vvEPA
United States
Environmental Protection
Agency
Environmental Criteria and
Assessment Office
Cincinnati OH 45268
EPA-600/9-84-014a
June 1984
Research and Development
Selected Approaches to
Risk Assessment for
Multiple Chemical
Exposures:
Progress Report on
Guideline Development at
ECAO-Cincinnati
-------
EPA-600/9-84-014a
June 1984
SELECTED APPROACHES TO RISK ASSESSMENT
FOR MULTIPLE CHEMICAL EXPOSURES
Progress Report on Guideline Development at ECAO-C1ndnnat1
Or. Jerry F. Stara, Director, ECAO
Editor
Environmental Criteria and Assessment Office
Cincinnati, Ohio 45268
Dr. Linda S. Erdrelch
Technical Editor
Environmental Criteria and Assessment Office
Cincinnati, Ohio 45268
Environmental Criteria and Assessment Office
Office of Health and Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, OH 45268
-------
NOTICE
This document has been reviewed 1n accordance with U.S. Environmental
Protection Agency policy and approved for publication. Mention of trade
names or commercial products does not constitute endorsement or recommenda-
tion for use.
The appendix to this document, consisting of post meeting comments and
selected references, will be available only from National Technical Informa-
tion Service, U.S. Department of Commerce, Springfield, VA 22161, as
"Selected Approaches to Risk Assessment for Multiple Chemical Exposures:
Appendix,- EPA-600/9-84-014b.
11
-------
FOREWORD
Methods must be developed for assessing the potential risks of simulta-
neous or sequential exposure to multiple chemicals 1n order to prepare valid
assessments of risks posed by environmental contamination. Many real-life
exposures have dynamic and chronic patterns that have no counterparts 1n
laboratory studies. This workshop and others 1n this series are Intended to
develop such methods, to promote discussion on alternative approaches, and
to stimulate research Into developing areas 1n risk assessment.
The need for risk assessment 1s driven by the presence of or threat of
environmental contamination and the absence of sufficient data for the human
health effects of these chemicals. The twenty-five peer reviewers listed 1n
this report were selected for their scientific expertise and Interest 1n
this pressing problem. The presentations and subsequent scientific Inter-
change among Agency and non-Agency scientists, and the summary statements of
the reviewers after the meeting are presented 1n this document to reflect
the dynamic nature of this contemporary problem. The document represents a
second progress report on the development of risk assessment methodologies
and guidelines which are 1n progress at the Environmental Criteria and
Assessment Office, Cincinnati.
Jerry F. Stara, DVM, OS
Director
111
-------
ABSTRACT
This report summarizes a workshop that addressed methods and Issues for
the assessment of human health risks from exposure to multiple chemicals.
The meeting was sponsored by the Environmental Criteria and Assessment
Office of the U.S. Environmental Protection Agency and was held 1n Cincin-
nati, Ohio, on July 12-13, 1983. The methods and Issues were selected from
those presented at a previous workshop held 1n September 1982, "Approaches
to Risk Assessment for Multiple Chemical Mixtures." Approaches 1n five
areas were discussed with the Intention of Identifying valid methods of
mult1chem1cal risk assessment that could be Implemented 1n the near future.
Consensus was reached among the participants on approaches to the Inter-
species conversion of dose and duration of exposure (use one-third power of
body weight and lifetime proportionality factors), the assessment of risk
for less than lifetime exposure to toxicants (consider dose, duration, and
no-effect levels for all available data), and multiple chemical assessment
(an add1t1v1ty model was considered adequate 1n the Interim). The other
topics proposed for consensus concerned the assessment of risk for less than
lifetime exposure to carcinogens, the determination of acceptable dally
Intakes based on quantal, continuous, or graded data, and the pharmaco-
klnetlc approach to route-to-route conversions.
Brief presentations were made by several participants on approaches
using structure-activity relationships, reproductive effects as endpolnts 1n
risk assessment, and use and biological justification of mathematical
models. Four working groups were also convened to consider approaches
proposed 1n four additional areas: consideration of high risk subgroups,
assessment of multiple routes of exposure, the ranking of the severity of
the effects, and the use of exposure and monitoring data 1n health risk
assessment. A statement Including recommendations was produced by each
group as a guide for future discussions. The results of this workshop will
provide a basis for developing guidelines for risk assessment from multiple
chemical exposures. This effort 1n guidelines development will culminate 1n
a symposium 1n the summer of 1984.
1v
-------
TABLE OF CONTENTS
Page
1. INTRODUCTION - MEETING OBJECTIVES AND STRUCTURE 1
2. CONSENSUS TOPICS (Session I) 4
2.1. INTERSPECIES CONVERSION OF DOSE AND DURATION OF
EXPOSURE — NONCARCINOGENIC TOXICANTS 4
2.2. HEALTH RISK ASSESSMENT FOR LESS THAN LIFETIME
EXPOSURE — TOXICANTS AND CARCINOGENS 9
2.3. ADIs BASED ON QUANTAL, CONTINUOUS, OR GRADED DATA 24
2.4. PHARMACOKINETIC APPROACH FOR ROUTE-TO-ROUTE CONVERSION . . 34
2.5. MULTIPLE CHEMICAL ASSESSMENT 44
3. PRESENTATION OF NEW TOPICS (Session II) 55
3.1. APPROACHES USING STRUCTURE-ACTIVITY RELATIONSHIPS 55
3.2. USE OF REPRODUCTIVE EFFECTS AS ENDPOINTS IN RISK
ASSESSMENT 60
3.3. USE AND BIOLOGICAL JUSTIFICATION OF MATHEMATICAL MODELS. . 63
4. WORKSHOPS (Session III) 72
4.1. CONSIDERATION OF HIGH RISK (SENSITIVE) SUBGROUPS IN
HEALTH RISK ASSESSMENT 72
4.2. ASSESSMENT OF MULTIPLE ROUTE EXPOSURE 81
4.3. RANKING THE SEVERITY OF EFFECTS 87
4.4. USE OF EXPOSURE DATA IN ASSESSING HEALTH RISK 94
REFERENCES 101
-------
LIST OF TABLES
No. Title Page
2-1 Various Effect Levels and Their Definitions 10
2-2 Constant Dose Rates Corresponding to Given Levels of
Extra Risk 21
2-3 Doses of 1-Week Duration that Correspond to Given Levels
of Extra Risk 22
2-4 Quantal Data Used To Illustrate Quantitative Dose Response
Methodology 29
2-5 Summary of Fits of Models to Quantal Data 1n Table 2-4. ... 30
2-6 Doses Corresponding to Given Levels of Extra Risk for
Quantal ETU Data 31
2-7 Comparison of Benchmark Doses with NOELs for Quantal Data . . 32
2-8 Magnitude of Rate Coefficients (t1/2) 42
3-1 Performance of SAR Equations 57
3-2 Predicted Clearance of Benzo(a)pyrene 1n Rats 66
4-1 Biological Factors Predisposing Individuals to Hyper-
suscept1b1Hty to Pollutants 76
4-2 Rating Values for NOELs, LOAELs and FELs for the
Determination of RVe 90
4-3 Relationship of Composite Score to RQ 91
v1
-------
LIST OF FIGURES
No. Title Page
2-1 Methoxychlor Oral Tox1c1ty Data 11
2-2 Methoxychlor Oral Toxldty Data - Option 1 13
2-3 Methoxychlor Oral Toxldty Data - Option 2 14
2-4 Male Rats Exposed to EDB by Gavage 18
2-5 Comparison of Estimates from Arm1tage-Doll Model with
Kaplan-Meier Estimates 19
2-6 Comparison of Estimates from ArmHage-Doll Model with
Kaplan-Meier Estimates Excluding Tumors Found at Sacrifice. . 20
2-7 Hypothetical Blood Concentrations for Three Exposure
Conditions 36
2-8 Hypothetical Blood Concentrations for Intermediate
Elimination Rates 37
2-9 Hypothetical Blood Concentrations for Long Elimination
Rates 38
2-10 Dose Addition 46
2-11 Response Addition 47
2-12 Isobole Diagram for Quanta! Response Data 48
2-13 Example of Dose and Response Addition 49
2-14 Models for Interaction 51
3-1 The Relationship Between a Measure of Internal Dose, the
Area Under the Concentration, Time Curve, and Either Oral
or Intravenous Metoprolol 1n Five Male Human Subjects .... 69
3-2 Mean Plasma Concentrations of Nortr1ptyl1ne and Us Major
Metabolite, I0-Hydroxynortr1ptyl1ne 70
4-1 Information Flow and Methodology Use 1n Conducting Site-
Specific MultUhemlcal Health Risk Assessment 75
4-2 Rating Values for Doses 89
-------
LIST OF PARTICIPANTS
ECAO-C1n AUTHORS AND REVIEWERS
Jerry F. Stara
Tox1colog1st (Director)
Linda S. Erdrelch
Epidemiologist
Randall J.F. Bruins
Environmental Scientist
Michael L. Dourson
Tox1colog1st
Richard C. Hertzberg
Blomathematlclan
William Pepelko
Tox1colog1st
EXTERNAL PEER REVIEWERS
Roy E. Albert
Institute of Environmental Medicine
Tuxedo, New York
Julian B. Andelman
University of Pittsburgh
Graduate School of Public Health
Pittsburgh, Pennsylvania
Eula Blngham
Ketterlng Laboratory
College of Medicine
University of Cincinnati
Cincinnati, Ohio
Edward J. Calabrese
Division of Public Health
University of Massachusetts
Amherst, Massachusetts
Thomas W. Clarkson
Division of Toxicology
School of Medicine and Dentistry
University of Rochester, New York
Herbert H. Cornish
School of Public Health
University of Michigan
Ann Arbor, Michigan
Kenny S. Crump
Science Research Systems, Inc.
