A EPA
EPA/635/R-19/201
IRIS Assessment Protocol
www.epa.gov/iris
Systematic Review Protocol for the Polychlorinated Biphenyls (PCBs)
Noncancer IRIS Assessment (Preliminary Assessment Materials)
[CASRN 1336-36-3]
December 2019
Integrated Risk Information System
Center for Public Health and Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
DISCLAIMER
This document is a public comment draft for review purposes only. This information is
distributed solely for the purpose of public comment It has not been formally disseminated by
EPA. It does not represent and should not be construed to represent any Agency determination or
policy. Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.
This document is a draft for review purposes only and does not constitute Agency policy.
ii DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
CONTENTS
CONTENTS iii
AUTHORS | CONTRIBUTORS | REVIEWERS viii
1. INTRODUCTION 1
2. SCOPING AND INITIAL PROBLEM FORMULATION SUMMARY 2
2.1. BACKGROUND 2
2.2. SCOPING SUMMARY 5
2.3. PROBLEM FORMULATION 6
2.4. ASSESSMENT APPROACH 11
2.5. KEY SCIENCE ISSUES 11
2.5.1. Impact of Congener Profile on the Toxicity of PCB Mixtures 11
2.5.2. Potential for Hazard Identification and Dose-Response Assessment for PCB
Exposure via Inhalation 13
2.5.3. Suitability of Available Pharmacokinetic Models for Reliable Route-to-Route,
Interspecies, or Intraspecies Extrapolation 13
3. OVERALL OBJECTIVES, SPECIFIC AIMS, AND POPULATIONS, EXPOSURES, COMPARATORS,
AND OUTCOMES (PECO) CRITERIA 14
3.1. SPECIFIC AIMS 14
3.2. POPULATIONS, EXPOSURES, COMPARATORS, AND OUTCOMES (PECO) CRITERIA 15
4. LITERATURE SEARCH AND SCREENING STRATEGIES 17
4.1. LITERATURE SEARCH STRATEGIES 17
4.2. NONPEER-REVIEWED DATA 18
4.3. LITERATURE SCREENING STRATEGY 19
4.3.1. Electronic Screening 20
4.3.2. Title and abstract-level screening 23
4.3.3. Full-text level screening 23
4.3.4. Multiple Publications of the Same Data 26
4.4.SUMMARY-LEVEL LITERATURE INVENTORIES 26
5. REFINED EVALUATION PLAN 28
6. STUDY EVALUATION (REPORTING, RISK OF BIAS, AND SENSITIVITY) STRATEGY 50
6.1.STUDY EVALUATION OVERVIEW FOR HEALTH EFFECT STUDIES 50
This document is a draft for review purposes only and does not constitute Agency policy.
iii DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
6.2. EPIDEMIOLOGY STUDY EVALUATION 54
6.3. EXPERIMENTAL ANIMAL STUDY EVALUATION 65
6.4. PHYSIOLOGICALLY BASED PHARMACOKINETIC (PBPK) MODEL DESCRIPTIVE SUMMARY
AND EVALUATION 77
6.4.1. Pharmacokinetic (PK)/Physiologically Based Pharmacokinetic (PBPK) Model
Descriptive Summary 77
6.4.2. Pharmacokinetic (PK)/Physiologically Based Pharmacokinetic (PBPK) Model
Evaluation 78
6.5. MECHANISTIC STUDY EVALUATION 80
7. ORGANIZING THE HAZARD REVIEW 81
8. DATA EXTRACTION OF STUDY METHODS AND RESULTS 83
8.1.STANDARDIZING REPORTING OF EFFECT SIZES 84
8.2.STANDARDIZING ADMINISTERED DOSE LEVELS/CONCENTRATIONS 84
9. SYNTHESIS OF EVIDENCE 85
9.1.SYNTHESES OF HUMAN AND ANIMAL HEALTH EFFECT EVIDENCE 89
9.2. MECHANISTIC INFORMATION 90
10. EVIDENCE INTEGRATION 98
10.1. EVALUATING THE STRENGTH OF THE HUMAN AND ANIMAL EVIDENCE STREAMS 101
10.2. OVERALL EVIDENCE INTEGRATION JUDGMENTS 104
10.3. HAZARD CONSIDERATIONS FOR DOSE-RESPONSE 110
11. DOSE-RESPONSE ASSESSMENT: STUDY SELECTION AND QUANTITATIVE ANALYSIS 112
11.1. SELECTING STUDIES FOR DOSE-RESPONSE ASSESSMENT 113
11.2. CONDUCTING DOSE-RESPONSE ASSESSMENTS 116
11.2.1. Dose-response Analysis in the Range of Observation 116
11.2.2. Extrapolation: Slope Factors and Unit Risks 118
11.2.3. Extrapolation: Reference Values 118
12. PROTOCOL HISTORY 121
REFERENCES 122
APPENDICES 128
APPENDIX A. ELECTRONIC DATABASE SEARCH STRATEGIES 128
APPENDIX B. DATA EXTRACTION FIELDS 131
This document is a draft for review purposes only and does not constitute Agency policy.
iv DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
TABLES
Table 1. EPA program and regional offices interest in a new assessment of PCBs (September 2018) 5
Table 2. Noncancer PCB toxicity values from U.S. federal agencies and international bodies for
exposures in the general population 8
Table 3. Preliminary literature survey: PCB studies by test system, route of exposure, and health effect
category 9
Table 4. Populations, exposures, comparators, outcomes (PECO) criteria 16
Table 5. Electronic prioritization of literature for hazard identification 22
Table 6. Sources of studies subjected to manual review for relevance to hazard identification 23
Table 7. Refined PECO criteria (cardiovascular, dermal, endocrine, gastrointestinal, hematopoietic,
immune, metabolic, musculoskeletal, nervous system, ocular, respiratory, and urinary
effects) 29
Table 8. Refined PECO criteria (developmental effects) 41
Table 9. Refined PECO criteria (hepatobiliary effects) 43
Table 10. Refined PECO criteria (reproductive effects) 46
Table 11. Questions to guide the development of criteria for each domain in epidemiology studies 56
Table 12. Information relevant to evaluation domains for epidemiology studies 65
Table 13. Questions to guide the development of criteria for each domain in experimental animal
toxicology studies 67
Table 14. Example descriptive summary for a physiologically based pharmacokinetic (PBPK) model
study 77
Table 15. Criteria for evaluating physiologically based pharmacokinetic (PBPK) models 79
Table 16. Querying the evidence to organize syntheses for human and animal evidence 81
Table 17. Information most relevant to describing primary considerations informing causality during
evidence syntheses 86
Table 18. Individual and social factors that could increase susceptibility to exposure-related health
effects 89
Table 19. Preparation for the analysis of mechanistic evidence 92
Table 20. Examples of iterative questions and considerations that focus the synthesis and application
of mechanistic information for evidence integration and dose-response analysis 95
Table 21. Evidence profile table template 100
Table 22. Considerations that inform evaluations of the strength of the human and animal evidence. 102
Table 23. Evidence integration judgments for characterizing potential human health hazards in the
evidence integration narrative 106
Table 24. Attributes used to evaluate studies for derivation of toxicity values 114
Table A-l. Database search strategy 128
Table B-l. Key data extraction elements to summarize study design, experimental model,
methodology, and results 131
This document is a draft for review purposes only and does not constitute Agency policy.
v DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
FIGURES
Figure 1. IRIS systematic review problem formulation and method documents 1
Figure 2. Chemical structure of PCBs (ATSDR, 2000) 3
Figure 3. Schematic illustration of electronic prioritization of literature depicting references clustered
by similarity using natural language processing 20
Figure 4. Illustration depicting clusters containing relevant seed references (circled blue clusters) 21
Figure 5. Visualization of identified clusters 22
Figure 6. Literature search flow diagram for PCBs 25
Figure 7. Number of human and animal studies of PCB exposure and cardiovascular effects 32
Figure 8. Number of human and animal studies of PCB exposure and dermal effects 33
Figure 9. Number of human and animal studies of PCB exposure and endocrine effects 33
Figure 10. Number of human and animal studies of PCB exposure and gastrointestinal effects 34
Figure 11. Number of human and animal studies of PCB exposure and hematopoietic effects 34
Figure 12. Number of human and animal studies of PCB exposure and immune effects 35
Figure 13. Number of human and animal studies of PCB exposure and metabolic effects 36
Figure 14. Number of human and animal studies of PCB exposure and musculoskeletal effects 37
Figure 15. Number of human and animal studies of PCB exposure and nervous system effects 38
Figure 16. Number of human and animal studies of PCB exposure and ocular effects 39
Figure 17. Number of human and animal studies of PCB exposure and respiratory effects 40
Figure 18. Number of human and animal studies of PCB exposure and urinary system effects 41
Figure 19. Number of human and animal studies of PCB exposure and developmental effects 43
Figure 20. Number of human and animal studies of PCB exposure and hepatobiliary effects 45
Figure 21. Number of human and animal studies of PCB exposure and reproductive effects 48
Figure 22. Number of studies evaluating health outcomes in each health effect category based on the
results of the literature search illustrated in Figure 6 49
Figure 23. Overview of IRIS study evaluation process 51
Figure 24. Process for evidence integration 99
This document is a draft for review purposes only and does not constitute Agency policy.
vi DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
ABBREVIATIONS
AD ME absorption, distribution, metabolism, and excretion
BMR benchmark response
CAS Chemical Abstracts Service
CASRN Chemical Abstracts Service registry number
CERCLA Comprehensive Environmental Response, Compensation, and Liability Act
CPAD Chemical and Pollutant Assessment Division
CPHEA Center for Public Health and Environmental Assessment
EPA Environmental Protection Agency
GLP good laboratory practices
HAWC Health Assessment Workspace Collaborative
HEEAD Health and Environmental Effects Assessment Division
HERO Health and Environmental Research Online
HPASB Hazardous Pollutant Assessment and Systems Branch
IRIS Integrated Risk Information System
LOAEC lowest-observed-adverse-effect concentration
LOAEL lowest-observed-adverse-effect level
MOA mode of action
NAM new approach method
NHANES National Health and Nutrition Examination Survey
NOAEC no-observed-adverse-effect concentration
NOAEL no-observed-adverse-effect level
NTP National Toxicology Program
OECD Organisation for Economic Co-operation and Development
OLEM Office of Land and Emergency Management
ORD Office of Research and Development
PCB polychlorinated biphenyl
PBPK physiologically based pharmacokinetic
PECO Populations, Exposures, Comparators, and Outcomes
PK pharmacokinetic
POD point of departure
RfC inhalation reference concentration
RfD oral reference dose
ROBINS-I Risk of Bias in Nonrandomized Studies of Interventions
TEAB-R Toxic Effects Assessment Branch-RTP
UF uncertainty factor
This document is a draft for review purposes only and does not constitute Agency policy.
vii DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
AUTHORS | CONTRIBUTORS | REVIEWERS
Assessment Team (U.S. EPA/Center for Public Health and Environmental Assessment
[CPHEA])
Geniece Lehmann (Co-Assessment Manager)
Laura Carlson (Co-Assessment Manager)
Xabier Arzuaga
Evan Coffman
Allen Davis
Jeff Gift
Dustin Kapraun
Danelle Lobdell
Anuradha Mudipalli
Paul Schlosser
John Stanek
Michele Taylor
Michael Wright
Erin Yost
Technical Experts/Contributors
Michael Bloom
Pam Factor-Litvak
ToddJusko
Aileen Keating
Carolyn Klocke
Pam Lein
John Meeker
Pradeep Rajan
Larry Robertson
Sharon Sagiv
Alexander Sergeev
Michal Toborek
Martina Velichkovska
University at Albany School of Public Health
Columbia University Medical Center
University of Rochester Medical Center
Iowa State University College of Agriculture and Life Sciences
UC Davis College of Biological Sciences
UC Davis College of Biological Sciences
University of Michigan School of Public Health
Pradeep Rajan LLC
University of Iowa College of Public Health
UC Berkeley School of Public Health
Ohio University College of Health Sciences and Professions
University of Miami Miller School of Medicine
University of Miami Miller School of Medicine
This document is a draft for review purposes only and does not constitute Agency policy.
viii DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Executive Direction
Wayne Cascio
Samantha Jones
Kristina Thayer
John Vandenberg
James Avery
Belinda Hawkins
Andrew Kraft
Paul White
Janice Lee
Andrew Hotchkiss
U.S. EPA/CPHEA/Director
U.S. EPA/CPHEA/Associate Director
U.S. EPA/CPHEA/Chemical and Pollutant Assessment Division
(CPAD)/Director
U.S. EPA/CPHEA/Health and Environmental Effects
Assessment Division (HEEAD)/Director
U.S. EPA/CPHEA/CPAD/Associate Director
U.S. EPA/CPHEA/CPAD/Senior Science Advisor
U.S. EPA/CPHEA/CPAD/Senior Science Advisor
U.S. EPA/CPHEA/CPAD/Senior Science Advisor
U.S. EPA/CPHEA/CPAD/Toxic Effects Assessment Branch-
RTP (TEAB-R)/Chief
U.S. EPA/CPHEA/HEEAD/Hazardous Pollutant Assessment
and Systems Branch (HPASB]/Chief
Production Team and Review (U.S. EPA/CPHEA)
Anna Champlin
Ryan Jones
Dahnish Shams
Vicki Soto
Contractor Support (ICF)
Robyn Blain
Dave Burch
Michelle Cawley
Anna Engstrom
Ali Goldstone
Brandall Ingle
Revathi Muralidharan
Sean Robins
Maryjane Selgrade
Kelly Shipkowski
Raquel Silva
Joanne Trgovcich
Arun Varghese
Nicole Vetter
This document is a draft for review purposes only and does not constitute Agency policy.
ix DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
1. INTRODUCTION
In April 2015, the U.S. Environmental Protection Agency (EPA) released scoping and
problem formulation materials for a new Integrated Risk Information System (IRIS) assessment to
address noncancer human health hazards associated with exposure to complex mixtures of
polychlorinated biphenyls (PCBs). An update of the existing evaluation of cancer risk from PCB
exposure (https://cfpub.epa.gov/ncea/iris2/chemicalLanding.cfm7substance nmbr=294) was not
identified as a priority need, and a new assessment of PCB cancer risk is not planned at this time.
The scoping and problem formulation materials were presented at a public science meeting on June
17-18, 2015 (https://www.epa.gov/iris/iris-public-meeting-iun-2015) to seek input from the
scientific community and interested parties on the IRIS Program's scoping and problem
formulation conclusions and identification of key areas of scientific complexity (U.S. EPA. 2015b).
This protocol document presents the objectives and specific aims of the assessment, the draft PECO
(Populations, Exposures, Comparators, and Outcomes) criteria, and methods for conducting the
systematic review and dose-response analysis. While the scoping and problem formulation
materials described what the assessment will cover, this protocol describes how the assessment
will be conducted (see Figure 1). The IRIS Program posts assessment protocols on its website and
in the Zenodo repository (https://zenodo.org/). Public input received is considered during
preparation of the draft assessment and any adjustments to the protocol will be reflected in an
updated version released in conjunction with the draft assessment Literature search results are
made available in EPA's Health and Environmental Research Online database (HERO). The PCB
project literature page will be updated annually with literature updates and can be found online
(https://hero.epa.gov/hero/index.cfm/proiect/page/proiect id/384).
Scoping
Systematic
Review Protocol
Literature
Inventory
Study
Evaluation
Data
Extraction
Evidence
Integration
Derive Toxicity
Values
Assessment
Initiated
Initial Problem
Formulation
Scoping and
Problem
Formulation:
What the
assessment
will cover
Literature Refined Organize Evidence Analysis Select and Model
Search Analysis Plan Hazard Review and Synthesis Studies
Protocols: How the assessment will be conducted
Assessment
Developed
Figure 1. IRIS systematic review problem formulation and method
documents.
This document is a draft for review purposes only and does not constitute Agency policy.
1 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
2. SCOPING AND INITIAL PROBLEM
FORMULATION SUMMARY
2.1. BACKGROUND
PCBs are a class of synthetic compounds characterized by a biphenyl structure with
chlorine substitutions atup to 10 positions, as shown in Figure 2. There are 209 possible PCB
congeners based on the various combinations of the numbers and positions of the chlorine
substitutions on the biphenyl molecule; PCB congeners vary in structure, stability, and toxicity
(Section 2.5.1). PCBs were manufactured and marketed in the United States between about 1930
and 1977 under the trade name Aroclor (e.g., Aroclors 1016,1242,1248,1254,1260). They were
used in many industrial applications because of their electrical insulating properties, chemical
stability, and relative inflammability. They were widely used in capacitors, transformers, and other
electrical equipment, and as coolants and lubricants. Other applications included use in
plasticizers, surface coatings, inks, adhesives, flame retardants, pesticide extenders, paints,
carbonless duplicating paper, and sealants and caulking compounds (ATSDR. 2000). EPA issued
final regulations banning the manufacture of PCBs and phasing out most PCB uses in 1979 under
the Toxic Substances Control Act (TSCA) (40 CFR 761) due to evidence that they persist and
accumulate in the environment and can cause toxic effects fhttp://www2.epa.gov/aboutepa/epa-
bans-pcb-manufacture-phases-out-usesl. Despite the ban on manufacturing, PCBs continue to be
present in environmental media (e.g., air, soil, sediment, food) and are redistributed from one
environmental compartment to another (ATSDR. 2000). They also can be released through the
continued use and disposal of PCB-containing products and as a result of inadvertent production
during certain manufacturing processes (Vorkamp. 2015). PCB-containing building materials such
as window glazes, fluorescent light ballasts, ceiling tile coatings, caulk, paints, and floor finishes are
potential sources of PCBs in the indoor environment fLehmann etal.. 20151. Over time, the
congener profile of PCB mixtures (e.g., Aroclors) can be transformed in the environment, leading to
diverse human exposures (Section 2.5.1).
This document is a draft for review purposes only and does not constitute Agency policy.
2 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
5
5'
6'
6
X = 1 to 5, Y = 1 to 5
Figure 2. Chemical structure of PCBs fATSDR. 20001.
Occupational exposure to PCBs can occur through inhalation and dermal contact at
workplaces where PCBs are present (e.g., handling PCB-containing electrical equipment, spills, or
waste-site materials without use of personal protective equipment to limit exposure) (ToxNet
Hazardous Substances Data. 20111. In the general population, PCB exposure occurs primarily via
dietary intake of contaminated food and inhalation of PCB-contaminated air fLehmann etal.. 2015:
ATSDR. 20001. The major contributors to dietary exposure to PCBs include fatty foods such as fish,
meat, and dairy products.
Inhalation also has been shown to be a contributor to total PCB exposure, especially in
indoor settings where PCB sources exist (Lehmann etal.. 2015: Harrad etal.. 20091. For example,
elevated indoor air PCB concentrations have been observed in some public school buildings. The
schools at highest risk of having elevated indoor air PCB concentrations are those that were built in
the 1950s-1970s and schools that were extensively remodeled during this period fMarek etal..
2017: Thomas etal.. 20121. Since September 2009, EPA has released several reports1 for school
administrators and building managers with important information about identifying and if present,
managing airborne PCBs, and tools to help minimize possible exposure.
General population exposure also can occur via dermal contact with PCBs in soil or other
media or through incidental ingestion of PCB-contaminated soil or dust fATSDR. 20001. The
presence of PCBs in blood, adipose tissue, and breast milk of non-occupationally exposed members
of the general population of the United States provides evidence of widespread exposure fXue etal..
2014: CDC. 2009: ATSDR. 20001.
Populations with potentially greater than average exposures include those who consume
PCB-contaminated fish or wild game or who eat a higher proportion of food grown in PCB-
1 Polychlorinated Biphenyls (PCBs) in School Buildings: Sources, Environmental Levels, and Exposures, EPA-600-
R-12-051 fhttps://www.epa.gov/sites/production/files/2015-08/documents/pch epa600rl2051 final.pdfl:
Fact Sheet for Schools: Caulk Containing PCBs May Be Present in Older Schools and Buildings, EPA-747-F-09-
003 (https://nepis.epa.gov/Exe/ZyPURL.cgi?Dockey=P1005ClD.TXT];
Proper Maintenance, Removal, and Disposal of PCB-Containing Fluorescent Light Ballasts (FLBs) in School
Buildings: A Guide for School Administrators and Maintenance Personnel
fhttps://www.epa.gov/sites/production/files/documents/PCBsInBallasts.pdfl: and
How to Test for PCBs and Characterize Suspect Material (https://www.epa.gov/pcbs/how-test-pcbs-and-
characterize-suspect-materials].
This document is a draft for review purposes only and does not constitute Agency policy.
3 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
contaminated areas flARC. 2015: ATSDR. 20001. Because many PCBs tend to accumulate in body
lipids and can be transferred to infants via breast milk, nursing infants are another potentially
highly exposed population. Certain occupational groups also can have elevated exposures through
inhalation, dermal contact, or incidental ingestion of PCB residues from contact with contaminated
materials in the workplace, during repair and maintenance of electrical equipment containing PCBs,
or from accidents or fires involving PCBs.
The IRIS database currently provides assessments for specific Aroclor mixtures:
quantitative assessments for Aroclor 1016 and Aroclor 1254 and a qualitative discussion for
Aroclor 1248. Although oral reference doses (RfDs) were derived for Aroclor 1016 and Aroclor
1254, no inhalation reference concentrations (RfCs) are available for PCBs.
Aroclor 1016 (posted in 1993;
https://cfpub.epa.gov/ncea/iris2/chemicalLanding.cfm7substance nmbr=462): The
Aroclor 1016 assessment derived an oral RfD based on data reported by fSchantz etal..
19891. in which perinatal toxicity and long-term neurobehavioral effects of Aroclor 1016
were evaluated in infant rhesus monkeys born to dams exposed at 0.007 or
0.028 mg/kg-day for 7 months prior to breeding until offspring were weaned at age
4 months. Based on reduced birth weights and neurobehavioral deficits of prenatally
exposed monkeys, the 0.028 mg/kg-day dose was identified as the lowest-observed-
adverse-effect level (LOAEL). The study no-observed-adverse-effect level (NOAEL) of
0.007 mg/kg-day was chosen for the point of departure (POD), yielding an RfD
of 7 x 10"5 mg/kg-day after application of a total uncertainty factor (UF) of 100, accounting
for intra- and interspecies variability, subchronic study duration, and database limitations.
Aroclor 1248 (posted in 1994;
https://cfpub.epa.gov/ncea/iris2/chemicalLanding.cfm7substance nmbr=6491: The
Aroclor 1248 assessment concluded that the health effect data were inadequate for the
derivation of an oral RfD. Derivation of an RfD was not recommended because a frank effect
(i.e., infant death) was noted at the lowest dose tested in rhesus monkeys. In the same set of
studies used for the Aroclor 1016 assessment, Schantzetal. T19891 evaluated
neurobehavioral performance in the offspring of rhesus monkeys exposed to 0.03, 0.1, and
0.2 mg/kg-day of dietary Aroclor 1248 for different durations. Infant death occurred at the
lowest dose and appeared to be a dose-responsive effect, leading to the identification of
0.03 mg/kg-day as a frank effect level.
Aroclor 1254 (posted in 1994;
https://cfpub.epa.gov/ncea/iris2 /chemicalLanding.cfm7substance nmbr=3891: Arnold et
al. f!993bl. Arnold etal. f!993al. Trvphonas etal. f!9891. Trvphonas etal. f!991al. and
Trvphonas etal. (1991b) were used in the derivation of the oral RfD for Aroclor 1254. In
these studies, mature female rhesus monkeys were exposed to 0.005, 0.02, 0.04, or
0.08 mg/kgbody weight Aroclor 1254 each day over 6.5 years. The low dose of
0.005 mg/kg-day was identified as the LOAEL based on immunotoxicity and observations of
eye exudate, inflammation or prominence of the eyelid tarsal glands, and nail lesions. The
RfD of2 x lO-5 mg/kg-day was calculated by applying a total UF of300, which accounted for
intra- and interspecies variability, subchronic study duration, and the use of a LOAEL as the
POD.
This document is a draft for review purposes only and does not constitute Agency policy.
4 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
2.2. SCOPING SUMMARY
Since the current IRIS assessments for noncancer health effects of Aroclor mixtures were
completed in 1993-1994, studies on the noncancer health effects of exposure to environmentally
relevant PCB mixtures (e.g., similar to those found in contaminated fish or human milk) have been
conducted, and new data are available.
The commercial manufacture of PCBs was banned in the United States in 1979. Since that
time, their use, manufacture, cleanup, and disposal have been regulated under TSCA (40 CFR 761).
However, as discussed above, because of the past widespread use and persistence of PCBs in the
environment, humans continue to be exposed to them by inhalation of volatilized PCBs, inhalation
of contaminated dust, contact with contaminated dust, contact with primary or secondary sources
of PCBs, and ingestion of foods contaminated with PCBs, including breast milk. In addition to
regulation under TSCA, PCBs are regulated under the Clean Water Act, the Safe Drinking Water Act,
and the Resource Conservation and Recovery Act Accordingly, PCBs are of interest to several EPA
program offices and regional offices due to widespread human exposure to PCBs from many
sources and through multiple environmental media.
During scoping, the IRIS program met with EPA program and regional offices interested in
an IRIS assessment for PCBs to discuss specific assessment needs. Table 1 provides a summary of
input from this outreach.
Table 1. EPA program and regional offices interest in a new assessment of
PCBs (September 2018)
EPA program or
regional office
Oral
Inhalation
Statutes/regulations and anticipated uses/interest
Office of Land and
Emergency
Management
(OLEM)
y
y
Comprehensive Environmental Response, Compensation, and Liability Act
(CERCLA) and Resource Conservation and Recovery Act (RCRA)
PCBs have been identified at numerous contaminated waste sites, including
492 CERCLA National Priority List (NPL) sites.3 CERCLA authorizes EPA to
conduct short- or long-term cleanups at Superfund sites and later recover
cleanup costs from potentially responsible parties under Section 107. PCB
toxicological information may be used to make risk determinations for
response actions (e.g., CERCLA short-term removals, CERCLA long-term
remedial response actions, or RCRA Corrective Action).
EPA Regions 1-10
a The Superfund Enterprise Management System (SEMS) database identified 492 NPL sites where PCBs were documented as a
contaminant in one or more media. SEMS is the official repository for site- and non-site-specific Superfund data in support of
CERCLA. It contains information on hazardous waste site assessment and remediation from 1982 to the present. These site
numbers are based on contaminant data from the remedy selection administrative records. SEMS only includes remedy data
from fiscal years 1982 to 2014 for final and deleted NPL sites, and for sites with a Superfund Alternative Approach (SAA)
agreement in place. NPL and SAA status is current as of September 17, 2018. The types of the 492 NPL and SAA sites include
Superfund Federal Facility sites and non-Federal Facility sites, Fund-lead sites, and Enforcement sites where CERCLA remedial
actions have been proposed. PCBs, identified as PCB, polychlorinated biphenyl, and Aroclor, were documented in SEMS as a
contaminant in the evaluation of human health and ecological risks. PCBs were documented as a contaminant in a wide range
of media, including air, soil, sediment, surface and ground water, sludge, fish tissue, and debris. Access to site documents or
additional information about individual sites can be found at https://cumulis.epa.eov/supercpad/cursites/srchsites.cfm. This
This document is a draft for review purposes only and does not constitute Agency policy.
5 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
website provides links to site profile pages, which typically include digital copies of site-related decision documents. Regarding
the SEMS database, EPA is providing this PCB site data as a public service and does not vouch for the accuracy, completeness,
or currency of data. Data provided by external parties is not independently verified by EPA. These data are made available to
the public strictly for informational purposes. Data do not represent EPA's official position, viewpoint, or opinion, express or
implied. This information is not intended for use in establishing liability or calculating Cost Recovery Statutes of Limitations
and cannot be relied upon to create any rights, substantive or procedural, enforceable by any party in litigation with the
United States or third parties. EPA reserves the right to change these data at any time without public notice.
A new IRIS assessment will identify noncancer human health hazards associated with
exposure to complex PCB mixtures (such as those found in the environment) through oral,
inhalation, and dermal routes, provided adequate data are available. Discussion of dose-response
information for identified hazards also will be included when feasible because this information can
be useful for characterizing risks at varying exposure levels and analyzing benefits associated with
reducing exposures. Derivation of an RfD for the dermal route of exposure is not planned at this
time because oral and inhalation exposure are generally considered the major exposure routes, and
relatively few studies of toxicological effects following dermal PCB exposure exist. However,
potential risks from dermal exposures can be evaluated using route-to-route extrapolation, and
toxicokinetic and other data relevant to dermal exposure will be included in the assessment to
support those evaluations. Furthermore, no new assessment for PCB cancer risk is planned. The
carcinogenicity of environmentally relevant PCB mixtures is addressed in the IRIS Carcinogenicity
Assessment for PCBs posted in 1996
(https: //cfpub.epa.gov/ncea/iris2 /chemicalLanding.cfm7substance nmbr=294). and an update of
the evaluation of cancer risk from PCB exposure has not been identified as a priority need.
2.3. PROBLEM FORMULATION
Problem formulation information pertaining to the new assessment of PCBs was included in
the scoping and problem formulation materials released to the public in April 2015 (U.S. EPA.
2015b); a public science meeting was held June 17-18, 2015 to obtain public input on these
materials (https://www.epa.gov/iris/iris-public-meeting-iun-2015).
As discussed in U.S. EPA f2015bl. a preliminary literature survey was performed to identify
noncancer health outcomes evaluated for possible associations with PCB exposure. This survey
consisted of a search for health assessment information produced by other federal, state, and
international health agencies (summarized in Table 2), and an additional broad search of PubMed
to locate more recent studies. The review of health assessment information was used to identify
health effect categories for consideration in the IRIS assessment and was supplemented by the
PubMed search covering dates after the publication of the health assessment. In addition, the
preliminary literature survey was used to identify key scientific issues important for assessing
human health risk associated with PCB exposure. The PubMed search was not intended to be a
comprehensive search of all available literature but was intended to identify noncancer health
outcomes that had not been evaluated in prior health assessments.
This document is a draft for review purposes only and does not constitute Agency policy.
6 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
The following health assessments, in addition to EPA's IRIS assessments for Aroclor 1016
fhttps://cfpub.epa.gov/ncea/iris2/chemicalLanding.cfm?substance nmbr=4621. Aroclor 1248
fhttps://cfpub.epa.gov/ncea/iris2/chemicalLanding.cfm?substance nmbr=649). and Aroclor 1254
(https://cfpub.epa.gov/ncea/iris2/chemicalLanding.cfm7substance nmbr=389). are available from
several federal and international health agencies:
1) Agency for Toxic Substances and Disease Registry. (ATSDR) (2011). Addendum to the
Toxicological Profile for Polychlorinated Biphenyls.
http://www.atsdr.cdc.gov/ToxProfiles/pcbs addendum.pdf
2) Agency for Toxic Substances and Disease Registry. ATSDR (2000). Toxicological Profile for
Polychlorinated Biphenyls (PCBs). http://www.atsdr.cdc.gov/ToxProfiles/tpiy.pdf
3) National Institute for Occupational Safety and Health. (NIOSH) (2019b). NIOSH Pocket
Guide to Chemical Hazards. RTECS. Chlorodiphenyl (54% chlorine).
http: / /www. cdc. gov/niosh /np g/np gdO 12 6.html
4) National Institute for Occupational Safety and Health. (NIOSH) (2019a). NIOSH Pocket
Guide to Chemical Hazards. RTECS. Chlorodiphenyl (42% chlorine).
http: //www, cdc. gov/niosh/np g/np gdO 12 5 .html
5) Occupational Safety and Health Administration. (OSHA) (2019). Chemical Sampling
Information, Chlorodiphenyl (42% CI).
https://www.osha. gov/chemicaldata/chemResulthtml?recNo=392
6) Occupational Safety and Health Administration. (OSHA) (2018). Chemical Sampling
Information, Chlorodiphenyl (54% CI).
https://www.osha.gov/chemicaldata/chemResulthtml?recNo=121
7) World Health Organization. (WHO) (2003). Concise International Chemical Assessment
Document 55. Polychlorinated Biphenyls: Human Health Aspects.
http://www.who.int/ipcs/publications/cicad/en/cicad55.pdf
This document is a draft for review purposes only and does not constitute Agency policy.
7 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 2. Noncancer PCB toxicity values from U.S. federal agencies and
international bodies for exposures in the general population
Reference
Risk value or
limit
Mixture
Rationale
ATSDR (2011)
ATSDR (2000)
Chronic MRL:
0.02 ng/kg-d
Intermediate MRL:
0.03 ng/kg-d
Aroclor 1254
Immunological (Trvohonas et al., 1991a;
Trvohonas et al., 1989)
Neurological (Rice, 1999; Rice and Havward,
1999; Rice, 1998,1997; Rice and Havward,
1997)
WHO (2003)
Chronic MRL:
0.02 ng/kg-d
Aroclor 1254
Based on assessment bv ATSDR (2000) and
ATSDR (2011)
U.S. EPA (1994a)
RfD: 0.02 ng/kg-d
Aroclor 1254
Immunological, dermal, ocular (Arnold et al.,
1993b; Arnold et al., 1993a; Trvohonas et al.,
1991a; Trvohonas et al., 1991b; Trvohonas et
al., 1989)
U.S. EPA (1993)
RfD: 0.07 ng/kg-d
Aroclor 1016
Reduced birth weight (Schantz et al., 1991;
Schantz et al., 1989; Levin et al., 1988;
Barsotti and van Miller, 1984)
Overall, the Toxicological Profile for Polychlorinated Biphenyls (PCBs) (ATSDR. 2011. 20001
was found to be the most comprehensive and current resource, including detailed information on
the widest array of health effects and synthesizing evidence from the largest number of primary
research articles. Information from other assessments listed above was included in the preliminary
literature survey to the extent that it added to the information already presented in (ATSDR. 2011.
20001.
The preliminary literature survey identified human, animal, and in vitro studies related to
multiple noncancer health outcomes, mechanisms of action, mode-of-action (MOA) hypotheses,
toxicokinetics, and susceptible lifestages or populations. Each row in Table 3 summarizes whether
data are available on a particular broad health effect category or other toxicologically relevant
topic. Although the checkmarks in Table 3 indicate the existence of studies that investigated certain
health effect categories in the context of PCB exposure, they do not indicate whether the data from
those studies support associations between PCB exposure and health effects in those categories.
Each column in Table 3 indicates the types of studies that are available with respect to test system
(i.e., human, animal, in vitro) and exposure route (i.e., oral or inhalation) for animal studies or
exposure setting (i.e., occupational, high fish or seafood consumption,2 general population) for
2 Studies of populations with "high fish or seafood consumption" were those in which the study authors
identified fish or seafood consumption, or both, as the PCB exposure source presumed to be dominant in the
study population.
This document is a draft for review purposes only and does not constitute Agency policy.
8 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
1 human studies. In addition, the table indicates whether animal studies of subchronic, chronic, or
2 developmental design3 are available.
Table 3. Preliminary literature survey: PCB studies by test system, route of
exposure, and health effect category3
Human studies
Animal studies
In vitro
studies
Occupational
High fish/
seafood
consumption13
General
population
Oral
Inhalation
Health effect categories
Cardiovascular
~
~
~
~
(Subchronic,
Chronic)
~
Dermal
V
V
(Subchronic,
Chronic,
Developmental)
Developmental
V
V
V
V
(Subchronic,
Chronic,
Developmental)
V
(Subchronic)
V
Endocrine
V
V
V
V
(Subchronic,
Developmental)
V
(Subchronic)
V
Gastrointestinal
V
V
V
(Subchronic,
Chronic)
Hematopoietic
V
V
(Subchronic,
Chronic)
V
(Subchronic)
Hepatobiliary
V
V
V
V
(Subchronic,
Chronic,
Developmental)
V
(Subchronic)
V
Immune System
V
V
V
V
(Subchronic,
Chronic,
Developmental)
V
(Subchronic)
V
3 In developmental studies, animals are exposed to a chemical during a critical window of development
(i.e., the developmental period of vulnerability during which adverse effects can be triggered by exposures to
environmental agents or other stressors). The critical windows of development for most biological systems
occur during the prenatal and early postnatal periods, but certain systems (e.g., nervous and reproductive
systems) do continue to develop throughout early life and adolescence. Studies conducted outside a critical
window of development can be characterized by exposure duration: acute (<24 hours), short-term
(>24 hours up to 30 days), subchronic (>30 days up to 10% of lifetime), and chronic (up to a lifetime).
This document is a draft for review purposes only and does not constitute Agency policy.
9 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Human studies
Animal studies
In vitro
studies
Occupational
High fish/
seafood
consumption13
General
population
Oral
Inhalation
Metabolic
V
V
V
V
(Subchronic,
Chronic)
V
Musculoskeletal
V
V
V
(Subchronic,
Chronic)
Nervous System
V
V
V
V
(Subchronic,
Developmental)
V
(Subchronic)
V
Ocular
V
V
(Subchronic,
Chronic,
Developmental)
Reproductive
V
V
V
V
(Subchronic,
Chronic,
Developmental)
V
Respiratory
V
V
V
(Subchronic,
Chronic)
V
(Subchronic)
Urinary System
V
V
V
(Subchronic,
Chronic)
V
(Subchronic)
V
Other data and analyses
ADME
V
V
V
V
V
V
Pharmacokinetic
models0
V
V
V
V
V
MOA
hypotheses
V
V
V
(Subchronic,
Chronic,
Developmental)
V
Susceptibility
datad
V
V
V
V
(Developmental)
Genotoxicity6
V
V
V
(Subchronic)
V
ADME = absorption, distribution, metabolism, and excretion; MOA = mode of action.
a Checkmarks indicate that one or more studies have been identified but do not indicate confidence in the methods used in
those studies, or if those studies support associations between PCB exposure and one or more health effect(s) in that
category; the absence of a checkmark indicates that no studies were identified for a given health effect category and study
design.
This document is a draft for review purposes only and does not constitute Agency policy.
