjork, J. A., J. L. Butenhoff and K. B. Wallace. 2011. Multiplicity of nuclear receptor activation
by PFOA and PFOS in primary human and rodent hepatocytes. Toxicology 288(l-3):8-
17. dot 10 101 t/i.tox.201 I 0 01.
97
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Blake, B E., H.A. Cope, S.M. Hall, R.D. Keys, B.W. Mahler, J. McCord, B. Scott, H.M.
Stapleton, M.J. Strynar, S.A. Elmore and S.E. Fenton. 2020. Evaluation of maternal,
embryo, and placental effects in CD-I mice following gestational exposure to
perfluorooctanoic acid (PFOA) or hexafluoropropylene oxide dimer acid (HFPO-DA or
GenX). Environmental Health Perspectives 128(2):27006.
Bliss, C.I. 1939. The toxicity of poisons applied jointly. Annals of Applied Biology 26:585-615.
Boobis, A., R. Budinsky, S. Collie, K. Crofton, M. Embry, S. Felter, R. Hertzberg, D. Kopp, G.
Mihlan, M. Mumtaz, P. Price, K. Solomon, L. Teuschler, R. Yang, and R. Zaleski. 2011.
Critical analysis of literature on low-dose synergy for use in screening chemical mixtures
for risk assessment. Critical Reviews in Toxicology 41(5):369-383.
Borg, D., B-E. Lund, N-G. Lindquist, and H. Hakansson. 2013. Cumulative health risk
assessment of 17 perfluoroalkylated and polyfluoroalkylated substances (PFASs) in the
Swedish population. Environment International 59:112-123.
Boulanger, B., J. Vargo, J.L. Schnoor, and K.C. Hornbuckle. 2004. Detection of perfluorooctane
surfactants in Great Lakes water. Environmental Science & Technology 38(15):4064-
4070.
Brendel S., E. Fetter, C. Staude, L. Vierke, and A. Biegel-Engler. 2018. Short-chain
perfluoroalkyl acids: environmental concerns and a regulatory strategy under REACH.
Environmental Sciences Europe 30(1):9.
Buck, R.C., J. Franklin, U. Berger, J.M. Conder, I.T. Cousins, P. de Voogt, A.A. Jensen, K.
Kannan, S.A. Mabury, and S.P. vanLeeuwen. 2011. Perfluoroalkyl and polyfluoroalkyl
substances in the environment: terminology, classification, and origins. Integrated
Environmental Assessment and Management 7(4): 513-541.
Buck, R.C., S. H. Korzeniowski, E. Laganis, and F. Adamsky. 2021. Identification and
classification of commercially relevant per- and poly-fluoroalkyl substances (PFAS).
Integrated Environmental Assessment and Management 17(5): 1045-1055.
Burkhard, L.P. 2021. Evaluation of published bioconcentration factor (BCF) and
bioaccumulation factor (BAF) data for per-and polyfluoroalkyl substances across aquatic
species. Environmental Toxicology and Chemistry 40(6): 1530-1543.
Burnham, K. P., D. R. Anderson and K. P. Huyvaert. 2010. AIC model selection and multimodel
inference in behavioral ecology: some background, observations, and comparisons.
Behavioral Ecology and Sociobiology 65(1):23—35. doi 10 100 ^002 ->010 1029-6.
Butenhoff, J.L., G.L. Kennedy, Jr., S.R. Frame, J.C. O'Connor, andR.G. York. 2004. The
reproductive toxicology of ammonium perfluorooctanoate (APFO) in the rat. Toxicology
196(l-2):95—116.
CAA (Clean Air Act). 1990. Clean Air Act Amendments of 1990. Public Law, November 15,
1990, 101-549.
98
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Cadwallader, A., A. Greene, H. Holsinger, A. Lan, M. Messner, M. Simic, and R. Albert. 2022.
A Bayesian hierarchical model for estimating national PFAS drinking water occurrence.
AWWA Water Science 4(3):e 1284. dc aws2.1284.
Calafat, A.M., L. Wong, Z. Kuklenyik, J.A. Reidy, and L.L. Needham. 2007. Polyfluoroalkyl
chemicals in the U.S. population: data from the National Health and Nutrition
Examination Survey (NHANES) 2003-2004 and comparisons with NHANES 1999-
2000. Environmental Health Perspectives 115(11): 1596-1602.
Calafat, A.M., K. Kato, K. Hubbard, T. Jia, J.C. Botelho, and L.Wong. 2019. Legacy and
alternative per and polyfluoroalkyl substances in the U.S. general population: paired
serum-urine data from the 2013-2014 National Health and Nutrition Examination Survey.
Environment International 131:105048.
Carlson, L.M., M. Angrish, A.V. Shirke, E.G. Radke, B. Schulz, A. Kraft, R. Judson, G.
Patlewicsz, R. Blain, C. Lin, N. Vetter, C. Lemeris, P. Hartman, H. Hubbard, X. Arzuaga,
A. Davis, L.V. Dishaw, I.L. Druwe, H. Hollinger, R. Jones, J.P. Kaiser, L. Lizarraga,
P.D. Noyes, M. Taylor, A.J. Shapiro, A.J. Williams, and K.A. Thayer. 2022. Systematic
evidence map for one hundred and fifty per- and polyfluoroalkylsubstances (PFAS).
Environmental Health Perspectives 130(5):056001. doi h« l „ 39/E H i 0 '43
CDPHE (Colorado Department of Public Health and Environment). 2020. Regulation No. 42:
Site-Specific Water Quality Classifications and Standards for Groundwater. CDPHE,
Water Quality Control Commission.
https://www.sos. state. co.us/CCR/GenerateRulePdf.do?mleVersionId=8815&fileName=5
%20CCR%201002-42.
CERCLA (Comprehensive Environmental Response Compensation and Liability Act). 2002.
Comprehensive Environmental response, compensation, and liability act of 1980
(Superfund). Public Law, December 31, 2002, 107-377.
CHAP (Chronic Hazard Avisory Panel). 2014. Chronic Hazard Advisory Panel on Phthalates
andPhthalate Alternatives. Prepared for the U.S. Consumer Product Saftey Commission,
Bethesda, MD, by CHAP, Bethesda, MD.
Clark, R. L., J.M. Antonello, S.J. Grossman, L.D. Wise, C. Anderson, W.J. Bagdon, S.
Prahalada, J.S. MacDonald, and R.T. Robertson. 1990. External genitalia abnormalities in
male rats exposed in utero to finasteride, a 5 alpha-reductase inhibitor. Teratology 42:91-
100.
Cochran, R.S. 2015. Evaluation of Perfluorinated Compounds in Sediment, Water, and Passive
Samplers Collected from the Barksdale Air Force Base. MS Thesis. Texas Tech
University, Lubbock, TX.
Conder, J.M., R.A. Hoke, W.D. Wolf, M.H. Russell, and R.C. Buck. 2008. Are PFCAs
bioaccumulative? A critical review and comparison with regulatory criteria and persistent
lipophilic compounds. Environmental Science & Technology 42(4):995-1003.
99
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Conley, J. M., B.R. Hannas, J.R. Furr, V.S. Wilson, and L.E. Gray, Jr. 2016. A demonstration of
the uncertainty in predicting the estrogenic activity of individual chemicals and mixtures
from an in vitro estrogen receptor transcriptional activation assay (T47D-KBluc) to the in
vivo uterotrophic assay using oral exposure. Toxicological Sciences 153(2):382-395.
Conley, J.M., C.S. Lambright, N. Evans, M. Cardon, J. Furr, V.S. Wilson, and L.E. Gray, Jr.
2018. Mixed "antiandrogenic" chemicals at low individual doses produce reproductive
tract malformations in the male rat. Toxicological Sciences 164:166-178.
Conley, J.M., C.S. Lambright, N. Evans, M.J. Strynar, J. McCord, B.S. Mclntyre, G.S. Travlos,
M.C. Cardon, E. Medlock-Kakaley, P.C. Hartig, V.S. Wilson and L.E. Gray, Jr. 2019.
Adverse maternal, fetal, and postnatal effects of hexafluoropropylene oxide dimer acid
(GenX) from oral gestational exposure in Sprague-Dawley rats. Environmental Health
Perspectives 127(3):37008. dc I89/EHP4372.
Conley, J.M., C. Lambright, N. Evans, E. Medlock Kakaley, D. Hill, A. Dixon, E. Hines, and
L.E. Gray. 2021a. Dose additive developmental toxicity of a mixture of three PFAS
(HFPO-DA, NBP2, PFOS) in the Sprague-Dawley rat. Abstract no. 2666. The
Toxicologist: Supplement to Toxicological Sciences 180(1).
Conley, J.M., C.S. Lambright, N. Evans, J. McCord, M.J. Strynar, D Hill, E. Medlock-Kakaley,
V.S. Wilson, and L.E. Gray, Jr. 2021b. Hexafluoropropylene oxide-dimer acid (HFPO-
DA or GenX) alters maternal and fetal glucose and lipid metabolism and produces
neonatal mortality, low birthweight, and hepatomegaly in the Sprague-Dawley rat.
Environment International 146:106024.