Ruston, Louisiana
Patrick R. Durkln
Syracuse Research Corporation
Syracuse, New York
Kurt Ensleln
Health Design, Inc.
Rochester, New York
Rolf Hartung
School of Public Health
University of Michigan
Ann Arbor, Michigan
Dale Hattls
Center for Policy Alternatives
Massachusetts Institute of Technology
Cambridge, Massachusetts
Marvin Legator
University of Texas Medical School
Dept. of Preventive Medicine and
Community Health
Galveston, Texas
Jeanne M. Manson
Dept. of Obstetrics and Gynecology
University of Cincinnati
Cincinnati, Ohio
Myron Mehlman
Mobil 011 Corporation
Princeton, New Jersey
Norton Nelson
New York University Medical Center
New York, New York
-------
EXTERNAL PEER REVIEWERS (cont.)
William J. Nicholson
Department of Environmental Health
Mt. S1na1 Hospital
New York, New York
Ian C. T. Nlsbet
Clement Associates
Arlington, Virginia
Ellen J. O'Flaherty
Institute of Environmental Health
Ketterlng Laboratory
University of Cincinnati
Cincinnati, Ohio
Magnus Plscator
Karollnska Institute
Stockholm, Sweden
Marvin A. Schnelderman
Clement Associates
Arlington, Virginia
Ellen K. Sllbergeld
Environmental Defense Fund
Washington, DC
James R. WUhey
Food Directorate
Bureau of Chemical Safety
Ottawa, Canada
Ronald E. Wyzga
Electric Power Research Institute
Palo Alto, California
OTHER GOVERNMENT SCIENTISTS AND
PARTICIPANTS
Irwln Baumel
U.S. EPA
Washington, DC
Judith Bellln
U.S. EPA
Washington, DC
Josephine Brecher
U.S. EPA
Washington, DC
Steven BrodeMus
U.S. EPA
Duluth, Minnesota
Robert Bruce
Procter & Gamble Company
Cincinnati, Ohio
C1pr1ano Cueto
Dynamac Corporation
Rockvllle, Maryland
Margaret Chu
U.S. EPA
Washington, DC
Chris Dlppel
Dynamac Corporation
Rockvllle, Maryland
Alan M. Ehrllch
U.S. EPA
Washington, DC
Len1 Field
Procter & Gamble Company
Cincinnati, Ohio
Gregory Kew
U.S. EPA
Washington, DC
Arnold Kuzmack
U.S. EPA
Washington, DC
William Lappenbusch
U.S. EPA
Washington, DC
Robert McGaughy
U.S. EPA
Washington, DC
Susan Moskowltz
U.S. EPA
Washington, DC
Edward Ohanlan
U.S. EPA
Washington, DC
1x
-------
OTHER GOVERNMENT SCIENTISTS AND
PARTICIPANTS (cont.)
Jean C. Parker
U.S. EPA
Washington, DC
Orvllle E. Paynter
U.S. EPA
Washington, DC
David J. Relsman
U.S. EPA
Cincinnati, Ohio
Rosemarle Russo
U.S. EPA
Duluth, Minnesota
Paul Seybold
Wright State University
Dayton, Ohio
Maryrose K. Smith
U.S. EPA
Cincinnati, Ohio
Janet Springer
U.S. EPA
Washington, DC
Glenn W. Suter, II
Oak Ridge National Laboratory
Oak Ridge, Tennessee
William W. Sutton
U.S. EPA
Las Vegas, Nevada
Deborah Taylor
U.S. EPA
Washington, DC
Todd Thorslund
U.S. EPA
Washington, DC
DOCUMENT DEVELOPMENT
The Initial draft was reviewed by the following personnel of the
Environmental Criteria and Assessment Office: Jerry F. Stara (Director),
Randall J.F. Bruins, Michael L. Dourson, Linda S. Erdrelch, Richard C.
Hertzberg and William Pepelko.
The ECAO Technical Support Staff Included Judith Olsen, Erma Durden,
Bette Zwayer, Patricia Daunt and Karen Mann.
-------
1. INTRODUCTION
MEETING OBJECTIVES AND STRUCTURE (J. Stara)
The U.S. EPA's Environmental Criteria and Assessment Office (ECAO) 1n
Cincinnati sponsored three major workshops that brought together expert
scientists to review and discuss guidelines for health risk assessment of
chemical mixtures. The purpose of these meetings was to revise and add to
the previously published health assessment guidelines (Federal Register,
1980) for assessing the chronic (or lifetime) health risk of single contami-
nants so that the more complex questions associated with exposures to chemi-
cal mixtures could be addressed.
The Agency's Interest 1n the development of risk assessment methodology
for multiple chemicals stems from recognition of the nature of actual expo-
sures. The Agency has already regulated public exposure to several mixtures
Including coke-oven emissions and auto and dlesel exhaust. Now, with the
Implementation of Superfund, the Agency must develop methods for determining
the risks posed by what may be thousands of unregulated sites In order to
prioritize and direct cleanup efforts. In addition, the Agency must respond
to Imminent hazards posed by chemical spills, and, toward that goal, ECAO
has developed a rapid response assessment system. Toxldty data are stored
1n a computer and used to make an assessment within 48 hours that can then
be provided to state or regional officials. Mult1chem1cal assessments also
have a role 1n enforcement, I.e., determining the reduction 1n risk after a
site's cleanup. More broadly, all Agency program offices realize that they
are dealing with mixtures, not only single pollutants.