10 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
b Studies of populations with "high fish/seafood consumption" were those in which the study authors identified fish or seafood
consumption, or both, as the PCB exposure source presumed to be dominant in the study population.
c Earliest physiologically based pharmacokinetic (PBPK) models for PCBs were based on intravenous exposure. Models also exist
for dermal exposure.
d Individuals who might be more susceptible to toxic effects include young children, especially those who are breastfed.
e Includes studies investigating potential epigenetic impacts of PCB exposure.
2.4. ASSESSMENT APPROACH
The overall objective of this assessment is to identify adverse human health effects and
characterize exposure-response relationships for the effects of PCB mixtures to support the
development of oral and inhalation noncancer toxicity values. This assessment will use systematic
review methods to evaluate the epidemiological and toxicological literature for PCBs; mechanistic
evidence will also be considered, focusing on data informative to analyses of the key science issues
identified in Section 2.5 (see Section 9.2). The evaluations conducted in this assessment will be
consistent with relevant EPA guidance.4
The specific approach taken to the assessment of the health effects of PCBs will be based on
input received during scoping, a survey of the literature describing the health effects of PCBs, and
consideration of the physicochemical properties of PCBs. The literature noted and screened as
described in Section 2.3 was used to identify broad categories of potential health effects considered
to be most relevant for assessment. U.S. EPA f2015bl proposed that the following list of broad
health effect categories be considered for further evaluation to identify specific health endpoints for
systematic review: cardiovascular, dermal, developmental, endocrine, gastrointestinal,
hematopoietic, hepatobiliary, immune, metabolic, nervous system, ocular and reproductive effects.
After consideration of stakeholder input collected at the public science meeting
fhttps: //www.epa.gov/iris/iris-public-meeting-iun-20151 on June 17-18, 2015, the decision was
made also to include musculoskeletal, respiratory, and urinary effects in this preliminary analysis,
the results of which are described in Section 5.
2.5. KEY SCIENCE ISSUES
As described in U.S. EPA f2015bl and discussed at the public science meeting on June 17-
18, 2015 fhttps: //www.epa.gov/iris/iris-public-meeting-iun-2015I the following key scientific
issues were identified that warrant further consideration in this assessment
2.5.1. Impact of Congener Profile on the Toxicity of PCB Mixtures
Humans are environmentally exposed to PCBs as complex mixtures of congeners. PCB
congeners differ not only structurally but also qualitatively and quantitatively with respect to
biological responses. Prior to human exposure, PCB mixtures in the environment undergo
4EPA guidance documents: http://www.epa.gov/iris/basic-information-about-integrated-risk-information-
svstem#guidance/. Note: the Agency has initiated a review of, and possible updates to, their guidance
documents.
This document is a draft for review purposes only and does not constitute Agency policy.
11 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
processes such as volatilization and preferential bioaccumulation, which can result in dramatic
differences in the congener profiles of PCB mixtures found in various exposure sources (e.g., human
milk, contaminated fish, indoor air) (IARC. 2015: ATSDR. 20001. Although environmental PCB
mixtures can be characterized analytically as if they were Aroclors, this can be imprecise given
weathering and degradation.5 Furthermore, of all possible congener combinations that might exist,
a relatively small subset of complex PCB mixtures has been tested in animal studies; important
differences might exist between these tested mixtures and the mixtures to which humans are
exposed in the environment Thus, methods for translating toxicological data from tested to
untested mixtures would be useful, including methods for addressing PCB mixtures with varying
proportions of congeners with diverse modes of action (e.g., "dioxin-like," "estrogenic," "anti-
estrogenic," "neurotoxic" (Wolff etal.. 1997: Wolff and Toniolo. 199511.
For these reasons, approaches for assessing chemical mixtures will be evaluated for use in
this assessment The Supplementary Guidance for Conducting Health Risk Assessment of Chemical
Mixtures (U.S. EPA. 20001 recommends several approaches to quantitative health risk assessment of
a chemical mixture, depending on the type of available data. The preferred approach is to use
toxicity data on the mixture of concern. Alternatively, when toxicity data are not available for the
mixture of concern, use of toxicity data on a "sufficiently similar" mixture is recommended.
Sufficient similarity, as discussed in U.S. EPA (20001. implies that the toxicological consequences of
exposure to the mixture of concern are expected to be identical or indistinguishable from those of
the mixture for which data are available. Sufficiently similar mixtures are of similar chemical
composition, or there is some understanding of chemical differences between the mixtures.
Methods for defining sufficient similarity have been developed and are an area of active
investigation fCatlinetal.. 2018: Rice etal.. 2018: Marshall etal.. 2013: Feder etal.. 20091. The
feasibility of using these new methods to support the derivation and application of toxicity values
for PCB mixtures is a current research area; if specific methods are identified for use in this
assessment, they will be described in a protocol update.
As described in U.S. EPA f2015bl. the assessment will review the available data and, as
feasible, will develop oral and inhalation toxicity values for environmental PCB mixtures by
evaluating (1) toxic potencies of complex PCB mixtures (e.g., environmental, commercial) tested for
various noncancer health effects in animal bioassays; (2) methods for using toxicological data from
a limited set of tested PCB mixtures for human health risk assessment in a wide variety of exposure
contexts (e.g., sufficient similarity testing); and (3) approaches for assessing health risk based on
5 There are diverse analytical methodologies for measuring PCBs in the environment and in biological
samples. Some of these analytical methods treat all environmental mixtures as Aroclor mixtures (with rigid
congener profiles), whereas others measure individual congeners (all 209 congeners or a subset of selected
congeners). Aroclor analyses are cost-effective but have limitations since chromatographic patterns and peak
ratios can change during environmental weathering. This can be especially challenging when multiple
Aroclors are present at a given site fErickson, 2018: U.S. EPA, 1996b: Alford-Stevens, 1986: Alford-Stevens et al.,
19851.
This document is a draft for review purposes only and does not constitute Agency policy.
12 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
measurements of PCB levels in the environment collected using various analytical techniques
(e.g., Aroclor analyses vs. congener-specific analyses).
2.5.2. Potential for Hazard Identification and Dose-Response Assessment for PCB Exposure
via Inhalation
As described in U.S. EPA f2015bl. evidence suggests that PCB inhalation can pose a hazard
to human health. However, the database of studies investigating health effects resulting from PCB
exposure consists primarily of oral exposure studies. Whether the existing database of inhalation
studies will be adequate to support human health risk assessment for inhalation exposure to PCBs
is not clear (Lehmann etal.. 2015). Based on the available data, feasible options for conducting a
dose-response assessment for PCB inhalation exposure will be evaluated, considering differences in
toxicity of congeners that are inhaled versus ingested and differences between the inhalation and
oral exposure routes. Potential options include the use of data from available PCB inhalation
studies or the use of kinetic models or default approaches for route-to-route extrapolation from
oral PCB exposure data.
2.5.3. Suitability of Available Pharmacokinetic Models for Reliable Route-to-Route,
Interspecies, or Intraspecies Extrapolation
Because the assessment will address noncancer hazards associated with exposure to
complex PCB mixtures, available pharmacokinetic models will be evaluated for their ability to
predict the dose metrics of such mixtures. Further information regarding pharmacokinetic model
evaluation can be found in Section 6.4. Lipophilicity, binding to liver proteins (e.g., cytochromes,
AhR), and rate of elimination (due to metabolism or fecal excretion) are the main determinants of
PCB toxicokinetics. Variation of these toxicokinetic determinants among individual PCBs limits the
application of congener-specific models in the assessment of a complex PCB mixture. A single set of
parameters to describe these determinants for the complex mixture might not be justifiable
because significant individual toxicokinetic variation has been observed for different PCB
congeners. Additionally, possibilities of toxicokinetic interaction, such as competition at binding
sites or synergy in the case of induction of enzymes, could exist between PCB congeners in a
complex mixture.
As described in U.S. EPA (2015b). the assessment will evaluate (1) existing pharmacokinetic
models for their potential ability to support reliable route-to-route, interspecies, or intraspecies
extrapolations, including the ability to quantitatively predict transfer of PCBs across the placenta or
via breast milk; (2) available information on toxicokinetic differences among PCB congeners and
mixtures; and (3) available information on inter- or intraspecies differences in the toxicokinetics of
PCBs, including differences across lifestages. All of this information will be carefully considered
during evidence synthesis (Section 9) and dose-response assessment (Section 11).
This document is a draft for review purposes only and does not constitute Agency policy.
13 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
3. OVERALL OBJECTIVES, SPECIFIC AIMS, AND
POPULATIONS, EXPOSURES, COMPARATORS,
AND OUTCOMES (PECO) CRITERIA
3.1. SPECIFIC AIMS
The aims of the assessment are to:
Identify epidemiological (i.e., human) and toxicological (i.e., experimental animal) literature
reporting effects of exposure to PCBs as outlined in the PECO. The assessment will include
evaluations of the evidence relevant to the following noncancer health effect categories:
cardiovascular, dermal, developmental, endocrine, gastrointestinal, hematopoietic,
hepatobiliary, immune, metabolic, musculoskeletal, nervous system, urinary, reproductive,
and respiratory. The systematic review will focus on the highest priority health effect
categories and outcomes (see Section 5).
Evaluate mechanistic information (including toxicokinetic understanding) associated with
exposure to PCBs to inform the interpretation of findings related to potential health effects
in studies of humans and animals. The scope of these analyses of mechanistic information
will be determined by the complexity and confidence in the phenotypic evidence in humans
and animals, the likelihood of the analyses to impact evidence synthesis conclusions for
human health, and the directness or relevance of the available model systems for
understanding potential human health hazards (Section 9.2). The mechanistic evaluations
will focus primarily on the key science issues identified in Section 2.5.
Conduct study evaluations for individual epidemiology and toxicology studies (evaluating
reporting quality, risk of bias, and sensitivity) and PBPK models (scientific and technical
review).
Extract data on relevant health outcomes from selected epidemiology and toxicology
studies based on the study evaluations. Full data extraction of low confidence studies might
not be performed for poorly studied health effects or for health effects for which extensive
medium and high confidence studies are available.
Synthesize the evidence across studies, assessing similar health outcomes using a narrative
approach.
For each health outcome (or grouping of outcomes), evaluate the strength of evidence
across studies (or subsets of studies) separately for studies of exposed humans and for
animal studies. If studies informing mechanisms are synthesized, mechanistic evidence will
be used to inform evaluations of the available health effect evidence (or lack thereof).
For each health outcome (or grouping of outcomes), develop an integrated expert judgment
across evidence streams as to whether the evidence is sufficient (or insufficient) to indicate
This document is a draft for review purposes only and does not constitute Agency policy.
14 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
that exposure to PCBs has the potential to be hazardous to humans (in rare instances, the
evidence may be judged as sufficient to indicate that a hazard is unlikely). The judgment
will be directly informed by the evidence syntheses and based on structured review of an
adapted set of considerations for causality first introduced by Austin Bradford Hill (Hill.
1965) (see Sections 9 and 10), including consideration (e.g., based on available mechanistic
information) and discussion of biological understanding. As part of the evidence
integration narrative, characterize the strength of evidence for the available database of
studies and its uncertainties, and identify and discuss issues concerning potentially
susceptible populations and lifestages.
Derive toxicity values (e.g., RfDs, RfCs) as supported by the available data (see Section 10.2),
considering the similarities and differences in toxicity among PCB mixtures (see Section
2.5.1). Evaluating the applicability and uncertainties of methods to address potential
differences will be a key consideration. The feasibility of using these methods to support
the derivation and application of PCB mixture-specific toxicity values is a current research
area; if specific methods are identified for use in this assessment, they will be described in a
protocol update.
Evaluate the feasibility and applicability of pharmacokinetic and dosimetric modeling to
account for interspecies differences and route-to-route extrapolation. In the absence of
appropriate models or data, apply default dosimetric adjustments and explore alternative
approaches to developing estimates across exposure routes. Given the differences in
toxicokinetic properties among PCB congeners (see Sections 2.5.2 and 2.5.3), evaluating the
applicability and uncertainties of methods to address these potential differences will be a
key consideration.
Characterize uncertainties and identify key data gaps and research needs such as
limitations of the available evidence, limitations of the systematic review, and consideration
of dose relevance and toxicokinetic differences when extrapolating findings from higher
dose animal studies to lower levels of human exposure.
3.2. POPULATIONS, EXPOSURES, COMPARATORS, AND OUTCOMES
(PECO) CRITERIA
The PECO is used to identify the evidence that addresses the specific aims of the assessment
and to focus the literature screening, including the inclusion/exclusion criteria, in a systematic
review. The PECO criteria for PCBs (see Table 4) are based on (1) nomination of the chemical for
assessment; (2) discussions with scientists in EPA program and regional offices to determine the
scope of the assessment that will best meet Agency needs; (3) preliminary review of the health
effect literature for PCBs (primarily reviews and authoritative health assessment documents as
described in Section 2.3) to identify the major health hazards associated with exposure to PCBs and
key areas of scientific complexity; and (4) input received during public discussion of preliminary
materials released to the public in 2015.
In addition to those studies meeting the PECO criteria, studies containing supplemental
material that are potentially relevant to the specific aims of the assessment were tracked during the
literature screening process. Although these studies did not meet PECO criteria, they were not
This document is a draft for review purposes only and does not constitute Agency policy.
15 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
1 excluded from further consideration. The categories used to track studies as "potentially relevant
2 supplemental material" during screening and to prioritize these studies for consideration in the
3 assessment based on likelihood to impact evidence synthesis conclusions for human health are
4 described in Section 4.3.
Table 4. Populations, exposures, comparators, outcomes (PECO) criteria
PECO
element
Evidence
Populations
Human: Anv population and lifestage (occupational or general population, including children and
other sensitive populations).
Animal: Nonhuman mammalian animal species (whole organism) of anv lifestage (including
preconception, in utero, lactation, peripubertal, and adult stages).
Exposures
Human: Anv exposure to PCBs (in vivo) as determined bv:
Controlled exposure
Measured concentration in contact medium (e.g., food, air, dust)
Biomarkers of exposure (e.g., serum PCB levels)
Occupation in a job involving exposure to PCBs (e.g., electric capacitor manufacturing)
Self-reported history of using commercial products containing PCBs (e.g., mixing
Aroclors into caulk).
Animal: One or more oral (gavage, diet, drinking water, intragastric), inhalation (aerosol, vapor,
or particle; whole-body or nose-only), dermal (occlusive, semi-occlusive, non-occlusive), or
injected (intravenous, subcutaneous, intraperitoneal) treatment(s) with any clearly quantified
dosage of PCBs alone administered to a whole animal (in vivo).
Comparators
Human: A comparison or referent population exposed to lower levels (or no exposure/exposure
below detection limits) of PCBs, or exposure to PCBs for shorter time periods. Case reports and
case series will be tracked as "potentially relevant supplemental information."
Animal: A concurrent control group exposed to vehicle-onlv treatment or untreated control.
Outcomes
Human: Anv examination of survival, bodv weight, or development, or of the structure or
function of dermatologic, cardiovascular, endocrine, gastrointestinal, hematologic, hepatic,
immune, nervous, ocular, musculoskeletal, renal, respiratory or reproductive cells, tissues or
systems.
Animal: Anv examination of survival, bodv weight, or development, or of the structure or
function of dermatologic, cardiovascular, endocrine, gastrointestinal, hematologic, hepatic,
immune, nervous, ocular, musculoskeletal, renal, respiratory or reproductive cells, tissues or
systems.
This document is a draft for review purposes only and does not constitute Agency policy.
16 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
4. LITERATURE SEARCH AND SCREENING
STRATEGIES
4.1. LITERATURE SEARCH STRATEGIES
The literature search strategy relied on terms to gather information on exposure to PCB
mixtures and individual PCB congeners ("E" component of PECO). Additional exposure terms were
used to identify studies that were not indexed by the chemical name (e.g., poisoning events
[Yusho/Yu-Cheng], capacitor manufacturing workers). These exposure terms were intentionally
broad and did not prioritize studies in which exposure was quantified; this was considered during
screening of the literature (Section 4.3). The search queries did not contain terms for the
population ("P"), comparison ("C"), or outcome ("0") components of the PECO statement; these
were also considered during screening of the literature.
The following databases were searched:
PubMed (National Library of Medicine)
Web of Science (Thomson Reuters)
Toxline (National Library of Medicine)
Searches were not restricted by publication date, and no language restrictions were applied.
The detailed search strategies are presented in Appendix A. Literature searches were conducted
using EPA's HERO database.6 The HERO page for the PCB assessment contains the literature search
results (https://hero.epa.gov/hero/index.cfm/proiect/page/proiect id/384). The literature
search will be periodically updated throughout development of the draft assessment to identify
literature published during the course of review.7 The last full literature search update is
anticipated to be conducted less than 1 year before the planned release of the draft assessment
document for public comment The results returned (i.e., the number of "hits" from each electronic
database or other literature source), including the results of any literature search updates, are
6 Health and Environmental Research Online: https: //hero.epa.gov/hero/.
7 The initial literature search was completed in July 2015 and is updated annually. References retrieved
through August 2016 are accounted for in this protocol. The literature is currently being updated and will be
updated regularly until several months prior to public release of the draft assessment. As such, the methods
for literature search and screening (and some of the approaches to refining the evaluation plan based on the
identified literature; see Section 5) are described in the protocol using the past tense, while the approaches
for other assessment methods are outlined in future tense.
This document is a draft for review purposes only and does not constitute Agency policy.
17 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
documented in the literature flow diagrams (see Figure 6, which also reflect the literature screening
decisions (see Section 4.3).
The IRIS Program takes extra steps to ensure identification of pertinent studies: by
encouraging the scientific community and the public to identify additional studies and ongoing
research; by searching for publicly available data submitted under the Toxic Substances Control Act
(TSCA) and the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA); and by considering
late-breaking studies that would impact the credibility of the conclusions, even during the review
process.8 Release of the PECO-screened literature in parallel with release of the protocol for public
comment provides an opportunity for stakeholders to identify any missing studies, which, if
identified, will be screened as outlined above for adherence to the PECO criteria. Studies identified
after peer review begins will be considered for inclusion only if they are directly relevant to the
PECO criteria and could fundamentally alter the assessment's conclusions.
4.2. NONPEER-REVIEWED DATA
IRIS assessments rely mainly on publicly accessible, peer-reviewed studies. However, it is
possible that nonpeer-reviewed data directly relevant to the PECO could be identified during
assessment development EPA might obtain external peer review if the owners of the data are
willing to have the study details and results made publicly accessible. Consistent with policies and
procedures outlined in U.S. EPA's Science Policy Council Peer Review Handbook (U.S. EPA. 2015a).
this independent, contractor-driven, peer review would include an evaluation of the study similar
to that for peer review of a journal publication. The contractor would identify and select two or
three scientists knowledgeable in scientific disciplines relevant to the topic as potential peer
reviewers. Persons invited to serve as peer reviewers would be screened for conflict of interest In
most instances, the peer review would be conducted by letter review. The study authors would be
informed of the outcome of the peer review and given an opportunity to clarify issues or provide
missing details. The study and its related information, if used in the IRIS assessment, would
become publicly available. In the assessment, EPA would acknowledge that the document
underwent peer review, and the names of the peer reviewers would be identified. In certain cases,
IRIS will conduct an assessment for utility and data analysis based on having access to a description
of study methods and to the raw data that have undergone rigorous quality assurance/quality
control review (e.g., ToxCast/Tox21 data, results of National Toxicology Program [NTP] studies)
but that have not yet undergone external peer review.
Unpublished data from personal communication by the author can supplement a
peer-reviewed study provided the information is made publicly available (typically through
documentation in HERO).
8 IRIS "stopping rules": https://www.epa.gov/sites/production/files/2014-
06 /documents /iris stoppingrules.pdf.
This document is a draft for review purposes only and does not constitute Agency policy.
18 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
4.3. LITERATURE SCREENING STRATEGY
1 The PECO criteria were used to determine inclusion or exclusion of a reference as a primary
2 source of health effect data. In addition to the inclusion of studies that meet the PECO criteria,
3 studies containing supplemental material that is potentially relevant to the specific aims were
4 tracked during the screening process. Although not considered to directly meet PECO criteria,
5 these studies are not strictly excluded unless otherwise specified. Unlike studies that meet PECO
6 criteria, supplemental studies might not be subject to additional consideration unless they help
7 address specific assessment aims (see Section 3.1). Studies that were categorized as "potentially
8 relevant supplemental material" include the following:
9 Study materials that have not been peer reviewed;
10 Study materials published in a language other than English;
11 Records that do not contain original data, such as other agency assessments, informative
12 scientific literature reviews, editorials, or commentaries;
13 Studies appearing only as abstracts (e.g., conference abstracts);
14 Mechanistic studies: Studies reporting measurements related to a health outcome that
15 informs the biological or chemical events associated with phenotypic effects, in both
16 mammalian and nonmammalian model systems, including in vitro, in vivo (by various
17 routes of exposure), ex vivo, and in silico studies;
18 ADME studies: Studies designed to capture information regarding absorption, distribution,
19 metabolism, and excretion, including toxicokinetic studies and studies describing PBPK
20 models for PCB congeners and mixtures;
21 Exposure characteristics: Exposure studies that include data unrelated to toxicological
22 endpoints, but which provide information on exposure sources or measurement properties
23 of the environmental agent (e.g., demonstrating a biomarker of exposure);
24 Susceptible populations: Studies that identify potentially susceptible populations and
25 lifestages, such as studies that focus on a specific demographic, lifestage, or genotype;
26 Human case reports or case series; and
27 Studies of PCB exposures and health effects in wildlife populations.
28 Because of the large size of the database and large number of health effects associated with
29 exposure to PCBs, the PECO criteria are expected to be narrowed to focus on the highest priority
30 health outcomes (see Section 5). As described below, the initial literature screen, conducted
31 according to the PECO criteria listed in Section 3.2, was used to identify the studies included in
32 summary-level literature inventories (see Section 4.4).
This document is a draft for review purposes only and does not constitute Agency policy.
19 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
4.3.1. Electronic Screening
The initial literature search described in Section 4.1 identified over 50,000 references.
Manual review of every record would have been time and resource intensive, so natural language
processing (NLP) and machine learning (ML) techniques were employed to identify the most
relevant literature for screening. Studies were prioritized using DoCTER, a publicly available
Document Classification and Topic Extraction Resource fhttps: //www.icfdocter.com/index].
Details of the NLP and ML methods are described elsewhere ("Varghese etal., 2019; Varghese etal.,
2017). Briefly, 484 studies
(https://hero.epa.gov/hero/index.cfm/proiect/page/usage id/18723/format/list/sort/year%
20desc/usage searchType/any/page/l/project id/384/rows/10) selected as meeting the PECO
criteria for inclusion in the assessment were designated as seed9 references and included in the
corpus of 53,801 references identified by the literature searches. In phase one of a two-phase
approach, schematically illustrated in Figure 3, titles and abstracts were represented in a
mathematical matrix and organized into clusters based on semantic similarity using NLP tools
(Figure 3).
Figure 3. Schematic illustration of electronic prioritization of literature
depicting references clustered by similarity using natural language
processing.
9 Seed references are a subset of the larger collection of unclassified references that are known to be topic
relevant. In a clustering analysis based on NLP, the distribution of seed references can be used to identify
clusters most likely to contain relevant references, as the references in the clusters containing seeds are
expected to be similar to the seeds fVarghese et al., 2017],
This document is a draft for review purposes only and does not constitute Agency policy.
20 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Two clustering algorithms (k-means, nonnegative matrix factorization) were applied using
cluster sizes of 10, 20, or 30 references for a total of 6 different clustering approaches. Clusters
harboring seed references were identified (Figure 4).
Figure 4. Illustration depicting clusters containing relevant seed references
(circled blue clusters). Clusters were ranked by the number of seed studies
included.
In each approach, clusters were ranked in decreasing order of the number of seed studies in
each cluster, and clusters were accepted in order until 90% of the total set of seed studies was
captured. This was repeated for all six approaches; thus, a given study could have appeared in one
of the accepted clusters (and thus appear with the greatest fraction of the seed studies) in 0 to 6 of
the approaches. Clusters containing seed references were grouped by the number of approaches in
which they were identified (groups A-F, Figure 5).
Studies that appeared in ranked clusters using 6, 5, 4, or 3 approaches were subjected to
title and abstract-level screening, as described below (groups A-D, Figure 5).
Then, for phase two, ML was used to predict relevance for those studies in the remaining
groups of clusters that appeared in one or two approaches (groups E and F, Figure 5). Also
included in this approach was one group of studies excluded from initial clustering until abstracts
were recovered and a second group of studies with titles only. The training dataset for this
secondary analysis included PECO-relevant and nonrelevant studies identified during screening in
phase one. Studies predicted to be relevant then were subjected to manual title and abstract-level
screening.
<1
r0T
This document is a draft for review purposes only and does not constitute Agency policy.
21 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Figure 5. Visualization of identified clusters. Clusters were organized into
groups (A-F) based on the number of approaches that identified the cluster such
that group A contains clusters harboring seed references identified by six
approaches and group F contains clusters harboring seed references identified by a
single approach. All references in the top four groups (A-D) were manually
screened for inclusion based on PECO criteria. Low-scoring groups (E, F) were
subjected to additional machine learning approaches to capture relevant references
for manual screening.
1 The number of studies identified using electronic prioritization methods is summarized in
2 Table 5. Studies not reviewed included those not identified by any of the clustering approaches or
3 those identified by one or two approaches but predicted to be nonrelevant during the ML phase. A
4 subset of nonprioritized studies was randomly selected for manual title and abstract-level review;
5 this additional review demonstrated that less than 10% of nonprioritized studies were relevant
6 based on PECO criteria.
Table 5. Electronic prioritization of literature for hazard identification
Group of studies
Prioritization approach
Number of prioritized studies
(of 53,798 retrieved in original search)
Groups A-D (Figure 5)
Supervised clustering
4,652
Groups E-F (Figure 5)
Supervised clustering and ML
3,428
Studies with titles only
ML
3,302
Total number electronically prioritized
11,382
This document is a draft for review purposes only and does not constitute Agency policy.
22 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 6. Sources of studies subjected to manual review for relevance to
hazard identification
Source
Number of references
DoCTER electronic prioritization (after duplicate removal) (Table 5)
11,382
Seed references used for priority ranking
484
Other sources
8
Stakeholder identified
29
2015-2016 literature update
1,818
Total manually screened
13,721
4.3.2. Title and abstract-level screening
Following a pilot phase to calibrate screening guidance, two screeners independently
conducted a manual title and abstract screen to identify records that appeared to meet the PECO
criteria for studies electronically prescreened as described above. Literature updates, references
identified as seed studies, and stakeholder-identified references yielded fewer references and were
not subjected to electronic prioritization prior to manual review (summarized in Table 6).
References retrieved through August 2016 were screened using structured forms developed for
DRAGON (a modular database with integrated literature evaluation and screening tools developed
for systematic review) (ICF. 20181. References identified in search updates after August 2016 will
be reviewed using SWIFT-Active Screener (Sciome; https: //www.sciome.com/swift-
activescreener/l. in which manual review is integrated with electronic prioritization using ML and
statistical approaches.
For citations with no abstract, articles were screened based on title relevance. Screening
conflicts were resolved by discussion among the primary screeners with consultation by a third
reviewer or technical advisor (if needed) to resolve any remaining disagreements.
Studies not meeting the PECO criteria but identified as "potentially relevant supplemental
material" were categorized (i.e., tagged) during the title and abstract screening process (further
described in Section 4.3). Conflict resolution is not required during the screening process to
identify supplemental information (i.e., tagging by a single screener is sufficient to identify the
study as potentially relevant supplemental material that could be considered during draft
development).
4.3.3. Full-text level screening
Records that were not excluded based on the title and abstract advanced to full-text review.
Full-text copies of these potentially relevant records were retrieved, stored in the HERO database,
and independently assessed by two screeners to confirm eligibility according to the PECO criteria.
This document is a draft for review purposes only and does not constitute Agency policy.
23 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
1 Screening conflicts were resolved by discussion between the primary screeners with consultation
2 by a third reviewer or technical advisor (as needed to resolve any remaining disagreements).
3 The results of this screening process were posted on the project page for this assessment in
4 the HERO database (https://hero.epa.gov/hero/index.cfm/proiect/page/project id/384), and
5 studies were "tagged" with appropriate category descriptors (e.g., studies eligible for study
6 evaluation, potentially relevant supplemental material, excluded). Results were also annotated
7 and reported in a literature flow diagram (see Figure 6). Figure 6 reflects literature searches
8 through August 2016. Literature search updates will be conducted, and the results will be
9 reflected in the draft assessment; the most current results can be viewed at any time in the HERO
10 project page provided above.
This document is a draft for review purposes only and does not constitute Agency policy.
24 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Database Searches (All studies to August 2016), n = 53,317 records after duplicate
removal
PubMed
(n =19,477)
WOS
(n = 37,701)
ToxLine
(n = 30,443)
Direct Manual Review, n=2,339
{1} Stakeholder Identified, n=29
(2) Seed studies, n=484
{3} Other sources, n=8
(4) 2016 Literature Update, n=l,818
DoCTER Electronic Prioritization
(1) Identification of potentially relevant studies based on
DoCTER Clustering and Machine Learning (excluding
seed studies used for supervised clustering and
machine learning training sets), n = 11,382
Excluded
39,596 studies not prioritized by
DoCTER or Manually Reviewed
~
TITLE AND ABSTRACT
Title & Abstract Screening
n=13,721
Excluded (n=10,609)
* Not relevant to PECO (n=10,606)
[ FUILTEXT
Full-Text Screening
n=3,112
Excluded (n=l,310)
* Not relevant to PECO (n=150)
\
Studies meeting PECO (n = 1,802)
* Human health effects studies (n=96G)
Animal health effects studies (n=836)
Tagged as potentially relevant
supplemental material (n=l,158)
n = the number of studies tagged as
supplemental across both title/abstract
and full text screening reviews
Figure 6. Literature search flow diagram for PCBs.
This document is a draft for review purposes only and does not constitute Agency policy,
25 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
4.3.4. Multiple Publications of the Same Data
For multiple publications using the same or overlapping data, all publications on the
research will be included, with one selected for use as the primary study; the others will be
considered as secondary publications with annotation indicating their relationship to the primary
record during data extraction. For epidemiology studies, the primary publication generally will be
the one with the longest follow-up, the largest number of cases, or the most recent publication date.
For animal studies, the primary publication typically will be the one with the longest duration of
exposure, or the one that assessed the outcome(s) most informative to the PECO. For both
epidemiology and animal studies, the assessment will include relevant data from all publications of
the study; however, if the same outcome is reported in more than one report, the data will be
extracted only once.
4.4. SUMMARY-LEVEL LITERATURE INVENTORIES
During full text-level screening, studies tagged based on PECO eligibility were further
categorized based on features such as evidence type (i.e., human or animal), health outcome(s), or
endpoint measure (s) included in the study. Based on the results of discussions with external
stakeholders at the public science meeting held to discuss scoping and problem formulation
materials for PCBs on June 17-18, 2015 (https://www.epa.gov/iris/iris-public-meeting-iun-2015).
studies were tagged to the following health outcome categories: cardiovascular, dermal,
developmental, endocrine, gastrointestinal, hematopoietic, hepatobiliary, immune, metabolic,
musculoskeletal, nervous system, ocular, reproductive, respiratory, and urinary. Literature
inventories for PECO-relevant studies were created to develop summary-level, sortable lists that
include some basic study design information (e.g., study population, exposure information such as
doses administered or biomarkers analyzed, age/lifestage10 of exposure, endpoints examined).
These literature inventories facilitate subsequent review of individual studies or sets of studies by
topic specific experts.
Inventories also will be created for studies that were tagged as "potentially relevant
supplemental material" during screening, including mechanistic studies (e.g., in vitro or in silico
models), ADME studies, and studies on endpoints or routes of exposure that do not meet the
specific PECO criteria but that might still be relevant to the research question(s). Here, the
objective is to create an inventory of studies that can be tracked and further summarized as
neededfor example, by model system, key characteristic [e.g., of carcinogens (Smith etal.. 2016)].
mechanistic endpoint, or key eventto support analyses of critical mechanistic questions that arise
at various stages of the systematic review (see Section 9.2 for a description of the process for
determining the specific questions and pertinent mechanistic studies to be analyzed). ADME data
10Age/lifestage of chemical exposure will be considered according to EPA's Guidance on Selecting Age Groups
for Monitoring and Assessing Childhood Exposures to Environmental Contaminants and EPA's A Framework for
Assessing Health Risk of Environmental Exposures to Children.
This document is a draft for review purposes only and does not constitute Agency policy.
26 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
1 and related information can be critical to the next steps of prioritizing or evaluating individual
2 PECO-specific studies and will be reviewed by subject matter experts early in the assessment
3 process. For example, the comprehensive identification of studies relevant to interpreting the
4 ADME or toxicokinetic characteristics of PCB congeners and mixtures will be prioritized.
This document is a draft for review purposes only and does not constitute Agency policy.
27 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
5. REFINED EVALUATION PLAN
The purpose of the refined evaluation plan is to describe refinements to the set of studies
meeting PECO criteria to be carried forward to study evaluation and to identify and group the
endpoints that will be the primary focus of the outcome-specific evaluations. The process also
helps determine which studies tagged as "potentially relevant supplemental material" might need
to be considered in the assessment. To focus on the studies most informative to this human health
assessment, refinements to the initial PECO criteria were developed based on the literature
inventories (shown in Tables 7-10) and explanations provided. The numbers of studies reporting
human and animal evidence associated with PCB exposure and specific health endpoints grouped
by hazard category are depicted in Figures 7-21.
Health outcome categories evaluated using the same P, E, and C criteria are combined into a
single table (Table 7). Of note, unique considerations are included in the P, E, or C criteria for
developmental, hepatobiliary, and reproductive effects; therefore, the refined PECO criteria for
these outcomes are presented as separate tables.
For developmental effects (Table 8), the human and animal populations considered are
restricted to those exposed during preconception, in utero, or as neonates, juveniles, or
adolescents because these are sensitive windows of exposure for developmental effects
fNTP. 2011: U.S. EPA. 19911.
The database of animal studies available to support evaluations of hepatobiliary effects is
particularly large (Figure 20). Although these effects have been evaluated in many
mammalian species exposed to PCBs, information from studies of nonhuman primates and
of well-characterized laboratory species (i.e., rats and mice) likely will be sufficient to
support hazard conclusions. Therefore, the animal populations considered for this health
effect category are restricted to these species (Table 9).
For animal studies of certain reproductive effects related to fertility and fecundity
(e.g., mating, conception, pregnancy rate), the exposure criteria require PCB exposure to
have been present prior to mating (Table 10); otherwise, observed effects on the number of
litters or offspring are more likely to result from effects on offspring development than on
fertility of the parental animals (Foster and Gray. 2013: NTP. 2011: U.S. EPA. 1996a).
This document is a draft for review purposes only and does not constitute Agency policy.
28 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 7. Refined PECO criteria (cardiovascular, dermal, endocrine,
gastrointestinal, hematopoietic, immune, metabolic, musculoskeletal, nervous
system, ocular, respiratory, and urinary effects)
PECO element
Evidence
Populations
Human: Adults and children with exoosure to PCBs at anv lifestage. The following studv designs will be
included: controlled-exposure and randomized intervention, cohort, case-control, and cross-sectional.
These studies include those conducted in the general population (e.g., NHANES), in cohorts specifically
assembled to assess PCB-related health effects, and in occupational settings, where the incidence of disease
is compared to a standard or reference population. Case reports, case series, and ecological studies will be
tracked as potentially relevant supplemental material.
Animal: Nonhuman mammalian animal species (whole organism) exposed during anv lifestage will be
considered (during any period from in utero through adulthood).
Exposures
Exposure to PCB mixtures containing 4a or more congeners, including at least one non-dioxin-like PCB. Such
"complex" mixtures include commercial PCB mixtures (e.g., Aroclors), mixtures found in the environment
(e.g., in contaminated fish or indoor air), and a range of mixtures administered to animals in the laboratory
setting. Because all humans are exposed to PCBs as complex mixtures in the environment, every study of
PCB exposure in humans is expected to meet this criterion.
Human: The following exposure assessment methods/exposure contexts will be considered informative:
controlled exposure; measured PCB concentration in contact medium (e.g., food, air, dust); biomarkers of
exposure (e.g., serum PCB levels); or occupation in a job involving exposure to PCBs (e.g., electrical
capacitor manufacturing).
The following exposure assessment methods/exposure contexts will not be considered in the absence of
biomarker measurements or estimates derived using scientifically sound methods:
Yusho/Yu-Cheng patient status; consumption of fish (or marine mammals or other wildlife); or residential
proximity to a PCB-contaminated site.
Animal: Exposure routes to be considered are anv oral, inhalation, dermal, or injection exposures: oral and
inhalation exposures will be judged the most informative. Studies employing exposures longer than 28 days
or short term, developmental exposures will be considered the most informative. Exposure to a single PCB
congener or to a mixture containing fewer than 4 congeners will be considered as mechanistic evidence in
support of hazard identification and development of toxicity value(s).
Comparators
Human: A comparison population exposed to lower levels (or no exposure/exposure below detection
levels). For a cohort study, comparisons are made: (1) between levels within a cohort, or (2) between the
cohort and an external cohort, presumed to be unexposed or exposed to a lesser degree. Comparisons are
made based on "Exposure" definitions above.
Animal: A concurrent control group exposed to vehicle-only treatment or untreated control.
Outcomes
Cardiovascular effects (Figure 7)
Human: Assessments of ischemic heart disease (IHD) and IHD mortality, myocardial infarction,
hypertension, atherosclerosis, heart failure, and cerebrovascular disease and cerebrovascular disease
mortality. Note: Studies that assessed "diseases of the heart" NOS (not otherwise specified) mortality (14
studies) and subjective complaints (4 studies) will be tracked as potentially relevant supplemental material.
Consideration of these outcomes was judged lower priority because they are ill-defined and capture a broad
range of conditions with potentially unrelated etiologies.