Conley, J.M., C.S. Lambright, N. Evans, E. Medlock-Kakaley, D. Hill, J. McCord, M.J. Strynar,
L.C. Wehmas, S Hester, D.K. MacMillan, and L.E. Gray, Jr. 2022a. Developmental
toxicity of Nafion byproduct 2 (NBP2) in the Sprague-Dawley rat with comparisons to
hexafluoropropylene oxide-dimer acid (HFPO-DA or GenX) and perfluorooctane
sulfonate (PFOS). Environment International 160:107056.
dot 10 101 /i.envint.20:i 10 '056.
Conley, J.M., C.S. Lambright, N. Evans, E. Medlock-Kakaley, A. Dixon, D. Hill, J. McCord,
M.J. Strynar, J. Ford, L.E. and Gray. 2022b. Cumulative maternal and neonatal effects of
combined exposure to a mixture of perfluorooctanoic acid (PFOA) and perfluorooctane
sulfonic acid (PFOS) during pregnancy in the Sprague-Dawley rat. Environment
International 170:107631. doi.off .envint.2022.107631.
Cousins, I.T., J.C. DeWitt, J. Gluge, G. Goldenman, D. Herzke, R. Lohmann, M. Miller, C.A.
Ng, Martin, Scheringer, L. Vierke, and Z. Wang. 2020. Strategies for grouping per-and
polyfluoroalkyl substances (PFAS) to protect human and environmental health.
Environmental Science: Processes and Impacts 22:1444-1460.
Crofton, K.M., E.S. Craft, J.M. Hedge, C. Gennings, J.E. Simmons, R.A. Carchman, W.H.
Carter, Jr., and M.J. DeVito. 2005. Thyroid-hormone-disrupting chemicals: evidence for
dose-dependent additivity or synergism. Environmental Health Perspectives
113(11): 1549—1554.
100
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Crofton, K.M. 2008. Thyroid disrupting chemicals: mechanisms and mixtures. International
Journal of Andrology 31(2):209-223.
Danish Environmental Protection Agency. 2015. PerfluoroalkylatedSubstances: PFOA, PFOS
and PFOSA. Evaluation of Health Hazards and Proposal of a Health Based Quality
Criterion for Drinking Water, Soil, and Groundwater. Environmental Project Number
1665. Danish Ministry of the Environment, Environmental Protection Agency,
Copenhagen, Denmark. https://www2.mst.dk/Udeiv/piiblications/2015/04/978-87-93283-
Das, K.P., B E. Grey, M B. Rosen, C.R. Wood, K.R. Tatum-Gibbs, R.D. Zehr, M.J. Strynar,
A.B. Lindstrom, and C. Lau. 2015. Developmental toxicity of perfluorononanoic acid in
mice. Reproductive Toxicology 51:133-144.
DE DHHS (Delaware Health and Social Services). 2021. Delaware PFOA and PFOS MCL
Implementation Plan. Proposed implementation plan for the maximum contaminant
levels (MCL) as required by Delaware House Bill 8 (HB-8), signed in October 2021.
https://www.dhss.delaware.gov/dhss/dph/hsp/files/MCLimplementationPlanPFAS.pdf
De Silva, A.O., C. Spencer, B.F. Scott, S. Backus, and D.C. Muir. 2011. Detection of a cyclic
perfluorinated acid, perfluoroethylcyclohexane sulfonate, in the Great Lakes of North
America. Environmental Science & Technology 45(19):8060-8066.
De Silva, A.O., J.M. Armitage, T.A. Bruton, C. Dassuncao, W. Heiger-Bernays, X.C. Hu, A.
Karrman, B. Kelly, C. Ng, A. Robuck, M. Sun, T.F. Webster, and E.M. Sunderland.
2020. PFAS exposure pathways for humans and wildlife, a synthesis of current
knowledge and key gaps in understanding. Environmental Toxicology and Chemistry
40:631-657.
Ding, G., J. Zhang, Y. Chen, L. Wang, M. Wang, D. Xiong and Y. Sun (2013). Combined effects
of PFOS and PFOA on zebrafish (Danio rerio) embryos. Archives of Environmental
Contamination and Toxicology 64(4):668-675. doi: 10 100 100 J II 0 i J 9864-2.
Dix, D.J., K.A. Houck, M.T. Martin, A.M. Richard, R.W. Setzer, and R.J. Kavlock. 2007. The
ToxCast program for prioritizing toxicity testing of environmental chemicals.
Toxicological Sciences 95( 1 ):5— 12. d /toxsci/kfl 103.
Dubois, K.P. 1961. Potentiation of the toxicity of organophosphorus compounds. Advances in
Pest Control Research 4:117-151.
EFSA (European Food Safety Authority), H.K. Knutsen, J. Alexander, L. Barregard, M.
Bignami, B. Bruschweiler, S. Ceccatelli, B. Cottrill, M. Dinovi, L. Edler, B. Grasl-
Kraupp, C. Hogstrand, L. R. Hoogenboom, C. S. Nebbia, I.P. Oswald, A. Petersen, M.
Rose, A.C. Roudot, C. Vleminckx, G. Vollmer, H. Wallace, L. Bodin, J. P. Cravedi, T. I.
Halldorsson, L. S. Haug, N. Johansson, H. van Loveren, P. Gergelova, K. Mackay, S.
Levorato, M. van Manen, and T. Schwerdtle. 2018. Risk to human health related to the
presence of perfluorooctane sulfonic acid and perfluorooctanoic acid in food. EFSA
Journal 16(12):e05194.
101
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
EFSA (European Food Safety Authority), D. Schrenk, M. Bignami, L. Bodin, J.K. Chipman, J.
Del Mazo, B. Grasl-Kraupp, C. Hogstrand, L.R. Hoogenboom, J.C. Leblanc, C.S.
Nebbia, E. Nielsen, E. Ntzani, A. Petersen, S. Sand, C. Vleminckx, H. Wallace, L.
Barregard, S. Ceccatelli, J.P. Cravedi, T.I. Halldorsson, L.S. Haug, N. Johansson, H.K.
Knutsen, M. Rose, A.C. Roudot, H. Van Loveren, G. Vollmer, K. Mackay, F. Riolo, and
T. Schwerdtle. 2020. Risk to human health related to the presence of perfluoroalkyl
substances in food. EFSA Journal 18(9):e06223.
Egeghy, P.P., and M. Lorber. 2010. An assessment of the exposure of Americans to
perfluorooctane sulfonate: a comparison of estimated intake with values inferred from
NHANES data. Journal of Exposure Science & Environmental Epidemiology 21:15 0—
168.
EPA (U.S. Environmental Protection Agency). 1986. Guidelines for the Health Risk Assessment
of Chemical Mixtures. EPA/630/R-98/002. EPA, Risk Assessment Forum, Washington,
DC. https://www.epa.gov/risk/giiidelines-health-risk-assessment-chemical-mixtiires.
EPA (U.S. Environmental Protection Agency). 1987. Interim Procedures for Estimating Risks
Associated with Exposures to Mixtures of Chlorinated Dibenzo-P-Dioxins and -
Dibenzofurans (CDDs and CDFs). EPA/625/3-87/012. EPA, Risk Assessment Forum,
Washington, DC
EPA (U.S. Environmental Protection Agency). 1989. Interim Procedures for Estimating Risks
Associated with Exposures to Mixtures of Chlorinated Dibenzo-P-Dioxins and -
Dibenzofurans (CDDs and CDFs) and 1989 Update. EPA/625/3-89/016. EPA, Risk
Assessment Forum, Washington, DC.
EPA (U.S. Environmental Protection Agency). 1991. Risk Assessment Guidance for Superfund,
Vol 1: Human Health Evaluation Manual (Part B, Development of Risk-Based
Preliminary Remediation Goals). EPA/540/R-92/003. EPA, Emergency and Remedial
Response, Washington, DC. https://www.epa.gov/risk/risk-assessment-guidance-
superfund-rags-part-b.
EPA (U.S. Environmental Protection Agency). 1999. Policy on a Common Mechanism of
Action: The Organophosphate Pesticides. Vol. 64. Federal Register, Washington, D.C.,
1999, pp. 5795-5799.
EPA (U.S. Environmental Protection Agency). 2000a. Methodology for Deriving Ambient Water
Quality Criteria for the Protection of Human Health. EPA/822/B-00/004. EPA, Office of
Water, Washington, DC. https://www.epa.gov/sites/default/flles/2018-
10/documents/methodology ~wqc~protection~hh~2000.pdf.
EPA (U.S. Environmental Protection Agency). 2000b. Supplementary Guidance for Conducting
Health Risk Assessment of Chemical Mixtures. EPA/63 0/R-00/002. EPA, Risk
Assessment Forum, Washington, DC.
https://cfpub. epa.gov/ncea/risk/recordisplav. cfm?deid=20533.
102
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
EPA (U.S. Environmental Protection Agency). 2003. Chapter 9. Toxic Equivalency Factors
(TEF) for Dioxin and Related Compounds in Part II: Exposure and Human Health
Reassessment for 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) and Related Compounds.
NAS review draft NCEA-I-0836. EPA, Office of Research and Development, National
Center for Environmental Assessment, Washington, DC.
EPA (U.S. Environmental Protection Agency). 2006. Organophosphorus Cumulative Risk
Assessment (2006 Update). EPA-HQ-OPP-2006-0618-0002. EPA, Office of Pesticide
Programs, Washington, DC.