1871A -1- 04/11/84
-------
The primary objective of this meeting, therefore, was the Injection of
scientific stimulus Into the development of the mult1chem1cal assessment
methodologies. To provide structure to this stimulus, this meeting was
divided Into three sessions.
In Session I, specific selected methodologlc approaches were proposed
for those subjects that had been discussed at length 1n previous meetings.
Discussion was focused on whether or not the approach was acceptable as
stated. If not, an attempt was made to amend the approach to achieve
consensus or to clearly Identify remaining problems or disagreements. The
following topics were presented for consensus: Interspedes conversion of
dose and duration of exposure; calculation of risk for partial-lifetime
exposure to carcinogens or toxicants; calculation of acceptable dally
Intakes (ADIs) based on quantal, continuous, or graded response data;
pharmacoklnetlc approaches that can be used for conversion among various
exposure routes; and methods for multlchemlcal assessment.
In Session II, several new topics were presented that had not been dis-
cussed or were only touched upon 1n previous meetings. One or two speakers
were asked to present a brief outline of the Issues, followed by a period of
discussion. The goal of this session was to elicit comments from reviewers
about the direction toward which these Issues should be developed. New
topics Included the use of structure-activity relationships, special methods
for reproductive endpolnts, and use and biological justification of mathe-
matical models.
1871A -2- 03/21/84
-------
Issues Introduced at previous meetings, but needing further development,
were the subject of four workgroups 1n Session III. Invited scientists and
ECAO staff discussed the strengths and limitations of alternative approaches
1n order to develop the guidelines 1n these areas. The objectives of the
workgroups were to eventually develop a scientifically defensible consensus
reflecting current knowledge and to Identify the options that could be best
developed Into a consensus approach. The following topics were discussed:
methods to account for sensitive subgroups, methods to assess effects of
multlroute exposures, methods for ranking the severity of various effects,
and the use of exposure data In assessing health risk.
1871A -3- 03/21/84
-------
2. CONSENSUS TOPICS (SESSION I)
2.1. INTERSPECIES CONVERSION OF DOSE AND DURATION OF EXPOSURE - NONCARCI-
NOGENIC TOXICANTS
2.1.1. Presentation (R. Htrtibtrg, T. Clarkson). The goal of this
session was to choose a method for determining how dose-response Information
from animal studies could be used to estimate human dose-response relation-
ships for noncardnogenlc toxicants. The primary limitation 1s that for
most chemicals, Insufficient data on absorption, metabolism, and elimination
are available to accurately extrapolate the dose-duration-effect relation-
ship from studies of animals to humans. Several methods, or options, were
proposed for accomplishing these 1nterspec1es conversions. To deal with the
duration question, a proposed method was to assume that equally toxic
effects 1n humans will result from exposures where duration 1s the same
fraction of the total lifetime as for the exposed animals. This approach 1s
currently used 1n assessing the risk posed by carcinogens and may be
applicable to most noncardnogenlc toxicants. However, 1t 1s probably not
applicable to certain types of effects, certainly not acute effects and most
likely not teratogenlc effects that result from exposures during a unique
susceptible period.
To deal with the question of equltoxlc dose rates, a number of conver-
sion methods were proposed:
1. Expression of dose as a function of body weight, I.e.,
mg/kg/day
2. Expression of dose as a function of metabolism or some surro-
gate, I.e., mg/cm2 body surface area
1871A -4- 03/21/84
-------
3. Use of a mg/kg brain weight per day conversion when extrapo-
lating central nervous system effects
4. Use of Options 1 or 2 when 1t 1s known that the substance 1s
detoxified or activated, respectively, 1n a manner that 1s
similar 1n the experimental species and humans
5. Use of Options 1 or 2 depending on the best fit of either
relationship to the available toxldty data.
None of these options 1s an obvious choice as the best conversion
method. Expression of dose on a straight surface area basis 1s said to be
more "protective" than the body weight-based conversion, because for doses
derived 1n studies with the jjsual test species, rodents, 1t results 1n lower
equltoxlc doses. Options 4 and 5 are substance-specific: Option 4 requires
metabolic data and Option 5 depends on the shape of the dose-response curve
derived from toxldty data. However, curve-fitting requires both substan-
tial data and the selection of a Justifiable model for the curve-fit.
2.2.2. Discussion. Dr. Clarkson Initiated the discussion of duration
conversion by stating that the most frequently used approach 1s to set a
proportion between the fraction of the lifetime of the test species for a
given effect and the fraction of a human lifetime expected to produce an
effect, but that the approach 1s not applicable to some types of effects,
I.e., acute or teratogenlc effects. Dr. Hartung pointed out that the
question of the validity of the use of proportionality 1s susceptible to
experimental resolution and that consideration of all three axes (dose,
duration, and response) should be made. Dr. Nelson Indicated that, 1n
general, the dividing line to be considered 1s between acute effects and
delayed effects, and that knowledge of the pharmacoklnetlcs of the substance
1s needed. Dr. Clarkson responded that pharmacoklnetlc data are usually
1871A _5- 03/21/84
-------
lacking and concluded that there seemed to be general agreement among the
participants that EPA should continue to use the lifetime proportionality
approach limited by Us applicability to only chronic types of effects. He
stated that, at this time, this approach 1s a "best guess" that needs
experimental resolution.
In the discussion of dose conversion approaches, Dr. Clarkson stated
that body weight, which 1s proportional to volume, 1s the preferred conver-
sion and that the rationale for using surface area (because 1t Is an Index
of metabolic rate) 1s now believed to be spurious. Dr. WHhey pointed out
that the Important factor to determine 1s the "effective dose," I.e., the
dose at the physiological level that 1s producing an effect, and that In his
published papers J1m Gillette has proposed a conversion approach using
effective dose. Dr. Schnelderman pointed out that a conversion could be
related to liver weight because the liver 1s the main site of activation and
detoxification; Dr. Clarkson supported the use of a target tissue relation-
ship. Dr. Hattls cited work by Boxenbaum (1982) that demonstrated how many
energy-consuming functions display Interspecific relations that are a 0.75
power of body weight and therefore the turnover time of the average body
constituent 1s
(body we1ght)V(body weight)0'75 = (body weight)0'25 (Hattls).
Dr. Clarkson stated that elimination time 1s Important. He presented data
on the elimination rates for methyl mercury 1n a number of species and
showed that the variation between the rates was reduced by a body weight to
the one-third adjustment. Several speakers noted that the overall metabo-
lism of a substance (whether 1t attains a steady state 1n the blood) would
Influence any conversion methods based on elimination rates (Hartung,
1871A -6- 04/11/84
-------
Nelson, and Nicholson). Dr. Plscator pointed out that when dose 1s ex-
pressed as parts per million of diet, conversions may not be needed because
food consumption 1s proportional to body weight. Difficulties 1n using
dietary concentration alone are that laboratory animals and humans have
diets with different composition (low water and fat vs. high water and fat,
respectively), and that 1n some studies the test material Influences food
consumption (Nlsbet).
A proposal for research Into the question was made by Dr. Hartung who
suggested using a wide range of dose-duration-effect data from different
species 1n an optimization program to determine the best-fitting exponent.
Dr. Ensleln suggested using structural parameters as covarlants, and Dr.
Crump proposed analyzing the data with the type of effect as the common
denominator rather than the chemical.
In their postmeetlng memoranda, several participants provided sugges-
tions for Improving Interspedes conversions. Dr. Hattls stated that more
sophistication could be used 1n determining a general scaling factor for
cases 1n which the application of Option 4 (pharmacoklnetlc approach) or
Option 5 (goodness-of-f1t for toxldty data) were limited by Insufficient
data. He suggested using the current data base or conducting research to
discern patterns 1n 1) which kinds of chemicals scale with larger vs.
smaller body weight exponents, and 2) which kinds of chemicals appear to
give rise to anomalous human data (1n which direction are the data anomalous
and what 1s the magnitude of the anomaly). In the absence of other Informa-
tion, Dr. Hattls advocated a one-fourth exponent as opposed to one-third.