Animal: Anv examination of changes in size, structure, or function of cardiovascular organs or tissues,
including the heart and blood vessels. Measures of cardiac enzyme induction and levels of metals in the
heart will be considered as supporting evidence for hazard identification or MOA analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
29 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
PECO element
Evidence
Outcomes
(continued)
Dermal effects (Figure 8)
Findings of dermatologic changes, including abnormal pigmentation, irritation, erythema, edema,
acne/chloracne, fingernail/toenail abnormalities, or alopecia.
Endocrine effects (Figure 9)
Assessments of hypothalamic-pituitary-thyroid and hypothalamic-pituitary-adrenal axis function (including
thyroid hormone levels, ultrasound thyroid nodules or thyroid volume, diagnosis of thyroid disease, thyroid
weight and histopathology, glucocorticoid and adrenal sex steroid hormone levels, and adrenal weight and
histopathology). Note: Measures of thyroid metabolizing enzymes/gene expression will be considered as
supporting evidence for hazard identification or MOA analysis. Studies that assessed hormones outside the
hypothalamic-pituitary-thyroid and hypothalamic-pituitary-adrenal axes (e.g., insulin-like growth factor,
vitamin D, parathyroid, growth hormone) will be tracked as potentially relevant supplemental material
because, compared to the outcomes selected for initial consideration, the number of studies evaluating
these outcomes is relatively small. However, if sensitivity evaluations (Section 5) identify these outcomes as
particularly sensitive to the effects of PCB exposure, they will be prioritized for further analysis.
Hypothalamic-pituitary-gonadal axis function was considered as a reproductive effect. Effects on insulin
levels and insulin resistance were considered metabolic effects grouped with diabetes.
Gastrointestinal effects (Figure 10)
Evaluations of gastrointestinal histopathology and abdominal ultrasonography. The database also contains
30 studies with evaluations of digestive system complaints and diseases, such as abdominal pain,
nausea/vomiting, changes in bowel habits, bloating, gastric ulcer, indigestion, and loss of appetite.
However, these studies will be tracked as potentially relevant supplemental material. Consideration of
these outcomes was judged lower priority because they are ill-defined and capture a broad range of
conditions with potentially unrelated etiologies.
Hematopoietic effects (Figure 11)
Assessments of red blood cells and associated endpoints (e.g., hemoglobin, mean corpuscular volume,
hematocrit), bone marrow histopathology, and platelets/clotting function. Note: Studies that assessed
blood disease mortality (5 studies) will be tracked as potentially relevant supplemental material; blood
disease mortality is ill-defined and captures a broad range of conditions with potentially unrelated
etiologies.
Immune effects (Figure 12)
Assessments of impaired immune function, as shown by changes in infectious morbidity, antigen-specific
antibody responses, or assays of white blood cell function/proliferation in human studies or by host
resistance or functional immune assays in animal studies, allergy and asthma, autoimmunity, and thymus
weight and histopathology (i.e., thymic atrophy). Measures of white blood cell phenotype counts and
percentages, non-specific Ig levels, cytokine production, serum complement, and thymosin levels will be
considered as supporting evidence for hazard identification or MOA analysis. Note: Studies that assessed
endotoxin sensitivity, spleen weight and histopathology, lymph node weight and histopathology, and sepsis
will be tracked as potentially relevant supplemental material. The number of studies evaluating endotoxin
sensitivity and sepsis is relatively small, and effects on spleen and lymph node weight and histopathology
were judged to have lower biological significance than the outcomes selected for consideration. However,
if sensitivity evaluations (Section 5) identify these outcomes as particularly sensitive to the effects of PCB
exposure, they will be prioritized for further analysis.
Metabolic effects (Figure 13)
Assessments of metabolic syndrome and related health outcomes in humans (e.g., blood triglycerides, body
mass index, obesity, waist circumference, and diabetes) and animals (e.g., blood triglycerides, adiposity,
pancreatic histopathology, and diabetes). Both human and animal studies of diabetes include those
evaluating effects on insulin levels, insulin resistance, and blood and urinary glucose levels. Note: Studies
that assessed resting metabolic rate, oxygen consumption, and body temperature will be tracked as
potentially relevant supplemental material; due to the relatively small number of studies on these
outcomes, the available evidence is unlikely sufficient to identify them as hazards of PCB exposure. Further
study might be needed to support associations between PCB exposure and these health effects.
This document is a draft for review purposes only and does not constitute Agency policy.
30 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
PECO element
Evidence
Outcomes
(continued)
Musculoskeletal effects (Figure 14)
Evaluations of osteoporosis, bone strength and density, bone histopathology, bone development, effects on
dentition, skeletal muscle histopathology, and arthritis. Note: Studies that assessed "musculoskeletal
complaints and diseases" will be tracked as potentially relevant supplemental material. Consideration of
these outcomes was judged lower priority because they are ill-defined and capture a broad range of
conditions with potentially unrelated etiologies. Studies that assessed muscle mass and tone will be tracked
as potentially relevant supplemental material; due to the relatively small number of studies on these
outcomes, the available evidence is unlikely sufficient to identify them as hazards of PCB exposure. Further
study might be needed to support associations between PCB exposure and these health effects.
Nervous system effects (Figure 15)
Human: Assessments of attention deficit hyperactivity disorder, autism spectrum disorders, and related
behaviors (primarily attention, impulse control, and hyperactivity; also executive function, social cognition);
cognitive function (includes general intelligence [IQ]; language/verbal skills, learning and memory,
school/academic performance, visual-spatial skills, executive function/attention); neonatal neurological and
behavioral function; brain aging disorders (e.g., Parkinson's disease, dementia [including Alzheimer's
disease], amyotrophic lateral sclerosis); sensory function (auditory function, olfactory function, visual
function); motor/cerebellar function, and emotional state (e.g., depression, anxiety symptoms). Deficits in
nerve activity (e.g., nerve conduction, electroencephalography) and structural abnormalities will be
considered as supporting evidence for hazard identification or MOA analysis. Note: Studies that assessed
mortality (caused by neurological disease), "neurological symptoms," and "play behavior" will be tracked as
potentially relevant supplemental material. Consideration of these outcomes was judged lower priority
because they are ill-defined or capture a broad range of conditions with potentially unrelated etiologies.
Animal: Assessments of changes in behavior (including motor, cognitive, sensory, attention and motivation,
impulse control and hyperactivity) and significant changes in brain structure. Measures of neurochemistry,
electrophysiology, neuropathology, and neurodevelopmental processes, including but not limited to
apoptosis, dendritic arborization, and neurogenesis in the brain will be considered as supporting evidence
for hazard identification or MOA analysis.
Ocular effects (Figure 16)
Findings of ocular changes, including abnormal pigmentation, irritation, erythema, periorbital edema,
Meibomian (tarsal) gland enlargement, or ocular discharge.
Respiratory effects (Figure 17)
Evaluations of pulmonary/lung weight and histopathology, pulmonary function, and chest radiography.
Note: Studies that assessed "respiratory disease mortality" or "respiratory complaints/illness history" will
be tracked as potentially relevant supplemental material. Consideration of these outcomes was judged
lower priority because they are ill-defined and capture a broad range of conditions with potentially
unrelated etiologies. Measures of respiratory sounds, sputum analysis, blood gas tension, and respiratory
rate will be considered as supporting evidence for hazard identification or MOA analysis.
Urinary effects (Figure 18)
Assessments of kidney weight, serum biomarkers of renal function, urinary system histopathology, and
kidney diseases or nephropathies (e.g., nephritis, diabetic nephropathy, nephrotic syndrome, gout, renal
failure). Note: Measures of urinalysis and urine output will be considered as supporting evidence for hazard
identification or MOA analysis.
NHANES = National Health and Nutrition Examination Survey; MOA = mode of action.
a As described in Section 2.4, a major goal of this assessment is to develop noncancer toxicity values for PCB
mixtures, especially those most relevant for human exposure (Section 2.5.1). Humans tend to be exposed to
complex PCB mixtures that contain many congeners of varied toxic potency and MOA. Studies of single
congeners and simple (i.e., binary or ternary) mixtures will be tracked as potentially relevant supplemental
material, but this assessment will focus primarily on studies of mixtures that better reflect a typical human
exposure scenario.
This document is a draft for review purposes only and does not constitute Agency policy.
31 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Heart Weight
Blood Pressure/Hypertension
Cardiovascular Histopathology
"Diseases of the Heart" -
Mortality
Ischemic Heart Disease -
Excluding Ml
Ischemic Heart Disease -
Mortality
Subjective Complaints
Myocardial Infarction (Ml)
Cerebrovascular Disease
Heart Rate
Atherosclerosis
Cerebrovascular Disease -
Mortality
Heart Failure
Cardiac Arrhythmia
35
28
20
14
6
4
4
4
3
2
2
2
1
1
~ Human
~ Animal
50 100 150
Number of Studies
200
250
Figure 7. Number of human and animal studies of PCB exposure and
cardiovascular effects. Labels indicate the total number of studies for each health
effect Database contains 57 human studies and 45 animal studies; some studies
evaluated more than one type of cardiovascular effect
This document is a draft for review purposes only and does not constitute Agency policy.
32 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Acne/Chloracne
I
Abnormal Pigmentation
22
Alopecia
| 22
Dermal Histopathology
I 19
Dermal Irritation
1
18
Finger/Toenail Deformity
17
Hyperkeratosis
]6
13
Subcutaneous Edema
Gingival Swelling/Recession
y 5
Erythema
I 5
Scar Formation
1
33
~ Human
~ Animal
50 100 150
Number of Studies
200
250
Figure 8. Number of human and animal studies of PCB exposure and dermal
effects. Labels indicate the total number of studies for each health effect Database
contains 28 human studies and 40 animal studies; some studies evaluated more
than one type of dermal effect
HPT Hormones
Thyroid Histopathology
Adrenal Weight
HPA Hormones
Thyroid Size/Weight
Adrenal Histopathology
Pituitary Weight
Pituitary Histopathology
'Thyroid Disease"
Insulin-Like Growth Factor
Parathyroid Histopathology
Parathyroid Hormone
Vitamin D
Growth Hormone
Melatonin
] 47
I 46
36
30
II 8
]7
25
J 5
3 5
I 4
2
2
1
1
161
~ Human
~ Animal
50 100 150
Number of Studies
200
250
Figure 9. Number of human and animal studies of PCB exposure and
endocrine effects. Labels indicate the total number of studies for each health
effect. Database contains 92 human studies and 155 animal studies; some studies
evaluated more than one type of endocrine effect.
This document is a draft for review purposes only and does not constitute Agency policy.
33 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Gastrointestinal
Histopathology
Digestive System
Com plaints/Diseases
Abdominal Ultrasonography
Gut Motility
31
29
E
50
~ Human
~ Animal
100 150
Number of Studies
200
250
Figure 10. Number of human and animal studies of PCB exposure and
gastrointestinal effects. Labels indicate the total number of studies for each health
effect. Database contains 19 human studies and 37 animal studies; some studies
evaluated more than one type of gastrointestinal effect
Red Blood Cells/Hemoglobin
50
Platelets
16
Bone Marrow Histopathology
Blood Disease Mortality
n
£
12
>
~ Human
~ Animal
Clotting Function
1
I
0
i
50
i I
100 150
Number of Studies
I
200
250
Figure 11. Number of human and animal studies of PCB exposure and
hematopoietic effects. Labels indicate the total number of studies for each health
effect. Database contains 18 human studies and 41 animal studies; some studies
evaluated more than one type of hematopoietic effect.
This document is a draft for review purposes only and does not constitute Agency policy.
34 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
White Blood Cell (WBC)
Counts
mmune Organ Size/Weight
Immune Organ Histopathology
49
Susceptibility to Infection
Antibody Responses
Immunoglobulin Levels -
Nonspecific
Ex Vivo WBC Function
Atopy (Allergy/Asthma)
Cytokines/Biomarkers
Autoantibody Levels
Delayed-Type Hypersensitivity
Endotoxin Sensitivity
Autoimmune Disease
Sepsis
~ Human
~Animal
50 100 150
Number of Studies
200
250
Figure 12. Number of human and animal studies of PCB exposure and immune
effects. Labels indicate the total number of studies for each health effect Database
contains 101 human studies and 131 animal studies; some studies evaluated more
than one type of immune effect
This document is a draft for review purposes only and does not constitute Agency policy.
35 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Blood Triglycerides
89
Diabetes
74
BM l/Obesrty/Waist
Circumference
49
Blood/Urinary Glucose
28
Pancreatic Histopathology
12
Body Temperature/Metabolic
Rate
Adipose Tissue Histopathology
-
9
7
~ Human
~ Animal
Adipose Tissue Weight
6
Metabolic Syndrome
6
Pancreas Weight
2
Skeletal Muscle Oxidative
Capacity
1
(
)
I
50
I I
100 150
Number of Studies
I
200
I
250
Figure 13. Number of human and animal studies of PCB exposure and
metabolic effects. Labels indicate the total number of studies for each health effect.
Database contains 148 human studies and 81 animal studies; some studies
evaluated more than one type of metabolic effect
This document is a draft for review purposes only and does not constitute Agency policy.
36 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Dental Effects
13
Bone Density/Strength &
Osteoporosis
13
Musculoskeletal
Complaints/Diseases
10
Skeletal Muscle Histopathology
8
Bone Histopathology
7
~ Human
Q Animal
Bone Development/Bone Age
5
Arthritis
3
Muscle Mass/Tone
1
0
50
100
l
150
200
250
Number of Studies
Figure 14. Number of human and animal studies of PCB exposure and
musculoskeletal effects. Labels indicate the total number of studies for each
health effect. Database contains 30 human studies and 24 animal studies; some
studies evaluated more than one type of musculoskeletal effect
This document is a draft for review purposes only and does not constitute Agency policy.
37 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Cognitive Function
Motor Function/Development
Brain Weight & Histopathology
Activity Level
Emotional State/Affective
Behavior
Sensory Function
Executive Function
Neurotransmitter Levels
Neurological Symptoms
Sexual Behavior
Neurological Condition
Social Behavior/Development
Neurological Disease Mortality
Brain Aging Disorders
Neurophysiology
ADD/ADHD Behaviors
Behavioral Performance
Autism/Autistic Behaviors
Play Behavior
Polyneuropathy
110
66
58
44
41
33
32
19
17
13
13
12
9
9
9
7
6
5
2
1
~ Human
~Animal
50 100 150
Number of Studies
200
250
Figure 15. Number of human and animal studies of PCB exposure and nervous
system effects. Labels indicate the total number of studies for each health effect
Database contains 156 human studies and 143 animal studies; some studies
evaluated more than one type of nervous system effect
This document is a draft for review purposes only and does not constitute Agency policy.
38 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Periorbital Edema
Ocular Discharge 18
Meibomian (Tarsal) Gland
Enlargement
Ocular Irritation
14
Ocular Histopathology 6
~ Human
~ Animal
Redness of Eyes
5
Abnormal Pigmentation 4
Conjunctivitis 3
0
50
100 150 200
Number of Studies
250
Figure 16. Number of human and animal studies of PCB exposure and ocular
effects. Labels indicate the total number of studies for each health effect Database
contains 16 human studies and 27 animal studies; some studies evaluated more
than one type of ocular effect.
This document is a draft for review purposes only and does not constitute Agency policy.
39
DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Pulmonary Histopathology
I 31
Lung Weight
18
Respiratory Disease Mortality
12
Respiratory Complaints/Illness
History
12
Pulmonary Function
Chest Radiography
]
7
5
~ Human
~ Animal
Respiratory Sounds
4
Blood Gas Tension
2
Respiratory Rate
2
Sputum Analysis
1
C
I
50
1 i
100 150
Number of Studies
l
200
I
250
Figure 17. Number of human and animal studies of PCB exposure and
respiratory effects. Labels indicate the total number of studies for each health
effect. Database contains 28 human studies and 42 animal studies; some studies
evaluated more than one type of respiratory effect.
This document is a draft for review purposes only and does not constitute Agency policy.
40 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Kidney Weight
Kidney Histopathology
Serum Biomarkers of Renal
Function
Urinalysis
Urinary Bladder
Histopathology
Renal Disease/Renal Failure
Urine Output
Diabetic Nephropathy
62
46
38
21
~-i ~ Human
O
J ~ Animal
6
2
2
[ 1 1 1 1
0 50 100 150 200 250
Number of Studies
Figure 18. Number of human and animal studies of PCB exposure and urinary
system effects. Labels indicate the total number of studies for each health effect
Database contains 24 human studies and 92 animal studies; some studies evaluated
more than one type of urinary system effect
Table 8. Refined PECO criteria (developmental effects)
PECO
element
Evidence
Populations
Human: Humans with exposure to PCBs during preconception, in utero, infancv, childhood, and
adolescence. The following study designs will be included: controlled-exposure and randomized
intervention, cohort, case-control, and cross-sectional. These studies include those conducted in
the general population (e.g., NHANES), in cohorts specifically assembled to assess PCB-related
health effects, and in occupational settings, where the incidence of disease is compared to a
standard or reference population. Case reports, case series, and ecological studies will be
tracked as potentially relevant supplemental material.
Animal: Nonhuman mammalian animal species (whole organism) exposed during preconception,
in utero, or as neonates, juveniles, or adolescents.
This document is a draft for review purposes only and does not constitute Agency policy.
41 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
PECO
element
Evidence
Exposures
Exposure to PCB mixtures containing 4 or more congeners, including at least one non-dioxin-like
PCB. Such "complex" mixtures include commercial PCB mixtures (e.g., Aroclors), mixtures found
in the environment (e.g., in contaminated fish or indoor air), and a range of mixtures
administered to animals in the laboratory setting. Because all humans are exposed to PCBs as
complex mixtures in the environment, every study of PCB exposure in humans is expected to
meet this criterion.
Human: The following exposure assessment methods/exposure contexts will be considered
informative: controlled exposure; measured PCB concentration in contact medium (e.g., food,
air, dust); biomarkers of exposure (e.g., serum PCB levels); or occupation in a job involving
exposure to PCBs (e.g., electrical capacitor manufacturing).
The following exposure assessment methods/exposure contexts will not be considered in the
absence of biomarker measurements or estimates derived using scientifically sound methods:
Yusho/Yu-Cheng patient status; consumption offish (or marine mammals or other wildlife); or
residential proximity to a PCB-contaminated site.
Animal: Exposure routes to be considered are anv oral, inhalation, dermal, or injection
exposures; oral and inhalation exposures will be considered the most informative. Studies
employing exposures longer than 28 days or short-term, developmental exposures will be
considered the most informative. Exposure to a single PCB congener or to a mixture containing
fewer than 4 congeners will be considered as mechanistic evidence in support of hazard
identification and development of toxicity values.
Comparators
Human: A comparison population exposed to lower levels (or no exposure/exposure below
detection levels). For a cohort study, comparisons are made: (1) between levels within a cohort,
or (2) between the cohort and an external cohort, presumed to be unexposed or exposed to a
lesser degree. Comparisons are made based on "Exposure" definitions above.
Animal: A concurrent control group exposed to vehicle-onlv treatment or untreated control.
Outcomes
Developmental effects (Figure 19)
Assessments of birth weight/small for gestational age (related anthropometrics measured in
utero (e.g., ultrasound) or at birth (e.g., head circumference) also will be considered); physical
growth (height, weight, BMI, overweight, obesity) at various stages of infancy, childhood, or
adolescence; birth defects; and pregnancy loss/offspring mortality. Note: Studies that assessed
sex ratio, anogenital distance, developmental milestones, and placental weight and
histopathology will be tracked as potentially relevant supplemental material. However, if
sensitivity evaluations (Section 5) identify these outcomes as particularly sensitive to the effects
of PCB exposure, they will be prioritized for further analysis. Due to the relatively small number
of studies on Apgar scores, the available evidence is unlikely sufficient to identify an associated
hazard of PCB exposure.
NHANES = National Health and Nutrition Examination Survey.
This document is a draft for review purposes only and does not constitute Agency policy.
42 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Weight and Size (Early Life)
220
Offspring Mortality
129
Sex Ratio
43
Birth Defects
30
Developmental Milestones
21
~ Human
Anogenital Distance
U
~ Animal
Placental Weight
11
8
Placental Histopathology
J 4
Apgar Score
j 3
(
)
50
100
1 f
150 200
250
Number of Studies
Figure 19. Number of human and animal studies of PCB exposure and
developmental effects. Labels indicate the total number of studies for each health
effect. Database contains 107 human studies and 161 animal studies; some studies
evaluated more than one type of developmental effect
Table 9. Refined PECO criteria (hepatobiliary effects)
PECO
element
Evidence
Populations
Human: Adults and children with exposure to PCBs at anv lifestage. The following studv designs
will be included: controlled-exposure and randomized intervention, cohort, case-control, and
cross-sectional. These studies include those conducted in the general population (e.g., NHANES),
in cohorts specifically assembled to assess PCB-related health effects, and in occupational
settings, where the incidence of disease is compared to a standard or reference population.
Case reports, case series, and ecological studies will be tracked as potentially relevant
supplemental material.
Animal: Rats, mice, and nonhuman primates (whole organism) exposed during anv lifestage will
be considered (anywhere during the period from in utero through adulthood). Hepatobiliary
effects have been evaluated in other nonhuman mammalian species exposed to PCBs but given
the large number of studies in this health effect category, studies of nonhuman primates and of
well-characterized laboratory species (i.e., rats and mice) have been prioritized.
This document is a draft for review purposes only and does not constitute Agency policy.
43 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
PECO
element
Evidence
Exposures
Exposure to PCB mixtures containing 4 or more congeners, including at least one non-dioxin-like
PCB. Such "complex" mixtures include commercial PCB mixtures (e.g., Aroclors), mixtures found
in the environment (e.g., in contaminated fish or indoor air), and a range of mixtures
administered to animals in the laboratory setting. Because all humans are exposed to PCBs as
complex mixtures in the environment, every study of PCB exposure in humans expected to meet
this criterion.
Human: The following exposure assessment methods/exposure contexts will be considered
informative: controlled exposure; measured PCB concentration in contact medium (e.g., food,
air, dust); biomarkers of exposure (e.g., serum PCB levels); or occupation in a job involving
exposure to PCBs (e.g., electrical capacitor manufacturing).
The following exposure assessment methods/exposure contexts will not be considered in the
absence of biomarker measurements or estimates derived using scientifically sound methods:
Yusho/Yu-Cheng patient status; consumption offish (or marine mammals or other wildlife); or
residential proximity to a PCB-contaminated site.
Animal: Exposure routes to be considered are anv oral, inhalation, dermal, or injection
exposures; oral and inhalation exposures will be considered the most informative. Studies
employing exposures longer than 28 days or short-term, developmental exposures will be
considered the most informative. Exposure to a single PCB congener or to a mixture containing
fewer than 4 congeners will be considered as mechanistic evidence in support of hazard
identification and development of toxicity value(s).
Comparators
Human: A comparison population exposed to lower levels (or no exposure/exposure below
detection levels). For a cohort study, comparisons are made: (1) between levels within a cohort,
or (2) between the cohort and an external cohort, presumed to be unexposed or exposed to a
lesser degree. Comparisons are made based on "Exposure" definitions above.
Animal: A concurrent control group exposed to vehicle-onlv treatment or untreated control.
Outcomes
Hepatobiliary effects (Figure 20)
Assessments of blood cholesterol levels, serum biomarkers of liver health and function, liver
weight and hepatomegaly, and liver disease, including cirrhosis and steatosis in humans and
studies evaluating liver histopathology or liver lipids in animals. Note: Studies that assessed bile
acid content/excretion, metabolic enzyme induction, liver cell proliferation,
porphyrins/porphyria, gall bladder histopathology, and hepatic levels of vitamins and
micronutrients will be tracked as potentially relevant supplemental material. The number of
studies evaluating bile acid content/excretion, liver cell proliferation, porphyrins/porphyria, and
gall bladder histopathology is relatively small, and effects on metabolic enzyme induction and
hepatic levels of vitamins and micronutrients were judged to have lower biological significance
than the outcomes selected for consideration. However, if sensitivity evaluations (Section 5)
identify these outcomes as particularly sensitive to the effects of PCB exposure, they will be
prioritized for further analysis.
NHANES = National Health and Nutrition Examination Survey.
This document is a draft for review purposes only and does not constitute Agency policy.
44 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Liver Weight/Hepatomegaly
Liver Enzyme Induction
Liver Histopathology
Cholesterol
Serum Biomarkers of Liver
Health & Function
Liver Lipids/Steatosis
Micronutrients
Porphyrins
Liver Disease/Cirrhosis
Liver Cell Proliferation
Bile Acid Content/Excretion
Gall Bladder Histopathology
18
5
3
3
128
113
109
88
69
48
243
~ Human
~ Animal
50 100 150
Number of Studies
200
250
Figure 20. Number of human and animal studies of PCB exposure and
hepatobiliary effects. Labels indicate the total number of studies for each health
effect. Database contains 64 human studies and 313 animal studies; some studies
evaluated more than one type of hepatobiliary effect
This document is a draft for review purposes only and does not constitute Agency policy.
45 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 10. Refined PECO criteria (reproductive effects)
PECO
element
Evidence
Populations
Human: Adults and children with exposure to PCBs at anv lifestage. The following studv designs
will be included: controlled-exposure and randomized intervention, cohort, case-control, and
cross-sectional. These studies include those conducted in the general population (e.g., NHANES),
in cohorts specifically assembled to assess PCB-related health effects, and in occupational
settings, where the incidence of disease is compared to a standard or reference population. Case
reports, case series, and ecological studies will be tracked as potentially relevant supplemental
material.
Animal: Nonhuman mammalian animal species (whole organism) exposed during anv lifestage
will be considered (anywhere during the period from in utero through adulthood).
Exposures
Exposure to PCB mixtures containing 4 or more congeners, including at least one non-dioxin-like
PCB. Such "complex" mixtures include commercial PCB mixtures (e.g., Aroclors), mixtures found
in the environment (e.g., in contaminated fish or indoor air), and a range of mixtures
administered to animals in the laboratory setting. Because all humans are exposed to PCBs as
complex mixtures in the environment, every study of PCB exposure in humans is expected to
meet this criterion.
Human: The following exposure assessment methods/exposure contexts will be considered
informative: controlled exposure; measured PCB concentration in contact medium (e.g., food,
air, dust); biomarkers of exposure (e.g., serum PCB levels); or occupation in a job involving
exposure to PCBs (e.g., electrical capacitor manufacturing).
The following exposure assessment methods/exposure contexts will not be considered in the
absence of biomarker measurements or estimates derived using scientifically sound methods:
Yusho/Yu-Cheng patient status; consumption offish (or marine mammals or other wildlife); or
residential proximity to a PCB-contaminated site.
Animal: Exposure routes to be considered are anv oral, inhalation, dermal, or injection
exposures; oral and inhalation exposures will be considered the most informative. Studies
employing exposures longer than 28 days or short term, developmental exposures will be
considered the most informative. Exposure to a single PCB congener or to a mixture containing
fewer than 4 congeners will be considered as mechanistic evidence in support of hazard
identification and development of toxicity values. For evaluation of fertility (e.g., mating and
pregnancy rate) in parental females, studies will be considered if the PCB exposure began prior to
mating.
Comparators
Human: A comparison population exposed to lower levels (or no exposure/exposure below
detection levels). For a cohort study, comparisons are made: (1) between levels within a cohort,
or (2) between the cohort and an external cohort, presumed to be unexposed or exposed to a
lesser degree. Comparisons are made based on "Exposure" definitions above.
Animal: A concurrent control group exposed to vehicle-onlv treatment or untreated control.
This document is a draft for review purposes only and does not constitute Agency policy.
46 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
PECO
element
Evidence
Outcomes
Reproductive effects (Figure 21)
Human: Assessments of time-to-pregnancv, gestation duration/preterm birth, endometriosis,
and semen quality (including sperm count, morphology, and motility). Note: Studies that
assessed male and female sexual development; menstrual cycle characteristics, including age at
menarche/menopause; and maternal body weight gain during pregnancy will be tracked as
potentially relevant supplemental material. However, if sensitivity evaluations (Section 5)
identify these outcomes as particularly sensitive to the effects of PCB exposure, they will be
prioritized for further analysis. Measures of reproductive hormone levels and function will be
considered as supporting evidence for hazard identification or MOA analysis.
Animal: Assessments of endpoints measuring characteristics needed for successful reproduction,
as defined by U.S. EPA (1996. Guidelines for Reproductive Toxicity Risk Assessment (EPA/630/R-
96/009) (U.S. EPA, 1996a). Washington, DC: U.S. Environmental Protection Agencv, Risk
Assessment Forum, https://www.epa.gov/sites/production/files/2014-
11/documents/guidelines repro toxicitv.pdf). Such endpoints include, but are not limited to,
mating and pregnancy rate, interval between offspring/litters, gestation duration, sperm count,
sperm morphology, sperm motility, and histopathology of reproductive organs (female and
male). Measures of reproductive hormone levels and function and reproductive organ weights
will be considered as supporting evidence for hazard identification or MOA analysis. Measures of
maternal body weight gain will be tracked as potentially relevant supplemental material and
could be considered as supporting evidence to inform interpretation of primary outcomes.
NHANES = National Health and Nutrition Examination Survey; MOA = mode of action.
This document is a draft for review purposes only and does not constitute Agency policy.
47 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
0)
to
E
a)
J?
(0
Time-to-Pregnancy & Couple Fertility
Pregnancy/Conception Rate
Maternal Weight Gain
Gestation Duration/Preterm Birth
Estrous/Menstrual Cycle
Reproductive Organ Size/Weight (F)
Sex Hormone Levels (F)
Endometriosis
Pubertal Development (F)
Reproductive Organ Histopathology (F)
Ovulation
Reproductive Aging
Preeclampsia
Gynecologic Disorders
Vaginal Bleeding During Pregnancy
Dystocia
Sex Hormone Levels (M)
Reproductive Organ Size/Weight (M)
Sperm/Semen Parameters
Reproductive Organ Histopathology (M)
Pubertal Development (M)
Male Fertility
Erectile Dysfunction
31
84
. 49
46
:
J 39
] 37
J 36
27
27
26
~ 17
6
2
1
1
1
64
61
46
J 25
20
~ 15
1
~ Human
~ Animal
0
50
100
150
200
250
Number of Studies
Figure 21. Number of human and animal studies of PCB exposure and
reproductive effects. Labels indicate the total number of studies for each health
effect Database contains 153 human studies and 205 animal studies; some studies
evaluated more than one type of reproductive effect
The next step in the refinement process will be to evaluate the potential sensitivity to PCB
exposure of each health outcome based on dose-response data from animal toxicology studies;
these sensitivity evaluations will consider differences in effect levels observed across PCB mixtures.
Health outcomes in each category then will be grouped and sorted into priority tiers based on
factors including potential sensitivity to PCB exposure, size of the evidence base (as a preliminary
measure of the potential strength of evidence), and relative biological significance. Outcomes
observed exclusively or primarily in humans will be assigned to tiers based on size of the evidence
base, relative biological significance, and expected etiological similarity to sensitive outcomes
observed in animal toxicology studies. Details of the sensitivity evaluation and resulting
refinements to the assessment approach will be documented in the updated protocol released with
the draft assessment.
In addition to prioritizing specific health outcomes within each health effect category, entire
health effect categories might also be prioritized based on relative sensitivity and potential strength
This document is a draft for review purposes only and does not constitute Agency policy.
48 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
of evidence. As shown in Figure 22, the following health effect categories contain over 100 studies
evaluating potential effects ofPCB exposure on outcomes: hepatobiliary, reproductive, nervous
system, developmental, endocrine, immune, metabolic, urinary system, and cardiovascular. Top-
tier outcomes from each of these health effect categories will be subject to full systematic review,
including additional, targeted literature searches (as described in Section 4.3) and steps described
below, beginning with study evaluation (Section 6). For the remaining health effect categories with
relatively small databases of supporting evidence (i.e., respiratory, dermal, hematopoietic,
gastrointestinal, musculoskeletal, ocular effects), the results of epidemiology and animal toxicology
studies will be tracked and recorded in inventories, but these outcomes likely will not be subject to
further analysis unless identified as particularly sensitive to the effects ofPCB exposure.
If discernable differences in potential sensitivity are observed across entire health effect
categories, this assessment might use a modular approach to evaluating health effects. The first
module would focus on the most sensitive health effect categories, which would form the basis for
reference values to assess overall health risk resulting from PCB exposure. While developing the
first module, EPA would also evaluate the potential utility of conducting hazard assessments and
deriving reference values for less sensitive health effect categories (e.g., to support cumulative risk
assessments of exposures to PCBs in combination with other agents or stressors). If identified as a
priority need, a second module would be developed and released separately from the first module.
Hepatobiliary
Reproductive
Nervous System
Developmental
Endocrine
Immune System
Metabolic
Urinary System
Cardiovascular
Respiratory
Dermal
Hematopoietic
Gastrointestinal
Musculoskeletal
Ocular
43
59
56
54
70
68
102
116
232
229
247
268
299
358
377
~ Human
~ Animal
0
100
200
Number of Studies
300
400
Figure 22. Number of studies evaluating health outcomes in each health effect
category based on the results of the literature search illustrated in Figure 6.
This document is a draft for review purposes only and does not constitute Agency policy.
49 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
6. STUDY EVALUATION (REPORTING, RISK OF
BIAS, AND SENSITIVITY) STRATEGY
The general approach for evaluating PECO-relevant primary health effect studies of all
study types is described in Section 6.1. However, the specifics of applying the approach differ; thus,
they are described separately for epidemiology and animal toxicology studies in Sections 6.2 and
6.3, respectively. Different approaches will be used for evaluation of PBPK models (see Section 6.4)
and mechanistic studies (see Sections 6.5 and 9.2).
6.1. STUDY EVALUATION OVERVIEW FOR HEALTH EFFECT STUDIES
Key concerns for the review of epidemiology and animal toxicology studies are risk of bias,
which is the assessment of internal validity (factors that affect the magnitude or direction of an
effect in either direction) and insensitivity (factors that limit the ability of a study to detect a true
effect; low sensitivity is a bias toward the null when an effect exists). Reporting quality is evaluated
to determine the extent the available information allows for evaluating these concerns. The study
evaluations are aimed at discerning the severity of any identified limitations (focusing on
limitations that could substantively change a result), considering also the expected direction of the
bias. The study evaluation considerations described below can be refined to address a range of
study designs, health effects, and chemicals. The general approach for reaching an overall judgment
for the study (or a specific analysis in a study) regarding confidence in the reliability of the results
is illustrated in Figure 23.
At least two reviewers will independently evaluate the studies to identify characteristics
that bear on the validity and sensitivity of the results and provide additional chemical- or
outcome-specific knowledge or methodological concerns.
Considerations for evaluating studies will be developed in consultation with topic-specific
technical experts. Existing guidance documents are used when available, including EPA guidance
for carcinogenicity, neurotoxicity, reproductive toxicity, and developmental toxicity fU.S. EPA.
2005.1998.1996a. 1991). The independent evaluations include a pilot phase to assess and refine
the evaluation process. During this phase, decisions will be compared and a consensus reached
between reviewers, and when necessary, differences will be resolved by discussion among the
reviewers, the chemical assessment team, or technical experts. As reviewers examine a group of
studies, additional chemical-specific knowledge or methodological concerns could emerge, and a
second pass might become necessary. Refinements to the study evaluation process made during
the pilot phase and subsequent implementation will be acknowledged as updates to the protocol.
This document is a draft for review purposes only and does not constitute Agency policy.
50 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
(a) Study evaluation process (b)
Individual evaluation domains
Refined evaluation plan
0
Criteria development
0
Pilot testing/refine criteria
£
Evaluation by two
reviewers
0
Conflict resolution
Final domain judgments
and overall study rating
Animal
Epidemiology
Selection and performance
¦ Allocation
Participant selection
¦ Observational bias/blinding
Confounding/variable control
Confounding
Selective reporting and attrition
Selective reporting
Exposure methods sensitivity
¦ Chemical administration and
Exposure measurement
characterization
Exposure timing, frequency, and duration
Outcome measures and results display
Outcome ascertainment
Endpoint sensitivity and specificity
Analysis
Results presentation
Reporting quality
Other sensitivity
Domain judgments
Judgment
0 Good
Adequate
Deficient
Critically
Deficient
Interpretation
Appropriate study conduct relating to the domain and
minor deficiencies not expected to influence results.
A study that may have some limitations relating to the
domain, but they are not likely to be severe or to
have a notable impact on results-
Identified biases or deficiencies interpreted as likely
to have had a notable impact on the results or
prevent reliable interpretation of study findings.
A serious flaw identified that makes the observed
effect(s) uninterpretable. Studies with a critical
deficiency will almost always be considered
*uninformative* overall
Overall study rating for an outcome
Rating
Interpretation
High
No notable deficiencies or concerns identified: potential
for bias unlikely or minimal: sensitive methodology.
Medium
Possible deficiencies or concerns noted, but resulting
bias or lack of sensitivity is unlikely to be of a notable
degree.
Low
Deficiencies or concerns were noted, and the potential
for substantive bias or inadequate sensitivity could have
a significant impact on the study results or their
interpretation.
Uninformative
Serious flaw(s) makes study results unusable for hazard
identification or dose response.
Figure 23. Overview of IRIS study evaluation process, (a) An overview of the
evaluation process, (b) The evaluation domains and definitions for ratings (i.e.,
domain and overall judgments, performed on an outcome-specific basis).
1 For studies that examine more than one outcome, the evaluation process will be performed
2 separately for each outcome because the utility of a study can vary for different outcomes. If a
3 study examines multiple endpoints for the same outcome,11 evaluations might be performed at a
4 more granular level if appropriate, but these measures could still be grouped for evidence
5 synthesis.
» "Outcome" will be used throughout these methods; the same methods apply to an endpoint within a larger
outcome.
This document is a draft for review purposes only and does not constitute Agency policy,
51 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
1 Authors might be queried either to obtain missing critical information, particularly when
2 reporting quality information or data are missing (e.g., content that would be required to conduct a
3 meta-analysis or other quantitative integration) or to provide additional analyses that could
4 address potential limitations. The decision on whether to seek missing information includes
5 consideration of what additional information would be useful, specifically with respect to any
6 information that could result in a reevaluation of the overall study confidence. Outreach to study
7 authors will be documented and considered unsuccessful if researchers do not respond to an email
8 or phone request within a month of the attempt to contact.