EPA (U.S. Environmental Protection Agency). 2007a. Concepts, Methods and Data Sources for
Cumulative Health Risk Assessment of Multiple Chemicals, Exposures and Effects: A
Resource Document. EPA/600/R-06/013F. EPA, Washington, DC.
https://cfpub.epa.gov/ncea/risk/recordisplav.cfm?deid=l 90187.
EPA (U.S. Environmental Protection Agency). 2007b. RevisedN-Methyl Carbamate Cumulative
Risk Assessment. EPA, Office of Pesticide Programs, Washington, DC.
EPA (U.S. Environmental Protection Agency). 2008. Framework for Application of the Toxicity
Equivalence Methodology for Polychlorinated Dioxins, Furans, and Biphenyls in
Ecological Risk Assessment. EPA/100/R-08/004. EPA, Risk Assessment Forum,
Washington, DC. https://www.epa.gov/risk/framework-application-toxicitv-equivalence-
methodology-polychlorinated-dioxins-furans-and.
EPA (U.S. Environmental Protection Agency). 2009. Endocrine Disruptor Screening Program
Test Guidelines OPPTS 890.1600: Uterotrophic Assay. EPA/740/C-09/0010. EPA,
Washington, DC.
EPA (U.S. Environmental Protection Agency). 2010. Recommended Toxicity Equivalence
Factors (TEFs) for Human Health Risk Assessments of 2,3,7,8- Tetrachlorodibenzo-p-
dioxin andDioxin-Like Compounds. EPA/100/R-10/005. EPA, Risk Assessment Forum,
Washington, DC. https://www.epa.gov/risk/documents-recommended-toxicitv-
equivalencv-factors-human-health-risk-assessments-dioxin-and.
EPA (U.S. Environmental Protection Agency). 2011a. Pyrethrins/Pyrethroid Cumulative Risk
Assessment. EPA-HQ-OPP-2011-0746-0003.
https://www.regiilations.gov/dociiment/EPA-HO-OPP- 03.
EPA (U.S. Environmental Protection Agency). 2011b. Recommended Use of Body Weight^4 as
the Default Method in Derivation of the Oral Reference Dose. EPA/100/R-11/0001. EPA,
Risk Assessment Forum, Washington, DC. https://www.epa.gov/risk/recommended-use-
bodv-weight-34-default-method-derivation-oral-reference-dose.
EPA (U.S. Environmental Protection Agency). 2012. Benchmark Dose Technical Guidance.
EPA/100/R-12/001. EPA, Risk Assessment Forum, Washington, DC.
https://www.epa.gov/risk/benchmark-dose-technical-guidance.
103
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
EPA (U.S. Environmental Protection Agency). 2016a. Drinking Water Health Advisory for
Perfluorooctanoic Acid (PFOA). EPA/822/R-16/005. EPA, Washington, DC.
https://www.epa.gov/sites/default/files/2Q16-
0 5/docum ents/pfoa health advisory final 5Q8.pdf.
EPA (U.S. Environmental Protection Agency). 2016b. Drinking Water Health Advisory for
Perfluorooctane Sulfonate (PFOS). EPA/822/R-16/004. EPA, Washington, DC.
https://www.epa.gov/sites/default/files/2Q16-
05/documents/pfos health advisory final 5Q8.pdf.
EPA (U.S. Environmental Protection Agency). 2018. Cumulative Assessment of Risk from
Pesticides. EPA, Washington, DC. https://www.epa.gov/pesticide-science-and-assessing-
pesticide-risks/cumulative-assessment-risk-pesticides.
EPA (U.S. Environmental Protection Agency). 2019a. Exposure Factors Handbook Chapter 3
(Update): Ingestion of Water and Other Select Liquids. EPA/600/R-18/259F. EPA,
Office of Research and Development, Washington, DC.
https://cfpub.epa.gov/ncea/efp/recordisplay.cfm?deid=343661.
EPA (U.S. Environmental Protection Agency). 2019b. Occurrence Data from the Third
Unregulated Contaminant Monitoring Rule (UCMR 3). https://www.epa.gov/monitoring-
unregulated-drinking-water-contaminants/occurrence-data-unregulated-contaminant#3.
EPA (U.S. Environmental Protection Agency). 2019c. Announcement of Preliminary Regulatory
Determinations for Contaminants on the Fourth Drinking Water Contaminant Candidate
List. Federal Register, March 10, 2020, 85(47): 14098-14142.
EPA (U.S. Environmental Protection Agency). 2021a. Human Health Toxicity Values for
Perfluorobutane Sulfonic Acid (CASRN 375-73-5) and Related Compound Potassium
Perfluorobutane Sulfonate (CASRN 29420-49-3). EPA/600/R-20/345F. EPA, Office of
Research and Development, Washington, DC.
http s: //cfpub. epa. gov/n cea/ri sk/recordi spl ay. cfm ? dei d=3 5 08 8 8.
EPA (U.S. Environmental Protection Agency). 2021b. Human Health Toxicity Values for
Hexafluoropropylene Oxide (HFPO) Dimer Acid and Its Ammonium Salt (CASRN 13252-
13-6 and CASRN 62037-80-3). Also Known as "GenXChemicals". EPA/822/R-21/010.
EPA, Office of Water, Health and Ecological Criteria Division, Washington, DC.
https://www.epa.gov/chemical-research/human-health-toxicitv-assessments-genx-
chemicals.
EPA (U.S. Environmental Protection Agency). 2021c. Assessing and Managing Chemicals under
TSCA: Fact Sheet PFOA Stewardship Program, https://www.epa.gov/assessing-and-
managing-chemicals-under-tsca/fact-sheet-20102015~pfoa~stewardship~program.
EPA (U.S. Environmental Protection Agency). 202Id. Fifth Unregulated Contaminant
Monitoring Rule. https ://www. epa. gov/dwucmr/fifth-unregulated-contaminant-
monitoring-rule.
104
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
EPA (U.S. Environmental Protection Agency). 202 le. External Peer Review Draft Framework
for Estimating Noncancer Health Risks Associated with Mixtures of Per- and
Polyfluoroalkyl Substances (PFAS). EPA 822D/21/003. EPA Office of Water and Office
of Research and Development, Washington, DC.
EPA (U.S. Environmental Protection Agency). 2021f. National PFAS Testing Strategy:
Identification of Candidate Per- and Polyfluoroalkyl Substances (PFAS) for Testing.
EPA, Office of Chemical Safety and Pollution Prevention, Office of Pollution Prevention
and Toxics, https://www.epa.gov/system/files/documents/2021-10/pfas-natl-test-
strategy.pdf
EPA (U.S. Environmental Protection Agency). 2022a. Interim Drinking Water Health Advisory:
Perfluorooctanoic Acid (PFOA) CASRN 335-67-1. EPA/822/R-22/003. EPA, Office of
Water, Health and Ecological Criteria Division, Washington, DC.
https://www.epa.eov/sdwa/drinkine-water-health-advisories-has.
EPA (U.S. Environmental Protection Agency). 2022b. Interim Drinking Water Health Advisory:
Perfluorooctane Sulfonic Acid (PFOS) CASRN 1763-23-1. EPA/822/R-22/004. EPA,
Office of Water, Health and Ecological Criteria Division, Washington, DC.
https://www.epa.eov/sdwa/drinkine-water-health-advisories-has.
EPA (U.S. Environmental Protection Agency). 2022c. Drinking Water Health Advisory:
Hexafluoropropylene Oxide (HFPO) Dimer Acid (CASRN 13252-13-6) andHFPO
Dimer Acid Ammonium Salt (CASRN 62037-80-3), Also Known as "GenX Chemicals. "
EPA/822/R-22/005. EPA, Office of Water, Health and Ecological Criteria Division,
Washington, DC. https://www.epa.eov/sdwa/drinkine-water-health-advisories-has.
EPA (U.S. Environmental Protection Agency). 2022d. Drinking Water Health Advisory:
Perfluorobutane Sulfonic Acid (CASRN 375-73-5) and Related Compound Potassium
Perfluorobutane Sulfonate (CASRN 29420-49-3). EPA/822/R-22/006. EPA, Office of
Water, Health and Ecological Criteria Division, Washington, DC.
https://www.epa.eov/sdwa/drinkine-water-health-advisories-has.
EPA (U.S. Environmental Protection Agency). 2022e. IRIS ToxicologicalReview of
Perfluorobutanoic Acid (PFBA, CASRN 375-22-4) and Related Salts. EPA/63 5/R-
22/277Fa. EPA, Office of Research and Development, Washington, DC.
https://iris.epa.eov/static/pdfs/0701tr.pdf.
EPA (U.S. Environmental Protection Agency). 2023a. EPA Response to Final Science Advisory
Board Recommendations (August 2021) on Four Draft Support Documents for the EPA's
Proposed PFAS National Primary Drinking Water Regulation. EPA-815-D23-001. EPA
Office of Water, Washington, DC.
EPA (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. EPA, Office of Water, Washington, DC.
105
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
EPA (U.S. Environmental Protection Agency). 2023c. PUBLIC COMMENT DRAFT Toxicity
Assessment and Proposed Maximum Contaminant Level Goal for Perfluorooctane
Sulfonic Acid (PFOS) in Drinking Water. EPA-822-P-23-007. EPA, Office of Water,
Washington, DC.