1871A -7- 03/21/84
-------
He also suggested placing the Hfespan scaling for duration conversion on a
firmer basis by making some comparisons of the "time x dose-rate tradeoff"
for a few defined toxic endpolnts for several species with different life-
spans. Dr. Schnelderman suggested (as did Drs. Hartung and Crump) determin-
ing the dose-conversion factor on the basis of a single toxic effect,
possibly using for an extrapolation the factor for a "crucial" or "key"
toxldty. In their postmeetlng memoranda, other participants pointed out
additional considerations 1n Interspedes conversions. Dr. O'Flaherty
stated that sufficient data were available for a more systematic approach
than that used by EPA thus far and raised two questions: 1) Should conver-
sions for short-term (nonsteady-state) exposures be different from conver-
sions for chronic conditions? and 2) how does one perform conversions for
compounds that are activated? Dr. Nicholson wrote that the current under-
standing of 1nterspec1es dose conversion for some endpolnts, for example
cancer, 1s Insufficient for a general policy. He pointed out the strong
Influence of metabolism on such conversions and cited as an example the 1-4
times higher observed risk 1n rats than 1n humans for hemanglosarcoma of the
liver caused by exposure to vinyl chloride. Dr. WHhey also expressed con-
cern that some species have metabolic deficiencies that alter the metabolism
of some classes of compounds such as hexabarbHol and phenols or have
different detoxification "priorities"; for example, rodents use glutathlone
as the primary detoxifying substrate while primates use water and epoxlde
hydrase before glutathlone.
2.1.3. Consensus. For Interspedes conversion of exposure duration, the
consensus was that the lifetime proportionality approach should be used
giving consideration to the type of effect being examined, and that the
1871A -8- 03/21/84
-------
validity of this approach could be determined by experimentation. For
Interspedes conversion of dose, the participants decided that the conver-
sion should be based on mg/body weight to the two-thirds power when Insuffi-
cient data were available to determine a more precise factor; they generally
acknowledged that the Ideal approach would Involve optimization of the
exponent 1n an equation relating dose, species, effect, and body weight
(chemical-specific) or dose, species, chemical, and body weight (effect-
specific).
2.2. HEALTH RISK ASSESSMENT FOR LESS THAN LIFETIME EXPOSURE - TOXICANTS
AND CARCINOGENS
2.2.1. Presentation - Toxicants (M. Dour son). The goal of this subses-
slon was to choose an approach to estimating ADIs for Individual substances
such as 1) exposure durations less than lifetime when subsequent exposure 1s
assumed to be zero, and 2) exposure duration of any length after known
previous exposure. Two options were proposed and Illustrated using data on
methoxychlor and mlrex. In both options, all the available data are plotted
on a graph of dose vs. duration. The dose 1s expressed on a mg/kg basis
adjusted by the dose (mg) per surface area conversion for species differ-
ences; the duration 1s the equivalent fraction of human llfespan, I.e.. the
exposure period of the study 1n days divided by the assumed or published
animal Hfespan times 70 years. The results from the study are assigned a
code corresponding to severity (Table 2-1), which 1s based on currently used
EPA methodology. The data for methoxychlor are presented 1n Figure 2-1.
The "RP," "GR," "SP," and "LV" notations Indicate effects on reproduction,
growth, spleen, and liver, respectively. The size of the symbol Indicates
the relative quality of the study. In Option 1, these data are divided Into
1871A -9- 03/21/84
-------
TABLE 2-1
Various Effect Levels and Their Definitions
Effect Level Symbol
Definition
AEL
PEL
NOAEL
o
NOEL
NOFEL
o
A
Adverse-Effect Level. That exposure level at
which there are statistically or biologically
significant Increases 1n frequency or severity of
adverse effects between the exposed population
and Us appropriate control.
Frank-Effect Level. That exposure level that
produces unmistakable adverse effects, such as
Irreversible functional Impairment or mortality,
at a statistically or biologically significant
Increase 1n frequency or severity between an
exposed population and Us appropriate control.
No-Observed-Adverse-Effect Level. That exposure
level at which there are no statistically or
biologically significant Increases 1n frequency
or severity of adverse effects between the
exposed population and Us appropriate control.
Effects are produced at this level, but they are
not considered to be adverse.
No-Observed-Effect Level. That exposure level at
which there are no statistically or biologically
significant Increases 1n frequency or severity of
effects between the exposed population and Us
appropriate control.
No-Observed-Frank-Effect Level. That exposure
level at which there are no statistically or
biologically significant Increases 1n frequency
or severity of frank effects between an exposed
population and Us appropriate control. Experi-
menters may or may not have looked for other
adverse effects.
1871A
-10-
03/21/84
-------
CD
100,000
10,000
I
1,000
Equivalent
Human
Dose
(mg/d)
100
-------
four areas corresponding to different durations: acute, short-term, sub-
chronic and chronic (Figure 2-2). Highest no-observed-adverse-effect levels
(NOAELs) are chosen 1n each category (Indicated by arrows 1n the figures),
and ADIs are estimated using U.S. EPA methods currently 1n use (Stara et
al., 1981). ADIs for subsequent exposures are estimated using the estimated
length (duration) of the exposure.
In Option 2, an estimate of the visually best-fitting, highest NOAEL
line 1s made and a corresponding ADI line 1s drawn. Figure 2-3 presents
this option as applied to the methoxychlor data. ADIs for future exposures
are determined using the time-weighted average of the previous and subse-
quent exposure levels. The primary assumptions 1n developing these plots
are that equivalent exposure durations for studies of different species are
adequately described by conversion to fractions of expected Hfespans and
that equivalent doses for experimental animals and humans are derived using
a dose equivalence based on surface area.
The advantage of the first option Is that 1t 1s similar to current U.S.
EPA procedures and 1s consistent with the methods used to estimate health
advisories by the Office of Drinking Water. The main disadvantage 1s that
this option does not estimate ADIs using previous exposure Information when
H 1s known, a distinct disadvantage when the previous exposure exceeds the
ADI for the current duration. Option 2 provides more flexibility 1n that
ADIs can be estimated for any duration once the NOAEL line 1s established,
and thus ADIs for .future exposures can be estimated when previous exposures
are known.
1871A -12- 03/21/84
-------
QO
00
I
I
100,000
10,000
1,000
Equivalent
Human
Dose
Ong/d)
10
A
A
ADI
ACUTE
0.07
O
ADI
O
SHORT TERM
CA
ADI
O
SUBCHRONIC
O
CHRONIC
0.7 7.0
Fraction of Lifespan (Yr)
70
o
co
FIGURE 2-2
Methoxychlor Oral Toxlclty Data - Option 1
-------
CO
~J
3»
100,000
10,000
1,000
Equivalent
Human
Dose
(mg/d)
NOAEL
100,1 —
10
o
o o
__ ADI
I
0.07
0.7 7.0
Fraction of Lifespan (Yr)
70
o
03
CO
FIGURE 2-3
Methoxychlor Oral Toxldty Data - Option 2
-------
2.2.2. Discussion - Toxicants. In the discussion, Dr. Schnelderman
pointed out that neither option uses all of the available data, rather only
the NOAEL points are used 1n estimating the ADIs. Dr. Hertzberg replied
that these options use more data than the current approach and allow one to
exercise more judgment, although a statistical method of drawing the NOAEL
and ADI lines 1s preferred.