9 For each outcome in a study,12 reviewers will reach a consensus judgment of good,
10 adequate, deficient, not reported, or critically deficient for each evaluation domain. If a consensus is
11 not reached, a third reviewer will perform conflict resolution. It is important to stress that these
12 evaluations are performed in the context of the study's utility for identification of individual
13 hazards. Although limitations specific to the usability of the study for dose-response analysis are
14 useful for later decisions, they do not contribute to the study confidence classifications. These
15 categories are applied to each evaluation domain for each study as follows:
16 Good represents a judgment that the study was conducted appropriately relative to the
17 evaluation domain, and any deficiencies, if present, are minor and would not be expected to
18 influence the study results.
19 Adequate indicates a judgment that methodological limitations relating to the evaluation
20 domain exist, but those limitations are not likely to be severe or to have a notable impact on
21 the results.
22 Deficient denotes identified biases or deficiencies that are interpreted as likely to have had a
23 notable impact on the results or that could prevent reliable interpretation of the study
24 findings.
25 Not reported indicates the information necessary to evaluate the domain question was not
26 available in the study. Generally, this term carries the same functional interpretation as
27 deficient for the purposes of the study confidence classification (described below).
28 Depending on the number and severity of other limitations identified in the study,
29 contacting the study authors to obtain this information might or might not be useful (see
30 discussion above).
Critically deficient reflects a judgment that the study conduct introduced a serious flaw that
makes the study uninterpretable. Studies with a determination of critically deficient in an
evaluation domain almost always will be considered overall uninformative.
12 "Study" is used instead of a more accurate term (e.g., "experiment") throughout these sections owing to an
established familiarity within the field for discussing a study's risk of bias, sensitivity, etc. However, all
evaluations discussed herein are explicitly conducted at the level of an individual outcome within an
(un)exposed group of animals or humans, or on a sample of the population within a study.
This document is a draft for review purposes only and does not constitute Agency policy.
52 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Once the evaluation domains have been rated, the identified strengths and limitations will
be considered to reach a study confidence classification of high, medium, or low confidence, or
uninformative for each specific health outcome. This classification is based on the reviewer
judgments across the evaluation domains and includes consideration of the likely impact the noted
deficiencies in bias and sensitivity or inadequate reporting have on the results. The classifications,
which reflect a consensus judgment between reviewers, are defined as follows:
High confidence: A well-conducted study with no notable deficiencies or concerns
identified; the potential for bias is unlikely or minimal, and the study used sensitive
methodology. High confidence studies generally reflect judgments of good across all or
most evaluation domains.
Medium confidence: A satisfactory (acceptable) study where deficiencies or concerns are
noted, but the limitations are unlikely to be notable. Generally, medium confidence studies
include adequate or good judgments across most domains, with the impact of any identified
limitation not being judged as severe.
Low confidence: A study where deficiencies or concerns are noted, and the potential for bias
or inadequate sensitivity could have a significant impact on the study results or their
interpretation. Typically, low confidence studies have a deficient evaluation for one or more
domains, although some medium confidence studies might have a deficient rating in
domain(s) considered to have less influence on the magnitude or direction of effect
estimates. Generally, low confidence results are given less weight compared to high or
medium confidence results during evidence synthesis and integration (see Section 10.1,
Tables 14 and 15) and are generally not used as the primary sources of information for
hazard identification or derivation of toxicity values unless they are the only studies
available. Studies rated as low confidence only because of sensitivity concerns about bias
toward the null would require additional consideration during evidence synthesis.
Observing an effect in these studies could increase confidence, assuming the study is
otherwise well conducted (see Section 9).
Uninformative: A study where serious flaw(s) make the study results unusable for informing
hazard identification. Studies with critically deficient judgments in any evaluation domain
are almost always classified as uninformative (see explanation above). Studies with
multiple deficient judgments across domains also might be considered uninformative.
Uninformative studies will not be considered further in the synthesis and integration of
evidence for hazard identification or dose-response but might be used to highlight possible
research gaps.
Study evaluation determinations reached by each reviewer and the consensus judgment
between reviewers will be documented in EPA's version of Health Assessment Workspace
Collaborative (HAWC), a free and open-source web-based software application.13 Final study
evaluations housed in HAWC will be made available when the draft is publicly released. The study
13HAWC: A modular web-based interface to facilitate development of human health assessments of chemicals
f https://hawcproiect.org/porta l/I
This document is a draft for review purposes only and does not constitute Agency policy.
53 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
confidence classifications and their rationales will be carried forward and considered as part of
evidence synthesis (see Section 9), to aid in interpreting results across studies.
6.2. EPIDEMIOLOGY STUDY EVALUATION
Evaluation of epidemiology studies of health effects to assess risk of bias and study
sensitivity will be conducted for the following domains: exposure measurement, outcome
ascertainment, participant selection, potential confounding, analysis, study sensitivity, and selective
reporting. Bias can result in false positives and false negatives, while study sensitivity is typically
concerned with identifying the latter.
The principles and framework used for evaluating epidemiology studies are adapted from
the principles in the Cochrane Risk of Bias in Nonrandomized Studies of Interventions [ROBINS-I;
(Sterne etal.. 20161], modified to address environmental and occupational exposures. The
underlying philosophy of ROBINS-I is to describe attributes of an "ideal" study relative to each
evaluation domain (e.g., exposure measurement, outcome classification). Core and prompting
questions are used to collect information to guide evaluation of each domain.
Core and prompting questions are presented in Table 11 with additional considerations
that apply to most outcomes for each domain. Core questions represent key concepts while the
prompting questions help the reviewer focus on relevant details under each key domain. Exposure-
and outcome-specific criteria to use during evaluation of studies will be developed using the core
and prompting questions and refined during a pilot phase with engagement from topic-specific
experts. The types of information that could be the focus of those criteria are listed in Table 12.
Several considerations will be applied when evaluating the exposure domain to assess the
exposure measurement and assessment methods for PCBs used in epidemiology studies. In
epidemiology studies of PCBs, exposure is commonly characterized using current measures of PCB
congeners or their metabolites in biological matrices, including blood serum/plasma, breast
milk/colostrum, and adipose tissue. These could be expressed on a whole-tissue basis
(e.g., ng of PCB/mL of serum) or might be lipid standardized (i.e., ngof PCB/g of lipid). These
studies often rely on a limited number of measured congeners or metabolites. For biomarkers of
exposure, exposure assessments that quantify multiple PCB congeners with a wide range of
chlorination levels characterizing current or previous exposure by various routes (e.g., oral,
inhalation, dermal) will receive higher ratings.
Interpretation of PCB exposure measurements will need to account for the following issues:
The half-life and elimination characteristics of PCB congeners vary significantly. Current
measures of PCBs in serum/plasma, adipose tissue, or breast milk/colostrum reflect
cumulative exposure to persistent PCB congeners, but only recent exposure to labile
congeners. In recent years, the full scope of human exposure to PCBs in the general
environment has become more appreciated, including the potential for significant
inhalation and dermal exposure to lower-chlorinated, less-persistent congeners. Single
time-point estimates of tissue PCB concentrations might therefore capture only a portion of
This document is a draft for review purposes only and does not constitute Agency policy.
54 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
past exposure levels and could impact the sensitivity of the study to detect associations that
might be present An exception would be studies that measure exposure during a discrete
period with relatively short duration (e.g., prenatal exposure where specific critical
exposure windows are known and correspond to sample collection).
Most studies do not measure all PCB congeners present in biological tissues; analyzed
congeners are generally selected because of their relative occurrence in biological samples
or the ability to detect them using a given analytical methodnot because of their
biological activity or their potential to induce a particular health effect Again, use of this
approach results in an incomplete exposure assessment that easily could miss important
relationships between exposure and effect. Nevertheless, tissue levels of PCB congeners
generally correlate with overall total PCB exposure (Devoto etal.. 1997): therefore, these
studies inform the potential for health hazard to result from exposure to PCBs, especially
when the analysis is based on a relatively large number of congeners.
Even for persistent congeners that are routinely measured in epidemiology studies
(e.g., PCBs 138,153,180), a current, single time-point estimate of tissue PCB levels might
not represent the composition of PCB congeners with biological activity during the relevant
period for the development of toxicity. Depending on the endpoint of concern, the timing of
exposure could be just as important as the magnitude. Therefore, assessing for exposure
during the relevant developmental window or within a relatively short time before or after
that period is important If PCB exposure is assessed years after the critical window has
passed, it is possible to envision several different exposure scenarios that could lead to the
observed PCB levels. And, for each scenario, although the resulting PCB levels are the same,
the toxicological implications could be very different
Limits of detection for PCB analytical methods vary. Regardless of the analytical method
used, confidence in exposure measures is increased by reporting on limits of detection and
methods used to account for values below those limits. PCB analytical methods also differ
in accuracy and in the type of information they provide.
PCBs are lipophilic compounds and, depending on the matrix, adjusting for lipids in these
studies could be important. General population studies differ in how they account for lipids:
studies lipid-standardize PCB exposure measures, include a measure of lipids as a covariate
in multivariable exposure-outcome models, or simply do not adjust for lipids. The most
appropriate method for addressing lipid levels in PCB analyses depends on the causal
structure of the exposure-outcome association fO'Brien etal.. 2016: Schisterman etal..
2005).
To address the issues listed above, criteria will be developed to evaluate the type of
analytical methods used by a study, the accuracy of the analytical method for the PCB congeners
assessed (e.g., total PCBs, individual congeners, Aroclor mixtures), and how the exposure
measurement will be interpreted in terms of the exposure period represented by the mixture, route
of exposure, and relevance to the window of susceptibility for each health effect. Criteria will be
developed to evaluate methods for lipid adjustment on an outcome-specific basis. These criteria
and others, both PCB specific and outcome specific, developed for use in this assessment, will be
documented in the updated protocol released with the draft assessment.
This document is a draft for review purposes only and does not constitute Agency policy.
55 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 11. Questions to guide the development of criteria for each domain in epidemiology studies
Domain and core
question
Prompting questions
Follow-up questions
Considerations that apply to most exposures and outcomes
Exposure measurement
Does the exposure
measure reliably
distinguish between
levels of exposure in a
time window considered
most relevant for a
causal effect with
respect to the
development of the
outcome?
For all:
Does the exposure measure
capture the variability in
exposure among the
participants, considering
intensity, frequency, and
duration of exposure?
Does the exposure measure
reflect a relevant time
window? If not, can the
relationship between
measures in this time and the
relevant time window be
estimated reliably?
Was the exposure
measurement likely to be
affected by a knowledge of
the outcome?
Was the exposure
measurement likely to be
affected by the presence of
the outcome (i.e., reverse
causality)?
For case-control studies of
occupational exposures:
Is exposure based on a
comprehensive job history
describing tasks, setting, time
Is the degree of exposure
misclassification likely to vary by
exposure level?
If the correlation between
exposure measurements is
moderate, is there an adequate
statistical approach to
ameliorate variability in
measurements?
If there is a concern about the
potential for bias, what is the
predicted direction or distortion
of the bias on the effect
estimate (if there is enough
information)?
These considerations require customization to the exposure and outcome
(relevant timing of exposure)
Good
Valid exposure assessment methods used, which represent the
etiologically relevant period of interest.
Exposure misclassification is expected to be minimal.
Adequate
Valid exposure assessment methods used, which represent the
etiologically relevant period of interest.
Exposure misclassification could exist but is not expected to greatly
change the effect estimate.
Deficient
Valid exposure assessment methods used, which represent the
etiologically relevant period of interest. Specific knowledge about
the exposure and outcome raise concerns about reverse causality,
but whether it is influencing the effect estimate is uncertain.
Exposed groups are expected to contain a notable proportion of
unexposed or minimally exposed individuals, the method did not
capture important temporal or spatial variation, or other evidence
of exposure misclassification exists that would be expected to
notably change the effect estimate.
Critically deficient
Exposure measurement does not characterize the etiologically
relevant period of exposure or is not valid.
Evidence indicates that reverse causality is very likely to account
for the observed association.
Exposure measurement was not independent of outcome status.
This document is a draft for review purposes only and does not constitute Agency policy.
56 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Domain and core
question
Prompting questions
Follow-up questions
Considerations that apply to most exposures and outcomes
period, and use of specific
materials?
For biomarkers of exposure, general
population:
Is a standard assay used?
What are the intra- and
interassay coefficients of
variation? Is the assay likely
to be affected by
contamination? Are values
less than the limit of
detection dealt with
adequately?
What exposure time period is
reflected by the biomarker?
If the half-life is short, what is
the correlation between serial
measurements of exposure?
Outcome ascertainment
Does the outcome
measure reliably
distinguish the presence
or absence (or degree of
severity) of the
outcome?
For all:
Is outcome ascertainment
likely to be affected by
knowledge of, or presence of,
exposure (e.g., consider
access to health care, if based
on self-reported history of
diagnosis)?
For case-control studies:
Is the comparison group
without the outcome
(e.g., controls in a
case-control study) based on
objective criteria with little or
no likelihood of inclusion of
people with the disease?
Is there a concern that any
outcome misclassification is
nondifferential, differential, or
both?
What is the predicted direction
or distortion of the bias on the
effect estimate (if there is
enough information)?
These considerations require customization to the outcome
Good
High certainty in the outcome definition (i.e., specificity and
sensitivity), minimal concerns with respect to misclassification.
Assessment instrument was validated in a population comparable
to the one from which the study group was selected.
Adequate
Moderate confidence that outcome definition was specific and
sensitive, some uncertainty with respect to misclassification but
not expected to greatly change the effect estimate.
Assessment instrument was validated but not necessarily in a
population comparable to the study group.
This document is a draft for review purposes only and does not constitute Agency policy.
57 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Domain and core
question
Prompting questions
Follow-up questions
Considerations that apply to most exposures and outcomes
For mortality measures:
How well does cause of death
data reflect occurrence of the
disease in an individual? How
well do mortality data reflect
incidence of the disease?
For diagnosis of disease measures:
Is the diagnosis based on
standard clinical criteria? If
based on self-report of the
diagnosis, what is the validity
of this measure?
For laboratory-based measures
(e.g., hormone levels):
Is a standard assay used?
Does the assay have an
acceptable level of interassay
variability? Is the sensitivity
of the assay appropriate for
the outcome measure in this
study population?
Deficient
Outcome definition was not specific or sensitive.
Uncertainty regarding validity of assessment instrument.
Critically deficient
Invalid/insensitive marker of outcome.
Outcome ascertainment is very likely to be affected by knowledge
of, or presence of, exposure.
Note: Lack of blinding will not be automatically construed to be critically
deficient.
Participant selection
Does evidence indicate
that selection into or out
of the study (or analysis
sample) was jointly
related to exposure and
to outcome?
For longitudinal cohort:
Did participants volunteer for
the cohort based on
knowledge of exposure or
preclinical disease symptoms?
Was entry into the cohort or
continuation in the cohort
related to exposure and
outcome?
For occupational cohort:
Were differences in participant
enrollment and follow-up
evaluated to assess bias?
If potential for bias is a concern,
what is the predicted direction
or distortion of the bias on the
effect estimate (if there is
enough information)?
Were appropriate analyses
performed to address changing
exposures over time relative to
symptoms?
These considerations could require customization to the outcome. This
might include determining what study designs effectively allow analyses of
associations appropriate to the outcome measures (e.g., design to capture
incident vs. prevalent cases, design to capture early pregnancy loss).
Good
Minimal concern for selection bias based on description of
recruitment process (e.g., selection of comparison population,
population-based random sample selection, recruitment from
sampling frame including current and previous employees).
Exclusion and inclusion criteria specified and would not induce
bias.
This document is a draft for review purposes only and does not constitute Agency policy.
58 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Domain and core
question
Prompting questions
Follow-up questions
Considerations that apply to most exposures and outcomes
Did entry into the cohort
begin with the start of the
exposure?
Was follow-up or outcome
assessment incomplete, and if
so, was follow-up related to
both exposure and outcome
status?
Could exposure produce
symptoms that would result
in a change in work
assignment/work status
("healthy worker survivor
effect")?
For case-control study:
Were controls representative
of populations and time
periods from which cases
were drawn?
Are hospital controls selected
from a group whose reason
for admission is independent
of exposure?
Could recruitment strategies,
eligibility criteria, or
participation rates result in
differential participation
relating to both disease and
exposure?
For population-based survey:
Was recruitment based on
advertisement to people with
Is there a comparison of
participants and nonparticipants
to address whether differential
selection is likely?
Participation rate is reported at all steps of study (e.g., initial
enrollment, follow-up, selection into analysis sample). If rate is not
high, there is appropriate rationale for why it is unlikely to be
related to exposure (e.g., comparison between participants and
nonparticipants or other available information indicates
differential selection is not likely).
Adequate
Sufficient description of the recruitment process to be comfortable
there is no serious risk of bias.
Inclusion and exclusion criteria specified and would not induce
bias.
Participation rate is incompletely reported but available
information indicates participation is unlikely to be related to
exposure.
Deficient
Little information on recruitment process, selection strategy,
sampling framework or participation; or aspects of these processes
raise the potential for bias (e.g., healthy worker effect, survivor
bias).
Critically deficient
Aspects of the processes for recruitment, selection strategy,
sampling framework, or participation result in concern that
selection bias resulted in a large impact on effect estimates
(e.g., convenience sample with no information about recruitment
and selection, cases and controls are recruited from different
sources with different likelihood of exposure, recruitment
materials stated outcome of interest and potential participants are
aware of or are concerned about specific exposures)
This document is a draft for review purposes only and does not constitute Agency policy.
59 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Domain and core
question
Prompting questions
Follow-up questions
Considerations that apply to most exposures and outcomes
knowledge of exposure,
outcome, and hypothesis?
Confounding
Is confounding of the
effect of the exposure
likely?
Is confounding adequately addressed
by considerations in:
Participant selection
(matching or restriction)?
Accurate information on
potential confounders and
statistical adjustment
procedures?
Lack of association between
confounder and outcome or
confounder and exposure in
the study?
Information from other
sources?
Is the assessment of confounders
based on a thoughtful review of
published literature, potential
relationships (e.g., as can be gained
through directed acyclic graphing),
minimizing potential overcontrol
(e.g., inclusion of a variable on the
pathway between exposure and
outcome)?
If potential for bias is a concern,
what is the predicted direction
or distortion of the bias on the
effect estimate (if there is
enough information)?
These considerations require customization to the exposure and outcome,
but this could be limited to identifying key covariates.
Good
Conveys strategy for identifying key confounders. This might
include: a priori biological considerations, published literature,
causal diagrams, or statistical analyses; with recognition that not
all "risk factors" are confounders.
Inclusion of potential confounders in statistical models not based
solely on statistical significance criteria (e.g., p < 0.05 from
stepwise regression).
Does not include variables in the models likely to be influential
colliders or intermediates on the causal pathway.
Key confounders are evaluated appropriately and considered
unlikely sources of substantial confounding. This often will
include:
o Presenting the distribution of potential confounders by levels
of the exposure of interest or the outcomes of interest (with
amount of missing data noted)
o Consideration that potential confounders were rare among
the study population, or were expected to be poorly
correlated with exposure of interest
o Consideration of the most relevant functional forms of
potential confounders
o Examination of the potential impact of measurement error or
missing data on confounder adjustment.
Adequate
Similar to good but might not have included all key confounders or
less detail might be available on the evaluation of confounders
This document is a draft for review purposes only and does not constitute Agency policy.
60 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Domain and core
question
Prompting questions
Follow-up questions
Considerations that apply to most exposures and outcomes
(e.g., sub-bullets in good). Residual confounding could explain part
of the observed effect, but concern is minimal.
Deficient
Does not include variables in the models likely to be influential
colliders or intermediates on the causal pathway.
And any of the following:
The potential for bias to explain some results is high based on an
inability to rule out residual confounding, such as a lack of
demonstration that key confounders of the exposure-outcome
relationships were considered.
Descriptive information on key confounders (e.g., their relationship
relative to the outcomes and exposure levels) is not presented.
Strategy of evaluating confounding is unclear or is not
recommended (e.g., based only on statistical significance criteria
or stepwise regression [forward or backward elimination]).
Critically deficient
Includes variables in the models that are colliders or intermediates
in the causal pathway, indicating that substantial bias is likely from
this adjustment.
Confounding is likely present and not accounted for, indicating all
of the results were most likely due to bias.
Presenting a progression of model results with adjustments for
different potential confounders, if warranted.
Analysis
Does the analysis
strategy and
presentation convey the
necessary familiarity with
the data and
assumptions?
Are missing outcome,
exposure, and covariate data
recognized, and if necessary,
accounted for in the analysis?
Does the analysis
appropriately consider
variable distributions and
modeling assumptions?
If there is a concern about the
potential for bias, what is the
predicted direction or distortion
of the bias on the effect
estimate (if there is enough
information)?
These considerations might require customization to the outcome. This
could include the optimal characterization of the outcome variable and
ideal statistical test (e.g., Cox regression).
Good
Use of an optimal characterization of the outcome variable.
This document is a draft for review purposes only and does not constitute Agency policy.
61 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Domain and core
question
Prompting questions
Follow-up questions
Considerations that apply to most exposures and outcomes
Does the analysis
appropriately consider
subgroups of interest
(e.g., based on variability in
exposure level or duration or
susceptibility)?
Is an appropriate analysis
used for the study design?
Is effect modification
considered, based on
considerations developed a
priori?
Does the study include
additional analyses
addressing potential biases or
limitations (i.e., sensitivity
analyses)?
Quantitative results presented (effect estimates and confidence
limits or variability in estimates) (i.e., not presented only as a
p-value or "significant"/"not significant").
Descriptive information about outcome and exposure provided
(where applicable).
Amount of missing data noted and addressed appropriately
(discussion of selection issuesmissing at random vs. differential).
Where applicable, for exposure, includes limits of detection (and
percentage below the limits of detection), and decision to use log
transformation.
Includes analyses that address robustness of findings,
e.g., examination of exposure-response (explicit consideration of
nonlinear possibilities, quadratic, spline, or threshold/ceiling
effects included, when feasible); relevant sensitivity analyses;
effect modification examined based only on a priori rationale with
sufficient numbers.
No deficiencies in analysis evident. Discussion of some details
might be absent (e.g., examination of outliers).
Adequate
Same as good, except:
Descriptive information about exposure provided (where
applicable) but might be incomplete; might not have discussed
missing data, cutpoints, or shape of distribution.
Includes analyses that address robustness of findings (examples in
good), but some important analyses are not performed.
Deficient
Does not conduct analysis using optimal characterization of the
outcome variable.
Descriptive information about exposure levels not provided (where
applicable).
This document is a draft for review purposes only and does not constitute Agency policy.
62 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Domain and core
question
Prompting questions
Follow-up questions
Considerations that apply to most exposures and outcomes
Effect estimate and p-value presented, without standard error or
confidence interval.
Results presented as statistically "significant"/"not significant."
Critically deficient
Results of analyses of effect modification examined without clear
a priori rationale and without providing main/principal effects
(e.g., presentation only of statistically significant interactions that
were not hypothesis driven).
Analysis methods are not appropriate for design or data of the
study.
Selective reporting
Is there reason to be
concerned about
selective reporting?
Were results provided for all
the primary analyses
described in the methods
section?
Is there appropriate
justification for restricting the
amount and type of results
that are shown?
Are only statistically
significant results presented?
If potential for bias is a concern,
what is the predicted direction
or distortion of the bias on the
effect estimate (if there is
enough information)?
These considerations generally do not require customization and might
have fewer than four levels.
Good
The results reported by study authors are consistent with the
primary and secondary analyses described in a registered protocol
or methods paper.
Adequate
The authors described their primary (and secondary) analyses in
the methods section and results were reported for all primary
analyses.
Deficient
Concerns were raised based on previous publications, a methods
paper, or a registered protocol indicating that analyses were
planned or conducted that were not reported, or that hypotheses
originally considered secondary were represented as primary in
the reviewed paper.
Only subgroup analyses were reported suggesting that results for
the entire group were omitted.
Only statistically significant results were reported.
This document is a draft for review purposes only and does not constitute Agency policy.
63 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Domain and core
question
Prompting questions
Follow-up questions
Considerations that apply to most exposures and outcomes
Sensitivity
Is there a concern that
sensitivity of the study is
not adequate to detect
an effect?
Is the exposure range
adequate to detect
associations and exposure-
response relationships?
Was the appropriate
population included?
Was the length of follow-up
adequate? Is the time/age of
outcome ascertainment
optimal, given the interval of
exposure and the health
outcome?
Are there other aspects
related to risk of bias or
otherwise that raise concerns
about sensitivity?
These considerations could require customization to the exposure and
outcome and might have fewer than four levels. Some study features that
affect study sensitivity might have already been included in the other
evaluation domains. Other features that have not been addressed will be
included here. Some examples include:
Adequate
The range of exposure levels provides adequate variability to
evaluate the relevant associations.
The population was exposed to levels expected to have an impact
on response.
The study population was sensitive to the development of the
outcomes of interest (e.g., ages, lifestage, sex).
The timing of outcome ascertainment was appropriate given
expected latency for outcome development (i.e., adequate
follow-up interval).
The study was adequately powered to observe an association
based on underlying population sensitivity and exposure contrasts.
No other concerns raised regarding study sensitivity.
Deficient
Concerns were raised about the issues described for adequate that
are expected to notably decrease the sensitivity of the study to
detect associations for the outcome.
This document is a draft for review purposes only and does not constitute Agency policy.
64 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 12. Information relevant to evaluation domains for epidemiology
studies
Domain
Types of information that might need to be collected or
are important for evaluating the domain
Exposure
measurement
Source(s) of exposure (e.g., consumer products, occupational, an industrial accident) and
source(s) of exposure data, blinding to outcome, level of detail for job history data, when
measurements were taken, type of biomarker(s), assay information, reliability data from repeat
measures studies, validation studies.
Outcome
ascertainment
Source of outcome (effect) measure, blinding to exposure status or level, how
measured/classified, incident vs. prevalent disease, evidence from validation studies, prevalence
(or distribution summary statistics for continuous measures).
Participant
selection
Study design, where and when was the study conducted, and who was included? Recruitment
process, exclusion and inclusion criteria, type of controls, total eligible, comparison between
participants and nonparticipants (or followed and not followed), and final analysis group. Does
the study include potential susceptible populations or lifestages (see discussion in Section 9)?
Confounding
Background research on key confounders for specific populations or settings; participant
characteristic data, by group; strategy/approach for consideration of potential confounding;
strength of associations between exposure and potential confounders and between potential
confounders and outcome; and degree of exposure to the confounder in the population.
Analysis
Extent (and if applicable, treatment) of missing data for exposure, outcome, and confounders;
approach to modeling; classification of exposure and outcome variables (continuous vs.
categorical); testing of assumptions; sample size for specific analyses; and relevant sensitivity
analyses.
Sensitivity
What are the ages of participants (e.g., not too young in studies of pubertal development)?
What is the length of follow-up (for outcomes with long latency periods)? Choice of referent
group, the exposure range, and the level of exposure contrast between groups (i.e., the extent to
which the "unexposed group" is truly unexposed, and the prevalence of exposure in the group
designated as "exposed").
Selective
reporting
Are results presented with adequate detail for all the endpoints and exposure measures
reported in the methods section, and are they relevant to the PECO? Are results presented for
the full sample and for specified subgroups? Were stratified analyses (effect modification)
motivated by a specific hypothesis?
1
2
3
4
5
6
7
This document is a draft for review purposes only and does not constitute Agency policy.
65 DRAFT-DO NOT CITE OR QUOTE
6.3. EXPERIMENTAL ANIMAL STUDY EVALUATION
The evaluation of experimental animal studies applies similar principles as those described
above for the evaluation of epidemiology studies. The evaluation process focuses on assessing
aspects of the study design and conduct through three broad types of evaluations: reporting quality,
risk of bias, and study sensitivity. A set of domains with accompanying core questions falls under
each evaluation type and directs individual reviewers to evaluate specific study characteristics. For
each domain and core question pairing, basic considerations provide additional guidance on how a
reviewer might evaluate and judge a study for that domain.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 13 provides the standard domains and core questions, along with some basic
considerations for guiding the evaluation. Some domain considerations will need to be tailored to
the chemical and endpoint/outcome, while others are generalizable across assessments
(e.g., considerations for reporting quality). Assessment teams work with subject matter experts to
develop the assessment-specific considerations. These specific considerations are determined
prior to performing study evaluation, although they might be refined as the study evaluation
proceeds (e.g., during pilot testing). Assessment-specific considerations are documented and made
publicly available with the assessment
Each domain receives a consensus judgment of good, adequate, deficient, not reportedor
critically deficient (as described in Section 6.1), accompanied by a rationale for the judgment Once
all domains are rated, an overall confidence classification of high, medium, or low confidence or
uninformative is assigned (as described in Section 6.1). The rationale for the classification,
including a brief description of any identified strengths or limitations from the domains and their
potential impact on the overall confidence determination, will be documented clearly and
consistently. This rationale will, to the extent possible, reflect an interpretation of the potential
influence on the results (including the direction or magnitude of influence).
One of the key uncertainties in this assessment relates to the impact of congener profile on
the toxicity of PCB mixtures. As a result, chemical-specific considerations for reporting quality will
be applied in evaluations of studies of PCB exposure. For example, studies that administer PCB
mixtures should provide the name, source, purity, and lot number of the mixture to receive the
highest evaluation rating {good) for chemical administration and characterization. The congener
profiles of different lots of Aroclor 1254 can vary, resulting in differences in biological activity
fKodavanti etal.. 20011: therefore, reporting the lot number of the administered PCB mixture is
important for fully characterizing its chemical composition. However, if the identity of the PCB
mixture (e.g., Aroclor 1254) is known, a lack of information on the lot number usually will not be
considered a significant limitation and, by itself, is unlikely to have a significant impact on overall
study confidence ratings.
A wide variety of outcomes have been assessed in animal studies of PCBs. Considerations
specific to each hazard domain and outcome are not included in this protocol; these will be
documented in the updated protocol released with the draft assessment However, examples of
specific considerations that could be applied include better domain ratings for studies that address
potential differences in timing (e.g., time of day) for evaluations of specific behaviors or for
evaluations of hormone levels (due to fluctuations with circadian rhythms) and for studies that
address fasting status for measurements related to metabolism.
This document is a draft for review purposes only and does not constitute Agency policy.
66 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 13. Questions to guide the development of criteria for each domain in experimental animal toxicology
studies
Evaluation concern
Domain -
core question
Prompting questions
General considerations
Reporting quality
Reporting quality
Does the study report information
for evaluating the design and
conduct of the study for the
endpoint(s)/outcome(s) of
interest?
Notes:
Reviewers will attempt to contact
authors to obtain missing
information when studies are
considered key for hazard
evaluation or dose-response.
This domain is limited to
reporting. Other aspects of
the exposure methods,
experimental design, and
endpoint evaluation
methods are evaluated
using the domains related
to risk of bias and study
sensitivity.
Does the study report the following?
Critical information necessarvto
perform study evaluation:
Species, test article name, levels
and duration of exposure, route
(e.g., oral; inhalation),
qualitative or quantitative
results for at least one endpoint
of interest
Important information for evaluating
the study methods:
Test animal: strain, sex, source,
and general husbandry
procedures
Exposure methods: source,
purity, method of administration
Experimental design: frequency
of exposure, animal age and
lifestage during exposure and at
endpoint/outcome evaluation
Endpoint evaluation methods:
assays or procedures used to
measure the
endpoints/outcomes of interest
These considerations typically do not need to be refined, although in
some instances the important information could be refined deoendins
on the endpoints/outcomes of interest or the chemical under
investigation.
A judgment and rationale for this domain will be given for the study.
Typically, these will not change, regardless of the endpoints/outcomes
investigated by the study. In the rationale, reviewers will indicate
whether the study adhered to good laboratory practices, OECD, or
other testing guidelines.
Good: All critical and important information is reoorted or
inferable for the endpoints/outcomes of interest.
Adequate: All critical information is reported but some
important information is missing. However, the missing
information is not expected to significantly impact the study
evaluation.
Deficient: All critical information is reoorted but important
information is missing that is expected to significantly reduce
the ability to evaluate the study.
Critically deficient: Study report is missing pieces of critical
information. Studies critically deficient for reoortins are
uninformative for the overall rating and not considered further
for evidence synthesis and integration.
This document is a draft for review purposes only and does not constitute Agency policy.
67 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Evaluation concern
Domain -
core question
Prompting questions
General considerations
Risk of
bias
Selection
and
performance
bias
Allocation
Were animals assigned to
experimental groups using a
method that minimizes selection
bias?
For each study:
Did each animal or litter have an
equal chance of being assigned
to any experimental group (i.e.,
random allocation3)?
Is the allocation method
described?
Aside from randomization, were
any steps taken to balance
variables across experimental
groups during allocation?
These considerations typically need not be refined.
A judgment and rationale for this domain will be given for each cohort
or experiment in the study.
Good: Experimental groups were randomized and any specific
randomization procedure was described or inferable
(e.g., computer-generated scheme). Note that normalization is
not the same as randomization [see response for adequate],
Adequate: Authors report that groups were randomized but do
not describe the specific procedure used (e.g., "animals were
randomized"). Alternatively, authors used a nonrandom
method to control for important modifying factors across
experimental groups (e.g., body-weight normalization).
Not reported (interpreted as deficient): No indication of
randomization of groups or other methods (e.g., normalization)
to control for important modifying factors across experimental
groups.
Critically deficient: Bias in the animal allocations was reported
or inferable.
This document is a draft for review purposes only and does not constitute Agency policy.
68 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Evaluation concern
Domain -
core question
Prompting questions
General considerations
Risk of
bias
(cont.)
Selection
and
performance
bias (cont.)
Observational bias/blinding
Did the study implement measures
to reduce observational bias?
For each endpoint/outcome or
grouping of endpoints/outcomes in a
study:
Does the study report blinding
or other methods/procedures
for reducing observational bias?
If not, did the study use a design
or approach for which such
procedures can be inferred?
What is the expected impact of
failure to implement (or report
implementation) of these
methods/procedures on results?
These considerations typically do not need to be refined. (Note that it
can be useful for teams to identify highly subjective measures of
endpoints/outcomes where observational bias might strongly
influence results prior to performing evaluations.)
A judgment and rationale for this domain will be given for each
endpoint/outcome or group of endpoints/outcomes investigated in the
study.
Good: Measures to reduce observational bias were described
(e.g., blinding to conceal treatment groups during endpoint
evaluation; consensus-based evaluations of histopathology
lesionsb).
Adequate: Methods for reducing observational bias
(e.g., blinding) can be inferred or were reported but described
incompletely.
Not reported: Measures to reduce observational bias were not
described.
o (Interpreted as adequate) The potential concern for bias
was mitigated based on the use of
automated/computer-driven systems; standard
laboratory kits; relatively simple, objective measures
(e.g., body or tissue weight); or screening-level
evaluations of histopathology.
o (Interpreted as deficient) The potential impact on the
results is major (e.g., outcome measures are highly
subjective).
Critically deficient: Strong evidence for observational bias that
impacted the results.
This document is a draft for review purposes only and does not constitute Agency policy.
69 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Evaluation concern
Domain -
core question
Prompting questions
General considerations
Risk of
bias
(cont.)
Confounding/
variable
control
Confounding
Are variables with the potential to
confound or modify results
controlled for and consistent
across all experimental groups?
For each study:
Are there differences across the
treatment groups
(e.g., co-exposures, vehicle, diet,
palatability, husbandry, health
status, etc.) that could bias the
results?
If differences are identified, to
what extent are they expected
to impact the results?
These considerations might need to be refined, as the specific variables
of concern can vary by experiment or chemical.
A judgment and rationale for this domain will be given for each cohort
or experiment in the study, noting when the potential for confounding
is restricted to specific endpoints/outcomes.
Good: Beyond the exposure of interest, variables likely to
confound or modify results appear to be controlled for and
consistent across experimental groups.
Adequate: Some concern that variables likely to confound or
modify results were uncontrolled or inconsistent across groups
but are expected to have a minimal impact on the results.
Deficient: Notable concern that potentially confounding
variables were uncontrolled or inconsistent across groups and
are expected to substantially impact the results.
Critically deficient: Confounding variables were presumed to
be uncontrolled or inconsistent across groups and are expected
to be a primary driver of the results.
This document is a draft for review purposes only and does not constitute Agency policy.
70 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Evaluation concern
Domain -
core question
Prompting questions
General considerations
Risk of
bias
(cont.)
Selective
reporting
and attrition
bias
Selective reporting and attrition
Did the study report results for all
prespecified outcomes and tested
animals?
Note:
This domain does not consider the
appropriateness of the
analysis/results presentation. This
aspect of study quality is evaluated
in another domain.
For each study:
Selective reporting bias;
Are all results presented for
endpoints/outcomes described
in the methods (see note)?
Attrition bias;
Are all animals accounted for in
the results?
If discrepancies exist, do authors
provide an explanation
(e.g., death or unscheduled
sacrifice during the study)?
If results omissions or attrition are
identified, what is the expected
impact on the interpretation of the
results?
These considerations typically do not need to be refined.
A judgment and rationale for this domain will be given for each cohort
or experiment in the study.
Good: Quantitative or qualitative results were reported for all
prespecified outcomes (explicitly stated or inferred), exposure
groups, and evaluation time points. Data not reported in the
primary article are available from supplemental material. If
results omissions or animal attrition are identified, the authors
provide an explanation, and these are not expected to impact
the interpretation of the results.
Adequate: Quantitative or qualitative results are reported for
most prespecified outcomes (explicitly stated or inferred),
exposure groups, and evaluation time points. Omissions or
attrition are not explained but are not expected to significantly
impact the interpretation of the results.
Deficient: Quantitative or qualitative results are missing for
many prespecified outcomes (explicitly stated or inferred),
exposure groups and evaluation time points or high animal
attrition; omissions or attrition could significantly impact the
interpretation of the results.