EPA (U.S. Environmental Protection Agency). 2023d. PUBLIC COMMENT DRAFT Appendix
- Toxicity Assessment and Proposed Maximum Contaminant Level Goal for
Perfluorooctanoic Acid (PFOA) in Drinking Water. EPA-822-P-23-006. EPA, Office of
Water, Washington, DC.
EPA (U.S. Environmental Protection Agency). 2023e. PUBLIC COMMENT DRAFT Appendix
- Toxicity Assessment and Proposed Maximum Contaminant Level Goal for
Perfluorooctane Sulfonic Acid (PFOS) in Drinking Water. EPA-822-P-23-008. EPA,
Office of Water, Washington, DC.
EPA SAB (U.S. Environmental Protection Agency Science Advisory Board). 2022. Review of
EPA's Analysis to Support EPA 's National Primary Drinking Water Rulemaking for
PFAS. EPA-SAB-22-008.
https:// sab. epa. eov/ords/ sab/f? i > 100 i s I i90947S * - * * - M * _601#report
EU (European Union). 2020. Commission Staff Working Document: Poly- andperfluoroalkyl
substances (PFAS). Accessed October 2021.
https://ec.eiiropa.eii/environment/pdf/chemicals/2020/10/SWD PFAS.pdf.
Evans, N., J.M. Conley, M. Cardon, P. Hartig, E. Medlock-Kakaley, and L.E. Gray. 2022. In
vitro activity of a panel of per- and polyfluoroalkyl substances (PFAS), fatty acids, and
pharmaceuticals in peroxisome proliferator-activated receptor (PPAR) alpha, PPAR
gamma, and estrogen receptor assays. Toxicology and Applied Pharmacology
449:116136. https://doi.org/10.1016/i.taap.2022
Evich, M.G., M.J.B. Davis, J.P. McCord, B. Acrey, J.A. Awkerman, D.R.U. Knappe, A.B.
Lindstrom, T.F. Speth, C. Tebes-Stevens, M.J. Stryner, Z. Wang, E.J. Weber, W.M.
Henderson, and J.W. Washington. 2022. Per- and polyfluoroalkyl substances in the
environment. Science 375(512). d i/science.abg9065.
Finney, D.J. 1971. Pr obit Analysis. 3rd ed. Cambridge University Press, New York, NY.
Florida Health. 2020. Per- and Polyfluoroalkyl Substances (PFAS).
https://www.floridahealth.gov/environmental-health/hazardous-waste-sites/contaminant-
facts/ documents/fmal-faq-pfas-upda 22.pdf.
FQPA (Food Quality Protection Act). 1996. Food Quality Protection Act of 1996. Public Law
August 3, 1996, 104-170.
Furr, J. R., C.S. Lambright, V.S. Wilson, P.M. Foster, and L.E. Gray, Jr. 2014. A short-term in
vivo screen using fetal testosterone production, a key event in the phthalate adverse
outcome pathway, to predict disruption of sexual differentiation. Toxicological Sciences
140(2):403-424.
106
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Gale, R. W., J.M. Bergeron, E.J. Willingham, and D. Crews. 2002. Turtle sex determination
assay: mass balance and responses to 2,3,7,8-tetrachlorodibenzo-p-dioxin and 3,3',4,4',5-
pentachlorobiphenyl. Environmental Toxicology and Chemistry 21:2477-2482.
Gennings, C., W.H. Carter, Jr., E.W. Carney, G.D. Charles, B.B. Gollapudi, and R.A. Carchman.
2004. A novel flexible approach for evaluating fixed ratio mixtures of full and partial
agonists. Toxicological Sciences 80(1): 134-150.
Gennings, C., W.H. Carter Jr., R.A. Carchman, L.K. Teuschler, J.E. Simmons, and E.W. Carney.
2005. A unifying concept for assessing toxicological interactions: changes in slope.
Toxicological Sciences 88(2):287-297.
Goodrow, S.M., B. Ruppel, R.L. Lippincott, G.B. Post, andN.A. Procopio. 2020. Investigation
of levels of perfluoroalkyl substances in surface water, sediment and fish tissue in New
Jersey, USA. Science of the Total Environment 729:138839.
Gray, L.E., A. Hotchkiss, C.J. Wolf, J.R. Furr, M.G. Price, C.R. Lambright, L.G. Parks, V.S.
Wilson, K.L. Bobseine, P.C. Hartig, and J.S. Ostby. 2001. Adverse effects of
Antiandrogenic Pesticide and Toxic Substances on Reproductive Development in the
Male. Presented at Society for the Study of Reproduction, July 28-August 1, 2001,
Ottawa, Ontario, Canada.
Gray, L.E., J.R. Furr, J.M. Conley, C.S. Lambright, N. Evans, M.C. Cardon, V.S.Wilson, P.M.
Foster, and P.C. Hartig. 2019. A conflicted tale of two novel AR antagonists in vitro and
in vivo: pyrifluquinazon vs. bisphenol C. Toxicological Sciences 168(2):632-643.
Gray, L. E., J. M. Conley, C. S. Lambright and J. R. Furr. 2022. In utero exposure to a mixture of
the perfluoroalkyl-isopropyl pesticide pyrifluquinazon with dibutyl phthalate
cumulatively disrupts male rat reproductive development via different mechanisms of
action. Toxicological Sciences 188(2):234-247. doi !3/toxsci/kfac059
Guelfo, J.L., and D.T. Adamson. 2018. Evaluation of a national data set for insights into sources,
composition, and concentrations of per- and polyfluoroalkyl substances (PFAS) in U.S.
drinking water. Environmental Pollution 236:505-513.
Gutleb, A. C., J. Appelman, M. Bronkhorst, J.H. van den Berg, and A.J. Murk. 2000. Effects of
oral exposure to polychlorinated biphenyls (PCBs) on the development and
metamorphosis of two amphibian species (Xenopus laevis and Rana temporaria). Science
of the Total Environment 262(1-2): 147-157.
Haas, U., M. Scholze, S. Christiansen, M. Dalgaard, A.M. Vinggaard, M. Axelstad, S.B.
Metzdorff, and A. Kortenkamp. 2007. Combined exposure to anti-androgens exacerbates
disruption of sexual differentiation in the rat. Environmental Health Perspectives
115(Suppl 1): 122-128.
Hannas, B.R., C.S. Lambright, J. Furr, K.L. Howdeshell, V.S. Wilson, and L.E. Gray, Jr. 2011.
Dose-response assessment of fetal testosterone production and gene expression levels in
rat testes following in utero exposure to diethylhexyl phthalate, diisobutyl phthalate,
diisoheptyl phthalate, and diisononyl phthalate. Toxicological Sciences 123(1):206-216.
107
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Hannas, B.R., K.L. Howdeshell, J. Furr, and L.E. Gray, Jr. 2013. In utero phthalate effects in the
female rat: amodel forMRKH syndrome. Toxicology Letters 223:315-321.
Hansen, K. J., H. Johnson, J. Eldridge, J. Butenhoff and L. Dick. 2002. Quantitative
characterization of trace levels of PFOS and PFOA in the Tennessee River.
Environmental Science & Technology 36(8): 1681-1685.
Hass, U., S. Christiansen, M. Axelstad, M. Scholze, and J. Boberg. 2017. Combined exposure to
low doses of pesticides causes decreased birth weights in rats. Reproductive Toxicology
72:97-105.
Health Canada. 2018. Guidelines for Canadian Drinking Water Quality: Guideline Technical
Document - Perjluorooctane Sulfonate (PFOS). Health Canada, Ottawa, Ontario.
https://www.canada.ca/content/dam/canada/health-canada/mieration/healthy-
canadians/publications/healthy-living-vie-saine/guidelines-canadian-drinking-water-
qualitv-guideline-technical-document-perfluorooctane-sulfonate/PFQS%202018-
'%20ENG.pdf.
Hodgson, E. 2012. Metabolic interactions of environmental toxicants in humans. Progress in
Molecular Biology and Translational Science 112:349-372.
Houck, K.A., G. Patlewicz, A.M. Richard, A.J. Williams, M.A. Shobair, M. Smeltz, M.S.
Clifton, B. Wetmore, A. Medvedev, and S. Makarov. 2021. Bioactivity profiling of per-
and polyfluoroalkyl substances (PFAS) identifies potential toxicity pathways related to
molecular structure. Toxicology 457:152789.
Houde, M., A. Trevor, D. Bujas, J. Small, R. S. Wells, P. A. Fair, G. D. Bossart, K.R. Solomon,
and D. Muir. 2006. Biomagnification of perfluoroalkyl compounds in the bottlenose
dolphin (Tursiops truncatus) food web. Environmental Science & Technology 40(13):
4138-4144.
Howdeshell, K. L., V.S. Wilson, J. Furr, C.R. Lambright, C.V. rider, C.R. Blystone, A.K.
Hotchkiss, and L.E. Gray, Jr. 2008. A mixture of five phthalate esters inhibits fetal
testicular testosterone production in the sprague-dawley rat in a cumulative, dose-additive
manner. ToxicologicalSciences 105(1): 153-165.
Howdeshell, K.L., C.V. Rider, V.S. Wilson, J.R. Furr, C.R. Lambright, and L.E. Gray. 2015.