Several participants provided additional comments 1n their postmeetlng
memoranda. Dr. Hattls noted that, although both options are similar, Option
2 1s preferable because any division or grouping of data leads to a loss of
Information. Dr. Sllbergeld suggested that for some toxic endpolnts,
physiologically similar life stages, such as pregnancy or lactation, be used
1n Option 1 to group the data rather than just the duration of exposure as a
fraction of Hfespan. Dr. Schnelderman wrote that neither option considers
the dose-response relationships observed 1n the Individual experiments, and
that a weakness of the approach 1s that the Identification of a NOEL 1s
highly dependent on the sample size, with the smaller sample sizes leading
to higher NOELs.
2.2.3. Consensus - Toxicants. In general, the consensus was that the
approach 1s acceptable; neither option was Identified as being preferable.
The main reservations were that the approach tended to be subjective and
that some weighting procedure was needed 1n drawing the ADI lines to account
for the quality of the data.
1871A -15- 04/11/84
-------
2.2.4. Presentation - Carcinogens (K. Crump). The goal of the subsesslon
on carcinogens was to Identify a method to estimate human lifetime cancer
risk from exposures of any duration regardless of age at the start of
exposure. The method should also be able to adjust for carcinogens known to
act at specific stages of the multistage process now being used by EPA to
model cardnogenesls. Three options were suggested:
1. Base all lifetime cancer risk estimates for durations less
than lifetime on a ratio of total dose based on exposure to
the total dose associated with a g1ve_n carcinogen potency
) and lifetime cancer level (e.g., 10~s).
2. Use the concept of the "Druckrey effect" (I.e., that lower
dose rates are more effective than higher dose rates 1n pro-
ducing cancer, at a given total dose) In establishing total
doses associated with different dose rates and durations for
estimating the lifetime cancer risk for durations less than
lifetime.
3. Modify the currently used multistage model to allow for esti-
mates of risk from partial lifetime exposure.
Option 1 1s currently used by EPA but does not account for the Druckrey
effect, the age of Initiation of exposure, or the stage at which a carcino-
gen may act. Option 2 accounts for the Druckrey effect but not the age of
Initiation or the carcinogen stage. Option 3 would Incorporate all three.
A method to Implement Option 3 was developed under a contract with ECAO
(Crump and Howe, 1983).
In his presentation, Dr. Crump pointed out that the basis for his proce-
dure 1s the Arm1tage-Doll multistage model of carclnogenesls as refined by
Whlttemore and Keller (1978) and Day and Brown (1980) to account for noncon-
tlnuous exposure and carcinogens that act on different stages. The primary
difficulties 1n the application of his procedure are the selection of the
number of stages and the selection of the stage at which the carcinogen
1871A -16- 03/21/84
-------
acts. Animal studies have provided few data to Identify the number of
stages 1n cardnogenesls, and their applicability to human cancer 1s 1n
question. Selection of the stage at which a carcinogen acts may be based on
the apparent mechanism, I.e., whether 1t acts as an Initiator or as a
promoter.
Or. Crump Illustrated his method using data from the NCI bloassay {NCI,
1978) of ethylene dlbromlde (EDB). The study Involved a pattern of high
mortality and tumor Incidence and an Interrupted pattern of dosing (Figure
2-4). Upon fitting the model using different numbers of stages and stages
of action, 1t appeared that the best fit to the data (the fit giving the
largest likelihood) had the assumption of six stages with the first stage
being dose-related. A comparison of the probability of death from carcinoma
of the forestomach (the primary tumor found) as predicted by the model and
the Kaplan-Meier estimates found good agreement using all of the data
(Figure 2-5) and using the data excluding tumors found at terminal sacrifice
(Figure 2-6), because 1t 1s questionable whether the tumors found at termi-
nal sacrifice should be used 1n predicting the probability of cancer death.
Table 2-2 Illustrates the effect of assuming different stages to be dose-
related. The effect on the doses 1s slight, suggesting that risk estimates
from constant dose rates are not extremely sensitive to the stage that 1s
assumed to be dose-related. Table 2-3 shows the effect of a single week of
dosing at either week 1, 14, or 30. A dose given at week 14 must be twice
as large as the dose given the first week to have the same carcinogenic risk
by week 105, and a dose given during week 30 must be about 5 times as large.
1871A -17- 03/21/84
-------
u, 10-1
« o
40-
ui
151
5-
* 0-
40-
3
IO
O
UJ
J 0
High Dose Animals
Low Dose Animals
10 20 30
Weeks
40 50
FIGURE 2-4
Male Rats Exposed to EDB by Gavage
1871A
-18-
03/21/84
-------
oo
UO
I
o
CO
IV)
00
I.O
.9
E E
e £
H- M
7-
£ £ 6
o
5
1 I 3
O C
* *
High dose:
Armilage- Doll
Kaplan-Meier
Low Dose:
rmitage - Doll
Kaplan- Meier
40
FIGURE 2-5
Comparison of Estimates from Arm1tage-Doll Model with Kaplan-Meier Estimates
(assumed six stages; first stage dose related)
50
-------
CO
u
o
6 i
o *-
t- W
H- 0>
I
ro
o
I
o
E E
x> o
o c
-o 'o
o t:
£• o
Q- o
High dose '
Armitage- Doll
Kaplan -Meier
Low dose:
Armitage - Doll
Kaplan-Meier
Weeks
FIGURE 2-6
Comparison of Estimates from Armltage-Doll Model with Kaplan-Meier Estimates
Excluding Tumors Found at Sacrifice
(assumed six stages; first stage dose related; tumors discovered during week 49 omitted)
-------
TABLE 2-2
Constant Dose Rates Corresponding to Given Levels of Extra Risk
Level of Extra
Model
Probability of death from tumor by
week 49; 6 stages; stage 1 dose
related
Probability of death from tumor by
week 49; 6 stages; stage 3 dose
related
Probability of death from tumor by
week 49; 6 stages; stage 1 dose
related; tumors found during last
week of study omitted
Probability of tumor by week 49;
6 stages; stage 1 dose related
Probability of death from tumor by
week 105; 6 stages; stage 1 dose
related; tumors found during last
week of study disregarded
10'1
3.3
(2.7)
2.5
(2.1)
4.7
(3.8)
0.91
(0.70)
0.048
(0.037)
10~2
0.31
(0.26)
0.24
(0.20)
0.45
(0.36)
0.086
(0.067)
0.0046
(0.0035)
Risk*
10""
0.0031
(0.0026)
0.0024
(0.0020)
0.0045
(0.0036)
0.00086
(0.00066)
4.6 x 10~5
(3.5 x 10~5)
*Max1mum likelihood estimates; 95X lower confidence limits 1n parentheses
1871A
-21-
03/21/84
-------
TABLE 2-3
Dose of 1-week Duration that Correspond to Given Levels of Extra R1ska
Beginning
of Exposure
Week 1
Week 14
Week 30
10'1
0.91b
(0.69)
1.8
(1.4)
4.7
(3.7)
Levels of Extra R1skb
10~2
0.087
(0.065)
0.17
(0.13)
0.45
(0.35)
10~«
0.00087
(0.00065)
0.0017
(0.0013)
0.0045
(0.0035)
aExtra probability of death from cancer by 105 weeks; tumors found during
last week of study disregarded; 6 stages assumed with stage 1 dose-related.
^Maximum likelihood estimates; 95% lower confidence limits 1n parentheses.
1871A -22- 03/21/84
-------
2.2.5. Discussion - Carcinogens. Dr. Kuzmack suggested that for the
Crump approach to be more practical 1t should be simplified, possibly by
reducing 1t to a table by making a set of plausible assumptions about likely
exposure conditions. Or. Nlsbet pointed out this 1s feasible because most
exposed populations will have both age extremes, and so Identification of
the carcinogen as early or late stage-acting may not be critical. Drs.