Critically deficient: Extensive results omission or animal
attrition are identified, preventing comparisons of results
across treatment groups.
This document is a draft for review purposes only and does not constitute Agency policy.
71 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Evaluation concern
Domain -
core question
Prompting questions
General considerations
Sensi-
tivity
Exposure
methods
sensitivity
Chemical administration and
characterization
Did the study adequately
characterize exposure to the
chemical of interest and the
exposure administration methods?
Note:
Consideration of the
appropriateness of the route of
exposure is not evaluated at the
individual study level. Relevance
and utility of the routes of
exposure are considered in the
PECO criteria for study inclusion
and during evidence synthesis.
For each study:
Does the study report the
source and purity or
composition (e.g., identity and
percent distribution of different
isomers) of the chemical? If not,
can the purity or composition be
obtained from the supplier
(e.g., as reported on the
website)?
Was independent analytical
verification of the test article
purity and composition
performed?
Did the authors take steps to
ensure the reported exposure
levels were accurate?
Are there concerns about the
methods used to administer the
chemical (e.g., inhalation
chamber type, gavage volume)?
For inhalation studies:
Were target concentrations
confirmed using reliable
analytical measurements in
chamber air?
It is essential that these criteria are considered, and potentially refined,
as the specific variables of concern can vary by chemical (e.g., stability
could be an issue for one chemical but not another).
A judgment and rationale for this domain will be given for each cohort
or experiment in the study.
Good: Chemical administration and characterization is
complete (i.e., source, purity, and analytical verification of the
test article are provided). There are no concerns about the
composition, stability, or purity of the administered chemical or
the specific methods of administration. For inhalation studies,
chemical concentrations in the exposure chambers are verified
using reliable analytical methods.
Adequate: Some uncertainties in the chemical administration
and characterization are identified but these are expected to
have minimal impact on interpretation of the results
(e.g., source and vendor-reported purity are presented, but not
independently verified; purity of the test article is suboptimal
but not concerning; for inhalation studies, actual exposure
concentrations are missing or verified with less reliable
methods).
This document is a draft for review purposes only and does not constitute Agency policy.
72 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Evaluation concern
Domain -
core question
Prompting questions
General considerations
Sensi-
tivity
(cont.)
Exposure
methods
sensitivity
(cont.)
Chemical administration and
characterization
(cont.)
For oral studies:
If necessary based on
consideration of
chemical-specific knowledge
(e.g., instability in solution;
volatility) or exposure design
(e.g., the frequency and
duration of exposure), were
chemical concentrations in the
dosing solutions or diet
analytically confirmed?
Deficient: Uncertainties in the exposure characterization are
identified and expected to substantially impact the results
(e.g., source of the test article is not reported; levels of
impurities are substantial or concerning; deficient
administration methods, such as the use of static inhalation
chambers or a gavage volume considered too large for the
species or lifestage at exposure).
Critically deficient: Uncertainties in the exposure
characterization are identified and there is reasonable certainty
that the results are largely attributable to factors other than
exposure to the chemical of interest (e.g., identified impurities
are expected to be a primary driver of the results).
Exposure timing, frequency, and
duration
Was the timing, frequency, and
duration of exposure sensitive for
the endpoint(s)/ outcome(s) of
interest?
For each endpoint/outcome or
grouping of endpoints/outcomes in a
study:
Does the exposure period
include the critical window of
sensitivity?
Was the duration and frequency
of exposure sensitive for
detecting the endpoint of
interest?
Considerations for this domain are highly variable depending on the
endpoint(s)/outcome(s) of interest and must be refined.
A judgment and rationale for this domain will be given for each
endpoint/outcome or group of endpoints/outcomes investigated in the
study.
Good: The duration and frequency of the exposure was
sensitive and the exposure included the critical window of
sensitivity (if known).
Adequate: The duration and frequency of the exposure was
sensitive and the exposure covered most of the critical window
of sensitivity (if known).
Deficient: The duration or frequency of the exposure is not
sensitive and did not include the majority of the critical window
of sensitivity (if known). These limitations are expected to bias
the results towards the null.
Critically deficient: The exposure design was not sensitive and
is expected to strongly bias the results toward the null. The
rationale will indicate the specific concern(s).
This document is a draft for review purposes only and does not constitute Agency policy.
73 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Evaluation concern
Domain -
core question
Prompting questions
General considerations
Sensi-
tivity
(cont.)
Outcome
measures
and results
display
Endpoint sensitivity and
specificity
Are the procedures sensitive and
specific for evaluating the
endpoint(s)/ outcome(s) of
interest?
Note:
Sample size alone is not a
reason to conclude an
individual study is critically
deficient.
Considerations related to
adjustments/corrections to
endpoint measurements
(e.g., organ weight
corrected for body weight)
are addressed under results
presentation.
For each endpoint/outcome or
grouping of endpoints/outcomes in a
study:
Are there concerns regarding
the specificity and validity of the
protocols?
Are there serious concerns
regarding the sample size?
Are there concerns regarding
the timing of the endpoint
assessment?
Considerations for this domain are highly variable, depending on the
endpoint(s)/outcome(s) of interest, and must be refined.
A judgment and rationale for this domain will be given for each
endpoint/outcome or group of endpoints/outcomes investigated in the
study.
Examples of potential concerns include:
Selection of protocols that are insensitive or nonspecific for the
endpoint of interest.
Use of unreliable methods to assess the outcome.
Assessment of endpoints at inappropriate or insensitive ages,
or without addressing known endpoint variation (e.g., due to
circadian rhythms, estrous cyclicity).
Decreased specificity or sensitivity of the response due to the
timing of endpoint evaluation, as compared to exposure
(e.g., short-acting depressant or irritant effects of chemicals;
insensitivity due to prolonged period of nonexposure prior to
testing).
This document is a draft for review purposes only and does not constitute Agency policy.
74 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Evaluation concern
Domain -
core question
Prompting questions
General considerations
Sensi-
tivity
(cont.)
Outcome
measures
and results
display
(cont.)
Results presentation
Are the results presented in a way
that makes the data usable and
transparent?
For each endpoint/outcome or
grouping of endpoints/outcomes in a
study:
Does the level of detail allow for
an informed interpretation of
the results?
Are the data analyzed,
compared, or presented in an
inappropriate or misleading
way?
Considerations for this domain are highly variable, depending on the
outcomes of interest, and must be refined.
A judgment and rationale for this domain will be given for each
endpoint/outcome or group of endpoints/outcomes investigated in the
study.
Examples of potential concerns include:
Nonpreferred presentation (e.g., developmental toxicity data
averaged across pups in a treatment group, when litter
responses are more appropriate; presentation of absolute
organ-weight data when relative weights are more
appropriate).
Failing to present quantitative results.
Pooling data when responses are known or expected to differ
substantially (e.g., across sexes or ages).
Failing to report on or address overt toxicity when exposure
levels are known or expected to be highly toxic.
Lack of full presentation of the data (e.g., presentation of mean
without variance data; concurrent control data are not
presented).
This document is a draft for review purposes only and does not constitute Agency policy.
75 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Evaluation concern
Domain -
core question
Prompting questions
General considerations
Overall confidence
Overall confidence
Considering the identified
strengths and limitations, what is
the overall confidence rating for
the endpoint(s)/outcome(s) of
interest?
Note:
Reviewers will mark studies that
are rated lower than high
confidence due only to low
sensitivity (i.e., bias toward the
null) for additional consideration
during evidence synthesis. If the
study is otherwise well conducted
and an effect is observed, the
confidence could be increased.
For each endpoint/outcome or
grouping of endpoints/outcomes in a
study:
Were concerns (i.e., limitations
or uncertainties) related to the
reporting quality, risk of bias, or
sensitivity identified?
If yes, what is their expected
impact on the overall
interpretation of the reliability
and validity of the study results,
including (when possible)
interpretations of impacts on
the magnitude or direction of
the reported effects?
The overall confidence rating considers the likely impact of the noted
concerns (i.e., limitations or uncertainties) in reporting, bias, and
sensitivity on the results.
A confidence rating and rationale will be given for each
endpoint/outcome or group of endpoints/outcomes investigated in the
study. Confidence ratings are described above (see Section 6.1.).
OECD = Organisation for Economic Cooperation and Development.
a Several studies have characterized the relevance of randomization, allocation concealment, and blind outcome assessment in experimental studies (Hirst et al., 2014; Krauth
et al., 2013) (Macleod, 2013; Higgins and Green, 2011).
b For nontargeted or screening-level histopathology outcomes often used in guideline studies, blinding during the initial evaluation of tissues is generally not recommended as
masked evaluation can make "the task of separating treatment-related changes from normal variation more difficult" and "there is concern that masked review during the
initial evaluation might result in missing subtle lesions." Generally, blinded evaluations are recommended for targeted secondary review of specific tissues or in instances when
a predefined set of outcomes is known or predicted to occur (Crissman et al., 2004).
This document is a draft for review purposes only and does not constitute Agency policy.
76 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
6.4. PHYSIOLOGICALLY BASED PHARMACOKINETIC (PBPK) MODEL
DESCRIPTIVE SUMMARY AND EVALUATION
PBPK (or classical pharmacokinetic [PK]) models might be used in an assessment when an
applicable one exists and no equal or better alternative for dosimetric extrapolation is available. As
described in Sections 2.5.2 and 2.5.3, pharmacokinetic models will be considered for use in this
assessment to support route-to-route (i.e., oral-to-inhalation) and interspecies extrapolations and
to quantitatively predict transfer of PCBs across the placenta or via breast milk. Any models used
will represent current scientific knowledge and accurately translate the science into computational
code reproducibly and transparently. For a specific target organ/tissue, employing or adapting an
existing PBPK model or developing a new PBPK model or an alternative quantitative approach
might be possible. Data for PBPK models could come from studies across various species and might
be in vitro or in vivo in design.
6.4.1. Pharmacokinetic (PK)/Physiologically Based Pharmacokinetic (PBPK) Model
Descriptive Summary
Key information from identified models will be summarized in tabular format (see example
Table 14 below).
Table 14. Example descriptive summary for a physiologically based
pharmacokinetic (PBPK) model study
Study detail
Description/notes
Author
LastName et al. (2003)
Contact email
xxxxxPemail.com
Contact phone
xxx-xxx-xxxx
Sponsor
N/A
Model summary
Species
Rat
Strain
F433
Sex
Male and female
Lifestage
Adult
Exposure routes
Inhalation
Oral
I.V.
Skin
Tissue dosimetry
Blood
Liver
Kidney
Urine
Lung
Model evaluation
Language
ACSL 11.8
Code available
YES
Effort to recreate model
COMPLETE
Code received
YES
Effort to migrate to open software
SIGNIFICANT
This document is a draft for review purposes only and does not constitute Agency policy.
11 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Study detail
Description/notes
Structure evaluated
YES
Math evaluated
YES
Code evaluated
YES. Issue (minor): Incorrect units listed in comments for liver metabolism (line 233).
Issue (major): Mass balance error in stomach compartment.
Available PK data
Urine (cumulative amount excreted) and blood (concentration) time course data for
oral (gavage) and inhalation (6 hr/day for 4 days) exposure. In vitro skin permeation.
6.4.2. Pharmacokinetic (PK)/Physiologically Based Pharmacokinetic (PBPK) Model
Evaluation
Once summarized, available PBPK models will be evaluated in accordance with criteria
outlined by U.S. EPA f20181. Judgments on the suitability of a model are separated into two
categories: scientific and technical (see Table 16). The scientific criteria focus on whether the
biology, chemistry, and other information available for chemical MOAs are justified (i.e., preferably
with citations to support use) and represented by the model structure and equations. The scientific
criteria are judged based on information presented in the publication or report that describes the
model and do not require evaluation of the computer code. Preliminary technical criteria include
availability of the computer code and completeness of parameter listing and documentation.
Studies that meet the preliminary scientific and technical criteria are then subjected to an in-depth
technical evaluation, which includes a thorough review and testing of the computational code. The
in-depth technical and scientific analyses focus on the accurate implementation of the conceptual
model in the computational code, use of scientifically supported and biologically consistent
parameters in the model, and reproducibility of model results reported in journal publications and
other documents. This approach stresses (1) clarity in the documentation of model purpose,
structure, and biological characterization; (2) validation of mathematical descriptions, parameter
values, and computer implementation; and (3) evaluation of each plausible dose metric. The
in-depth analysis is used to evaluate the potential value and cost of developing a new model or
substantially revising an existing one. PBPK models developed by EPA during the course of the
assessment will be peer reviewed, either as a component of the draft assessment or by publication
in a journal article.
This document is a draft for review purposes only and does not constitute Agency policy.
78 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 15. Criteria for evaluating physiologically based pharmacokinetic
(PBPK) models
Category
Specific criteria
Scientific
Biological basis for the model is accurate.
Consistent with mechanisms that significantly impact dosimetry.
Predicts dose metric(s) expected to be relevant.
Applicable for relevant route(s) of exposure.
Consideration of model fidelity to the biological system strengthens the scientific basis of the
assessment relative to standard exposure-based extrapolation (default) approaches.
Ability of model to describe critical behavior, such as nonlinear kinetics in a relevant dose
range, better than the default (i.e., BW3/4 [body weight scaling to the % power] scaling).
Model parameterization for critical lifestages or windows of susceptibility. Evaluation of
these criteria also will consider the model's fidelity vs. default approaches and possible
use of an intraspecies UF in conjunction with the model to account for variations in
sensitivity between lifestages.
Predictive power of model-based dose metric vs. default approach, based on exposure.
o Specifically, model-based metrics might correlate better than the applied doses with
animal/human dose-response data.
o The degree of certainty in model predictions vs. default is also a factor. For example,
although target tissue metrics are generally considered better than blood
concentration metrics, lack of data to validate tissue predictions when blood data are
available might lead to choosing the latter.
Principle of parsimony
Model complexity or biological scale, including number and parameterization of
(sub)compartments (e.g., tissue or subcellular levels) will be commensurate with data
available to identify parameters.
Model describes existing PKdata reasonably well, both in "shape" (matches curvature, inflection
points, peak concentration time, etc.) and quantitatively (e.g., within factor of 2-3).
Model equations are consistent with biochemical understanding and biological plausibility.
Initial
Well-documented model code is readily available to EPA and public.
technical
Set of published parameters is clearly identified, including origin/derivation.
Parameters do not vary unpredictably with dose (e.g., any dose dependence in absorption
constants is predictable across the dose ranges relevant for animal and human modeling).
Sensitivity and uncertainty analysis has been conducted for relevant exposure levels (local
sensitivity analysis is sufficient, but global analysis provides more information).
If a sensitivity analysis was not conducted, EPA could decide to conduct this additional
work independently before using the model in the assessment.
A sound explanation will be provided when sensitivity of the dose metric to model
parameters differs from what is reasonably expected based on experience.
This document is a draft for review purposes only and does not constitute Agency policy.
79 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
6.5. MECHANISTIC STUDY EVALUATION
Sections 9 and 10 outline an approach for considering information from mechanistic studies
(including in vitro, in vivo, ex vivo, and in silico studies) where the specific analytical approach is
targeted to the assessment needs, depending in part on the extent and nature of the phenotypic
human and animal evidence. In this way, the mechanistic synthesis for a given health effect might
range from a high-level summary (or detailed analysis) of potential mechanisms of action to
specific, focused questions needed to address important and impactful assessment uncertainties
unaddressed by the available phenotypic studies (e.g., expected shape of the dose-response curve in
the low-dose region, applicability of the animal evidence to humans, addressing susceptible
populations). Individual study-level evaluation of mechanistic endpoints typically will not be
pursued. However, for some chemical assessments, it may be necessary to identify assay-specific
considerations for study endpoint evaluations on a case-by-case basis to provide a more detailed
summary and evaluation for the most relevant individual studies. This might be done, for example,
when the scientific understanding of a critical mechanistic event or MOA is less established or lacks
scientific consensus, the reported findings on a mechanistic endpoint are conflicting, the available
mechanistic evidence addresses a complex and influential aspect of the assessment, or in vitro or in
silico data make up the bulk of the evidence base and little or no evidence from epidemiology
studies or animal bioassays is available. Any considerations used to evaluate mechanistic studies
will be documented in the assessment
This document is a draft for review purposes only and does not constitute Agency policy.
80 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
7. ORGANIZING THE HAZARD REVIEW
The organization and scope of the hazard evaluation is determined by the available
evidence for the chemical regarding routes of exposure, metabolism and distribution, outcomes
evaluated, and number of studies pertaining to each outcome and by the results of the evaluation of
sources of bias and sensitivity. The hazard evaluations will be organized around organ systems
(e.g., respiratory, nervous system) informed by one or multiple related outcomes, and a decision
will be made as to what level (e.g., organ system or subsets of outcomes within an organ system) to
organize the synthesis.
Table 16 lists some questions that might be asked of the evidence to aid this decision. These
questions extend from considerations and decisions made during development of the refined
evaluation plan to include review of the concerns raised during individual study evaluations and
the direction and magnitude of the study-specific results. Resolution of these questions then will
inform critical decisions about the organization of the hazard evaluation and what studies might be
useful in dose-response analyses.
Table 16. Querying the evidence to organize syntheses for human and animal
evidence
Evidence
Questions
Follow-up questions
ADME
Are absorption, distribution, metabolism, or excretion
different by route of exposures studied, lifestage when
exposure occurred, or dosing regimens used?
Will separate analyses be needed by route of
exposure, or by methods of dosing within a
route of exposure (e.g., are large differences
expected between gavage and dietary
exposures)?
Which lifestages and what dosing regimens
are more relevant to human exposure
scenarios?
Is there toxicity information for metabolites that also
should be evaluated for hazard?
What exposures will be included in the
evaluation?
Is the parent chemical or metabolite also produced
endogenously?
Outcomes
What outcomes are reported in studies? Are the data
reported in a comparable manner across studies
(similar output metrics at similar levels of specificity,
such as adenomas and carcinomas quantified
separately)?
At what level (hazard, grouped outcomes, or
individual outcomes) will the synthesis be
conducted?
This document is a draft for review purposes only and does not constitute Agency policy.
81 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Evidence
Questions
Follow-up questions
Are there interrelated outcomes? If so, consider
whether some outcomes are more useful or of greater
concern than others.
By what commonalities will the outcomes be
grouped:
health effect,
exposure levels,
functional or population-level
consequences (e.g., endpoints all
ultimately leading to decreased
fertility or impaired cognitive
function),
involvement of related biological
pathways?
How well do the assessed human and animal
outcomes relate within a level of grouping?
Does the evidence indicate greater sensitivity to
effects (at lower exposure levels or severity) in certain
subgroups (by age, sex, ethnicity, lifestage)? Should
the hazard evaluation include a subgroup analysis?
Does incidence or severity of an outcome increase
with duration of exposure or a particular window of
exposure? What exposure time frames are relevant to
development or progression of the outcome?
Is there mechanistic evidence that informs any of the
outcomes and how might they be grouped?
How robust is the evidence for specific outcomes?
What outcomes are reported by both human
and animal studies and by one or the other?
Were different animal species and sexes (or
other important population-level differences)
tested?
In general, what are the study confidence
conclusions of the studies (high, medium, low,
not informative) for the different outcomes? Is
there enough evidence from high and medium
confidence studies for particular outcomes to
draw conclusions about causality?
What outcomes should be highlighted?
Should the others be synthesized at all?
Would comparisons by specific limitations be
informative?
Dose-
response
Did some outcomes include better coverage of
exposure ranges that could be most relevant to human
exposure than others?
What outcomes and study characteristics are
informative for development of toxicity
values?
Which outcomes have sufficient data available to draw
conclusions about dose-response? Are any outcomes
with study results sufficiently similar (e.g., an
established linkage in a biological pathway) to allow
examination or calculation of common measures of
effect across studies? Do the mechanistic data identify
surrogate or precursor outcomes sufficient for
dose-response analysis?
Do some subgroups exhibit responses at lower
exposure levels than others?
Could findings from ADME studies inform data-derived
extrapolation factors, or link toxicity observed via
different routes of exposure, or link effects between
humans and experimental animals?
Can a common internal dose metric be used
to compare species or routes of exposure?
This document is a draft for review purposes only and does not constitute Agency policy.
82 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
8. DATA EXTRACTION OF STUDY METHODS AND
RESULTS
Data extraction and content management will be carried out using HAWC or DRAGON. Data
extraction elements that might be collected from epidemiology, controlled human exposure, animal
toxicology, and in vitro studies are listed in Appendix B. Data extraction elements that might be
collected from PBPK studies are listed in Table 14. The content of the data extraction might be
revised following the identification of the studies included in the review as part of a pilot phase to
assess the data extraction workflow. Not all studies that meet the PECO criteria will be subject to
data extraction. Studies evaluated as being uninformative are not considered further and would,
therefore, not undergo data extraction. In addition, outcomes determined to be less relevant during
PECO refinement might not go through data extraction or could have only minimal data extraction.
The same could be true for low confidence studies if sufficient medium and high confidence studies
are available. All findings are considered for extraction, regardless of statistical significance,
although the level of extraction for specific outcomes within a study might differ (i.e., ranging from
a narrative to full extraction of dose-response effect size information). Similarly, decisions about
data extraction for low confidence studies are typically made during implementation of the protocol
based on consideration of the quality and extent of the available evidence. The version of the
protocol released with the draft assessment will outline how low confidence studies were treated
for extraction and evidence synthesis.
The data extraction results for included studies will be presented in the assessment and
made available for download from EPA HAWC in Excel format when the draft is publicly released.
Note that the following browsers are fully supported for accessing HAWC: Google Chrome
(preferred), Mozilla Firefox, and Apple Safari; errors in functionality occur when viewed with
Internet Explorer. Data extraction will be performed by one member of the evaluation team and
checked by one or two other members. Discrepancies in data extraction will be resolved by
consultation with a third member of the evaluation team. Once the data have been verified, they
will be "locked" to prevent accidental changes. Digital rulers, such as WebPlotDigitizer
(https://automeris.io/WebPlotDigitizer/). are used to extract numerical information from figures.
Use of digital rulers is documented during extraction.
As previously described, routine attempts will be made to obtain information missing from
human and animal health effect studies if missing information is considered influential during study
evaluations (see Section 6) or when required to conduct a meta-analysis (e.g., missing group size or
variance descriptors such as standard deviation or confidence interval). Missing data from
individual mechanistic (e.g., in vitro) studies generally will not be sought. Outreach to study
This document is a draft for review purposes only and does not constitute Agency policy.
83 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
authors will be documented and considered unsuccessful if researchers do not respond to email or
phone requests after one or two attempts.
8.1. STANDARDIZING REPORTING OF EFFECT SIZES
In addition to providing quantitative outcomes in their original units for all study groups,
results from outcome measures will be transformed to a common metric, when possible, to help
compare distinct but related outcomes measured with different scales. These standardized effect
size estimates facilitate systematic evaluation and evidence integration for hazard identification
and meta-analysis when feasible for an assessment (see Section 9.1). Based on metrics across the
available studies, a common metric might be used and the calculation presented in the assessment.
For epidemiology studies, the typical approach is to extract adjusted statistical estimates
when possible, rather than unadjusted or raw estimates.
It is important to consider the variability associated with effect size estimates, with stronger
studies generally showing more precise estimates. However, effect size estimation can be
influenced by such factors as variances that differ substantially across treatment groups or by lack
of information to characterize variance, especially for animal studies in biomedical research
(Vesterinen etal.. 20141.
8.2. STANDARDIZING ADMINISTERED DOSE LEVELS/CONCENTRATIONS
Exposures will be standardized to common units. Exposure levels in oral studies will be
expressed in units of mg PCB/kg-day. When study authors provide exposure levels as
concentrations in the diet, dose conversions will be made using study-specific food consumption
rates and body weights when available. Otherwise, EPA defaults will be used fU.S. EPA. 19881.
addressing age and study duration as relevant for the species/strain and sex of the animal of
interest Exposure levels in inhalation studies will be expressed in units of mg/m3. Assumptions
used in performing dose conversions will be documented.
Unless otherwise reported by study authors, the background level in experimental animal
studies is assumed 0 ppm (0 mg/kg-day).
This document is a draft for review purposes only and does not constitute Agency policy.
84 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
9. SYNTHESIS OF EVIDENCE
For the purposes of this assessment, evidence synthesis and integration are considered
distinct, but related, processes. The syntheses of separate bodies of evidence (i.e., human, animal,
and mechanistic evidence) described in this section will directly inform the integration across all
evidence to draw an overall judgment for each assessed human health effect (described in
Section 10). The phrase "evidence integration" used here is analogous to the phrase "weight of
evidence" used in some other assessment processes (EFSA. 2017: U.S. EPA. 2017: NRC. 2014: U.S.
EPA. 20051.14
For each potential health hazard or smaller subset of related outcomes, the available
phenotypic human and animal health effect evidence will be synthesized separately. Mechanistic
evidence also will be considered, although the specific analytical approach is targeted to the
assessment needs, depending on the extent and nature of the phenotypic human and animal
evidence (see Sections 9.2 and 10). The results of the analyses of mechanistic evidence will be used
to inform key uncertainties; as a result, the scope of the mechanistic analyses will generally depend
on the extent and nature of the human and animal evidence (see Sections 9.2 and 10). Thus, apart
from the pre-defined mechanistic analyses (see Sections 9.2.1-9.2.3), the human and animal
evidence syntheses (or the lack of phenotypic data in humans and animals) help determine the
approach to be taken in synthesizing the available mechanistic evidence (see Section 9.2.4). In this
way, a mechanistic evidence synthesis might range from a high level summary of potential toxicity
mechanisms discussed in the published literature to a detailed analysis of multiple potential
modes-of-action, or it might evaluate specific, focused questions that inform key uncertainties
unaddressed by the phenotypic human and animal evidence (e.g., shape of the dose response curve
at low doses, applicability of the animal evidence to humans, addressing susceptible populations).
Each synthesis will provide a summary discussion of the available evidence that addresses
considerations adapted from considerations for causality introduced by Austin Bradford Hill (Hill.
19651: consistency, exposure-response relationship, strength of the association, temporal
relationship, biological plausibility, coherence, and "natural experiments" in humans [(U.S. EPA.
2005.1994b): see Table 17]. Importantly, the evidence synthesis process explicitly considers and
incorporates the conclusions from the individual study evaluations (see Section 6).
14 This revision has been adopted primarily based on the 2014 NAS review of IRIS (NRC, 2014): 'The present
committee found that the phrase weight of evidence has become far too vague as used in practice today and thus
is of little scientific use. In some accounts, it is characterized as an oversimplified balance scale on which evidence
supporting hazard is placed on one side and evidence refuting hazard on the other... The present committee found
the phrase evidence integration to be more useful and more descriptive of what is done at this point in an IRIS
assessmentthat is, IRIS assessments must come to a judgment about whether a chemical is hazardous to human
health and must do so by integrating a variety of evidence."
This document is a draft for review purposes only and does not constitute Agency policy.
85 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 17. Information most relevant to describing primary considerations
informing causality during evidence syntheses
Consideration
Description of the consideration and its application in IRIS syntheses
Study confidence
Description: Incorporates decisions about study confidence within each consideration.
Application: In evaluating the evidence for each causality consideration described in
the following rows, syntheses will consider study confidence decisions. High
confidence studies carry the most weight. Syntheses will consider specific limitations
and strengths of studies and how they inform each consideration.
Consistency
Description: Examines the similarity of results (e.g., direction, magnitude) across
studies.
Application: Syntheses will evaluate the homogeneity of findings on a given outcome
or endpoint across studies. When inconsistencies exist, the syntheses consider
whether results were "conflicting" (i.e., unexplained positive and negative results in
similarly exposed human populations or in similar animal models) or "differing"
(i.e., mixed results explained by differences between human populations, animal
models, exposure conditions, or study methods) (U.S. EPA, 2005) based on analyses of
potentially important explanatory factors such as:
Confidence in the studies' results, including study sensitivity (e.g., some study
results that appear to be inconsistent could be explained by potential biases or
other attributes that affect sensitivity).
Exposure, including route (if applicable) and administration methods, levels,
duration, timing with respect to outcome development, and exposure
assessment methods (i.e., in epidemiology studies).
Specificity and sensitivity of the endpoint for evaluating the health effect in
question (e.g., functional measures can be more sensitive than organ weights).
Populations or species, including consideration of potential susceptible groups
or differences across lifestage at exposure or endpoint assessment.
Toxicokinetic information explaining observed differences in responses across
route of exposure, other aspects of exposure, species, or lifestages.
The interpretation of consistency will emphasize biological significance, to the extent it
is understood, over statistical significance. Statistical significance from suitably applied
tests adds weight when biological significance is not well understood. Consistency in
the direction of results increases confidence in that association even in the absence of
statistical significance. In some cases, considering the potential for publication bias
and providing context to interpretations of consistency could be helpful.3
This document is a draft for review purposes only and does not constitute Agency policy.
86 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Consideration
Description of the consideration and its application in IRIS syntheses
Strength (effect
magnitude) and
precision
Description: Examines the effect magnitude or relative risk, based on what is known
about the assessed endpoint(s) and considers the precision of the reported results
based on analyses of variability (e.g., confidence intervals, standard error). This might
include consideration of the rarity or severity of the outcomes.
Application: Syntheses will analvze results both within and across studies and could
consider the utility of combined analyses (e.g., meta-analysis). Although larger effect
magnitudes and precision (e.g., p < 0.05) help reduce concerns about chance, bias, or
other factors as explanatory, syntheses also will consider the biological or
population-level significance of small effect sizes.
Biological gradient/
dose-response
Description: Examines whether the results (e.g., response magnitude, incidence,
severity) change in a manner consistent with changes in exposure (e.g., level,
duration), including consideration of changes in response after cessation of exposure.
Application: Syntheses will consider relationships both within and across studies,
acknowledging that the dose-response (e.g., shape) can vary depending on other
aspects of the experiment, including the biology underlying the outcome and the
toxicokinetics of the chemical. Thus, when dose-response is lacking or unclear, the
synthesis also will consider the potential influence of such factors on the response
pattern.
Coherence
Description: Examines the extent to which findings are cohesive across different
endpoints that are related to, or dependent on, one another (e.g., based on known
biology of the organ system or disease, or mechanistic understanding such as
toxicokinetic/dynamic understanding of the chemical or related chemicals). In some
instances, additional analyses of mechanistic evidence from research on the chemical
under review or related chemicals that evaluate linkages between endpoints or
organ-specific effects might be needed to interpret the evidence. These analyses
could require additional literature search strategies.
Application: Syntheses will consider potentially related findings, both within and
across studies, particularly when relationships are observed within a cohort or within a
narrowly defined category (e.g., occupation, strain or sex, lifestage of exposure).
Syntheses will emphasize evidence indicative of a progression of effects, such as
temporal- or dose-dependent increases in the severity of the type of endpoint
observed. If an expected coherence between findings is not observed, possible
explanations will be explored including the biology of the effects and the sensitivity
and specificity of the measures used.
This document is a draft for review purposes only and does not constitute Agency policy.
87 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Consideration
Description of the consideration and its application in IRIS syntheses
Mechanistic evidence
related to biological
plausibility
Description: There are multiple uses for mechanistic information (see Section 9.2) and
this consideration overlaps with "coherence." This examines the biological support (or
lack thereof) for findings from the human and animal health effect studies and
becomes more impactful on the hazard conclusions when notable uncertainties in the
strength of those sets of studies exist. These analyses can also improve understanding
of dose- or duration-related development of the health effect. In the absence of
human or animal evidence of apical health endpoints, the synthesis of mechanistic
information could drive evidence integration judgments (when such information is
available).
Application: Syntheses can evaluate evidence on precursors, biomarkers, or other
molecular or cellular changes related to the health effect(s) of interest to describe the
likelihood the observed effects result from exposure. This will be an analysis of
existing evidence, and not simply whether a theoretical pathway can be postulated.
This analysis might not be limited to evidence relevant to the PECO but also could
include evaluations of biological pathways (e.g., for the health effect, established for
other, possibly related, chemicals). The synthesis will consider the sensitivity of the
mechanistic changes and the potential contribution of alternative or previously
unidentified mechanisms of toxicity.
Natural experiments
Description: Specific to epidemiology studies and rarelv available, this consideration
examines effects in populations that have experienced well-described, pronounced
changes in chemical exposure (e.g., lead exposures before and after banning lead in
gasoline).
Application: Compared to other observational designs, one benefit of natural
experiments is that people are divided into exposed and unexposed groups without
influencing their own exposure status. During synthesis, associations in medium and
high confidence natural experiments can substantially reduce concerns about residual
confounding.
PECO = populations, exposures, comparators, and outcomes.
a Publication bias involves the influence of the direction, magnitude, or statistical significance of the results on the
likelihood of a paper being published; it can result from decisions made, consciously or unconsciously, by study
authors, journal reviewers, and journal editors (Dickersin, 1990). When evidence of publication bias is present for
a set of studies, less weight might be placed on the consistency of the findings for or against an effect during
evidence synthesis and integration.
1
2 Data permitting, the syntheses will also discuss analyses relating to potential susceptible
3 populations.15 These analyses will be based on knowledge about the health outcome or organ
4 system affected, demographics, genetic variability, lifestage, health status, behaviors or practices,
15 Various terms have been used to characterize populations that could be at increased risk of developing
health effects from exposure to environmental chemicals, including "susceptible," "vulnerable," and
"sensitive." Further, these terms have been inconsistently defined across the scientific literature.
The term susceptibility is used in this protocol to describe populations at increased risk, focusing on
biological (intrinsic) factors and social and behavioral determinants that can modify the effect of a specific
exposure. However, certain factors resulting in higher exposures to specific groups (e.g., proximity,
occupation, housing) might not be analyzed to describe potential susceptibility among specific populations or
subgroups.
This document is a draft for review purposes only and does not constitute Agency policy.
88 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
social determinants, and exposure to other pollutants (see Table 18). This information will be used
to describe potential susceptibility among specific populations or lifestages in a separate section
(see Section 10.3). This summary will describe concerns across the available evidence for all
potential human health effects and will inform hazard identification and dose-response analyses.
Table 18. Individual and social factors that could increase susceptibility to
exposure-related health effects
Factor
Examples
Demographic
Gender, age, race/ethnicity, education, income, occupation, geography
Genetic variability
Polymorphisms in genes regulating cell cycle, DNA repair, cell division, cell
signaling, cell structure, gene expression, apoptosis, and metabolism
Lifestage
In utero, childhood, puberty, pregnancy, women of childbearing age, elderly
Health status
Preexisting conditions or disease such as psychosocial stress, elevated body
mass index, frailty, nutritional status, chronic disease
Behaviors or practices
Diet, mouthing, smoking, alcohol consumption, pica, subsistence or
recreational hunting and fishing
Social determinants
Income, socioeconomic status, neighborhood factors, health care access, and
social, economic, and political inequality
9.1. SYNTHESES OF HUMAN AND ANIMAL HEALTH EFFECT EVIDENCE
The syntheses of the human and animal health effect evidence will focus on describing
aspects of the evidence that best inform causal interpretations, including the exposure context
examined in the sets of studies. These syntheses (or the lack of data within these bodies of
evidence) help determine the approach to be taken in synthesizing the available mechanistic
evidence (see Section 9.2).
Evidence synthesis will be based primarily on studies of high and medium confidence. Low
confidence studies might be used, if few or no studies with higher confidence are available, to help
evaluate consistency, or if the study designs of the low confidence studies address notable
uncertainties in the set of high or medium confidence studies on a given health effect. If low
confidence studies are used, a careful examination of risk of bias and sensitivity with potential
impacts on the evidence synthesis conclusions will be included in the narrative.
As previously described, these syntheses will articulate the strengths and the weaknesses of
the available evidence organized around the considerations described in Table 17 and issues that
stem from the evaluation of individual studies (e.g., concerns about bias or sensitivity). If possible,
results across studies will be compared using graphs and charts or other data visualization
strategies. The analysis typically will include examination of results stratified by any or all of the
following: study confidence classification (or specific issues within confidence evaluation domains);
population or species; exposures (e.g., level, patterns [intermittent or continuous]; duration;
This document is a draft for review purposes only and does not constitute Agency policy.
89 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
intensity); sensitivity (e.g., low vs. high); and other factors that might have been identified in the
refined evaluation plan (e.g., sex, lifestage, other demographic). The number of studies and the
differences encompassed by the studies will determine the extent to which specific types of factors
can be examined to stratify study results. Additionally, for both the human and animal evidence
syntheses, if supported by the available data, additional analyses across studies (such as
meta-analysis) also might be conducted.
9.2. MECHANISTIC INFORMATION
Mechanistic information includes any experimental measurement related to a health
outcome that informs the biological or chemical events associated with phenotypic effects; these
measurements can improve understanding of the mechanisms involved in the biological effects
following exposure to a chemical but generally are not, by themselves, considered adverse
outcomes. Mechanistic data are reported in a diverse array of observational and experimental
studies across species, model systems, and exposure paradigms, including in vitro, in vivo (by
various routes of exposure), ex vivo, and in silico studies. The evidence available to describe
mechanistic events or MOAs fU.S. EPA. 20051 is typically aggregated from numerous studies, often
involving a diverse range of exposure paradigms and models and a wide spectrum of diverse
endpoints. In addition, a chemical could operate through multiple mechanistic pathways (U.S. EPA.
2005). Similarly, multiple mechanistic pathways might interact to cause an adverse effect In
contrast to the defined scope of the evaluation and syntheses of PECO-specific human or animal
health effect studies, the potential utility and interpretation of mechanistic information can be quite
broad and hard to define. Thus, to be pragmatic and provide clear and transparent syntheses of the
most useful information, the mechanistic syntheses for most health outcomes will focus on a subset
of the most relevant mechanistic studies. It should be stressed that the process of evaluating
mechanistic information differs fundamentally from evaluations of the other evidence streams.
More specifically, the mechanistic analysis for any specific substance will depend on evaluating the
confidence that the relevant data are consistent with a plausible biological understanding of how a
chemical exposure might generate an adverse outcome, rather than focusing on evaluations of
individual studies.