Dose addition models based on biologically relevant reductions in fetal testosterone
accurately predict postnatal reproductive tract alterations by a phthalate mixture in rats.
Toxicological Sciences 148(2):488-502.
Howdeshell, K.L., A.K. Hotchkiss, and L.E. Gray, Jr. 2017. Cumulative effects of
antiandrogenic chemical mixtures and their relevance to human health risk assessment.
International Journal of Hygiene and Environmental Health 220:179-188.
Ishibashi, H., M. Hirano, E. Y. Kim, and H. Iwata. 2019. In vitro and in silico evaluations of
binding affinities of perfluoroalkyl substances to Baikal seal and human peroxisome
proliferator-activated receptor alpha. Environmental Science & Technology 53 (4) :2181—
2188.
108
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
ITRC (Interstate Technology Regulatory Council). 2022. PFAS Technical and Regulatory
Guidance Document. ITRC, Washington, DC. https://pfas-l .itrcweb.org.
Jacob, C.C., R. Reimschuessel, L.S. Von Tungeln, G.R. Olson, A.R. Warbritton, D.G. Hattan,
F.A. Beland, and G.G. da Costa. 2011. Dose-response assessment of nephrotoxicity from
a 7-day combined exposure to melamine and cyanuric acid in F344 rats. Toxicological
Sciences 119:391-397.
Jarvis, A.L., J.R. Justice, M.C. Elias, B. Schnitker, and K. Gallagher. 2021. Perfluorooctane
sulfonate in US ambient surface waters: a review of occurrence in aquatic environments
and comparison to global concentrations. Environmental Toxicology and Chemistry
40(9):2425-2442.
Johnson, K.J., S.S. Auerbach, and E. Costa. 2020. A rat liver transcriptomic point of departure
predicts a prospective liver or non-liver apical point of departure. Toxicological Sciences
176(1)86-102.
Jonker, D, R.A. Woutersen, and V.J. Feron. 1996. Toxicity of mixtures of nephrotoxi cants with
similar or dissimilar mode of action. Food and Chemical Toxicology 34:1075-1082.
Judson, R.S., R.J. Kavlock, R.W. Setzer, E.A. Cohen Hubal, M.T. Martin, T.B Knudsen, K.A.
Houck, R.S. Thomas, B.A. Wetmore, and D.J. Dix. 2011. Estimating toxicity-related
biological pathway altering doses for high-throughput chemical risk assessment.
Chemical Research in Toxicology 24(4):451-462. d 2.
Kahn, H.D., K. Stralka, and P.D. White. 2013. Estimates of water ingestion in formula by infants
and children based on USDA's 1994-1996 and 1998 continuing survey of food intakes by
individuals. Human and Ecological Risk Assessment: An International Journal
19(4):888-889
Kannan, K., L. Tao, E. Sinclair, S.D. Pastva, D.J. Judem and J.P. Giesy. 2005. Perfluorinated
compounds in aquatic organisms at various trophic levels in a Great Lakes food chain.
Archives of Environmental Contamination and Toxicology 48(4):559-566.
Kavlock, R., K. Chandler, K. Houck, S. Hunter, . Judson, N. Kleinstreuer, T. Knudsen, M.
Martin, S. Padilla, D. Reif, A. Richard, D. Rotroff, N. Sipes, and D. Dix. 2012. Update on
EPA's ToxCast program: providing high throughput decision support tools for chemical
risk management. Chemical Research in Toxicology 25(7): 1287-1302.
d 021/1x3000939.
Kjeldsen, L.S., and E.C. Bonefeld-Jorgensen. 2013. Perfluorinated compounds affect the
function of sex hormone receptors. Environmental Science and Pollution Research
International 20(11):8031-8044. dot:
Konwick, B.J., G.T. Tomy, N. Ismail, J.T. Peterson, R.J. Fauver, D. Higginbotham, and A.T.
Fisk. 2008. Concentrations and patterns of perfluoroalkyl acids in Georgia, USA surface
waters near and distant to a major use source. Environmental Toxicology and Chemistry:
An International Journal 27(10):2011-2018.
109
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Kortenkamp, A., M. Faust, M. Scholze, and T. Backhaus. 2007. Low level exposure to multiple
chemicals: Reason for human health concerns? Environmental Health Perspectives
115(Suppl. 1): 106-114.
Kortenkamp, A., and U. Haas. 2009. Expert Workshop on Combination Effects of Chemicals, 28-
20 January 2009. Workshop Report. Hornbaek, Denmark.
Kotlarz, N., J. McCord, D. Collier, C.S. Lea, M. Strynar, A.B. Lindstrom, A.A. Wilkie, J.Y.
Islam, K. Matney, P. Tarte, M.E. Polera, K. Burdette, J. DeWitt, K. May, R.C. Smart, D.
R.U. Knappe, and J. A. Hoppin. 2020. Measurement of novel, drinking water-associated
pfas in blood from adults and children in Wilmington, North Carolina. Environmental
Health Perspectives 128(7). d 289/EHP6837.
Lau, C., J.R. Thibodeaux, R. G. Hanson, J.M. Rogers, B. E. Grey, M.E. Stanton, J.L. Butenhoff,
and L.A. Stevenson. 2003. Exposure to perfluorooctane sulfonate during pregnancy in rat
and mouse. II: postnatal evaluation. Toxicological Sciences 74(2):382-392.
Lau, C., J. R. Thibodeaux, R. G. Hanson, M. G. Narotsky, J. M. Rogers, A. B. Lindstrom and M.
J. Strynar. 2006. Effects of perfluorooctanoic acid exposure during pregnancy in the
mouse. Toxicological Sciences 90(2): 510-518. doi: 3/toxsci/kfj 105.
Lioy, P. J., R. Hauser, C. Gennings, H.M. Koch, P.E. Mirkes, B.A. Schwetz, and A.
Kortenkamp. 2015. Assessment of phthalates/phthalate alternatives in children's toys and
childcare articles: review of the report including conclusions and recommendation of the
Chronic Hazard Advisory Panel of the Consumer Product Safety Commission. Journal of
Exposure Science and Environmental Epidemiology 25:343-353.
Lorber, M., and P.P. Egeghy. 2011. Simple intake and pharmacokinetic modeling to characterize
exposure of Americans to perfluorooctanoic acid, PFOA. Environmental Science &
Technology 45(19):8006-8014.
LSRO (Life Sciences Research Office, Federation of American Societies for Experimental
Biology). 1995. Third Report on Nutrition Monitoring in the United States: Volume 1.
Prepared for the Interagency Board for Nutrition Monitoring and Related Research. US
Government Printing Office, Washington, DC.
Luebker, D.J., M.T. Case, R.G. York, J.A. Moore, K.J. Hansen and J.L. Butenhoff. 2005a. Two-
generation reproduction and cross-foster studies of perfluorooctanesulfonate (PFOS) in
rats. Toxicology 215(1-2): 126-148.
Luebker, D.J., R.G. York, K.J. Hansen, J.A. Moore and J.L. Butenhoff. 2005b. Neonatal
mortality from in utero exposure to perfluorooctanesulfonate (PFOS) in Sprague-Dawley
rats: dose-response, and biochemical and pharamacokinetic parameters. Toxicology
215(1-2): 149—169.
Maine DEP (Maine Department of Environmental Protection). 2021. Maine PFAS Screening
Levels. Accessed July 2021. https://www.maine.gov/dep/spills/topics/pfas/Maine-PFAS-
Screening-Levels-Rev-6.2 [f.
110
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Marques, E.S., J. Agudelo, E.M. Kaye, S.M.S. Modaresi, M. Pfohl, J. Becanova, W. Wei, M.
Polunas, M. Goedken, and A.L. Slitt. 2021. The role of maternal high fat diet on mouse
pup metabolic endpoints following perinatal PFAS and PFAS mixture exposure.
Toxicology 462:152921.
Marques, E., M. Pfohl, W. Wei, G. Tarantola, L. Ford, O. Amaeze, J. Alesio, S. Ryu, X. Jia, H.
Zhu, G. D. Bothun and A. Slitt. 2022. Replacement per- and polyfluoroalkyl substances
(PFAS) are potent modulators of lipogenic and drug metabolizing gene expression
signatures in primary human hepatocytes. Toxicology and Applied Pharmacology
442:115991. dot: 10.1016/i.taap.2022
Martin, O., M. Scholze, S. Ermler, J. McPhie, S.K. Bopp, A. Kienzler, N. Parissis, and A.
Kortenkamp. 2021. Ten years of research on synergisms and antagonisms in chemical
mixtures: a systematic review and quantitative reappraisal of mixture studies.
Environment International 146:106206.
Mass DEP (Massachusetts Department of Environmental Protection). 2019. Technical Support
Document, Per- and Polyfluoroalkyl Substances (PFAS): An Updated Subgroup
Approach to Groundwater and Drinking Water Values. Accessed July 2019.
https://www.mass.eov/files/documents/2019/12/27/ o20TSD%202019-12-
26%2 OF IN AI.. pdf.
Medvedev, A,L.A. Medvedeva, E. Martsen, M. Moeser, K.L. Gorman, B. Lin, B. Blackwell,
D.L. Villeneuve, K.A. Houck, K.M. Crofton, and S.S. Makarov. 2020. Harmonized cross
species assessment of endocrine disruptors by Ecotox FACTORIAL assay.