Thorslund and Andelman noted that the multistage model does not fit all the
sets of data on human cancer, for Instance, leukemia. Dr. Crump acknowl-
edged that the multistage model 1s most applicable to solid tumors, and not
leukemlas, and that Incorporation of pharmacoklnetlc Information would be an
Improvement In his approach. Dr. Nlsbet pointed out that 1f the assumption
Is made that most carcinogens act at the first stage, then the risk for
promoters with late life exposures will be seriously underestimated. Dr.
Nicholson also noted a problem 1n using animal data to develop models for
estimating human risk 1n that animal studies are not designed to determine
the effect of promoters 1n the same environment of Initiators 1n which
humans live. He suggested that animal studies of a suspected promoter
should be conducted with concurrent exposure to Initiators.
Additional discussion was provided by participants 1n their postmeetlng
memoranda. Dr. Hartung believes that the multistage model for cancer Induc-
tion has been accepted without sufficient scrutiny of Its scientific basis.
Dr. Hattls stated that he believes the Crump-Howe approach 1s preferable to
the other two options because at least 1t 1s based on a hypothesized mecha-
nism of action. Dr. Andelman asked 1f the concentration (dose) term 1n the
model can be effectively dealt with when 1t Is not a first order variable,
I.e., concentration to the 1.2 power, which he felt was not an uncommon
situation.
1871A -23- 03/21/84
-------
Dr. Schnelderman made both general and specific comments on the Crump-
Howe approach 1n his memorandum. In general, he noted that animal experi-
ments have thus far been capable of Identifying only two stages {Initiation
and promotion) and recommended that both EPA and NTP should design animal
models capable of distinguishing more stages and at which stage a carcinogen
may be acting. He also stated the Crump-Howe report should be modified to
show the effects of a "remaining lifetime" exposure resulting from modifying
a dose, rather than showing a fixed number of years of exposure. This would
more closely follow the exposure pattern 1n which the U.S. EPA Is Interest-
ed, where exposures are to be reduced. He also suggested constructing
tables from data on carcinogens that appear to act at more than one stage.
Dr. Schnelderman made some specific points about the Crump-Howe paper (see
Appendix to this report) concerning the expression of cumulative dose 1n
some tables, the 100-fold difference 1n week 49 stage 1 dose and week 105 1n
Table 2-2, and the Importance of clarifying reasons for nonllnearHy.
2.2.6. Consensus - Carcinogens. Dr. Crump's approach was considered to
be acceptable 1n principle with some reservations about Us applicability to
cancers for which the multistage model of cancer may not apply. Childhood
leukemia was given as an example. For these cancers, 1t was suggested that
an ad hoc model be developed based on the available human data.
2.3. ADIs BASED ON QUANTAL, CONTINUOUS, OR GRADED DATA
2.3.1. Presentation (M. Dourson, K. Crump). The goal of this session was
to choose a method for estimating ADIs or risk levels for toxicants from
1871A -24- 04/11/84
-------
quantal or continuous toxldty data. EPA's current methodology for estimat-
ing ADIs 1s recommended for use with graded toxldty data. Definitions of
these three classes of effects data are as follows: graded (or ranked) data
are 1n the form of severity of adverse effects (e.g., fatty Infiltration of
the liver, single cell liver necrosis, liver necrosis, liver flbrosls) vs.
dose without reference to the percentage of animals with the given effect.
The data are often considered to be biologically significant by a patholo-
gist, but cannot always be statistically tested. Graded data are frequently
reported 1n the literature. Contlnous data are 1n the form of an Increase
or decrease 1n a measured biological parameter (I.e., body weight) vs. dose.
The data may be expressed only as an overall average of the experimental
animals within a dose group. Quantal effects data are 1n the form of
percentage of animals with a specific endpolnt (I.e., death) vs. dose.
Two methods were proposed for using quantal or continuous toxldty data
for estimating ADIs. The first option 1s to fit a mathematical model to
dose response data and, using statistical confidence limits, extrapolate
downward to doses appropriately safe to humans (I.e., doses corresponding to
risks of 10~5 or less). The advantage of this approach 1s that 1t rewards
good experimentation 1n that experiments with larger numbers of animals
(larger number of quantal or continuous data points) tend to produce more
narrow confidence limits and consequently higher values for the lower
confidence limits of safe doses. It also explicitly takes Into account the
shape of the dose-response curve because a mathematical model 1s fit to all
of the dose-response data. It provides estimates of risk corresponding to
any dose, along with associated confidence limits, and therefore can be
1871A -25- 04/11/84
-------
conveniently used 1n rlsk-benefU analyses. The method does not, 1n prin-
ciple, rule out thresholds because a model that Incorporates a threshold can
be used for the extrapolation.
The chief disadvantage of an extrapolation approach 1s related to the
selection of the models; different models that fit the observed data about
equally well can yield vastly different results when extrapolated to doses
corresponding to very small risks. When used 1n estimating carcinogenic
risks, Information on the nature of the carcinogenic process has been used
1n the selection of a model. Most Information has Indicated that the dose-
response curve 1s apt to be nonthreshold and approximately linear at low
doses whenever a chemical Initiates cancer through a change In the DNA of a
single cell, or whenever background cardnogenesls 1s present. However, for
nongenotoxlc events, a linear nonthreshold dose response seems more
unlikely. Although a linear model still would define upper bounds to low
dose risks, these upper bounds might overestimate the true risk by large
amounts 1n some Instances. Thus, the uncertainty as to the true shape of
the dose-response curve 1n the low dose range for nongenotoxlc effects and
the fact that different models can give vastly different results constitute
a disadvantage to the model extrapolation approach for nongenotoxlc effects.
A second disadvantage to the extrapolation approach 1s that toxlcologlcal
experiments are frequently not designed or reported 1n a manner that facili-
tates the use of model-fitting methods. Frequently, the doses are selected
too far apart to adequately describe the dose-response curve; sometimes the
response data necessary for fitting a model are not reported. However, the
experimental design and reporting of data 1n future toxlcologlcal experi-
ments could be Improved 1f a model-fitting approach were adopted.
1871A -26- 03/21/84
-------
The second option 1s to fit a mathematical model to dose-response data
and, using statistical confidence limits, calculate a lower confidence limit
for the dose corresponding to a preset risk of 10"1 or 10~2; then apply
a safety or uncertainty factor to this dose that reflects the severity of
the toxic effect, the thoroughness of the toxlcologlcal study, and possibly
the mechanism of action. This approach combines features of the current EPA
methodology and Option 1. It shares some of their advantages while avoiding
some of their disadvantages. Like Option 1, but unlike EPA methodology, 1t
takes the shape of the dose-response curve explicitly Into account. It
would reward good experimentation 1n that larger, better designed experi-
ments should yield higher levels at the lower confidence limits and thereby
higher allowable human exposures. Also, 1t would reduce the problem of the
selection of the mathematical model because the estimates from the various
models diverge less If extrapolated only down to a risk of 10"1 or 10~2.
The disadvantages of this approach are that selection of a safety factor
could be considered arbitrary and, like Option 1, Its experimental design
and reporting are often Inadequate.