The synthesis of mechanistic information informs the integration of health effect evidence
for both hazard identification (i.e., biological plausibility or coherence of the available human or
animal evidence, inferences regarding human relevance, or the identification of susceptible
populations and lifestages across the human and animal evidence) and dose-response evaluation.
As introduced in Section 2.5, several key science issues essential to consider in this
assessment will involve a focused analysis and synthesis of mechanistic information. One such
issue is the identification of toxicokinetic parameters for use in pharmacokinetic models of PCBs,
particularly PCB congener half-life values needed to support route-to-route, interspecies, and
intraspecies extrapolations. Half-lives are known to vary significantly among the congeners for
This document is a draft for review purposes only and does not constitute Agency policy.
90 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
which they have been determined and are critical determinants of PCB body burden given long-
term exposure. Another key issue is the evaluation of the relative contributions of individual PCB
congeners to the toxicity of complex PCB mixtures. Because toxic potency can vary independently
from half-life, the mechanistic analyses will need to identify both half-life data and toxicity data
available for specific congeners and to estimate (e.g., using QSAR methods) half-lives and relative
potencies of congeners for which no data are available.
Other analyses within the syntheses of mechanistic information will focus on the evidence
most useful for informing key uncertainties in the human or animal health effect evidence. This
means that, for example, if extensive and consistent high confidence human or animal evidence is
available, the need to synthesize all available mechanistic evidence will likely be diminished. In
such cases, the synthesis will focus on the analysis and interpretation of smaller sets of mechanistic
studies that specifically address controversial issues that are anticipated to have a substantial
impact on the assessment conclusions. For example, data related to applicability of animal evidence
to humans when the human evidence is weak, or the shape of the dose-response curve at low
exposure levels when this understanding is highly uncertain and data informing this uncertainty
are available. Thus, consideration of biological understanding represents an important component
of the evidence analysis. However, mechanistic understanding is not a prerequisite for drawing a
conclusion that a chemical causes a given health effect fNTP. 2015: NRC. 20141.
To identify the focused set(s) of studies for use in analyses of critical mechanistic questions,
the synthesis will apply a phased approach that progressively focuses the scope of the mechanistic
information to be considered. This stepwise focusing, which begins during the literature search
and screening steps based on problem formulation decisions, depends primarily on the potential
hazard signals that arise from the human or animal health effect studies, or from mechanistic
studies that signal potential hazards that have not been examined in health effect studies (Table
19). Table 20 lists examples of the focused questions or scenarios triggering these mechanistic
evaluations and when, during the systematic review, they are likely to apply. Although the specific
methods for evaluating the sets of studies relevant to each question will vary, some general
considerations are provided below.
This document is a draft for review purposes only and does not constitute Agency policy.
91 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 19. Preparation for the analysis of mechanistic evidence
Assessment stages of
identifying
mechanistically
relevant information
Examples of evidence to review and key considerations
Scoping and problem
formulation materials
For the chemical under review, identify existing chemical-specific MOAs from
other agency assessments or review articles. If summary information is lacking,
are there structurally similar chemicals that are better studied mechanistically?
Are there indications that a specific mechanistic analysis will be warranted? For
example, are there recognized areas of scientific controversy or predefined
assessment questions that are already known to require a mechanistic
evaluation (e.g., chemicals with a potential mutagenic MOA)?
o If so, consider whether additional, targeted literature searches would be
informative.
o If mechanistic information relevant to a key scientific controversy or to
address a mutagenic MOA is lacking, consider whether inferences can be
drawn from structure-activity relationships or other "data-poor"
approaches.
What is the active moiety of the agent? Are there metabolites that should be
considered? Are there indications that the purity is critically important? Is the
chemical endogenously produced?
Literature inventory of
toxicokinetic, ADME, and
physicochemical
information
Based on ADME differences across species, does information exist that suggests
a lack of relevance of the animal exposure scenarios to human situations? Is
there evidence that the active moiety would not be expected to reach the
target tissue(s) in some species?
If exposure and risk need to be evaluated for routes of exposure not included in
existing PBPK models, how should this disconnect be addressed?
Literature inventories of
human, animal, and
mechanistic information
(including in vitro and in
silico studies)
Which human health hazards (both cancer and noncancer) appear to be well
studied in the mechanistic inventory? For cancer, which key characteristics of
carcinogens are indicated by the database?
o Are there mechanistic studies on an organ system, hazard, or key
characteristic that were not examined by human or animal studies meeting
the PECO criteria? If so, consider evidence mapping or similar approaches
to highlight these knowledge gaps.
Are there mechanistic endpoints identified from human and animal studies
meeting PECO criteria that could be added to the mechanistic inventory?
This document is a draft for review purposes only and does not constitute Agency policy.
92 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Assessment stages of
identifying
mechanistically
relevant information
Examples of evidence to review and key considerations
Human and animal
evidence syntheses
For the health effects of primary concern, is an in-depth mechanistic
evaluation(s) warranted to inform the available evidence in humans or animals?
Typically, this consideration would focus on health effects that show some
indication of an association in epidemiology studies or causality in experimental
studies. Based on the literature inventory, consider whether mechanistic data
are available to inform the specific, key uncertainties that remain. Examples of
specific scenarios for evaluation could include:
o If cancer has been observed and tumor types appear to differ across
populations (e.g., species or sex), can mechanistic evaluations inform
potential explanations (noting that site concordance is not a requirement
for determining the relevance of animal data for humans)?
o When notable uncertainties in the human or animal findings occur for a
health effect (e.g., outstanding methodological limitations), is evidence of
biological precursors in humans or animals linked to the observed
outcome? Precursors in the same studies or populations provide stronger
evidence.
o Were questions of relevance raised that could be addressed by an
evaluation of the mechanistic evidence to establish the human relevance of
effects observed in animal studies?
o Were pronounced, unexplained differences in susceptibility observed that
might be explained by an analysis of toxicokinetic or toxicodynamic
differences across lifestages or populations (e.g., animal strain, human
demographic)?
ADME = absorption, distribution, metabolism, and excretion; MOA = mode of action; PBPK = physiologically based
pharmacokinetic; PECO = populations, exposures, comparators, and outcomes.
The information collected (e.g., in sortable inventories) will be used to identify studies
available for consideration in addressing the specific gaps in understanding identified as critical to
address through the application of the questions in Table 2 0, including postulated mechanistic
pathways or MOAs that might be involved in the toxicity of the chemical. Subsequently, from the
studies available to potentially address the identified gaps in understanding, the synthesis will
focus on those considered most impactful to the specified evaluation based on study design
characteristics (which might or might not encompass all studies relevant for a particular question),
with transparent documentation of the rationale for the focusing. As the potential influence of the
information provided by these studies can vary depending on the hazard question(s) or the
associated mechanistic events or pathways, the level of rigor also will depend on the potential
impact of increased understanding to hazard identification or dose-response decisions, and could
range from overviews of potential mechanisms or cursory insights drawn from sets of unanalyzed
results to detailed evaluations of a subset of the most relevant mechanistic studies.
This document is a draft for review purposes only and does not constitute Agency policy.
93 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Although the application of this approach cannot be predefined, for the small subsets of
studies that best address the key mechanistic questions, the synthesis will first prioritize the
studies based on their toxicological relevance to answering the specific question (e.g., model
system, specificity of the assay for the effect of interest). The path for focusing the mechanistic
database will be documented in the updated protocol released with the draft assessment
More rigorous analyses will be particularly important when the sets of studies available to
inform influential mechanistic conclusions are inconsistent and potentially conflicting, or when the
studies include experiments that directly challenge the necessity of proposed mechanistic
relationships between exposure and an apical effect (e.g., altering a receptor-mediated pathway
through chemical intervention or using knock-out animals). More detailed analyses also might be
useful when the study design aspects in the available studies are likely to have significant flaws or
introduce important uncertainties (e.g., potential shortcomings identified during the evaluation of
exposure methods might be clarified using mechanistic studies). In some instances, additional
literature searches could be warranted, targeting mechanistic events or biological pathways that
are not specific to one chemical.
For the more rigorous mechanistic analyses, the review will be facilitated using pathway-
based organizational methods and established evidence evaluation frameworks. These approaches
provide transparency and objectivity to the integration and interpretation of mechanistic events
and pathways anchored to the specific questions that have been identified (e.g., anchored to a
specific health effect) across diverse sets of relevant data (e.g., human, animal, in vitro studies).
The mechanistic analyses will inform the evidence integration and dose-response analyses,
described in Sections 11 and 12. Examples of how mechanistic information can inform these steps
are summarized in Table 20.
This document is a draft for review purposes only and does not constitute Agency policy.
94 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 20. Examples of iterative questions and considerations that focus the
synthesis and application of mechanistic information for evidence integration
and dose-response analysis
Systematic review step
Mechanistic synthesis triggers and example actions
Human and animal evidence
syntheses (see Section 9.1).
Did the sets of studies report findings that appear to be biologically related
to the health effects of interest? Consider whether these findings might
serve as precursors informing an association between exposure and effect;
if the set of studies has notable uncertainties (e.g., they are all low
confidence), consider a focused analysis of precursors to inform strength of
evidence; if the data amenable to dose-response analysis are weaka or if
responses are observed only at high exposure levels, consider evaluating the
precursor data for quantitative analysis.
Do the results appear to differ by categories that indicate the apparent
presence of susceptible populations (e.g., across demographics, species,
strains, sexes, or lifestages)? Consider analyses to better characterize the
sources and impact of potential susceptibilities that might be explained by
mechanistic information (e.g., due to genetic polymorphisms or metabolic
differences).
Were other key uncertainties or data gaps identified during the analyses of
the sets of available human or animal health effect studies? If so, does the
literature inventory of mechanistic studies indicate the likelihood of a
reasonable number of studies on the topic? If yes, a focused analysis of
these studies could be informative. If no, consider whether an additional
focused search of mechanistic information might be worthwhile (i.e., to
identify other informative studies not captured by the initial PECO).
Evidence integration (see
Section 10.1): Information
relating to biological
plausibility
Are there notable uncertainties in the sets of human or animal health effect
studies for which related mechanistic information is available? An
understanding of mechanistic pathways (e.g., by identifying mechanistic
precursor events linked qualitatively or quantitatively to apical health
effect[s]) can increase the strength of the evidence integration judgments.
This document is a draft for review purposes only and does not constitute Agency policy.
95 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Systematic review step
Mechanistic synthesis triggers and example actions
Evidence integration (see
Section 10.2): Considering
human relevance of animal
findings
When human evidence is lacking or has results that differ from animal
studies, is there evidence that the mechanisms underlying the effects in
animals operate in humans? Analyses of the mechanisms underlying the
animal response in relation to those presumed to operate in humans, or the
suitability of the animal models to a specific human health outcome, can
inform the extent to which the animal response is likely to be directly
relevant to humans.
The analysis will focus on evaluations of the following issues. The extent of
the analysis will vary depending on the impact of the animal evidence to
the conclusions.
o Evidence for a plausible mechanistic pathway or MOA, within which
the key events and relationships are evaluated regarding the likelihood
of similarities (e.g., in presence or function) across species.
o Coherence of mechanistic changes observed in exposed humans (or a
demonstrated lack of changes that would be expected, e.g., that are
known to be linked to the health effect) with animal evidence of
mechanistic/toxicological changes.
o ADME information describing similarities across species, primarily
relating to distribution (e.g., to the likely target tissue).
Evidence integration (see
Sections 10.2 and 10.3):
Characterizing potential
susceptible populations or
lifestages
A mechanistic understanding of how a health outcome develops, even
without a full MOA, can clarify characteristics of important events
(e.g., their presence or sensitivity across lifestages or across genetic
variations) and helps identify susceptible populations.
Identification of lifestages or groups likely to be at greatest risk can clarify
hazard descriptions and identify key data gaps including whether the most
susceptible populations or lifestages have been adequately tested. If a
proposed mechanistic pathway or MOA indicates a sensitive population or
lifestage in humans, consider whether the appropriate analogous exposures
and populations or lifestages were adequately represented in the human or
animal database.
When there is evidence of susceptibilities, but specific studies addressing
these susceptibilities are unavailable for quantitative analysis, susceptibility
data might support refined human variability UFs or probabilistic
uncertainty analyses.
This document is a draft for review purposes only and does not constitute Agency policy.
96 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Systematic review step
Mechanistic synthesis triggers and example actions
Dose-response analysis (see
Section 11):
Biological understanding,
including the identification of
precursor events
A biological understanding of mechanistic events/MOAs, including the
identification of precursor events in humans and the exposure conditions
expected to result in these effects, can inform the use of
o particular dose-response models (e.g., models integrating data across
several related outcomes or incorporating toxicokinetic knowledge)
o proximal measures of exposure (e.g., external vs. internal metrics)
o surrogate endpoints (e.g., use of well-established precursors in lieu of
direct observation of apical endpoints)
o improved characterization of responses (e.g., combination of related
outcomes, such as benign and malignant tumors resulting from the
same MOA).
PECO = populations, exposures, comparators, and outcomes; MOA = mode of action; UF = uncertainty factor.
a Note that "weak" here refers to the study's usability for dose-response analysis specifically. Such studies might be
judged to be of medium or high confidence for the purposes of identifying potential hazards but possess
limitations preventing their use for deriving reliable quantitative estimates.
This document is a draft for review purposes only and does not constitute Agency policy.
97 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
10. EVIDENCE INTEGRATION
For the analysis of human health outcomes that might result from chemical exposure, IRIS
assessments draw integrated judgments across human, animal, and mechanistic evidence for each
assessed health effect (see Section 9). During evidence integration, a structured and documented
process will be used, as follows (and depicted in Figure 24):
Building from the separate syntheses of the human and animal evidence (see Section 9.1),
the strength of the evidence from the available human and animal health effect studies will
be summarized in parallel, but separately, using a structured evaluation of an adapted set of
considerations first introduced by Sir Bradford Hill fHill. 19651. Table 22 describes these
structured evaluations and the explicit consideration of study confidence within each
evaluation domain. Based on the approaches and considerations described in Section 9.2,
these summaries will incorporate mechanistic evidence (or MOA understanding) that
informs the biological plausibility and coherence within the available human or animal
health effect studies.
The strength of the animal and human evidence will be considered together in light of
inferences across evidence streams. Specifically, the inferences considered during this
integration include the human relevance of the animal and mechanistic evidence, coherence
across the separate bodies of evidence, and other important information (e.g., judgments
regarding susceptibility). Without evidence to the contrary, the human relevance of animal
findings is assumed.
A summary judgment is drawn as to whether the available evidence base for each potential
human health effect as a whole is sufficient (or insufficient) to indicate that PCB exposure
has the potential to be hazardous to humans.16
16 Due to the expected rarity of scenarios where there is "sufficient evidence to judge that a hazard is unlikely"
(see description in Table 23 and section 10.2) and to improve readability, this judgment is not specified in
some instances.
This document is a draft for review purposes only and does not constitute Agency policy.
98 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Stronger bodies of evidence: for example, consistent
among high or medium confidence studies, and may
have additional support; little or some uncertainty
Weaker bodies of evidence: for example, conflicting
or low confidence studies, with no additional
support; a large amount of uncertainty
Incorporating confidence in individual studies (risk of
bias; insensitivity) into review of each consideration
Sufficient
evidence
Insufficient
evidence
Evidence Integration
Summary Judgment
Overall judgments across
evidence for each
potential human health
effect, including evidence
basis rationale
Consistency
Dose-response
Magnitude & Precision
Coherence
Mechanistic evidence on biological
plausibility
Inference Across Evidence Streams
Information on the human relevance of
the animal and mechanistic evidence
Coherence across bodies of evidence or
with related effects
Other (e.g., read-across; susceptibility)
Evidence Stream Evaluation
Based on the structured review of adapted Hill
considerations (including biological understanding),
as part of the evidence integration narrative:
Qualitatively summarize the strength of the
evidence from studies in humans.
Qualitatively summarize the strength of the
evidence from animal studies.
Figure 24. Process for evidence integration. Note that "sufficient evidence"
could indicate a judgment of "sufficient evidence for hazard" or "sufficient evidence
to judge that a hazard is unlikely", depending on the nature and extent of the
available evidence (see Table 23).
The decision points within the structured evidence integration process will be summarized
in an evidence profile table for each health effect category (see Table 21 for a preliminary template
version) in support of the evidence integration narrative. The specific decision frameworks for the
structured evaluation of the strength of the human and animal evidence streams and for drawing
the overall evidence integration judgment are described in Section 10.1. This process is similar to
that used by the Grading of Recommendations Assessment, Development and Evaluation (Morgan et
al.. 2016: Guvattetal.. 2011: Schiinemann et al.. 20111. which arrives at an overall integration
conclusion based on consideration of the body of evidence. As described in Section 9, the human,
animal, and mechanistic syntheses serve as inputs providing a foundation for the evidence
integration decisions; thus, the major conclusions from these syntheses will be summarized in the
evidence profile table (see Table 21 for a preliminary template version) supporting the evidence
integration narrative. The evidence profile tables for each potential human health effect evaluated
will summarize the judgments and their evidence basis for each step of the structured evidence
integration process. Separate sections are included for summarizing the human and animal
evidence, for the inference drawn across evidence streams, and for the overall evidence integration
judgment The table presents the key information from the different bodies of evidence that
informs each decision.
This document is a draft for review purposes only and does not constitute Agency policy.
99 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 21. Evidence profile table template
Studies and
interpretation
Factors that increase
strength
Factors that
decrease strength
Summary of evidence streams
Inferences across evidence
streams
Overall evidence integration
judgment
Evidence from studies of humans (may be presented by exposure route)3
Human relevance of
findings in animals
Cross-stream coherence
Other inferences:
o Information on
susceptibility
o MOA analysis
inferences
o Relevant information
from other sources
(e.g., read across)
Describe judgment regarding
whether there is sufficient (or
insufficient) evidence to
identify a potential human
health hazard, integrating
evidence across streams and
including a summary of the
models and range of dose
levels upon which the
judgment is primarily reliant.
References
Study confidence
Study design
description
Consistency
Dose-response
gradient
Coherence of
observed effects
Effect size
Mechanistic
evidence providing
plausibility
Medium or high
confidence studies'5
Unexplained
inconsistency
Imprecision
Low confidence
studies'5
Evidence
demonstrating
implausibility
Qualitative summary of the strength
of the evidence from human studies
based on the factors at left, including
the primary evidence basis and
considering:
Results across human
epidemiological and controlled
exposure studies
Human mechanistic evidence
informing biological plausibility
(e.g., precursor events linked to
adverse outcomes)
Evidence from animal studies (may be presented by exposure route)3
References
Study confidence
Study design
description
Consistency or
replication
Dose-response
gradient
Coherence of
observed effects
Effect size
Mechanistic
evidence providing
plausibility
Medium or high
confidence studies'5
Unexplained
inconsistency
Imprecision
Low confidence
studies'5
Evidence
demonstrating
implausibility
Qualitative summary of the strength
of the evidence for an effect in
animals based on the factors at left,
including the primary evidence basis
and considering:
Results across animal toxicology
studies
Animal mechanistic evidence
informing biological plausibility
(e.g., precursor events linked to
adverse outcomes)
MOA = mode of action.
a In addition to exposure route, the summaries of the strength of each evidence stream may include multiple rows (e.g., by study confidence, population, or species) if this
informs the analysis of results heterogeneity.
b Study confidence, based on evaluation of risk of bias and study sensitivity (see Section 6), and information on susceptibility will be considered when evaluating each of the
other factors that increase or decrease strength (e.g., consistency). Notably, lack of findings in studies deemed insensitive neither increases nor decreases strength.
This document is a draft for review purposes only and does not constitute Agency policy.
100 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
10.1. EVALUATING THE STRENGTH OF THE HUMAN AND ANIMAL
EVIDENCE STREAMS
As summarized above, prior to drawing overall evidence integration judgments about
whether exposure to PCBs has the potential to cause certain health effect(s) in humans given
relevant exposure circumstances, the strength of evidence for the available human and animal
evidence will be evaluated and summarized. For each assessed health effect or health effect
grouping (see Section 5 for examples of the endpoints that will be considered within each health
effect category), the relevant mechanistic evidence in exposed humans and animals (or in their
cells, relevant new approach methods [NAMs], or in silico models), which will be synthesized based
on the approaches and considerations in Section 9.2, will be integrated with the evidence from the
available studies of phenotypic effects in humans and animals. The considerations outlined in
Table 17 (the different features of the evidence considered and summarized during evidence
synthesis; see Section 9) will be evaluated in the context of how they impact judgements of the
strength of evidence (see Table 22), which will directly inform the overall evidence integration
judgment (see section 10.2). The evaluation of the strength of the human or animal health effects
evidence (i.e., based on the considerations in Table 22) will preferably occur at the most specific
health outcome level possible (e.g., an analysis at the level of decreased pulmonary function is
generally preferable to an analysis of respiratory system effects), if there is an adequate set of
studies for analyses at this level and considering the interrelatedness of the available outcomes. If
studies on a target system are sparse or varied, or if the interpretation of evidence strength relies
largely on the consideration of coherence across related outcomes, then the analyses may need to
be conducted at a broader health effect level. The factors judged to increase or decrease the
strength of the evidence will be summarized in tabular format using the evidence profile table
template in Table 21 to transparently convey, for each health effect or outcome grouping, expert
judgments made throughout the evidence synthesis and integration processes. The evidence
profile table allows for consistent documentation of the supporting rationale for each decision.
This document is a draft for review purposes only and does not constitute Agency policy.
101 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 22. Considerations that inform evaluations of the strength of the human and animal evidence
Consideration
Increased evidence strength
(of the human or animal evidence)
Decreased evidence strength
(of the human or animal evidence)
The structured categories and criteria in Table 23 (section 10.2) will guide the application of strength of evidence judgments for an outcome or health effect. Evidence synthesis
scenarios that do not warrant an increase or decrease in evidence strength for a given consideration will be considered "neutral" and are not described in this table (and, in general, will
not be captured in the assessment-specific evidence profile tables).
Risk of bias;
sensitivity (across
studies)
An evidence base of high or medium confidence studies increases
strength.
An evidence base of mostly low confidence studies decreases strength. An
exception is an evidence base of studies where the primary issues resulting in low
confidence are related to insensitivity. This might increase evidence strength in
cases where an association is identified because the expected impact of study
insensitivity is toward the null.
Decisions to increase strength for other considerations in this table will generally
not be made if there are serious concerns for risk of bias.
Consistency
Similarity of findings for a given outcome (e.g., of a similar
magnitude, direction) across independent studies or experiments
increases strength, particularly when consistency is observed across
populations (e.g., location) or exposure scenarios in human studies,
and across laboratories, populations (e.g., species), or exposure
scenarios (e.g., duration; route; timing) in animal studies.
Unexplained inconsistency (conflicting evidence) decreases strength. Generally,
strength will not be decreased if discrepant findings can be explained reasonably
by factors including study confidence conclusions, variation in population or
species, sex, lifestage, exposure patterns (e.g., intermittent or continuous),
exposure levels (low or high), exposure duration, or exposure intensity.
Strength (effect
magnitude) and
precision
Evidence of a large-magnitude effect (considered either within or
across studies) can increase strength. Effects of a concerning rarity
or severity can also increase strength, even if the magnitude is small.
Precise results from individual studies or across the set of studies
increases strength, noting that biological significance is prioritized
over statistical significance.
Strength might be decreased if effect sizes that are small in magnitude are
concluded not to be biologically significant, or if there are only a few studies with
imprecise results.
This document is a draft for review purposes only and does not constitute Agency policy.
102 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Consideration
Increased evidence strength
(of the human or animal evidence)
Decreased evidence strength
(of the human or animal evidence)
Biological gradient/
dose-response
Evidence of dose-response increases strength. Dose-response can be
demonstrated across studies or within studies and can be dose or
duration dependent. The dose-response need not be monotonic
(monotonicity is not necessarily expected, e.g., different outcomes
could be expected at low vs. high doses due to activation of different
mechanistic pathways or induction of systemic toxicity at very high
doses).
Decreases in a response after cessation of exposure (e.g., symptoms
of current asthma) also might increase strength by increasing
certainty in a relationship between exposure and outcome (this is
especially useful for interpreting evidence drawn from epidemiology
studies because of their observational nature).
A lack of dose-response when expected based on biological understanding and
having a wide range of doses/exposures evaluated in the evidence base can
decrease strength.
If the data are not adequate to evaluate a dose-response pattern, strength is
neither increased nor decreased.
Coherence
Biologically related findings within an organ system, or across
populations (e.g., sex) increase strength, particularly when a
temporal- or dose-dependent progression of related effects is
observed within or across studies, or when related findings of
increasing severity are observed with increasing exposure.
An observed lack of expected coherent changes (e.g., well-established biological
relationships) typically will decrease evidence strength. However, the biological
relationships between the endpoints being compared and the sensitivity and
specificity of the measures used need to be carefully examined. The decision to
decrease depends on the availability of evidence across multiple related endpoints
for which changes would be anticipated, and it considers factors (e.g., dose and
duration of exposure, strength of expected relationship) across the studies of
related changes.
Mechanistic evidence
related to biological
plausibility
Mechanistic evidence of precursors or health effect biomarkers in
well-conducted studies of exposed humans or animals, in
appropriately exposed human or animal cells, or other relevant
human, animal, or in silico models (including NAMs) increases
strength, particularly when this evidence is observed in the same
cohort/population exhibiting the phenotypic health outcome.
Evidence of changes in biological pathways or support for a proposed
MOA in appropriate models also increases strength, particularly
when support is provided for rate-limiting or key events or is
conserved across multiple components of the pathway or MOA.
Mechanistic understanding is not a prerequisite for drawing a conclusion that a
chemical causes a given health effect; thus, an absence of knowledge will not be
used as a basis for decreasing strength (NTP, 2015; NRC, 2014).
Mechanistic evidence in well-conducted studies (see examples of evidence types at
left) that demonstrates the health effect(s) are unlikely to occur, or likely to occur
only under certain scenarios (e.g., above certain exposure levels), can decrease
evidence strength. A decision to decrease depends on an evaluation of the
strength of the mechanistic evidence supporting vs. opposing biological plausibility
and on the strength of the health effect-specific findings (e.g., stronger health
effect data require more certainty in mechanistic evidence opposing plausibility).
MOA = mode of action; NAM = new approach method.
a Publication bias can result in strength of evidence judgments that are stronger than would be merited if the entire body of research were available. However, the existence of
publication bias can be difficult to determine. If strong evidence of publication bias exists for an outcome, the increase in evidence strength resulting from considering the
consistency of the evidence across studies could be reduced.
This document is a draft for review purposes only and does not constitute Agency policy.
103 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
For human and animal evidence, the analyses of each consideration in Table 22 will be used
to qualitatively summarize the strength of evidence for the separate evidence streams in the
evidence integration narrative. Table 23 provides the criteria that will guide how to develop the
judgment for each health effect, and the terms that will be used to summarize those evidence
integration judgments.
10.2. OVERALL EVIDENCE INTEGRATION JUDGMENTS
Evidence integration combines decisions regarding the strength of the animal and human
evidence with considerations regarding mechanistic information on the human relevance of the
animal evidence, relevance of the mechanistic evidence to humans (especially in cases where
animal evidence is lacking), coherence across bodies of evidence, and information on susceptible
populations and lifestages. This evidence integration decision process will culminate in an
evidence integration narrative that summarizes the judgments regarding the evidence for each
potential health effect evaluated. For each health effect, this narrative will include
A descriptive summary of the primary judgments about the evidence informing the
potential for health effects in exposed humans, based on the following analyses:
o evaluations of the strength of the available human and animal evidence (see Section
10.1);
o consideration of the coherence of findings (i.e., the extent to which the evidence for
health effects and relevant mechanistic changes are similar) across human and animal
studies;
o other information on the human relevance of findings in animals (see Section 9.2); and
o conclusions drawn based on mechanistic analyses (see Section 9.2).
A summary of key evidence supporting these judgments, highlighting the evidence that was
the primary driver of these judgments and any notable issues (e.g., data quality; coherence
of the results), and a narrative expression of confidence (a summary of strengths and
remaining uncertainties) for these judgments.
Information on the general conditions of expression of these health effects (e.g., exposure
routes and levels in the studies that were the primary drivers of these judgments), noting
that these conditions will be clarified during dose response analysis (see Section 11).
Indications of potentially susceptible populations or lifestages (i.e., an integrated summary
of the available evidence on potential susceptible populations and lifestages drawn across
the syntheses of the human, animal, and mechanistic evidence).
A summary of key assumptions used in the analysis, which are generally based on EPA
guidelines and which are largely captured in this protocol.
Strengths and limitations of the evidence integration judgments, including key uncertainties
and data gaps, as well as the limitations of the systematic review.
This document is a draft for review purposes only and does not constitute Agency policy.
104 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
In short, the evidence integration narrative will present a qualitative summary of the
strength of each evidence stream and an overall judgment across all relevant evidence, with
exposure context provided. For each health effect, the first sentence of the evidence integration
narrative will include the summary judgment The assessment will also include evidence profile
tables (see Table 21) to support the evidence integration narrative by providing the major
decisions and supporting rationale. Table 23 describes the categories of evidence integration
judgments that will be used in this assessment and provides examples of database scenarios that fit
each category of evidence. These summary judgments provide a succinct and clear representation
of the decisions from the more detailed analyses of whether the evidence strength indicates that
PCB exposure has the potential to cause the human health effect(s) under specified exposure
conditions. Consistent with EPA guidelines, a judgment that the evidence supports an apparent lack
of an effect of PCB exposure on the health effect(s) will only be used when the available data are
considered robust for deciding that there is no basis for human hazard concern; lesser levels of
evidence suggesting a lack of an effect will be characterized as "insufficient"
This document is a draft for review purposes only and does not constitute Agency policy.
105 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 23. Evidence integration judgments for characterizing potential human health hazards in the evidence
integration narrative
Evidence
integration
judgment3
Evidence in studies of humans
Evidence in animal studies
Inferences across evidence streams
Sufficient
evidence for
hazard
A judgment of sufficient evidence for hazard requires that a scenario below is met for either the evidence in studies of humans OR evidence in animal studies,
incorporating the considerations outlined under inferences across evidence streams. The scenarios justifying this judgment span a broad range of overall
evidence strength, and examples are provided below, starting with the weakest evidence.15
Strong mechanistic evidence in well-conducted
studies of exposed humans (medium or high
confidence) or human cells (including NAMs), in
the absence of other substantive data, where an
informed evaluation has determined that the
data are reliable for assessing toxicity relevant to
humans and the mechanistic events have been
causally linked to the development of the health
effect of interest.0
A single high or medium confidence study
without supporting coherent evidence or concern
for unexplained inconsistency. Specifically, there
are no comparable studies of similar confidence
and sensitivity providing conflicting evidence, or
the differences can be reasonably explained by,
e.g., the populations or exposure levels studied
(U.S. EPA, 2005).
Multiple studies showing generally consistent
findings, including at least one high or medium
confidence study and supporting evidence, but
with some residual uncertainty due to potential
chance, bias, or confounding (e.g., effect
estimates of low magnitude or small effect sizes
given what is known about the endpoint;
uninterpretable patterns with respect to
exposure levels). Alternatively, a single high or
medium confidence study with a large magnitude
or severity of the effect, a dose-response
gradient, or other factors that increase the
evidence strength, without serious residual
Strong mechanistic evidence in well-conducted
studies of animals or animal cells (including
NAMs), in the absence of other substantive
data, where an informed evaluation has
determined the assays are reliable for assessing
toxicity relevant to humans and the
mechanistic events have been causally linked
to the development of the health effect.0
A single high or medium confidence
experiment in the absence of comparable
experiment(s) of similar confidence and
sensitivity providing conflicting evidence (i.e.,
evidence that cannot be reasonably explained,
e.g., by respective study designs or differences
in animal model (U.S. EPA, 2005)).d
At least one high or medium confidence study
with supporting information increasing the
strength of the evidence. Although the results
are largely consistent, notable uncertainties
remain. However, in scenarios when
inconsistent evidence or evidence indicating
nonspecific effects exist, it is not judged to
reduce or discount the level of concern
regarding the positive findings, or it is not from
a comparable body of higher confidence,
sensitive studies.d The additional support
provided includes either consistent effects
across laboratories or species; coherent effects
across multiple related endpoints; an unusual
magnitude of effect, rarity, age at onset, or
Supplemental evidence (e.g., structure-
activity data; chemical class
information; other NAMs) is judged to
increase the strength of limited or near-
equivocal, chemical-specific human or
animal evidence to sufficient evidence
for hazard.
Coherent or biologically consistent
findings across evidence streams
increases the strength of limited or
near-equivocal human or animal
evidence (e.g., single or few high or
medium confidence studies with some
conflicting evidence) to sufficient
evidence for hazard.
The strength of the evidence is
decreased because mechanistic
information (even if it does not provide
MOA understanding) raises
uncertainties regarding the human
and/or animal evidence, but overall the
evidence is still considered strong
enough to result in a judgment of
sufficient evidence for hazard.
The strength of the evidence is
decreased because findings across
evidence streams are conflicting (U.S.
EPA, 2005) or biologically inconsistent,
but a judgment of sufficient evidence for
hazard is
Stronger Evidence Stream Scenarios
7
This document is a draft for review purposes only and does not constitute Agency policy.
106 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
uncertainties. In both scenarios, associations
with related endpoints, including mechanistic
evidence from exposed humans, can address
uncertainties relating to exposure response,
temporality, coherence, and biological
plausibility, and any conflicting evidence is not
from a comparable body of higher confidence,
sensitive studies.d
A set of high or medium confidence independent
studies reporting an association between the
exposure and the health outcome, with
reasonable confidence that alternative
explanations, including chance, bias, and
confounding, can be ruled out across studies.
The set of studies is primarily consistent, with
reasonable explanations when results differ; and
an exposure response gradient is demonstrated.
Supporting evidence, such as associations with
biologically related endpoints in human studies
(coherence) or large estimates of risk or severity
of the response, may help to rule out alternative
explanations. Similarly, mechanistic evidence
from exposed humans may serve to address
uncertainties relating to exposure-response,
temporality, coherence, and biological
plausibility (i.e., providing evidence consistent
with an explanation for how exposure could
cause the health effect based on current
biological knowledge).
severity; a strong dose response relationship;
or consistent observations across exposure
scenarios (e.g., route, timing, duration), sexes,
or animal strains. Mechanistic evidence in
animals may serve to provide this support or
otherwise address residual uncertainties.
A set of high or medium confidence
experiments with consistent findings of
adverse or toxicologically significant effects
across multiple laboratories, exposure routes,
experimental designs (e.g., a subchronic study
and a two-generation study), or species; and
the experiments reasonably rule out the
potential for nonspecific effects to have caused
the effects of interest. Any inconsistent
evidence (i.e., evidence that cannot be
reasonably explained based on study design or
differences in animal model) is from a set of
experiments of lower confidence or sensitivity.
To reasonably rule out alternative
explanations, multiple additional factors in the
set of experiments exist, such as: coherent
effects across biologically related endpoints; an
unusual magnitude of effect, rarity, age at
onset, or severity; a strong dose response
relationship; or consistent observations across
animal lifestages, sexes, or strains. Similarly,
mechanistic evidence (e.g., precursor events
linked to adverse outcomes) in animal models
may exist to address uncertainties in the
evidence base.
supported by review of the adversity
and human relevance (prioritizing
findings relevant to human toxicity) of
the effects.
The strength of the evidence is neither
increased nor decreased due to a lack of
experimental information on the human
relevance of the animal evidence or
mechanistic understanding (mechanistic
evidence may exist, but it is
inconclusive); in these cases, the animal
data are judged not to conflict with
current biological understanding and
thus are assumed to be relevant, while
findings in humans and animals are
presumed to be real unless proven
otherwise.
For the strongest animal evidence,
there is mechanistic understanding that
the findings are expected to occur and
progress in humans. Most notably, an
MOA interpreted with reasonable
certainty would rule out alternative
explanations.
For the strongest evidence, there is
adequate testing of potentially
susceptible lifestages and populations,
based on the effect(s) of interest and
chemical knowledge (e.g.,
toxicokinetics).
Stronger Evidence Stream Scenarios
7
Insufficient
evidence
A judgment of insufficient evidence requires that a scenario below is met for both the evidence in studies of humans AND evidence in animal studies,
incorporating the considerations outlined under inferences across evidence streams.
A body of evidence, including scenarios with one or
more high or medium confidence studies reporting an
association between exposure and the health outcome,
where either (1) conflicting evidence exists in studies of
similar confidence and sensitivityd e OR (2) considerable
methodological uncertainties remain across the body
of evidence (typically related to exposure or outcome
A body of evidence, including scenarios with one or
more high or medium confidence experiments
reporting effects but without supporting coherent
evidence that increases the overall evidence strength,
where conflicting evidence exists from a set of
sensitive experiments of similar or higher confidence
(can include mechanistic evidence).d e
The evidence in animal studies meets a
scenario for sufficient evidence for
hazard, but strong experimental
evidence (e.g., an MOA interpreted with
reasonable certainty) indicates the
findings in animals are unlikely to be
relevant to humans.
This document is a draft for review purposes only and does not constitute Agency policy.
107 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
ascertainment, including temporality), AND there is no
supporting coherent evidence that increases the overall
evidence strength.
A set of only low confidence studies.
No studies of exposed humans or well-conducted
studies of human cells.
A set of largely null studies that does not meet a
scenario for sufficient evidence to judge that a hazard is
unlikely.
A set of only low confidence experiments.
No animal studies or well-conducted studies of
animal cells.
The available endpoints are not informative to the
hazard question under evaluation.
A set of largely null studies that does not meet the
criteria for sufficient evidence to judge that a hazard
is unlikely.
The evidence meets a scenario for
sufficient evidence to judge that a
hazard is unlikely, but there is
inadequate testing of susceptible
populations and lifestages, or data
conflict across evidence streams.
The evidence in animal studies meets a
scenario for sufficient evidence to judge
that a hazard is unlikely, but the
database lacks experimental support
that the models are relevant to humans
for the effect of interest.