Environmental Science and Technology 54(19): 12142-12153.
Menger, F., J. Pohl, L. Ahrens, G. Carlsson and S. Orn. 2020. Behavioural effects and
bioconcentration of per- and polyfluoroalkyl substances (PFASs) in zebrafish (Danio
rerio) embryos. Chemosphere 245:125573. C- 10 101 )/i.chemosphere.201° I
McCord, J., and M. Strynar. 2019. Identification of per-and polyfluoroalkyl substances in the
cape fear river by high resolution mass spectrometry and nontargeted screening.
Environmental Science & Technology 53(9):4717-4727.
McCord, J.P., M.J. Strynar, J.W. Washington, E.L. Bergman, and S.W. Goodrow. 2020.
Emerging chlorinated polyfluorinated polyether compounds impacting the waters of
southwestern New Jersey identified by use of nontargeted analysis. Environmental
Science & Technology Letters 7(12):903-908.
Metzdorff, S.B., M. Dalgaar, S. Christiansen, M. Axelstad, U. Hass, M.K. Kiersgaard, M.
Scholze, A. Kortenkamp, and A.M. Vinggaard. 2007. Dysgenesis and histological
changes of genitals and perturbations of gene expression in male rats after in utero
exposure to antiandrogen mixtures. Toxicological Sciences 98:87-98.
MN DOH (Minnesota Department of Health). 2021. Human Health-Based Water Guidance
Table. Accessed August 27, 2021.
https://www.health.state.mn.iis/commimities/environment/risk/euidance/ew/table.html.
Ill
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Monosson, E. 2000. Reproductive and developmental effects of PCBs in fish: a synthesis of
laboratory and field studies. Reviews in Toxicology 3(1):25—75.
Moser, V.C., M. Casey, A. Hamm, W.H. Carter, J.E. Simmons, and C. Gennings. 2005.
Neurotoxicological and statistical analyses of a mixture of five organophosphorus
pesticides using a ray design. Toxicological Sciences 86(1): 101-115.
Moser, V.C., J.E. Simmons, and C. Gennings. 2006. Neurotoxicological interactions of a five-
pesticide mixture in preweanling rats. Toxicological Sciences 92:235-245.
Moser, V.C., S. Padilla, J.E. Simmons, L.T. Haber, and R.C. Hertzberg. 2012. Impact of
chemical proportions on the acute neurotoxicity of a mixture of seven carbamates in
preweanling and adult rats. Toxicological Sciences 129(1): 126-134.
MT DEQ (Montana Department of Environmental Quality). 2020. PFAS and Your Health.
https://deq.mt.eov/files/DEQAdmin/PFAS/JulY%202021%20PFAS%20Fact%20Sheet.pdf.
Mumtaz, M., M.C. Buser, and H.R Pohl. 2021. Per- and polyfluoroalkyl mixtures toxicity
assessment "proof-of-concept" illustration for the hazard index approach. Journal of
Toxicology and Environmental Health, Part A 84(13):553-567.
doi 10 1080/15287394.2011 l ^ M _ l
Murphy, S. D., and K.P. Dubois. 1957. Quantitative measurement of inhibition of the enzymatic
detoxification of malathion by EPN (ethyl p-nitrophenyl thionobenzenephosphonate).
Proceedings of the Society for Experimental Biology and Medicine 96:813-818.
Mwanza, J. C., D.F. Lyke, R.C. Hertzberg, L. Haber, M. Kohrman-Vincent, R. Li, Y. Pan, R.H.
Lyles, J.E. Simmons, D.K. Macmillan, R.D. Zehr, A.E. Swank, and D.W. Herr. 2012.
Cholinesterase inhibition and depression of the photic after discharge of flash evoked
potentials following acute or repeated exposures to a mixture of carbaryl and
propoxur. Neurotoxicology 33(3):332-346.
Nakayama, S., M.J. Strynar, L. Helfant, P. Egeghy, X. Ye, and A.B. Lindstrom. 2007.
Perfluorinated compounds in the Cape Fear drainage basin in North Carolina.
Environmental Science & Technology 41 (15):5271-5276.
NASEM (National Academies of Sciences, Engineering, and Medicine. 2022. Guidance on
PFAS Exposure, Testing, and Clinical Follow-up. Washington, DC: The National
Academies Press, https://doi.org/10.17226/26156.
NATO (North Atlantic Treaty Organization). 1988. Pilot Study on International Information
Exchange on Dioxins and Related Compounds, Waste Disposal Sites Contaminated with
Dioxins and Related Compounds. NATO Report 175. NATO, Brussels, Belgium.
NC DEQ (North Carolina Department of Environmental Quality). 2021. Managing Emerging
Compounds in Water. NC DEQ, Raleigh, NC. https://deq.nc.gov/news/kev-
issues/emerging-compounds/managing-emerging-compounds-water#groundwater-and-
surface-water-qualitv-standards-actions
112
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
NCHS (National Center for Health Statistics). 1993. Joint policy on variance estimation and
statistical reporting standards on NHANES III and CSFII reports: HNIS/NCHS Analytic
Working Group Recommendations. Human Nutrition Information Service
(HNIS)/Analytic Working Group. Agricultural Research Service, Survey Systems/Food
Consumption Laboratory, Riverdale, MD.
Nielsen, G., W.J. Heiger-Bernays, J.J. Schlezinger, and T.F. Webster. 2022. Predicting the
effects of per- and polyfluoroalkyl substance mixtures on peroxisome proliferator-
activated receptor alpha activity in vitro. Toxicology 465:153024.
d /i.tox.2021.153024.
NRC (National Research Council). 2008. Phthalates and Cumulative Risk Assessment: The
Tasks Ahead. The National Academies Press, Washington, DC. dc ?226/12528.
NTP (National Toxicology Program). 2016. NTP Monograph: Immunotoxicity Associated with
Exposure to Perjluorooctanoic Acid (PFOA) or Perfluorooctane Sulfonate (PFOS). U.S.
Department of Health and Human Services, NTP, Research Triangle Park, NC.
NTP (National Toxicology Program). 2019a. NTP Technical Report on the Toxicity Studies of
Perfluoroalkyl Carboxylates (Perfluorohexanoic Acid, Perfluorooctanoic Acid,
Perfluorononanoic Acid, andPerfluorodecanoic Acid) Administered by Gavage to
Spr ague-Daw ley (Hsd: Spr ague-Daw ley SD) Rats. NTP Toxicity Report 97. U.S.
Department of Health and Human Services, NTP, Research Triangle Park, NC.
NTP (National Toxicology Program). 2019b. NTP Technical Report on the Toxicity Studies of
Perfluoroalkyl Sulfonates (Perfluorobutane Sulfonic Acid, Perfluorohexane Sulfonate
Potassium Salt, and Perfluorooctane Sulfonic Acid) Administered by Gavage to Sprague-
Dawley (Hsd: Spr ague-Daw ley SD) Rats. NTP Toxicity Report 96. U.S. Department of
Health and Human Services, NTP, Research Triangle Park, NC.
OECD (Organisation for Economic Co-operation and Development). 2018. Toward a New
Comprehensive Global Database of Per- and Polyfluoroalkyl Substances (PFASs):
Summary Report on Updating the OECD 2007 List of Per and Polyfluoroalkyl
Substances (PFASs). Series on Risk Management No. 39. OECD, Paris, France.
https://www.oecd.org/offi cialdocuments/publicdisplaydocumentpdf/?cote=ENV-JM-
MQNO(2Q18)7&doclanguage=en.
OECD (Organisation for Economic Co-operation and Development). 2021. Reconciling
Terminology of the Universe of Per- and Polyfluoroalkyl Substances: Recommendations
and Practical Guidance. Series on Risk Management No. 61. OECD, Paris, France.
https://www.oecd.org/offi cialdocuments/publicdisplavdocumentpdf/?cote=ENV/CBC/M
ONO(2Q21 )25&docLamguage=em.
Ohio EPA (Environmental Protection Agency). 2019. Per- and Polyfluoroalkyl Substances
(PFAS). https://epa.ohio.gov/pfas ublic-drinking-water.
Ojo, A.F., C. Peng and J.C. Ng. 2021. Assessing the human health risks of per- and
polyfluoroalkyl substances: a need for greater focus on their interactions as mixtures.
Journal of Hazardous Materials 407:124863. dc 'i.ihazmat.2020.124863.
113
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Olmstead, A.W., and G.A. Leblanc. 2005. Toxicity assessment of environmentally relevant
pollutant mixtures using a heurisitc model. Integrated Environmental Assessment and
Management 1 (2): 114— 122.
Padilla, P. 2006. Cumulative Effects of Organophosphorus or Carbamate Pesticides. Chapter 41
in Toxicology of Organophosphate and Carbamate Compounds, pp. 607-615. EPA,
Research Triangle Park, NC.
Parish, S.T., M. Aschner, W. Casey, M. Corvaro, M.R. Embry, S. Fitzpatrick, D. Kidd, N.C.
Kleinstreuer, B.S. Lima, R.S. Settivari, D.C. Wolf, D. Yamazaki, and A. Boobis. 2020.