Dr. Crump presented a summary of work he has done on refining the Option
2 approach. He pointed out that the traditional method of determining a
NOAEL to which safety factors were applied to estimate an ADI has several
limitations: 1} a NOAEL 1s an experimental dose that results from the study
design and does not reflect the entire dose-response relationship; 2) a
smaller sample size, I.e., a less reliable experiment, makes 1t more likely
to find a NOAEL; 3) Identifying a NOAEL 1s difficult, especially with
continuous data; and 4) some studies are adequately conducted but do not
have a NOEL and therefore are not used. In his approach, Dr. Crump proposed
1871A -27- 03/21/84
-------
applying one or more models to experimentally derived dose-response data
(either quantal or continuous) to Identify a "benchmark" dose. The bench-
mark was defined as a lower statistical confidence limit for the dose corre-
sponding to a specified Increase 1n the level of health effects over the
background level. To Illustrate this approach, four models (quantal linear
regression, quantal polynomial regression, quantal Welbull, and log-normal)
were applied to five sets of quantal data (Table 2-4). The results (Table
2-5) Indicated good fits (p-value close to 1) for the models with the excep-
tion of quantal Welbull and log-normal as applied to the botullnus toxin
data, Implying that the mechanism of action of the toxin Involves a thresh-
old. When maximum likelihood estimates (MLE) and lower confidence limits
were calculated for the different models and levels of risk for one data set
(Table 2-6), 1t can be seen that these estimates diverge at the lower levels
of risk and yield similar results at the higher levels. Dr. Crump stated
that he has applied the approach to continuous data with satisfactory
results but did not have time to present them. As a summary, he displayed a
comparison of benchmark doses and NOELs for the five data sets (Table 2-7)
and pointed out that for four sets the NOEL was between the estimated doses
corresponding to 1% and 5% extra risk and that, for the fifth set, his
procedure determined a benchmark dose when no NOEL had been Identified.
2.3.2. Discussion. Dr. Hartung commented that for any approach to fully
utilize all the data and be applicable to all types of responses 1t must be
multidimensional. As an example of data for which the proposed approach may
underestimate risk, he selected a continuous response, such as a change 1n
blood pressure, 1n which a small proportion of the population responded at a
low dose of a chemical with severe effects. Other participants also were
1871A -28- 03/21/84
-------
TABLE 2-4
Quantal Data Used to Illustrate Quantitative Dose Response Methodology
Ethylenethlourea (ETU) (Khera. 1973)
Fetal Anomalies 1n Rats
Doses (mg/kg): 0 5 10 20 40 80
No. affected/total no.: 0/167 0/132 1/138 14/81 142/178 24/24
2.3.7.8-Tetrach1orod1benzo-p-d1ox1n (TCDD) (Khera and Ruddlck. 1973)
Intestinal Anomalies 1n Rat Fetuses
Doses (ug/kg): 0 0.125 0.25 0.5 1.0
No. affected/total no.: 0/24 0/38 1/33 3/31 3/10
2.3.7.8-Tetrachlorod1benzo-p-d1ox1n (TCDD) (Hurray, et al. 1979)
Rats Dead at Birth
Doses (ug/kg/day): 0 0.001 0.01
No. affected/total no.: 22/318 16/224 17/100
Hexachlorobenzene (HCB) (Khera. 1974)
14th R1b Anomaly 1n Rat Fetuses
Doses (mg/kg): 0 10 20 40 60
No. affected/total no.: 0/80 4/79 8/91 15/87 25/96
Botullnus toxin - Type A (Food Research Institute. Univ. of Wisconsin;
see Food Safety Council. 1978)
Death Due to Botullnus
Doses (ng):
No. dead/total no.:
Doses (ng):
No. dead/total no.:
0.01
0/30
0.034
11/30
0.015
0/30
0.037
10/30
0.020
0/30
0.040
16/30
0.024
0/30
0.045
26/30
0.027
0/30
0.050
26/30
0.030
4/30
1871A -29- 03/21/84
-------
TABLE 2-5
Summary of Fits of Models to Quantal Data 1n Table 2-4
Data
ETU
TCDD
TCDD
HCB
Botullnus toxin
*QLR = Quantal linear
Model*
QLR
QPR
QW
LN
QLR
QPR
QW
LN
QLR
QPR
QW
LN
QLR
QPR
QW
LN
QLR
QPR
QW
LN
regression (w1
QPR = Quantal polynomial regression
QW = Quantal Welbul
LN = Log-normal (w1
Ch1-square
17
0.0
1.3
0.46
0.17
0.014
0.32
0.23
0
0
0
0.11
0.09
0.11
0.31
7.0
4.4
162
159
th a threshold)
(with a threshold)
d.f.
3
2
2
2
2
1
3
3
2
1
2
2
8
7
8
8
p-value
0.0007
1.0
0.73
0.93
0.92
0.91
0.85
0.89
0.95
0.76
0.95
0.86
0.54
0.73
0.00001
0.00001
1 (without a threshold)
thout a threshold)
1871A
-30-
03/21/84
-------
TABLE 2-6
Doses Corresponding to Given Levels of Extra Risk for Quantal ETU Data
Model Extra Risk
QLR 0.1
QPR
QW
LN
QLR 0.05
QPR
QW
LN
QLR 0.01
QPR
QW
LN
QLR 1 x 10~6
QPR
QW
LN
MLEa
12.2
16.4
17.9
17.1
11.0
13.2
14.5
14.8
10.1
10.2
8.9
11.2
9.9
9.4
5.9-1
4.2
Doses (mg/kq)
95X Lower
11.8
13.9
15.7
15.3
10.6
11.6
12.2
13.0
9.7
7.2
7.0
9.5
9.5
4.0-1
2.9-1
3.1
99% Lower
11.6
13.2
14.7
14.6
10.4
11.0
11.3
12.2
9.5
6.0
6.2
8.8
9.2
1.5-3&
2.1-1
4.7
aMax1mum likelihood estimates
b1.5 x 10"3
1871A
-31-
03/21/84
-------
TABLE 2-7
Comparison of Benchmark Doses with NOELs for Quantal Data
Benchmark Doses*
Corresponding to Extra Risk of
Data Set
ETU
TCDD
TCDD
HCB
Botullnus toxin
Units
mg/kg
ug/kg
ug/kg/day
mg/kg
ng
NOEL
5
0.125
1.0-3
ND
0.027
10%
13.9
0.32
5.3-3
17.4
0.029
5%
11.6
0.22
2.6-3
8.5
0.027
1%
7.2
0.049
5.1-4
1.7
0.025
*95X lower limits derived from QPR model
ND = Not determined
1871A
-32-
03/21/84
-------
concerned that the approach may Imply the existence of a threshold below
which no effect would be expected, especially considering the wide variation
1n Individual thresholds 1n the human population (Kuzmack, Nelson, Hattls).
Dr. Crump responded that the models he used Incorporating thresholds were
found to fit some data sets the best, and that his method was not designed
to determine a threshold but a dose that corresponds to a low level of risk.
Dr. Sllbergeld and others (Nlsbet, O'Flaherty) felt that more discussion was
needed on the application of the model to continuous data and on the conver-
sion of one type of data to another, I.e., continuous to quantal. Or. Stara
suggested that Dr. Crump's complete paper be distributed by ECAO to the
participants to provide the details of the applicability of the approach to
quantal and continuous data.
Some of the participants provided written comments on the proposed
approach 1n their postmeetlng memoranda. Dr. Schnelderman noted that the
benchmark approach (Option 2) has the same difficulties as those of modeling
cardnogenesls data, for example, how to choose a model that accurately
reflects the data Including the presence or absence of an observed thresh-
old. Dr. Sllbergeld also cited difficulties with the proposed benchmark
approach 1n that the wide variation 1n the severity of effects would compli-
cate the selection of the safety factors that are applied to the benchmark
values. Dr. Nlsbet wrote that for "truly QCG [quanta!, continuous, graded]
data" derivation of an ADI 1s Inappropriate because such data can often be
fitted to a nonthreshold model.