Sufficient
evidence to
judge that a
hazard is
unlikelyf
A judgment of sufficient evidence to judge that a hazard is unlikely requires that a scenario below is met for either the evidence in studies of humans OR
evidence in animal studies, incorporating the considerations outlined under inferences across evidence streams.
Several high confidence studies showing null results
(e.g., an odds ratio of 1.0), ruling out alternative
explanations including chance, bias, and confounding
with reasonable confidence. Each of the studies will
have used an optimal outcome and exposure
assessment and adequate sample size (specifically for
higher exposure groups and for susceptible
populations). The overall set will include the full range
of levels of exposures that human beings are known to
encounter and an evaluation of an exposure-response
gradient.
A set of high confidence experiments examining a
reasonable spectrum of endpoints relevant to a type
of toxicity that demonstrate a lack of biologically
significant effects across multiple species, both sexes
(if applicable), and a broad range of exposure levels.
The data are compelling in that the experiments have
examined the range of scenarios across which health
effects in animals could be observed, and an
alternative explanation (e.g., inadequately controlled
study designs; inadequate sample sizes) for the
observed lack of effects is not available. The
experiments were designed to specifically test for the
effects of interest, including suitable exposure timing
and duration, post-exposure latency, and endpoint
evaluation procedures.
There is adequate testing of susceptible
populations and lifestages.
When the evidence in animal studies
meets a scenario for this judgment,
there is experimental support that the
models are relevant to humans for the
effect of interest and no conflicting
human evidence exists.
When the evidence in studies of humans
meets a scenario for this judgment and
conflicting animal data exist,
mechanistic information indicates that
the animal data are unlikely to be
relevant to humans.
When multiple high confidence animal
experiments and studies in humans
indicate lack of an effect, but the
evidence does not meet a scenario for
sufficient evidence to judge that a
hazard is unlikely, strong mechanistic
evidence in models relevant to humans
supports lack of an effect such that the
totality of evidence supports this
judgment.
This document is a draft for review purposes only and does not constitute Agency policy.
108 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
NAM = new approach method.
a These categories are based on those indicated for use in hazard characterization from the existing EPA guidelines for noncancer health effects (U.S. EPA, 1998. 1996a. 1991)
and, as described in those guidance documents, they depend heavily on expert judgment (note: as applied herein, the process of 'evidence integration' is synonymous with
'weight of evidence'). The evidence integration judgment for each assessed health effect will be included as part of an evidence integration narrative, the specific
documentation of the various expert decisions and evidence-based (or default) rationales are summarized in an evidence profile table, and the judgement will be
contextualized based on the primary supporting evidence (experimental model or observed population, and exposure levels tested or estimated). Importantly, as discussed in
Section 10.1, these judgments may be based on analyses of grouped outcomes at different levels of granularity (e.g., motor activity versus neurobehavioral effects versus
nervous system effects) depending on the specifics of the health effect evidence base. Health effects characterized as having sufficient evidence for hazard will be evaluated for
use in dose-response assessment.
b Qualitative descriptions of differences in the strength of the evidence across different health effects judged as having sufficient evidence for hazard are useful for other
assessment decisions, including prioritizing outcomes in quantitative analyses and characterizing assessment uncertainties. Thus, for all evidence scenarios, but particularly for
those in the lower end of this range, it is important to characterize the uncertainties in the evidence base within the evidence integration narrative and to convey the evidence
strength to subsequent steps, including toxicity values developed based on those effects. Existing guidance defines the minimum evidence necessary to judge that a health
hazard could exist as one adverse endpoint from one well-conducted study (U.S. EPA, 1998); this has been expanded in this table to better incorporate mechanistic evidence,
including NAM data.
cScientific understanding of toxicity mechanisms and of the human implications of new toxicity testing methods (e.g., from high throughput screening, from short term in vivo
testing of alternative species, or from new in vitro and in silico testing and other NAMs) will continue to increase. Thus, the sufficiency of mechanistic evidence alone for
identifying potential human health hazards is expected to increase as the science evolves. The decision to identify a potential human hazard based on these data is an expert
judgment dependent on the state-of-the-science at the time of review.
d Scenarios with unexplained heterogeneity across sets of studies with similar confidence and sensitivity can be considered either sufficient evidence for hazard or insufficient
evidence, depending on the expert judgment of the overall weight of evidence. Specifically, this judgment considers the level of support (or lack thereof) provided by
evaluations of the magnitude or severity of the effects, coherence of related findings (including mechanistic evidence), dose-response, and biological plausibility, as well as the
comparability of the supporting and conflicting evidence (e.g., the specific endpoints tested, or the methods used to test them; the specific sources of bias or insensitivity in the
respective sets of studies). The evidence-specific factors supporting either evidence integration judgment will be clearly articulated in the evidence integration narrative.
e When the database includes at least one well-conducted study and a hazard characterization judgment of insufficient evidence is drawn, quantitative analyses may still be
useful for some purposes (e.g., providing a sense of the magnitude and uncertainty of estimates for health effects of potential concern, ranking potential hazards, or setting
research priorities), but not for others (see related discussions in U.S. EPA (2005)). It is critical to transparently convey the extreme uncertainty in any such estimates.
fThe criteria for this category are intentionally more stringent than those justifying a conclusion of sufficient evidence for hazard, consistent with the "difficulty of proving a
negative" (as discussed in U.S. EPA (1991). U.S. EPA (1996a). and U.S. EPA (1998)).
This document is a draft for review purposes only and does not constitute Agency policy.
109 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
10.3. HAZARD CONSIDERATIONS FOR DOSE-RESPONSE
This section provides a transition from hazard identification to the dose-response section,
highlighting (1) information that will inform the selection of outcomes or broader health effect
categories for which toxicity values will be derived, (2) whether toxicity values can be derived to
protect specific populations or lifestages, (3) how dose-response modeling will be informed by
toxicokinetic information, and (4) information aiding the identification of biologically based
benchmark response (BMR) levels. The pool of outcomes and study-specific endpoints will be
discussed to identify which categories of effects and study designs are considered the strongest and
most appropriate for quantitative assessment of a given health effect. Health effects analyzed
relative to exposure levels within or closer to the range of exposures encountered in the
environment are particularly informative. When multiple endpoints are available for an
organ/system, considerations for characterizing the overall impact on this organ/system will be
discussed. For example, if multiple histopathological alterations relevant to changes in liver
function are indicated, liver necrosis might be selected as the most representative endpointto
consider for dose-response analysis. This section can review or clarify which endpoints or
combination of endpoints in each organ/system characterize the overall effect for dose-response
analysis.
Biological considerations important for dose-response analysis (e.g., that could help with
selection of a BMR) will be discussed. The impact of route of exposure on toxicity to different
organs/systems will be examined, if appropriate. The existence and validity of PBPK models or
toxicokinetic information that might allow the estimation of internal dose for route-to-route
extrapolation will be presented. In addition, mechanistic evidence presented in Section 9 that will
influence the dose-response analyses will be highlighted, for example, evidence related to
susceptibility or potential shape of the dose-response curve (i.e., linear, nonlinear, threshold
model). Mode(s) of action will be summarized including any interactions between them relevant to
understanding overall risk.
This section will also draw from Sections 9 and 10 to describe the evidence (i.e., human,
animal, mechanistic) regarding populations and lifestages susceptible to the hazards identified and
factors that increase risk of the hazards. This section will include a discussion of the populations
that, in general, could be susceptible to the health effects identified as hazards of exposure to the
assessed chemical, even if no specific data on effects of exposure to that chemical in the potentially
susceptible population are available. Background information about biological mechanisms or
ADME and biochemical and physiological differences among lifestages can be used to guide the
selection of populations and lifestages to consider. At a minimum, particular consideration will be
given to infants and children, pregnant women, and women of childbearing age. Evidence on
factors that contribute to increased responses to chemical exposure in some population groups or
factors that contribute to increases in exposure or dose will be summarized and evaluated relative
This document is a draft for review purposes only and does not constitute Agency policy.
110 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
1 to patterns across studies pertinent to consistency, coherence, and the magnitude and direction of
2 effect measures. Relevant factors could include intrinsic factors (e.g., age, sex, genetics, health
3 status, behaviors); extrinsic factors (e.g., socioeconomic status, access to health care); and
4 differential exposure levels or frequency (e.g., occupation-related exposure, residential proximity to
5 locations with greater exposure intensity).
6 The section will consider options for using data related to susceptible populations to impact
7 dose-response analysis. In particular, an attempt will be made to highlight when it might be
8 possible to develop separate risk estimates for a specific population or lifestage or to determine
9 whether evidence is available to select a data-derived UF.
This document is a draft for review purposes only and does not constitute Agency policy.
Ill DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
11. DOSE-RESPONSE ASSESSMENT: STUDY
SELECTION AND QUANTITATIVE ANALYSIS
The previous sections of this protocol describe how systematic review principles are
applied to evaluate study quality (potential bias and sensitivity) and reach evidence integration
judgments on health outcomes (or hazard identification) associated with exposure to the chemical
of interest Selection of specific data sets for dose-response assessment and performance of the
dose-response assessment is conducted after hazard identification is complete and involves
database- and chemical-specific biological judgments. Several EPA guidance and support
documents detail data requirements and other considerations for dose-response modeling,
especially EPA's Benchmark Dose Technical Guidance fU.S. EPA. 2012bl and EPA's Review of the
Reference Dose and Reference Concentration Processes (U.S. EPA. 2005. 2002). This section of the
protocol provides an overview of considerations for conducting the dose-response assessment,
particularly statistical considerations specific to dose-response analysis that support quantitative
risk assessment Importantly, these considerations do not supersede existing EPA guidance.
For IRIS assessments, dose-response assessments are typically performed for both
noncancer and cancer hazards, and for both oral and inhalation routes of exposure following
chronic exposure17 to the chemical of interest, if supported by existing data. For noncancer
hazards, an RfD and an RfC are usually derived. An RfD or an RfC is an estimate, with uncertainty
spanning perhaps an order of magnitude, of an exposure to the human population (including
susceptible populations or lifestages) likely to be without an appreciable risk of deleterious health
effects over a lifetime (U.S. EPA. 2002). Reference values are not predictive risk values; that is, they
provide no information about risks at higher or lower exposure levels.
As discussed in Section 2 ("Scoping and Initial Problem Formulation Summary") of this
assessment, IRIS will conduct the assessment with a goal of developing oral and inhalation
reference values for noncancer toxicity from exposure to complex PCB mixtures. The derivation of
noncancer reference values might also depend on the nature of the hazard conclusions. Specifically,
for noncancer outcomes, dose-response assessment generally will be conducted when the evidence
integration judgments indicate there is "sufficient evidence for hazard", with preference given to
health effects with stronger evidence scenarios within that category (Section 10.2), and quantitative
analyses generally will not be attempted for "insufficient evidence."
17 Dose-response assessments can also be conducted for shorter durations, particularly if the evidence base
for a chemical indicates risks associated with shorter exposures to the chemical (U.S. EPA. 20021
This document is a draft for review purposes only and does not constitute Agency policy.
112 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
11.1. SELECTING STUDIES FOR DOSE-RESPONSE ASSESSMENT
The dose-response assessment begins with a review of the important health effects
highlighted in the hazard identification step (see Section 10), particularly among the studies of
highest quality and studies that exemplify the attributes summarized in Table 24. This review also
considers whether opportunities for quantitative evidence integration exist Examples of
quantitative integration, from simplest to more complex, include (1) combining results for an
outcome across sex (within a study); (2) characterizing overall toxicity, as in combining effects that
comprise a syndrome, or occur on a continuum (e.g., precursors and overt toxicity, benign tumors
that progress to malignant tumors); and (3) conducting a meta-analysis or metaregression of all
studies addressing a category of important health effects.
Some studies used qualitatively for hazard identification might or might not be useful
quantitatively for dose-response assessment due to such factors as the lack of quantitative
measures of exposure or lack of variability measures for response data. If the needed information
cannot be located (see Section 7), semiquantitative analysis might be feasible (e.g., using
NOAEL/LOAEL). Studies of low sensitivity might be less useful if they fail to detect a true effect or
yield PODs with wide confidence limits, but such studies would be considered for inclusion in a
meta-analysis.
Among the studies that support the hazard conclusions, those most useful for
dose-response analysis generally have at least one exposure level in the region of the
dose-response curve near the BMR (the response level to be used for deriving toxicity values), to
minimize low-dose extrapolation, and more exposure levels and larger sample sizes overall fU.S.
EPA. 2012b). These attributes support a more complete characterization of the shape of the
exposure-response curve and decrease the uncertainty in the associated exposure-response metric
(e.g., RfC) by reducing statistical uncertainty in the POD and minimizing the need for low-dose
extrapolation. In addition to these more general considerations, specific issues that could impact
the feasibility of dose-response modeling for individual data sets are described in more detail in the
Benchmark Dose Technical Guidance fU.S. EPA. 2012bl.
This document is a draft for review purposes only and does not constitute Agency policy.
113 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Table 24. Attributes used to evaluate studies for derivation of toxicity values
Study attributes
Considerations
Human studies
Animal studies
Study confidence
High or medium confidence studies are highly preferred over low confidence studies. The available high and medium confidence studies are
further differentiated based on the study attributes below and on a reconsideration of the specific limitations identified and their potential
impact on dose-response analyses.
Rationale for choice of
species
Human data are preferred over animal data to eliminate interspecies
extrapolation uncertainties (e.g., in toxicodynamics, relevance of
specific health outcomes to humans).
Animal studies provide supporting evidence when adequate human
studies are available and are considered principal studies when
adequate human studies are not available. For some hazards, studies of
particular animal species known to respond similarly to humans would
be preferred over studies of other species.
Relevance of
exposure
paradigm
Exposure
route
Studies involving human environmental exposures (oral, inhalation).
Studies by a route of administration relevant to human environmental
exposure are preferred. A validated pharmacokinetic model can also be
used to extrapolate across exposure routes.
Exposure
durations
When developing a chronic toxicity value, chronic or subchronic studies are preferred over studies of acute exposure durations. Exceptions exist,
such as when a susceptible population or lifestage is more sensitive in a particular time window (e.g., developmental exposure).
Exposure
levels
Exposures near the range of typical environmental human exposures are preferred. Studies with a broad exposure range and multiple exposure
levels are preferred to the extent that they can provide information about the shape of the exposure-response relationship [see the EPA
Benchmark Dose Technical Guidance: (U.S. EPA, 2012b)l and facilitate extrapolation to more relevant (generally lower) exposures.
Subject selection
Studies that provide risk estimates in the most susceptible groups are preferred.
Controls for possible
confounding3
Studies with a design (e.g., matching procedures, blocking) or analysis (e.g., covariates or other procedures for statistical adjustment) that
adequately address the relevant sources of potential critical confounding for a given outcome are preferred.
Measurement of exposure
Studies that can reliably distinguish between levels of exposure in a
time window considered most relevant for development of a causal
effect are preferred. Exposure assessment methods that provide
measurements at the level of the individual and that reduce
measurement error are preferred. Measurements of exposure should
not be influenced by knowledge of health outcome status.
Studies providing actual measurements of exposure (e.g., analytical
inhalation concentrations vs. target concentrations) are preferred.
Relevant internal dose measures could facilitate extrapolation to
humans, as would availability of a suitable animal PBPK model in
conjunction with an animal study reported in terms of administered
exposure.
Measurement of health
outcome(s)
Studies that can reliably distinguish the presence or absence (or degree of severity) of the outcome are preferred. Outcome ascertainment
methods using generally accepted or standardized approaches are preferred.
Studies with individual data are preferred in general. Examples include: to characterize experimental variability more realistically, to characterize
overall incidence of individuals affected by related outcomes.
This document is a draft for review purposes only and does not constitute Agency policy.
114 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Study attributes
Considerations
Human studies
Animal studies
Among several relevant health outcomes, preference is generally given to those with greater biological significance.
Study size and design
Preference is given to studies using designs reasonably expected to have power to detect responses of suitable magnitude.15 This does not mean
that studies with substantial responses but low power would be ignored, but that they will be interpreted in light of a confidence interval or
variance for the response. Studies that address changes in the number at risk (through decreased survival, loss to follow-up) are preferred.
a An exposure or other variable associated with both exposure and outcome but not an intermediary between the two.
b Power is an attribute of the design and population parameters, based on a concept of repeatedly sampling a population; it cannot be inferred post hoc using data from one
experiment (Hoenig and Heisev, 2001).
This document is a draft for review purposes only and does not constitute Agency policy.
11S DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
11.2. CONDUCTING DOSE-RESPONSE ASSESSMENTS
EPA uses an approach for dose-response assessment that distinguishes analysis of the
dose-response data in the range of observation from any inferences about responses at lower
environmentally relevant exposure levels (U.S. EPA. 2012b. 20051:
1) Within the observed dose range, the preferred approach is to use dose-response modeling
to incorporate as much of the data set as possible into the analysis. This modeling yields a
POD, an exposure level ideally near the lower end of the range of observation, without
significant extrapolation to lower exposure levels. See Section 11.2.1 for more details.
2) Derivation of reference values nearly always involves extrapolation to exposures lower than
the POD and is described in more detail in Section 11.2.3.
When sufficient and appropriate human data and laboratory animal data are both available
for the same outcome, human data are generally preferred for the dose-response assessment
because their use eliminates the need to perform interspecies extrapolations.
For reference values, IRIS assessments typically derive a candidate value from each suitable
data set, whether for human or animal (see Section 11.1). Evaluating these candidate values
grouped within a particular organ/system yields a single organ-/system-specific value for each
organ/system under consideration. Next, evaluation of these organ-/system-specific values results
in the selection of a single overall reference value to cover all health outcomes across all
organs/systems. Although this overall reference value is the focus of the assessment, the
organ-/system-specific values can be useful for subsequent cumulative risk assessments that
consider the combined effect of multiple agents acting on a common organ/system.
11.2.1. Dose-response Analysis in the Range of Observation
For conducting a dose-response assessment, toxicodynamic ("biologically based") modeling
can be used when data are sufficient to ascertain the MOA and quantitatively support model
parameters that represent rates and other quantities associated with the key precursor events of
the MOA. Toxicodynamic modeling is potentially the most comprehensive way to account for the
biological processes involved in a response. Such models seek to reflect the sequence of key
precursor events that lead to a response. Toxicodynamic models can contribute to dose-response
assessment by revealing and describing nonlinear relationships between internal dose and
response. Such models can provide a useful approach for analysis in the range of observation,
provided the purpose of the assessment justifies the effort involved.
When a toxicodynamic model is not available for dose-response assessment or when the
purpose of the assessment does not warrant developing such a model, empirical modeling will be
used to fit the data (on the apical outcome or a key precursor event) in the range of observation.
For this purpose, EPA has developed a standard set of models f http: //www.epa.gov/ncea/bmds]
that can be applied to typical data sets, including those that are linear and nonlinear. When
This document is a draft for review purposes only and does not constitute Agency policy.
116 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
alternative models with significant biological support are available, the decision maker can be
informed by the presentation of these alternatives along with the models' strengths and
uncertainties. EPA has developed guidance on modeling dose-response data, assessing model fit,
selecting suitable models, and reporting modeling results [see the EPA Benchmark Dose Technical
Guidance (U.S. EPA. 2012b)]. Additional judgment or alternative analyses are used if the procedure
fails to yield reliable results, for example, if the fit is poor, modeling might be restricted to the lower
doses, especially when competing toxicity occurs at higher doses.
For each modeled response, a POD from the observed data will be estimated to mark the
beginning of extrapolation to lower doses. The POD is an estimated dose (expressed in
human-equivalent terms) near the lower end of the observed range without significant
extrapolation to lower doses. The POD is used as the starting point for subsequent extrapolations
and analyses. For nonlinear extrapolation, the POD is used in calculating an RfD or RfC.
Due to the biopersistent nature of many PCB congeners, the relationship between the
exposure rate (mg/kg-day administered, absorbed, or inhaled) and the concentration of the
congener in the body is a complex function of the exposure rate, exposure duration, and lifestage(s)
over which the exposure occurs. Thus, interpretation and comparison of studies with different
exposure designs is facilitated by using a pharmacokinetic model that tracks the accumulation of
each congener over time and accounts for transfer to offspring during gestation and via lactation
for studies that include developmental exposures. Except when comparing studies with otherwise
identical exposure designs, comparisons and analyses described in the assessment, including POD
estimations, will be based on measures of internal dose averaged over the exposure duration or
over the critical window of exposure for the health effect of interest (if known).
The response level at which the POD is calculated is guided by the severity of the endpoint
If linear extrapolation is used, selection of a response level corresponding to the POD is not highly
influential, so standard values near the low end of the observable range are generally used (for
example, 10% extra risk for experimental animal histopathology data, 1% for epidemiological
data). Nonlinear approaches account for both statistical and biological considerations. For
dichotomous data, a response level of 10% extra risk is generally used for minimally adverse
effects, 5% or lower for more severe effects. For continuous data, a response level is ideally based
on an established definition of biological significance. In the absence of such definition, one control
standard deviation from the control mean is often used for minimally adverse effects, and one-half
standard deviation for more severe effects. The POD is the 95% lower bound on the dose
associated with the selected response level.
EPA has developed standard approaches for determining the relevant dose to be used in the
dose-response modeling in the absence of appropriate pharmacokinetic modeling. These standard
approaches also facilitate comparison across exposure patterns and species:
Intermittent study exposures are standardized to a daily average over the duration of
exposure. For chronic effects, daily exposures are averaged over the lifespan. Exposures
This document is a draft for review purposes only and does not constitute Agency policy.
117 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
during a critical period, however, are not averaged over a longer duration (U.S. EPA. 2005.
19911.
Doses are standardized to equivalent human terms to facilitate comparison of results from
different species. Oral doses are scaled allometrically using mg/kg3/4-day as the equivalent
dose metric across species. Allometric scaling pertains to equivalence across species, not
across lifestages, and is not used to scale doses from adult humans or mature animals to
infants or children (U.S. EPA. 2011a. 2005). Inhalation exposures are scaled using
dosimetry models that apply species-specific physiological and anatomical factors and
consider whether the effect occurs at the site of first contact or after systemic circulation
fU.S. EPA. 2012a. 1994bl.
Converting doses across exposure routes can be informative. If this is done, the assessment
describes the underlying data, algorithms, and assumptions (U.S. EPA. 20051.
In the absence of study-specific data on, for example, intake rates or body weight, EPA has
developed recommended values for use in dose-response analysis (U.S. EPA. 19881.
These standard approaches will be augmented through the use of pharmacokinetic
modeling because of the indirect relationship between exposure and internal concentration or
dose, as briefly described above. In particular:
Standardization of intermittent exposures will be conducted by determining the internal
concentration averaged over the appropriate time period, or area-under-the-concentration
curve for effects that are presumed to result from cumulative damage.
The human equivalent internal dose (concentration) for a given response level is assumed
identical to the animal internal dose for exposure over the biologically equivalent period
(e.g., human vs. animal gestation). Because data exist to determine the half-lives of
congeners in humans independent of animal species, those congener-specific values will be
used rather than those derived from allometric scaling. Equivalent human inhalation
exposures likewise will be estimated using a human version of a pharmacokinetic model
that includes inhalation uptake and exhalation.
Route-to-route extrapolation likewise will be conducted using a pharmacokinetic model
capable of describing oral and inhalation (and possibly dermal) exposure.
The pharmacokinetic modeling might use either study-specific intake or body weight or
recommended standard values.
11.2.2. Extrapolation: Slope Factors and Unit Risks
A cancer assessment is not included in the scope of the current assessment for PCBs.
Accordingly, this assessment will not derive an oral slope factor or inhalation unit risk.
11.2.3. Extrapolation: Reference Values
Reference value derivation is EPA's most frequently used type of nonlinear extrapolation
method. It is most commonly used for noncancer effects.
This document is a draft for review purposes only and does not constitute Agency policy.
118 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
For each data set selected for reference value derivation, reference values are estimated by
applying relevant adjustments to the PODs to account for the conditions of the reference value
definitionfor human variation, extrapolation from animals to humans, extrapolation to chronic
exposure duration, and extrapolation to a minimal level of risk (if not observed in the data set).
Increasingly, data-based adjustments (U.S. EPA. 20141 and Bayesian methods for characterizing
population variability (NRC. 20141 might be feasible and might be distinguished from the UF
considerations outlined below. The assessment will discuss the scientific bases for applying these
data-based adjustments and UFs:
Animal-to-human extrapolation: If animal results are used to make inferences about
humans, the reference value derivation incorporates the potential for cross-species
differences, which could arise from differences in toxicokinetics or toxicodynamics. If
available, a biologically based model that adjusts fully for toxicokinetic and toxicodynamic
differences across species could be used. Otherwise, the POD is standardized to equivalent
human terms or is based on pharmacokinetic or dosimetry modeling, that might range from
detailed chemical-specific to default approaches (U.S. EPA. 2014. 2011a), and a factor of
101/2 (rounded to 3) is applied to account for the remaining uncertainty involving
toxicokinetic and toxicodynamic differences.
Human variation: The assessment accounts for variation in susceptibility across the human
population and the possibility that the available data might not represent individuals who
are most susceptible to the effect, by using a data-based adjustment or UF or a combination
of the two. When appropriate data or models for the effect or for characterizing the internal
dose are available, the potential for data-based adjustments for toxicodynamics or
toxicokinetics is considered (U.S. EPA. 2014. 20021.18.19 When sufficient data are available,
an intraspecies UF either less than or greater than 10-fold might be justified (U.S. EPA.
20021. This factor can be reduced if the POD is derived from or adjusted specifically for
susceptible individuals [not for a general population that includes both susceptible and
nonsusceptible individuals; (U.S. EPA. 2002.1998.1996a. 1994b. 19911], When the use of
such data or modeling is not supported, a UF with a default value of 10 is considered.
LOAEL to NOAEL: If a POD is based on a LOAEL, the assessment includes an adjustment to an
exposure level where such effects are not expected. This can be a matter of great
uncertainty if no evidence is available at lower exposures. A factor of 3 or 10 generally is
applied to extrapolate to a lower exposure expected to be without appreciable effects. A
factor other than 10 can be used depending on the magnitude and nature of the response
and the shape of the dose-response curve fU.S. EPA. 2002.1998.1996a. 1994b. 19911.
18 Examples of adjusting the toxicokinetic portion of interhuman variability include the IRIS boron
assessment's use of nonchemical-specific kinetic data [e.g., glomerular filtration rate in pregnant humans as a
surrogate for boron clearance (U.S. EPA. 20041] and the IRIS trichloroethylene assessment's use of population
variability in trichloroethylene metabolism, via a PBPK model, to estimate the lower 1st percentile of the dose
metric distribution for each POD (U.S. EPA. 2011b).
19 Note that when a PBPK model is available for relating human internal dose to environmental exposure,
relevant portions of this UF might be more usefully applied prior to animal-to-human extrapolation,
depending on the correspondence of any nonlinearities (e.g., saturation levels) between species.
This document is a draft for review purposes only and does not constitute Agency policy.
119 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Subchronic-to-chronic exposure: When using subchronic studies to make inferences about
chronic/lifetime exposure, the assessment considers whether lifetime exposure could have
effects at lower levels of exposure. A factor up to 10 can be applied to the POD, depending
on the duration of the studies and the nature of the response fU.S. EPA. 2002.1998.1994b).
Database deficiencies-. In addition to the adjustments above, if database deficiencies raise
concern that further studies might identify a more sensitive effect, organ system, or
lifestage, the assessment can apply a database UF fU.S. EPA. 2002.1998.1996a. 1994b.
19911. The size of the factor depends on the nature of the database deficiency. For
example, EPA typically follows the recommendation that a factor of 10 be applied if both a
prenatal toxicity study and a two-generation reproduction study are missing and a factor of
101/2 (i.e., 3) if either is missing fU.S. EPA. 20021.
The POD for a reference value is divided by the product of these factors. U.S. EPA f20021
recommends that any composite factor that exceeds 3,000 represents excessive uncertainty, and
recommends against relying on the associated reference value. The derivation of oral and
inhalation reference values for PCBs conducted as part of the current assessment will be performed
consistent with EPA guidance summarized above.
This document is a draft for review purposes only and does not constitute Agency policy.
120 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
12. PROTOCOL HISTORY
1 [This section is a placeholder for tracking information on the original protocol release and any
2 potential protocol updates.]
This document is a draft for review purposes only and does not constitute Agency policy.
121 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
REFERENCES
Alford-Stevens. AL. (1986). Analyzing PCBs. Environ Sci Technol 20: 1194-1199.
http://dx.doi.Org/10.1021/es00154a001
Alford-Stevens. AL: Budde. WL: Bellar. TA. (1985). Interlaboratory study on determination of
polychlorinated biphenyls in environmentally contaminated sediments. Anal Chem 57:
2452-2457. http://dx.doi.Org/10.1021/ac00290a007
Arnold. PL: Brvce. F: Karpinski. K: Mes. 1: Fernie. S: Trvphonas. H: Truelove. 1: Mcguire. PF: Burns. D:
Tanner. 1. R.: Staplev. R: Zawidzka. ZZ: Basford. D. (1993a). Toxicological consequences of
Aroclor 1254 ingestion by female rhesus (Macaca mulatta) monkeys. Part IB. Prebreeding
phase: Clinical and analytical laboratory findings. Food Chem Toxicol 31: 811-824.
http://dx.doi.org/10.1016/0278-6915r93190219-0
Arnold. PL: Brvce. F: Staplev. R: Mcguire. PF: Burns. D: Tanner. 1. R: Karpinski. K. (1993b).
Toxicological consequences of Aroclor 1254 ingestion by female rhesus (Macaca mulatta)
monkeys. Part 1A. Prebreeding phase: clinical health findings. Food Chem Toxicol 31: 799-
810. http://dx.doi.org/10.1016/0278-6915r93190218-N
ATSDR. (2000). Toxicological profile for polychlorinated biphenyls (PCBs) [ATSDRTox Profile],
Atlanta, GA: U.S. Department of Health and Human Services, Public Health Service.
http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=142&tid=26
ATSDR. (2011). Addendum to the toxicological profile for polychlorinated biphenyls. Atlanta, GA.
Barsotti. DA: van Miller. TP. (1984). Accumulation of a commercial polychlorinated biphenyl mixture
(Aroclor 1016) in adult rhesus monkeys and their nursing infants. Toxicology 30: 31-44.
Catlin. NR: Collins. BT: Auerbach. SS: Ferguson. SS: Harnlv. TM: Gennings. C: Waidvanatha. S: Rice. GE:
Smith-Roe. SL: Witt. KL: Rider. CV. (2018). How similar is similar enough? A sufficient
similarity case study with Ginkgo biloba extract. Food Chem Toxicol 118: 328-339.
http://dx.doi.Org/10.1016/i.fct.2018.05.013
CDC. (2009). Fourth national report on human exposure to environmental chemicals. Atlanta, GA:
U.S. Department of Health and Human Services, Centers for Disease Control and Prevention.
http: / /www, cdc.gov/exposurereport/
Crissman. TW: Goodman. DG: Hildebrandt. PK: Maronpot. RR: Prater. DA: Riley. TH: Seaman. WT:
Thake. DC. (2004). Best practices guideline: Toxicologic histopathology. Toxicol Pathol 32:
126-131. http://dx.doi.org/10.1080/01926230490268756
Devoto. E: Fiore. BT: Millikan. R: Anderson. HA: Sheldon. L: Sonzogni. WC: Longnecker. MP. (1997).
Correlations among human blood levels of specific PCB congeners and implications for
epidemiologic studies. Am J Ind Med 32: 606-613.
Dickersin. K. (1990). The existence of publication bias and risk factors for its occurrence. JAMA 263:
1385-1389.
EFSA. (2017). Guidance on the use of the weight of evidence approach in scientific assessments.
EFSAJ15: 1-69. http://dx.doi.Org/10.2903/i.efsa.2017.4971
Erickson. MP. (2018). Aroclor misidentification in environmental samples: how do we
communicate more effectively between the laboratory and the data user? Environ Sci Pollut
Res Int 25: 16291-16299. http://dx.doi.org/10.1007/sll356-017-0160-6
Feder. PI: Ma. ZT: Bull. RT: Teuschler. LK: Schenck. KM: Simmons. IE: Rice. G. (2009). Evaluating
sufficient similarity for disinfection by-product (DBP) mixtures: multivariate statistical
procedures [Review], J Toxicol Environ Health A 72: 468-481.
http://dx.doi.Org/10.1080/15287390802608965
This document is a draft for review purposes only and does not constitute Agency policy.
122 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Foster. P: Gray. LE. Tr. (2013). Toxic responses of the reproductive system. In CD Klaassen (Ed.),
(8th ed., pp. 861-906). New York, NY: McGraw-Hill Education.
Guyatt, G; Oxman. AD; Akl, EA; Kunz, R; Vist, G; Brozek, 1; Norris, S; Falck-Ytter, Y; Glasziou, P;
DeBeer. H: Taeschke. R: Rind. D: Meerpohl. 1: Dahm. P: Schiinemann. HI. (2011). GRADE
guidelines: 1. Introduction-GRADE evidence profiles and summary of findings tables. J Clin
Epidemiol 64: 383-394. http://dx.doi.Org/10.1016/i.iclinepi.2010.04.026
Harrad. S: Ibarra. C: Robson. M: Melvmuk. L: Zhang. X: Diamond. M: Douwes. I. (2009).
Polychlorinated biphenyls in domestic dust from Canada, New Zealand, United Kingdom
and United States: Implications for human exposure. Chemosphere 76: 232-238.
http://dx.doi.Org/10.1016/i.chemosphere.2009.03.020
Hill. AB. (1965). The environment and disease: Association or causation? Proc R Soc Med 58: 295-
300.
Hirst. TA: Howick. I: Aronson. IK: Roberts. N: Perera. R: Koshiaris. C: Heneghan. C. (2014). The need
for randomization in animal trials: an overview of systematic reviews [Review], PLoS ONE
9: e98856. http://dx.doi.org/10.1371/iournal.pone.0098856
Hoenig. TM: Heisev. DM. (2001). The abuse of power: The pervasive fallacy of power calculations for
data analysis. Am Stat 55: 19-24.
I ARC. (2015). Polychlorinated and polybrominated biphenyls: Volume 107. In IARC Monographs on
the Evaluation of Carcinogenic Risks to Humans. Geneva, Switzerland: World Health
Organization. http://monographs.iarc.fr/ENG/Monographs/voll07/index.php
ICF. (2018). DRAGON: An online tool for systematic review [Website],
https://www.icf.com/resources/solutions-and-apps/dragon-online-tool-systematic-review
Kodavanti. PRS: Kannan. N: Yamashita. N: Derr-Yellin. EC: Ward. TR: Burgin. DE: Tilson. HA:
Birnbaum. LS. (2001). Differential effects of two lots of Aroclor 1254: Congener-specific
analysis and neurochemical end points. Environ Health Perspect 109: 1153-1161.
Krauth. D: Woodruff. TT: Bero. L. (2013). Instruments for assessing risk of bias and other
methodological criteria of published animal studies: a systematic review [Review], Environ
Health Perspect 121: 985-992. http://dx.doi.org/10.1289/ehp.1206389
Lehmann. GM: Christensen. K: Maddaloni. M: Phillips. LI. (2015). Evaluating health risks from
inhaled polychlorinated biphenyls: research needs for addressing uncertainty. Environ
Health Perspect 123: 109-113. http://dx.doi.org/10.1289/ehp.1408564
Levin. ED: Schantz. SL: Bowman. RE. (1988). Delayed spatial alternation deficits resulting from
perinatal PCB exposure in monkeys. Arch Toxicol 62: 267-273.
http://dx.doi.Org/10.1007/BF00332486
Marek. RF: Thorne. PS: Herkert. NT: Awad. AM: Hornbuckle. KC. (2017). Airborne PCBs and OH-PCBs
inside and outside urban and rural U.S. schools. Environ Sci Technol 51: 7853-7860.
http://dx.doi.org/10.1021/acs.est.7b01910
Marshall. S: Gennings. C: Teuschler. LK: Stork. LG: Tornero-Velez. R: Crofton. KM: Rice. GE. (2013).
An empirical approach to sufficient similarity: combining exposure data and mixtures
toxicology data. Risk Anal 33: 1582-1595. http: / /dx.doi.or g/10.1111 /risa. 12015
Morgan. RL: Thayer. KA: Bero. L: Bruce. N: Falck-Ytter. Y: Ghersi. D: Guyatt. G: Hooiimans. C:
Langendam. M: Mandrioli. D: Mustafa. RA: Rehfuess. EA: Roonev. AA: Shea. B: Silbergeld. EK:
Sutton. P: Wolfe. MS: Woodruff. TT: Verbeek. TH: Hollowav. AC: Santesso. N: Schiinemann. HI.
(2016). GRADE: Assessing the quality of evidence in environmental and occupational health.
Environ Int 92-93: 611-616. http://dx.doi.Org/10.1016/i.envint.2016.01.004
NIOSH. (2019a). NI0SH pocket guide to chemical hazards. RTECS. Chlorodiphenyl (42% chlorine).
http: / /www. cdc. gov /niosh /np g/np gdO 12 5 .html
NIOSH. (2019b). NIOSH pocket guide to chemical hazards. RTECS. Chlorodiphenyl (54% chlorine).
http: / /www, cdc. gov/niosh /np g/np gdO 12 6.html
This document is a draft for review purposes only and does not constitute Agency policy.
123 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
NRC. (2014). Review of EPA's Integrated Risk Information System (IRIS) process. Washington, DC:
The National Academies Press, http: //www.nap.edu/catalog.php7record id=18764
NTP. (2011). Specifications for the conduct of studies to evaluate the reproductive and
developmental toxicity of chemical, biological and physical agents in laboratory animals.
https://ntp.niehs.nih.gov/ntp/test info/finalntp reprospecsmav2011 508.pdf
NTP. (2015). Handbook for conducting a literature-based health assessment using 0HAT approach
for systematic review and evidence integration. U.S. Dept of Health and Human Services,
National Toxicology Program.
https://ntp.niehs.nih.gov/ntp/ohat/pubs/handbookian2015 508.pdf
O'Brien. KM: Upson. K: Cook. NR: Weinberg. CR. (2016). Environmental chemicals in urine and
blood: Improving methods for creatinine and lipid adjustment Environ Health Perspect
124: 220-227. http://dx.doi.org/10.1289/ehp.1509693
OSHA. (2018). Chemical sampling information, chlorodiphenyl (54% CI).
https://www.osha.gov/chemicaldata/chemResult.html?recNo=121
OSHA. (2019). Chemical sampling information, chlorodiphenyl (42% CI).
https://www.osha.gov/chemicaldata/chemResult.html?recNo=392
Rice. DC. (1997). Effect of postnatal exposure to a PCB mixture in monkeys on multiple fixed
interval-fixed ratio performance. Neurotoxicol Teratol 19: 429-434.