An evaluation framework for new approach methodologies (NAMs) for human health
safety assessment. Regulatory Toxicology and Pharmacology 112:104592.
d /i.vrtph.2020.104592.
Paul-Friedman, K., M. Gagne, L. Loo, P. Karamertzanis, T. Netzeva, T. Sobanski, J.A. Franzosa,
A.M. Richard, R.R. Lougee, A. Gissi, J.J. Lee, M. Angrish, J.L. Dome, S. Foster, K.
Raffaele, T. Bahadori, M.R. Gwinn, J. Labert, M. Whelan, M. Rasenberg, T. Barton-
Maclaren, and R.S. Thomas. 2020. Utility of in vitro bioactivity as a lower bound
estimate of in vivo adverse effect levels and in risk-based prioritization. Toxicological
Sciences 173(l):202-225.
Podder, A., A.H.M.A. Sadmani, D. Reinhart, N.B. Chang, and R. Goel. 2021. Per and poly-
fluoroalkyl substances (PFAS) as a contaminant of emerging concern in surface water: a
transboundary review of their occurrences and toxicity effects. Journal of Hazardous
Materials 419:126361.
Poothong, S., E. Papadopoulou, J.A. Padilla-Sanchez, C. Thomsen, and L.S. Haug. 2020.
Multiple pathways of human exposure to poly- and perfluoroalkyl substances (PFASS):
from external exposure to human blood. Environment International 134:105244.
Remucal, C. K. 2019. Spatial and temporal variability of perfluoroalkyl substances in the
Laurentian Great Lakes. Environmental Sciences: Processes and Impacts 21(11): 1816-
1834.
Restum, J. C., J.P. Giesy, J.A. Render, E.B. Shipp, D.A Vergrugge, and R.J. Aulerich. 1998.
Multigenerational study of the effects of consumption of PCB-contaminated carp from
Saginaw Bay, Lake Huron, on mink. 1. Effects on mink reproduction, kit growth and
survival, and selected biological parameters. Journal of Toxicological and Environmental
Health, Part A 54(5):343-375.
Richard, A.M., R.S. Judson, K.A. Houck, C.M. Grulke, P. Volarath, I. Thillainadarajah, C. Yang,
J. Rathman, M.T. Martin, J.F. Wambaugh, T.B. Knudsen, J. Kancherla, K. Mansouri, G.
Patlewicz, A.J. Williams, S.B. Little, K.M. Crofton, and R.S. Thomas. 2016. ToxCast
Chemical Landscape: Paving the Road to 21st Century Toxicology. Chemical Research
in Toxicology 29(8): 1225-1251. doi: 10.1021 /acs.chemrestox.6b00135.
114
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
RIDEM (Rhode Island Department of Environmental Management). 2017. Management
Determination of a Groundwater Quality Standardfor: Perfluorooctanoic Acid (PFOA)
and Perfluorooctane Sulfonate (PFOS).
http ://www. dem .ri. gov/program s/benviron/water/ qualitv/pdf/pfoa.pdf.
Rider, C.V., and G.A. LeBlanc. 2005. An integrated addition and interaction model for assessing
toxicity of chemical mixtures. Toxicological Sciences 87(2):520-528.
d /toxsci/kfi247.
Rider, C.V., J. Furr, V.S. Wilson, and L.E. Gray, Jr. 2008. A mixture of seven antiandrogens
induces reproductive malformations in rats. International Journal of
Andrology 31(2):249-262.
Rider, C.V., V.S. Wilson, K.L. Howdeshell, A.K. Hotchkiss, J.R. Furr, C.R. Lambright, and L.E.
Gray, Jr. 2009. Cumulative effects of in utero administration of mixtures of
"antiandrogens" on male rat reproductive development. Toxicologic Pathology
37(1): 100—113.
Rider, C.V., J. Furr, V.S. Wilson, and L.E. Gray, Jr. 2010. Cumulative effects of in utero
administration of mixtures of reproductive toxicants that disrupt common target tissues
via diverse mechanisms of toxicity. International Journal of Andrology 33(2):443-462.
RIVM. 2018. Mixture exposure to PFAS: A Relative Potency Factor Approach. RIVM Report
2018-0070. National Institute for Public Health and the Environment, Bilthoven, The
Netherlands, https://www.rivm.nl/bibliotheek/rapporten/2018-0070.pdf.
Rosen, M. B., J.R. Thibodeaux, C.R. Wood, R.D. Zehr, J.E. Schmid, and C. Lau. 2007. Gene
expression profiling in the lung and liver of PFOA-exposed mouse fetuses. Toxicology
239(1-2): 15-33.
Rosen, M. B., K.P. Das, J. Rooney, B. Abbott, C. Lau, and J.C. Corton. 2017. PPARalpha-
independent transcriptional targets of perfluoroalkyl acids revealed by transcript
profiling. Toxicology 387:95-107.
Roth, K., Z. Yang, M. Agarwal, W. Liu, Z. Peng, Z. Long, J. Birbeck, J. Westrick, W. Liu, and
M.C. Petriello. 2021. Exposure to a mixture of legacy, alternative, and replacement per-
and polyfluoroalkyl substances (PFAS) results in sex-dependent modulation of
cholesterol metabolism and liver injury. Environment International 157:106843.
Rotroff, D.M., B.A. Wetmore, D.J. Dix, S.S. Ferguson, H.J. Clewell, K.A. Houck, E.L.
Lecluyse, M.E. Anderson, R.S. Judson, C.M. Smith, M.A. Sochaski, R.J. Kavlock, F.
Boellmann, M.T. Martin, D. Reif, J.F. Wambaugh, and R.S. Thomas. 2010. Incorporating
human dosimetry and exposure into high-throughput in vitro toxicity screening.
Toxicological Sciences 117:348-358.
Ruyle, B.J., C.P. Thackray, J.P. McCord, M.J. Strynar, K.A. Mauge-Lewis, S.E. Fenton, and
E.M. Sunderland. 2021. Reconstructing the composition of per- and polyfluoroalkyl
substances in contemporary aqueous film-forming foams. Environmental Science &
Technology Letters 8( 1): 59-65. d /acs.estlett.QcQQ798.
115
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Safe, S.H. 1994. Polychlorinated biphenyls (PCBs): environmental impact, biochemical and
toxic responses, and implications for risk assessment. Critical Reviews in Toxicology
24(2):87-149.
SARA (Superfund Amendments and Reauthorization Act). 2002. Superfund Amendments and
Reauthorization Act of 1986. Public Law December 31, 2002, 107-377.
SDWA (Safe Drinking Water Act). (1996). Safe Drinking Water Act Amendments of 1996.
Public Law August 6, 1996, 104-182.
Sedlak, M.D., J.P. Benskin, A. Wong, R. Grace, and D.J. Greig. 2017. Per- and polyfluoroalkyl
substances (PFASs) in San Francisco Bay wildlife: temporal trends, exposure pathways,
and notable presence of precursor compounds. Chemosphere 185:1217-1226.
Silva, E., N. Rajapakse, and A. Kortenkamp. 2002. Something from "nothing"—eight weak
estrogenic chemicals combined at concentrations below NOECs produce significant
mixture effects. Environmental Science & Technology 36:1751-1756.
Simcik, M. F., and K.J. Dorweiler. 2005. Ratio of perfluorochemical concentrations as a tracer of
atmospheric deposition to surface waters. Environmental Science & Technology
39(22): 8678-8683.
Simmons, J.E., L.K. Teuschler, C. Gennings, T.F. Speth, S.D. Richardson, R.J. Miltner, M.G.
Narotsky, K.D. Schenck, E.S. Hunter, R.C. Hertzberg, and G. Rice. 2004. Component-
based and whole-mixture techniques for addressing the toxicity of drinking-water
disinfection by-product mixtures. Journal of Toxicology and Environmental Health, Part
A 67(8-10):741-754. d 080/15287390490428215.
Sinclair, E., and K. Kannan. 2006. Mass loading and fate of perfluoroalkyl surfactants in
wastewater treatment plants. Environmental Science & Technology 40(5): 1408-1414.
Svendsgaard, D.J., and R.C. Hertzberg. 1994. Statistical methods for the toxicological evaluation
of the additivity assumption as used in the EPA chemical mixture risk assessment
guidelines. Toxicology of Chemical Mixtures: Case Studies, Mechanisms, and Novel
Approaches 23:599-642.
Swedish Food Agency. 2021. PFAS in Drinking Water and Self-Caught Fish - Risk
Management, https://www.livsmedelsverket.se/en/production-control-and-trade/drinking-
water-production-and-control/t.
Takacs, M.L., and B.D. Abbott. 2007. Activation of mouse and human peroxisome proliferator-
activated receptors (alpha, beta/delta, gamma) by perfluorooctanoic acid and
perfluorooctane sulfonate. Toxicological Sciences 95(1): 108-117.
Thibodeaux, J R., R.G. Hanson, J.M. Rogers, B E. Grey, B.D. Barbee, J.H. Richards, J.L.
Butenhoff, L.A. Stevenson, and C. Lau. 2003. Exposure to perfluorooctane sulfonate
during pregnancy in rat and mouse. I: maternal and prenatal evaluations. Toxicological
Sciences 74(2):369-381.
116
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Thomas R.S., H.J. Clewell 3rd, B.C. Allen, S.C. Wesselkamper, N.C.Y. Wang, J.C. Lambert, J.K.
Hess-Wilson, Q.J. Zhao, and M.E. Anderson. 2011. Application of transcriptional
benchmark dose values in quantitative cancer and noncancer risk assessment.
Toxicological Sciences 120:194-205.
Thomas R.S., S.C. Wesselkamper, N.C. Wang, Q.J. Zhao, D.D. Petersen, J.C. Lambert, I. Cote,
L. Yang, E. Healy, M.B. Black, H.J. Clewell 3rd, B.C. Allen, and M.E. Andersen. 2013.
Temporal concordance between apical and transcriptional points of departure for
chemical risk assessment. Toxicological Sciences 134(1): 180-194.
Tinwell, H., and J. Ashby. 2004. Sensitivity of the immature rat uterotrophic assay to mixtures of
estrogens. Environmental Health Perspectives 112:575-582.
Van den Berg, M., L.S. Birnbaum, M. Denison, M.D. Vito, W. Farland, M. Feeley, H. Fiedler, H.
Hakansson, A. Hanberg, L. Haws, M. Rose, S. Safe, D. Schrenk, C. Tohyama, A.
Tritscher, J. Tuomisto, M. Tysklind, N. Walker, andR.E. Peterson. 2006. The 2005 World
Health Organization reevaluation of human and mammalian toxic equivalency factors for
dioxins and dioxin-like compounds. Toxicological Sciences 93(2):223-241.
Van Der Ven, L.T.M., P.V. Ommeren, E.P. Zwart, E.R. Gremmer, H.M. Odemaekers, H.J.
Heusinkveld, J.D. van Klaveren, and E. Rorije. 2022. Does addition in the induction of
craniofacial malformations in zebrafish embryos exposed to a complex mixture of food-
relevant chemicals with dissimilar modes of action. Environmental Health Perspectives
130(4).
Vanden Heuvel, J.P., J.T. Thompson, S.R. Frame and P.J. Gillies. 2006. Differential activation of
nuclear receptors by perfluorinated fatty acid analogs and natural fatty acids: a
comparison of human, mouse, and rat peroxisome proliferator-activated receptor-alpha, -
beta, and -gamma, liver X receptor-beta, and retinoid X receptor-alpha. Toxicological
Sciences 92(2):476-489.
VT DEC (Vermont Department of Environmental Conservation). 2021. PFAS & Drinking Water
Information Page. VT DEC, Drinking Water & Groundwater Protection Division,
Montpelier, VT. Accessed July 2021. https://dec.vermont.gov/water/drinking-water/pfas.
Walker, N.J., P.W. Crockett, A. Nyska, A.E. Brix, M.P. Jokinen, D.M. Sells, J.R. Hailey, M.
Easterling, J.K. Haseman, M. Yin, J.R. Bucher, and C.J. Portier. 2005. Dose-additive
carcinogenicity of a defined mixture of "dioxin-like compounds". Environmental Health
Perspectives 113(1):43—48.
Wambaugh, J.F., B.A. Wetmore, R. Pearce, C. Strope, R. Goldsmith, J.P. Sluka, A. Sedykh, A.
Tropsha, S. Bosgra, I. Shah, R. Judson, R.S. Thomas, and R.W. Setzer. 2015.
Toxicokinetic triage for environmental chemicals. Toxicological Sciences 147(l):55-67.
Wambaugh, J., M. Hughes, C. Ring, D. MacMillan, J. Ford, T. Fennell, S. Black, R. Snyder, N.
Sipes, B. Wetmore, J. Westerhout, W. Setzer, R. Pearce, J. Simmons, and R. Thomas.
2018. Evaluating in vitro-in vivo extrapolation of toxicokinetics. Toxicological Sciences
163(1): 152—169.
117
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Wang, N.C.Y., Q.J. Zhao, S.C. Wesselkamper, J.C. Lambert, D. Petersen, and J.K. Hess-Wilson.
2012. Application of computational toxicological approaches in human health risk
assessment. I. A tiered surrogate approach. Regulatory Toxicology and Pharmacology
63(1): 10—19. dc i.vrtph.2012.02.006.
Wang, Z., J.C. DeWitt, C.P. Higgins, and I.T. Cousins. 2017. A never-ending story of per- and
polyfluoroalkyl substances (PFASs)? Environmental Science & Technology 51(5):2508-
2518.
Wetmore, B.A., J.F. Wambaugh, S.S. Ferguson, M.A. Sochaski, D.M. Rotroff, K. Freeman, H.J.
Clewell 3rd, D.J. Dix, M.E. Andersen, K.A. Houck, B. Allen, R.S. Judson, R. Singh, R.J.
Kavlock, A.M. Richard, and R.S. Thomas. 2012. Integration of dosimetry, exposure, and
high-throughput screening data in chemical toxicity assessment. Toxicological Sciences
125(1): 157-74.
Wetmore, B.A., B. Allen, H.J. Clewell 3rd, T. Parker, J.F. Wambaugh, L.M. Almond, M.A.
Sochaski, and R.S. Thomas. 2014. Incorporating population variability and susceptible
subpopulations into dosimetry for high-throughput toxicity testing. Toxicological
Sciences 142(l):210-224.
Wetmore, B.A., J.F. Wambaugh, B. Allen, S.S. Ferguson, M.A. Sochaski, R.w. Setzer, K.A.
Houck, C.L Strope, K. Cantwell, R.S. Judson, E. LeCluyse, H.J. Clewell, R.S. Thomas,
and M.E. Andersen. 2015. Incorporating high-throughput exposure predictions with
dosimetry-adjusted in vitro bioactivity to inform chemical toxicity testing. Toxicological
Sciences 148(1)121-36.WHO (World Health Organization). 2008. Melamine and
Cyanuric acid: Toxicity, Preliminary Risk Assessment and Guidance on Levels in Food.
WHO, Geneva, Switzerland.
https://web.archive.Org/web/20080927021953/http://www.who.int/foodsafetv/fs manage
ment/infosan events/en/indexl .html.
WIDHS (Wisconsin Department of Health Services). 2019a. Perfluorooctanoic acid (PFOA)
Scientific Support Document. WI DHS, Madison, WI.
https://dnr.wi.gov/topic/Contaminants/documents/pfas/ >ci entificSupport.pdf.
WI DHS (Wisconsin Department of Health Services). 2019b. Perfluorooctane Sulfonic acid
(PFOS) Scientific Support Document. WI DHS, Madison, WI.
https://dnr.wi.gov/topic/Contaminants/dociiments/pfas/PFOSScientificSiipport.pdf.
Wolansky, M.J., C. Gennings, M.J. DeVito, and K.M. Crofton. 2009. Evidence for dose-additive
effects of pyrethroids on motor activity in rats. Environmental Health Perspectives
117:1563-1570.
Wolf, C.J., J.E. Schmid, C. Lau and B.D. Abbott. 2012. Activation of mouse and human
peroxisome proliferator-activated receptor-alpha (PPARalpha) by perfluoroalkyl acids
(PFAAs): further investigation of C4-C12 compounds. Reproductive Toxicology
33(4^:546-551. doi: 6/i.reprotox.2011.09.009.
118
-------
PUBLIC REVIEW DRAFT
FEBRUARY 2023
Wolf, C.J., C.V. Rider, C. Lau, and B.D. Abbott. 2014. Evaluating the additivity of
perfluoroalkyl acids in binary combinations on peroxisome proliferator-activated
receptor-alpha activation. Toxicology 316:43-54.
Wu, S., K. Blackburn, J. Amburgey, J. Jaworska, and T. Federle. 2010. A framework for using
structural, reactivity, metabolic and physicochemical similarity to evaluate the suitability
of analogs for SAR-based toxicological assessments. Regulatory Toxicology and
Pharmacology 56(1):67-81.
Yamazoe, Y., T. Yamada, and K. Nagata. 2020. Prediction and characterization of CYP3A4-
mediated metabolisms of azole fungicides: an application of the fused-grid template*
system. Food Safety (Tokyo) 8:34-51.
Yasunaga, R., J. Ikuta, Y. Murata, K. Inoue, H. Koga, T. Masaki, and H. Tamura. 2013. Ligand-
independent androgen receptor antagonism caused by the newly developed pesticide
pyrifluquinazon (PFQ). Reproductive Toxicology 35(1): 1-6.
Yu, C.H., C.D. Riker, S. Lu, and Z. Fan. 2020. Biomonitoring of emerging contaminants,
perfluoroalkyl and polyfluoroalkyl substances (PFAS), in New Jersey adults in 2016-
2018. International Journal of Hygiene and Environmental Health 223(1): 34-44.
Zareitalabad, P., J. Siemens, M. Hamer, and W. Amelung. 2013. Perfluorooctanoic acid (PFOA)
and perfluorooctanesulfonic acid (PFOS) in surface waters, sediments, soils and
wastewater - a review on concentrations and distribution coefficients. Chemosphere
91(6):725-732.
Zhao, Y., K. Zhang, J.P. Giesy, and J. Hu. 2015. Families of Nuclear Receptors in Vertebrate
Models: Characteristic and Comparative Toxicological Perspective. Scientific Reports
5:8554. doi: 8/srep08554.
119
------- |