Rice. DC. (1998). Effects of postnatal exposure of monkeys to a PCB mixture on spatial
discrimination reversal and DRL performance. Neurotoxicol Teratol 20: 391-400.
Rice. DC. (1999). Effect of exposure to 3,3',4,4',5-pentachlorobiphenyl (PCB 126) throughout
gestation and lactation on development and spatial delayed alternation performance in rats.
Neurotoxicol Teratol 21: 59-69. http://dx.doi.org/10.1016/S0892-0362r98100031-2
Rice. DC: Havward. S. (1997). Effects of postnatal exposure to a PCB mixture in monkeys on
nonspatial discrimination reversal and delayed alternation performance. Neurotoxicology
18: 479-494.
Rice. DC: Havward. S. (1999). Effects of exposure to 3,3',4,4',5-pentachlorobiphenyl (PCB 126)
throughout gestation and lactation on behavior (concurrent random interval-random
interval and progressive ratio performance) in rats. Neurotoxicol Teratol 21: 679-687.
http://dx.doi.org/10.1016/S0892-0362r99100021-5
Rice. GE: Ingvar. E: Feder. EI: Gennings. C. (2018). Assessing human health risks using information
on whole mixtures. In C Rider; J Simmons (Eds.), (pp. 421-463). Cham: Springer.
http://dx.doi. org/10.1007/978-3-319-56234-6 15
Schantz. SL: Levin. ED: Bowman. RE. (1991). Long-term neurobehavioral effects of perinatal
polychlorinated biphenyl (PCB) exposure in monkeys. Environ Toxicol Chem 10: 747-756.
http://dx.doi.Org/10.1002/etc.5620100606
Schantz. SL: Levin. ED: Bowman. RE: Heironimus. MP: Laughlin. NK. (1989). Effects of perinatal PCB
exposure on discrimination-reversal learning in monkeys. Neurotoxicol Teratol 11: 243-
250. http://dx.doi.Org/10.1016/0892-0362r89190066-4
Schisterman. EF: Whitcomb. BW: Louis. GM: Louis. TA. (2005). Lipid adjustment in the analysis of
environmental contaminants and human health risks. Environ Health Perspect 113: 853-
857. http://dx.doi.org/10.1289/ehp.7640
Schiinemann. H: Hill. S: Guvatt. G: Akl. EA: Ahmed. F. (2011). The GRADE approach and Bradford
Hill's criteria for causation. J Epidemiol Community Health 65: 392-395.
http://dx.doi.org/10.1136/iech.2010.119933
Smith. MT: Guvton. KZ: Gibbons. CF: Fritz. TM: Portier. CI: Rusvn. I: DeMarini. DM: Caldwell. TC:
Kavlock. RT: Lambert. PF: Hecht. SS: Bucher. TR: Stewart. BW: Baan. RA: Cogliano. VI: Straif.
K. (2016). Key characteristics of carcinogens as a basis for organizing data on mechanisms
of carcinogenesis [Review], Environ Health Perspect 124: 713-721.
http://dx.doi.org/10.1289/ehp.1509912
This document is a draft for review purposes only and does not constitute Agency policy.
124 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Sterne. TAC: Hernan. MA: Reeves. BC: Savovic. T: Berkman. ND: Viswanathan. M: Henry. D: Altman.
DG: Ansari. MT: Boutron. I: Carpenter. TR: Chan. AW: Churchill. R: Peeks. TT: Hrobiartsson. A:
Kirkham. 1: Tiini. P: Loke. YK: Pigott. TP: Ramsay. CR: Regidor. D: Rothstein. HR: Sandhu. L:
Santaguida. PL: Schiinemann. HI: Shea. B: Shrier. I: Tugwell. P: Turner. L: Valentine. TC:
Waddington. H: Waters. E: Wells. GA: Whiting. PF: Higgins. 1PT. (2016). ROBINS-I: A tool for
assessing risk of bias in non-randomised studies of interventions. Br Med J 355: i4919.
Thomas. K: Xue. 1: Williams. R: Tones. P: Whitaker. D. (2012). Polychlorinated biphenyls (PCBs) in
school buildings: Sources, environmental levels, and exposure [EPA Report], (EPA/600/R-
12/051). U.S. Environmental Protection Agency, https://www.epa.gov/sites/production/
files/2015-08/documents/pcb epa600rl2051 final.pdf
ToxNet Hazardous Substances Data. B. (2011). HSDB: Polychlorinated biphenyls [Website],
https://to xnet.nlm.nih.gov/cgi-bin/sis/search/a?dbs+hsdb:@term+@DQCNO+3945
Trvphonas. H: Havward. S: O'Gradv. L: Loo. TC: Arnold. PL: Brvce. F: Zawidzka. ZZ. (1989).
Immunotoxicity studies of PCB (Aroclor 1254) in the adultrhesus (Macaca mulatta)
monkey-preliminary report. International Journal of Immunopharmacology 11: 199-206.
http://dx.doi.org/10.1016/0192-056ir89190072-6
Trvphonas. H: Luster. MI: Schiffman. G: Dawson. LL: Hodgen. M: Germolec. D: Havward. S: Brvce. F:
Loo. TC: Mandv. F. (1991a). Effect of chronic exposure of PCB (Aroclor 1254) on specific and
nonspecific immune parameters in the rhesus (Macaca mulatta) monkey. Fundam Appl
Toxicol 16: 773-786. http://dx.doi.Org/10.1093/toxsci/16.4.773
Trvphonas. H: Luster. MI: White. KL. Tr.: Navlor. PH: Erdos. MR: Burleson. GR: Germolec. D: Hodgen.
M: Havward. S: Arnold. PL. (1991b). Effects of PCB (Aroclor 1254) on non-specific immune
parameters in rhesus (Macaca mulatta) monkeys. International Journal of
Immunopharmacology 13: 639-648. http://dx.doi.org/10.1016/0192-056ir91190176-8
U.S. EPA. (1988). Recommendations for and documentation of biological values for use in risk
assessment [EPA Report] (pp. 1-395). (EPA/600/6-87/008). Cincinnati, OH: U.S.
Environmental Protection Agency, Office of Research and Development, Office of Health and
Environmental Assessment http: //cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=34855
U.S. EPA. (1991). Guidelines for developmental toxicity risk assessment (pp. 1-71). (EPA/600/FR-
91/001). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment Forum.
http: / /cfpub.epa.gov/ncea/cfm /recordisplav.cfm?deid=2 3162
U.S. EPA. (1993). IRIS chemical assessment summary for aroclor 1016 (CASRN 12674-11-2).
https://cfpub.epa.gov/ncea/iris2/chemicalLanding.cfm7substance nmbr=462
U.S. EPA. (1994a). IRIS chemical assessment summary for aroclor 1254 (CASRN 11097-69-1).
https://cfpub.epa.gov/ncea/iris2/chemicalLanding.cfm7substance nmbr=389
U.S. EPA. (1994b). Methods for derivation of inhalation reference concentrations and application of
inhalation dosimetry [EPA Report], (EPA/600/8-90/066F). Research Triangle Park, NC: U.S.
Environmental Protection Agency, Office of Research and Development, Office of Health and
Environmental Assessment, Environmental Criteria and Assessment Office.
https://cfpub.epa.gov/ncea/risk/recordisplay.cfm?deid=71993&CFID=51174829&CFT0KE
N=25006317
U.S. EPA. (1996a). Guidelines for reproductive toxicity risk assessment (pp. 1-143). (EPA/630/R-
96/009). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment Forum.
https://www.epa.gov/sites/production/files/2014-
11/documents/guidelines repro toxicitv.pdf
U.S. EPA. (1996b). PCBs: Cancer dose-response assessment and application to environmental
mixtures [EPA Report], (EPA/600/P-96/001). Washington, DC: U.S. Environmental
Protection Agency, Office of Research and Development, National Center for Environmental
Assessment. http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=12486
This document is a draft for review purposes only and does not constitute Agency policy.
125 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
U.S. EPA. (1998). Guidelines for neurotoxicity risk assessment [EPA Report] (pp. 1-89).
(EPA/630/R-95/001F). Washington, DC: U.S. Environmental Protection Agency, Risk
Assessment Forum, http: //www.epa.gov/risk/guidelines-neurotoxicity-risk-assessment
U.S. EPA. (2000). Supplementary guidance for conducting health risk assessment of chemical
mixtures (pp. 1-209). (EPA630R00002). Washington, DC: U.S. Environmental Protection
Agency, Risk Assessment Forum.
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=20533
U.S. EPA. (2002). A review of the reference dose and reference concentration processes.
(EPA/630/P-02/002F). Washington, DC: U.S. Environmental Protection Agency, Risk
Assessment Forum, https: //www.epa.gov/sites/production/files/2014-12 /documents/rfd-
final.pdf
U.S. EPA. (2004). Toxicological review of boron and compounds. In support of summary
information on the Integrated Risk Information System (IRIS) [EPA Report],
(EPA/635/04/052). Washington, DC: U.S. Environmental Protection Agency, IRIS.
http://nepis.epa.gov/exe/ZyPURL.cgi?Dockey=Pl 006CK9.txt
U.S. EPA. (2005). Guidelines for carcinogen risk assessment [EPA Report] (pp. 1-166). (EPA/630/P-
03/001F). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment Forum.
https://www.epa.gov/sites/production/files/2Q13-
09/documents/cancer guidelines final 3-25-05.pdf
U.S. EPA. (2011a). Recommended use of body weight 3/4 as the default method in derivation of the
oral reference dose (pp. 1-50). (EPA/100/R-11/0001). Washington, DC: U.S. Environmental
Protection Agency, Risk Assessment Forum, Office of the Science Advisor.
https://www.epa.gov/risk/recommended-use-body-weight-34-default-method-derivation-
oral-reference-dose
U.S. EPA. (2011b). Toxicological review of trichloroethylene (CASRN 79-01-6) in support of
summary information on the Integrated Risk Information System (IRIS) [EPA Report],
(EPA/635/R-09/011F). Washington, DC.
https: //cfpub.epa.gov/ncea/iris /iris documents/documents/toxreviews /0199tr /0199tr.p
df
U.S. EPA. (2012a). Advances in inhalation gas dosimetry for derivation of a reference concentration
(RfC) and use in risk assessment (pp. 1-140). (EPA/600/R-12/044). Washington, DC.
https://cfpub.epa. gov/ncea/risk/recordisplay.cfm?deid=244650&CFID=50524762&CFTOK
EN=17139189
U.S. EPA. (2012b). Benchmark dose technical guidance. (EPA/100/R-12/001). Washington, DC: U.S.
Environmental Protection Agency, Risk Assessment Forum.
https://www.epa.gov/risk/benchmark-dose-technical-guidance
U.S. EPA. (2014). Guidance for applying quantitative data to develop data-derived extrapolation
factors for interspecies and intraspecies extrapolation [EPA Report], (EPA/100/R-
14/002F). Washington, DC: Risk Assessment Forum, Office of the Science Advisor.
https://www.epa.gOv/sites/production/files/2015-01/documents/ddef-final.pdf
U.S. EPA. (2015a). Science policy council peer review handbook [EPA Report] (4th ed.).
(EPA/100/B-15/001). Washington, DC: U.S. Environmental Protection Agency, Science
Policy Council, https: //www.epa.gov/osa/peer-review-handbook-4th-edition-2015
U.S. EPA. (2015b). Scoping and problem formulation for the toxicological review of polychlorinated
biphenyls (PCBS): Effects other than cancer [CASRN 1336-36-3] (Draft). (EPA/635/R-
14/198). Washington, DC. http://ofmpub.epa.gov/eims/eimscomm.getfile?
p download id=524917
U.S. EPA. (2017). Guidance to assist interested persons in developing and submitting draft risk
evaluations under the Toxic Substances Control Act. (EPA/740/R17/001). Washington, DC:
U.S Environmental Protection Agency, Office of Chemical Safety and Pollution Prevention.
This document is a draft for review purposes only and does not constitute Agency policy.
126 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
https://www.epa.gov/sites/production/files/2017-
06/documents/tsca ra guidance final.pdf
U.S. EPA. (2018). An umbrella Quality Assurance Project Plan (QAPP) for PBPK models [EPA
Report], (0RD QAPP ID No: B-0030740-QP-1-1). Research Triangle Park, NC.
Varghese. A: Cawlev. M: Hong. T. (2017). Supervised clustering for automated document
classification and prioritization: A case study using toxicological abstracts. Environ Syst
Decis 38: 398-414. http://dx.doi.org/l 0.1007/sl 0669-017-9670-5
Varghese. A: Hong. T: Hunter. C: Agveman-Badu. G: Cawlev. M. (2019). Active learning in automated
text classification: a case study exploring bias in predicted model performance metrics.
Environ SystDecisl-12. http://dx.doi.org/10.1007/sl0669-019-09717-3
Vesterinen. HM: Sena. ES: Egan. KT: Hirst. TC: Churolov. L: Currie. GL: Antonic. A: Howells. DW:
Macleod. MR. (2014). Meta-analysis of data from animal studies: a practical guide. J
Neurosci Methods 221: 92-102. http://dx.doi.org/10.1016/i.ineumeth.2013.09.010
Vorkamp. K. (2015). An overlooked environmental issue? A review of the inadvertent formation of
PCB-11 and other PCB congeners and their occurrence in consumer products and in the
environment [Review], Sci Total Environ 541: 1463-1476.
http://dx.doi.Org/10.1016/i.scitotenv.2015.10.019
WHO. (2003). Polychlorinatedbiphenyls: Human health aspects. (Concise International Chemical
Assessment Document 55). Geneva.
Wolff. MS: Camann. D: Gammon. M: Stellman. SD. (1997). Proposed PCB congener groupings for
epidemiological studies [Letter], Environ Health Perspect 105: 13-14.
Wolff. MS: Toniolo. PG. (1995). Environmental organochlorine exposure as a potential etiologic
factor in breast cancer. Environ Health Perspect 103: 141-145.
Xue. I: Liu. SV: Zartarian. VG: Geller. AM: Schultz. BP. (2014). Analysis of NHANES measured blood
PCBs in the general US population and application of SHEDS model to identify key exposure
factors. J Expo Sci Environ Epidemiol 24: 615-621. http: //dx.doi.org/10.1038/ies.2013.91
This document is a draft for review purposes only and does not constitute Agency policy.
127 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
APPENDICES
APPENDIX A. ELECTRONIC DATABASE SEARCH STRATEGIES
Table A-l. Database search strategy
Search
Search strategy
Date and results
Pub Med
Chemical terms
«((pcb[tw] OR pcb's OR "pcbs") NOT ("printed circuit board" OR "printed circuit
boards")) OR "polychlorinated biphenyl" OR "polychlorinated biphenyls" OR
"aroclor" OR "aroclors" OR "arochlor" OR "arochlors" OR "chlophen" OR
"chlophens" OR "chlorinated biphenyl" OR "chlorinated biphenyls" OR "chlorinated
diphenyl" OR "chloro biphenyl" OR "chloro biphenyls" OR clophen OR "clophens"
OR fenclor OR "fenclors" OR inerteen OR "inerteens" OR kanechlor OR
"kanechlors" OR phenochlor OR "phenochlors" OR phenoclor OR "phenoclors" OR
"polychlorobiphenyl" OR polychlorobiphenyls OR "pyralene" OR "pyranol" OR
"sovol"[tw] OR "sovols" OR therminol OR "therminols" OR "polychloro biphenyl"
OR "polychloro biphenyls" OR "polychlorodiphenyls" OR "polychlorinated
diphenyls" OR delor[tw] OR delors[tw] OR chlorofen OR "chlorofens" OR
monochlorobiphenyl OR monochlorobiphenyls OR chlorobiphenyl OR
"chlorobiphenyls" OR chlorodiphenyl OR "chlorodiphenyls" OR
monochlorodiphenyl OR "monochlorodiphenyls" OR dichlorobiphenyl OR
dichlorobiphenyls OR dichlorodiphenyl OR "dichlorodiphenyls" OR
"bichlorobiphenyl" OR "bichlorobiphenyls" OR trichlorobiphenyl OR
trichlorobiphenyls OR trichlorodiphenyl OR "trichlorodiphenyls" OR
tetrachlorobiphenyl OR tetrachlorobiphenyls OR tetrachlorodiphenyl OR
"tetrachlorodiphenyls" OR "tetrachloro biphenyl" OR pentachlorobiphenyl OR
pentachlorobiphenyls OR pentachlorodiphenyl OR "pentachlorodiphenyls" OR
"pentachloro biphenyl" OR hexachlorobiphenyl OR hexachlorobiphenyls OR
"hexachloro biphenyl" OR "hexachloro biphenyls" OR heptachlorobiphenyl OR
heptachlorobiphenyls OR octachlorobiphenyl OR octachlorobiphenyls OR
nonachlorobiphenyl OR nonachlorobiphenyls OR decachlorobiphenyl OR
decachlorobiphenyls)) or "capacitor manufacturing workers" or "Yu-Cheng" or
Yucheng or Yusho or "polychlorinated-biphenyls" or "Lake Michigan fish" or "North
Carolina Breast Milk and Formula Project" or "Great Lakes fish" or "Lake Ontario
fish" or "European Background PCB Study" or "Great Lakes Consortium" or "New
York State Angler Cohort" or "Lake Michigan Aging Population Study" or "Michigan
Anglers Cohort" or "Michigan Long-Term PBB Study"
7/29/2015: 19,089
8/31/2016: 388
Health effect terms
Not applicable
Other concepts
Health effect literature was prioritized using supervised clustering and machine
learning (DoCTOR). A total of 487 health effect-related publications derived from
the 2012 Toxicological Review of Polychlorinated Biphenyls (PCBs): Effects Other
Than Cancer (EPA/635/R-11/079C) were used as seed studies for clustering and
machine learning.
Web of Science
Chemical terms
((((TS="pcb" OR TS="pcbs") NOT (TS="printed circuit board" OR TS="printed circuit
boards")) OR TS="polychlorinated biphenyl" OR TS="polychlorinated biphenyls" OR
TS="aroclor" OR TS="aroclors" OR TS="arochlor" OR TS="arochlors" OR
7/29/15: 35,962
8/31/2016: 1,739
This document is a draft for review purposes only and does not constitute Agency policy.
128 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Search
Search strategy
Date and results
TS="chlophen" OR TS="chlophens" OR TS="chlorinated biphenyl" OR
TS="chlorinated biphenyls" OR TS="chlorinated diphenyl" OR TS="chlorinated
diphenyls" OR TS="chloro biphenyl" OR TS="chloro biphenyls" OR TS="clophen" OR
TS="clophens" OR TS="fenclor" OR TS="fenclors" OR TS="inerteen" OR
TS="inerteens" OR TS="kanechlor" OR TS="kanechlors" OR TS="phenochlor" OR
TS="phenochlors" OR TS="phenoclor" OR TS="phenoclors" OR
TS="polychlorobiphenyl" OR TS="polychlorobiphenyls" OR TS="pyralene" OR
TS="pyranol" OR TS="sovol" OR TS="sovols" OR TS="therminol" OR
TS="therminols" OR TS="polychloro biphenyl" OR TS="polychloro biphenyls" OR
TS="polychlorodiphenyls" OR TS="polychlorinated diphenyl" OR
TS="polychlorinated diphenyls" OR TS="delor" OR TS="delors" OR TS="chlorofen"
OR TS="chlorofens" OR TS="monochlorobiphenyl" OR TS="monochlorobiphenyls"
OR TS="chlorobiphenyl" OR TS="chlorobiphenyls" OR TS="chlorodiphenyl" OR
TS="chlorodiphenyls" OR TS="monochlorodiphenyl" OR TS="monochlorodiphenyls"
OR TS="monochloro biphenyl" OR TS="monochloro biphenyls" OR
TS="dichlorobiphenyl" OR TS="dichlorobiphenyls" OR TS="dichlorodiphenyl" OR
TS="dichlorodiphenyls" OR TS="dichloro biphenyl" OR TS="dichloro biphenyls" OR
TS="bichlorobiphenyl" OR TS="bichlorobiphenyls" OR TS="trichlorobiphenyl" OR
TS="trichlorobiphenyls" OR TS="trichlorodiphenyl" OR TS="trichlorodiphenyls" OR
TS="trichloro biphenyl" OR TS="trichloro biphenyls" OR TS="tetrachlorobiphenyl"
OR TS="tetrachlorobiphenyls" OR TS="tetrachlorodiphenyl" OR
TS="tetrachlorodiphenyls" OR TS="tetrachloro biphenyl" OR TS="tetrachloro
biphenyls" OR TS="pentachlorobiphenyl" OR TS="pentachlorobiphenyls" OR
TS="pentachlorodiphenyl" OR TS="pentachlorodiphenyls" OR TS="pentachloro
biphenyl" OR TS="pentachloro biphenyls" OR TS="hexachlorobiphenyl" OR
TS="hexachlorobiphenyls" OR TS="hexachloro biphenyl" OR TS="hexachloro
biphenyls" OR TS="heptachlorobiphenyl" OR TS="heptachlorobiphenyls" OR
TS="heptachloro biphenyl" OR TS="heptachloro biphenyls" OR
TS="octachlorobiphenyl" OR TS="octachlorobiphenyls" OR TS="octachloro
biphenyl" OR TS="octachloro biphenyls" OR TS="nonachlorobiphenyl" OR
TS="nonachlorobiphenyls" OR TS="nonachloro biphenyl" OR TS="nonachloro
biphenyls" OR TS="decachlorobiphenyl" OR TS="decachlorobiphenyls" OR
TS="decachloro biphenyl" OR TS="decachloro biphenyls"))
Health effect terms
Not applicable
Other concepts
Health effect literature was prioritized using supervised clustering and machine
learning (DoCTOR). A total of 487 health effect-related publications cited in the
2012 Toxicological Review Of Polychlorinated Biphenyls (PCBs): Effects Other Than
Cancer (EPA/635/R-11/079C) were used as seed studies for clustering and machine
learning.
Toxline
Chemical terms
@OR+(@TERM+@rn+1336-36-3+@TERM+@rn+12767-79-2+@TERM+@rn+64093-
59-0+@TERM+@rn+12521-86-7+@TERM+@rn+55945-68-l+@TERM+@rn+62251-
ll-0+@TERM+@rn+37353-77-8+@TERM+@rn+11137-46-5+@TERM+@rn+27323-
18-8+@TERM+@rn+25512-42-9+@TERM+@rn+25323-68-6+@TERM+@rn+26914-
33-0+@TERM+@rn+25429-29-2+@TERM+@rn+26601-64-9+@TERM+@rn+28655-
71-2+@TERM+@rn+55722-26-4+@TERM+@rn+53742-07-7+@TERM+@rn+2051-
24-3)+@NOT+@org+"nih+reporter"
7/29/15: 30,443
8/31/2016: 0
@OR+(@TERM+@rn+38444-93-8+@TERM+@rn+52663-59-
9+@TERM+@rn+36559-22-5+@TERM+@rn+70362-46-8+@TERM+@rn+41464-39-
5+@TERM+@rn+70362-45-7+@TERM+@rn+41464-47-5+@TERM+@rn+2437-79-
8+@TERM+@rn+70362-47-9+@TERM+@rn+41464-40-8+@TERM+@rn+62796-65-
0+@TERM+@rn+68194-04-7+@TERM+@rn+35693-99-3+@TERM+@rn+41464-41-
9+@TERM+@rn+15968-05-5+@TERM+@rn+74338-24-2+@TERM+@rn+41464-43-
This document is a draft for review purposes only and does not constitute Agency policy.
129 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Search
Search strategy
Date and results
l+@TERM+@rn+70424-67-8+@TERM+@rn+41464-49-7+@TERM+@rn+74472-33-
6+@TERM+@rn+33025-41-l+@TERM+@rn+33284-53-6+@TERM+@rn+54230-22-
7+@TERM+@rn+74472-34-7+@TERM+@rn+52663-58-8+@TERM+@rn+33284-54-
7+@TERM+@rn+32598-10-0+@TERM+@rn+73575-53-8+@TERM+@rn+73575-52-
7+@TERM+@rn+60233-24-l+@TERM+@rn+32598-ll-l+@TERM+@rn+41464-46-
4+@TERM+@rn+41464-42-0+@TERM+@rn+74338-23-l+@TERM+@rn+32690-93-
0+@TERM+@rn+32598-12-2+@TERM+@rn+70362-48-0+@TERM+@rn+32598-13-
3+@TERM+@rn+70362-49-l+@TERM+@rn+41464-48-6+@TERM+@rn+33284-52-
5+@TERM+@rn+70362-50-4)+@NOT+@org+"nih+reporter"
@OR+(@TERM+@rn+2051-60-7 +@TERM+@rn+2051-61-8+@TERM+@rn+2051-
62-9+@TERM+@rn+13029-08-8+@TERM+@rn+16605-91-7+@TERM+@rn+25569-
80-6+@TERM+@rn+33284-50-3+@TERM+@rn+34883-43-7+@TERM+@rn+34883-
39-l+@TERM+@rn+33146-45-l+@TERM+@rn+2050-67-l+@TERM+@rn+2974-92-
7+@TERM+@rn+2974-90-5+@TERM+@rn+34883-41-5+@TERM+@rn+2050-68-
2+@TERM+@rn+38444-78-9+@TERM+@rn+37680-66-3+@TERM+@rn+37680-65-
2+@TERM+@rn+38444-73-4+@TERM+@rn+38444-84-7+@TERM+@rn+55702-46-
0+@TERM+@rn+38444-85-8+@TERM+@rn+55720-44-0+@TERM+@rn+55702-45-
9+@TERM+@rn+55712-37-3+@TERM+@rn+38444-81-4+@TERM+@rn+38444-76-
7+@TERM+@rn+7012-37-5+@TERM+@rn+15862-07-4+@TERM+@rn+35693-92-
6+@TERM+@rn+16606-02-3+@TERM+@rn+38444-77-8+@TERM+@rn+38444-86-
9+@TERM+@rn+37680-68-5+@TERM+@rn+37680-69-6+@TERM+@rn+38444-87-
0+@TERM+@rn+38444-90-5+@TERM+@rn+53555-66-l+@TERM+@rn+38444-88-
l)+@NOT+@org+"nih+reporter"
@OR+(@TERM+@rn+38380-07-3+@TERM+@rn+55215-18-
4+@TERM+@rn+52663-66-8+@TERM+@rn+61798-70-7+@TERM+@rn+38380-05-
l+@TERM+@rn+35694-04-3+@TERM+@rn+52704-70-8+@TERM+@rn+52744-13-
5+@TERM+@rn+38411-22-2+@TERM+@rn+35694-06-5+@TERM+@rn+35065-28-
2+@TERM+@rn+56030-56-9+@TERM+@rn+59291-64-4+@TERM+@rn+52712-04-
6+@TERM+@rn+41411-61-4+@TERM+@rn+68194-15-0+@TERM+@rn+68194-14-
9+@TERM+@rn+74472-40-5+@TERM+@rn+51908-16-8+@TERM+@rn+68194-13-
8+@TERM+@rn+74472-41-6+@TERM+@rn+38380-04-0+@TERM+@rn+68194-08-
l+@TERM+@rn+52663-63-5+@TERM+@rn+68194-09-2+@TERM+@rn+35065-27-
l+@TERM+@rn+60145-22-4+@TERM+@rn+33979-03-2+@TERM+@rn+38380-08-
4+@TERM+@rn+69782-90-7+@TERM+@rn+74472-42-7+@TERM+@rn+39635-35-
3+@TERM+@rn+41411-62-5+@TERM+@rn+74472-43-8+@TERM+@rn+39635-34-
2+@TERM+@rn+74472-44-9+@TERM+@rn+74472-45-0+@TERM+@rn+74472-46-
l+@TERM+@rn+41411-63-6+@TERM+@rn+52663-72-6+@TERM+@rn+59291-65-
5+@TERM+@rn+32774-16-6)+@NOT+@org+"nih+reporter"
@OR+(@TERM+@rn+52663-62-4+@TERM+@rn+60145-20-
2+@TERM+@rn+52663-60-2+@TERM+@rn+65510-45-4+@TERM+@rn+55312-69-
l+@TERM+@rn+38380-02-8+@TERM+@rn+55215-17-3+@TERM+@rn+73575-57-
2+@TERM+@rn+68194-07-0+@TERM+@rn+68194-05-8+@TERM+@rn+52663-61-
3+@TERM+@rn+73575-56-l+@TERM+@rn+73575-55-0+@TERM+@rn+38379-99-
6+@TERM+@rn+73575-54-9+@TERM+@rn+41464-51-l+@TERM+@rn+60233-25-
2+@TERM+@rn+38380-01-7+@TERM+@rn+39485-83-l+@TERM+@rn+37680-73-
2+@TERM+@rn+68194-06-9+@TERM+@rn+60145-21-3+@TERM+@rn+56558-16-
8+@TERM+@rn+32598-14-4+@TERM+@rn+70424-69-0+@TERM+@rn+70424-68-
9+@TERM+@rn+70362-41-3+@TERM+@rn+74472-35-8+@TERM+@rn+38380-03-
9+@TERM+@rn+39635-32-0+@TERM+@rn+74472-36-9+@TERM+@rn+68194-10-
5+@TERM+@rn+74472-37-0+@TERM+@rn+74472-38-l+@TERM+@rn+18259-05-
7+@TERM+@rn+68194-ll-6+@TERM+@rn+31508-00-6+@TERM+@rn+56558-17-
9+@TERM+@rn+68194-12-7+@TERM+@rn+56558-18-0+@TERM+@rn+76842-07-
4+@TERM+@rn+65510-44-3+@TERM+@rn+70424-70-3+@TERM+@rn+74472-39-
2+@TERM+@rn+57465-28-8+@TERM+@rn+39635-33-
l)+@NOT+@org+"nih+reporter"
This document is a draft for review purposes only and does not constitute Agency policy.
130 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
APPENDIX B. DATA EXTRACTION FIELDS
Table B-l. Key data extraction elements to summarize study design,
experimental model, methodology, and results
Field label
Data extraction elements
HUMAN
Funding
Funding source(s)
Reporting of conflict of interest by authors
Subjects
Study population name/description
Dates of study and sampling time frame
Geography (country, region, state, etc.)
Demographics (sex, race/ethnicity, age or lifestage at exposure and at outcome assessment)
Number of subjects (target, enrolled, n per group in analysis, and participation/follow-up
rates)
Inclusion/exclusion criteria/recruitment strategy
Description of reference group
Methods
Study design (e.g., prospective or retrospective cohort, nested case-control study,
cross-sectional, population-based case-control study, intervention, case report)
Length of follow-up
Health outcome category (e.g., cardiovascular)
Health outcome (e.g., blood pressure)
Diagnostic or methods used to measure health outcome
Confounders or modifying factors and how considered in analysis (e.g., included in final model,
considered for inclusion but determined not needed)
Chemical/Mixture name
Exposure assessment (e.g., blood, urine, hair, air, drinking water, job classification, residence,
administered treatment in controlled study)
Methodological details for exposure assessment (e.g., analytical method, limit of detection)
Statistical methods
Results
Exposure levels (e.g., mean, median, measures of variance as presented in paper, such as SD,
SEM, 75th/90th/95th percentile, minimum/maximum); range of exposure levels, number of
exposed cases
This document is a draft for review purposes only and does not constitute Agency policy.
131 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Field label
Data extraction elements
Statistical findings (e.g., adjusted p, standardized mean difference, adjusted odds ratio,
standardized mortality ratio, relative risk) or description of qualitative results. When possible,
convert measures of effect to a common metric with associated 95% confidence intervals.
Most often, measures of effect for continuous data are expressed as mean difference,
standardized mean difference, and percentage control response. Categorical data are typically
expressed as odds ratio, relative risk (also called risk ratio), or p values, depending on the
metric most commonly reported in the included studies and ability to obtain information for
effect conversions from the study or through author query.
Observations on dose-response (e.g., trend analysis, description of whether dose-response
shape appears to be monotonic, nonmonotonic)
Other
Documentation of author queries, use of digital rulers to estimate data values from figures,
exposure unit, and statistical result conversions, etc.
ANIMAL
Funding
Funding source(s)
Reporting of conflict of interest by authors
Animal model
Sex
Species
Strain
Source of animals
Age or lifestage at start of dosing and at health outcome assessment
Diet and husbandry information (e.g., diet name/source)
Treatment
Chemical/Mixture name and CAS (Chemical Abstracts Service) number
Source of chemical
Purity or Lot # of chemical
Dose levels or concentration (as presented and converted to mg/kg bw-day when possible)
Other dose-related details, such as whether administered dose level was verified by
measurement, information on internal dosimetry
Vehicle used for exposed animals
Route of administration (e.g., oral, inhalation, dermal, injection)
Duration and frequency of dosing (e.g., hours, days, weeks when administration was ended,
days per week)
Methods
Study design (e.g., single treatment, acute, subchronic [e.g., 90 days in a rodent], chronic,
multigenerational, developmental, other)
Guideline compliance (i.e., use of EPA, OECD, NTP, or another guideline for study design,
conducted under GLP guideline conditions, non-GLP but consistent with guideline study,
nonguideline peer reviewed publication)
Number of animals per group (and dams per group in developmental studies)
Randomization procedure, allocation concealment, blinding during outcome assessment
This document is a draft for review purposes only and does not constitute Agency policy.
132 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Field label
Data extraction elements
Method to control for litter effects in developmental studies
Use of negative controls and whether controls were untreated, vehicle-treated, or both
Report on data from positive controlswas expected response observed?
Endpoint health category (e.g., reproductive)
Endpoint (e.g., infertility)
Diagnostic or method to measure endpoint
Statistical methods
Results
Measures of effect at each dose or concentration level (e.g., mean, median, frequency, and
measures of precision or variance) or description of qualitative results. When possible,
convert measures of effect to a common metric with associated 95% confidence intervals.
Most often, measures of effect for continuous data will be expressed as mean difference,
standardized mean difference, and percentage control response. Categorical data will be
expressed as relative risk (also called risk ratio).
NOAEL, LOAEL, benchmark dose analysis, statistical significance of other dose levels, or other
estimates of effect presented in paper.
Note: The NOAEL and LOAEL are highly influenced by study design, do not give any
quantitative information about the relationship between dose and response, and can be
subject to author's interpretation (e.g., a statistically significant effect might not be considered
biologically important). Also, a NOAEL does not necessarily mean zero response. Ideally, the
response rate at specific dose levels is used as the primary measure to characterize the
response.
Observations on dose-response (e.g., trend analysis, description of whether dose-response
shape appears to be monotonic, nonmonotonic)
Data on internal concentration, toxicokinetics, or toxicodynamics (when reported)
Other
Documentation of author queries, use of digital rulers to estimate data values from figures,
exposure unit, and statistical result conversions, etc.
IN VITRO
Funding
Funding source(s)
Reporting of conflict of interest by authors (reporting bias)
Cell/tissue
Cell line, cell type, or tissue
model
Source of cells/tissue (and validation of identity)
Sex of human/animal of origin
Species
Strain
Treatment
Chemical name and CAS number
Concentration levels (as presented and converted to nM when possible)
Source of chemical
Purity of chemical
This document is a draft for review purposes only and does not constitute Agency policy.
133 DRAFT-DO NOT CITE OR QUOTE
-------
Systematic Review Protocol for the PCBs Noncancer IRIS Assessment
Field label
Data extraction elements
Vehicle used for experimental/control conditions
Duration and frequency of dosing (e.g., hours, days, weeks when administration was ended,
times per day or week)
Methods
Guideline compliance (i.e., use of EPA, OECD, NTP, or another guideline for study design,
conducted under GLP guideline conditions, non-GLP but consistent with guideline study,
nonguideline peer reviewed publication)
Randomization procedure, allocation concealment, blinding during outcome assessment
(selection bias)
Number of replicates per group (information bias)
Percentage serum/plasma in medium
Use of negative controls and whether controls were untreated, vehicle-treated, or both
Report on data from positive controlswas expected response observed? (information bias)
Endpoint health category (e.g., endocrine)
Endpoint or assay target (e.g., estrogen receptor binding or activation)
Name and source of assay kit
Diagnostic or method to measure endpoint (e.g., reporter gene; information bias)
Statistical methods (information bias)
Results
No-observed-adverse-effect concentration (NOAEC), lowest-observed-adverse-effect
concentration (LOAEC), statistical significance of other concentration levels, AC50, or other
estimates of effect presented in paper.
Note: The NOAEC and LOAEC are highly influenced by study design, do not give any
quantitative information about the relationship between dose and response, and can be
subject to author's interpretation (e.g., a statistically significant effect might not be considered
biologically important). Also, a NOAEC does not necessarily mean zero response.
Observations on dose-response (e.g., trend analysis, description of whether dose-response
shape appears to be monotonic, nonmonotonic)
Other
Documentation of author queries, use of digital rulers to estimate data values from figures,
exposure unit, and statistical result conversions, etc.
AC50 = 50% activity concentration; EPA = U.S. Environmental Agency; GLP = Good Laboratory Practice;
LOAEC = lowest-observed-adverse-effect concentration; LOAEL = lowest-observed-adverse-effect level;
NOAEC = no-observed-adverse-effect concentration; NOAEL = no-observed-adverse-effect level; NTP = National
Toxicology Program; OECD = Organisation for Economic Co-operation and Development.
This document is a draft for review purposes only and does not constitute Agency policy.
134 DRAFT-DO NOT CITE OR QUOTE
------- |