Dr. Hattls pointed out 1n his memorandum that a major problem 1s that
the steepness of the dose-response curve depends on the genetic diversity of
1871A -33- 03/21/84
-------
the exposed population and therefore animal data are likely to produce a
steeper curve than that expected for humans and lead to an underestimation
of risk. He also proposed modifying Option 2 by the use of different safety
factors for different effects depending on the likely distribution of
susceptibility 1n the exposed human population and on policy judgments about
the severity and socially desirable protective posture for different
effects. Dr. O'Flaherty also suggested a modification to make the approach
relatively model Independent. When working with quantal data, she proposed
the calculation of a confidence limit on the dose corresponding to a speci-
fied low risk (I.e., 10~2) and then the application of appropriate
uncertainty factors. She also pointed out that 1n the application of the
approach to quantal data, which vary from 0 to 1.0, selection of the
uncertainty factors should take this variability Into account.
2.3.3. Consensus. The consensus of the participants was that Dr. Crump's
benchmark approach was conceptually acceptable but that no decision could be
made on how to go from a benchmark dose to an ADI, which would be certain to
provide adequate protection for all members of a human population. Addi-
tional consideration of the applicability of the approach to continuous data
was also Indicated.
2.4. PHARNACOKINETIC APPROACH FOR ROUTE-TO-ROUTE CONVERSION
2.4.1. Presentation (W. Pepelko, J. WUhey). Dr. Pepelko proposed that
consensus be sought 1n three areas. First, that the primary routes of
concern for human exposure are Inhalation and oral and that data from animal
studies using different routes be used when such factors as absorption
1871A -34- 03/21/84
-------
efficiency, dose frequency, and hepatic metabolism are known and taken Into
consideration. Second, that extrapolation, especially from oral to Inhala-
tion and vice versa, 1s feasible under the following conditions: 1) 1f the
portal of entry 1s not the target organ, 2) 1f the toxicant chemical 1s not
totally Inactivated before reaching the target organ, and 3) 1f elimination
rates are sufficiently slow that blood levels do not fluctuate greatly
during periodic dosing. Third, that the extrapolation procedure account for
any factor that affects the concentration of the toxicant at the target
organ and 1s different for different routes, Including absorption efficiency
and the degree of activation or deactlvatlon before reaching the target
organ. Dr. Pepelko stated that the primary considerations are the condi-
tions under which the results from one route of exposure can be converted to
another and how that conversion 1s to be accomplished.
In his presentation, Dr. WHhey reviewed graphical comparisons of the
dose levels 1n blood achieved by Inhalation exposure, 4 oral doses, and 20
oral doses 1n a one-compartment pharmacoklnetlc model 1n which total uptakes
for each exposure were equal. In the first case, where a short elimination
time 1s assumed (Figure 2-7), 1t was evident that the four oral doses lead
to much higher blood levels, which could greatly affect the type or severity
of the effects observed. In the two other cases, 1n which Intermediate
elimination rates (Figure 2-8) and very long elimination rates (Figure 2-9)
were assumed, the blood levels for the three administrations were seen to
converge. From these data, Dr. WHhey concluded that Increasing the fre-
quency of the dose gives a closer approximation of oral uptake to steady-
state Inhalation systemic levels, and fast elimination rates (short t )
give larger fluctuations about the steady-state level.
1871A -35- 04/11/84
-------
Dose Interval
00
I
CO
Case 1
TIME (hr)
o
CO
ro
oo
FIGURE 2-7
Hypothetical Blood Concentrations for Three Exposure Conditions:
Inhalation, 4 Oral Doses, and 20 Oral Doses
-------
CD
I
CO
Dose Interval
I
3>
30
§M
DC
Z
Ul
o
/"\
1 \
1
1
1
m
\
\
f
!\
l\
Case 2
Dose
Interval
8
CD
1
) 2
1
4
1
6
1
8
1
10
I
12
14
TIME (hr)
o
CO
00
FIGURE 2-8
Hypothetical Blood Concentrations for Intermediate Elimination Rates
-------
-8C-
VU81
BLOOD CONCENTRATION kg/ml)
DO
o
o.
o
o
(Jl
o
=3
I/I
O
3
(O
m
I
vO
on
TO
Oi
<*»
o
-------
In the text of his presentation, Or. WUhey also stated: "It 1s worth-
while to point out that, where a compound 1s absorbed relatively rapidly and
eliminated slowly, the dose Interval and regimen do not Influence the
systemic concentration or their temporal relationship, as shown 1n case 3.
In such circumstances 1t 1s the amount rather than the kinetics of systemic
uptake and elimination that would be expected to Influence the systemic
toxldty. In these cases, provided that the Inhalation TLV 1s soundly based
(I.e., there 1s little Involvement of first pass effects, organohepatlc or
site of uptake effects are not the Index of toxldty, etc.) and absorption
coefficients are known, there would appear to be no reason why the oral
equivalent could not be calculated using the Stoklnger-Woodward approach."
Or. Wlthey pointed out that at the previous meeting, Or. O'Flaherty had
discussed an Important consideration 1n route-to-route conversions: the
special case of the deposition of aerosols and partlculates 1n the lung.
Deposition varies with particle size but, once deposition occurs, uptake (at
least of aerosols) 1s remarkably similar, regardless of particle size. The
amounts cleared from the lung are not proportional to lung deposition.
These considerations are Important 1n the assessment of certain kinds of
pulmonary uptake and 1n addressing special mechanisms associated with route.
Dr. Wlthey also mentioned the Importance of the first-pass effect
because 1t 1s very much dependent on route. "Substances Injected Intra-
venously or absorbed via the lung are carried, and distributed, more or less
equally to the vessel-rich organs of the body. Substances administered
orally pass from the gastrointestinal tract via the portal system and Into
the liver. Thus, for substances that are subject to extensive hepatic
1871A -39- 03/21/84
-------
metabolism, the effective dose reaching the systemic circulation 1s consid-
erably reduced (1n the case of Udocalne, for example, 100% of an oral dose
1s metabolized on first pass through the liver)." Similarly, compounds that
are activated by the liver may have exposure route-specific patterns of
toxldty because of the first-pass effect.
Dr. WHhey reviewed the need for route-to-route conversion methods.
"First, the absence or lack of adequate studies for the route of Interest 1n
humans may necessitate the use of pharmacoklnetlc principles for an estima-
tion of toxic Insult. This will Include experiments that have been poorly
designed or executed, 1n terms of dose, duration, numbers, etc., or practi-
cal considerations like low aqueous solubility, which preclude toxldty
studies 1n drinking water, the use of an Inappropriate dosing vehicle, for
example, vegetable or mineral oil as opposed to water, and animal data where
the pharmacoklnetlc rates and mechanisms are species-, sex-, or strain-
dependent and extrapolation 1s precluded." Second, potentially useful data
may be available from human or animal studies 1n which the route has been
Intravenous, subcutaneous, Intramuscular, IntrapeHtoneal, etc. He pointed
out that the data from such studies may be acceptable 1n every other sense
and, 1f there 1s no reason to suppose that metabolic processes or other
factors are affecting the blood levels, the data should be extrapolated to a
more conventional route. However, he noted that enormous amounts of data
may be needed 1n order to derive a defensible estimate.
1871A -40- 04/11/84
-------
Dr. WHhey also cited the following as Information requirements for
undertaking the application of any route extrapolation methodology:
1. Nature of the pharmacoklnetic model.
(one-, two-, or muIt1compartment, M1chael1s-Menten)
2. Nature of exposure.
(Dose and dose Interval/duration)
3. Absorption efficiency.
4. Hybrid rate coefficients for absorption and elimination and
their variation with dose.
5. First-pass effect (magnitude).
6. Are metabolites the active toxicant?
7. Are metabolic pathways the same?
8. Are metabolic rates the same?
9. Is systemic concentration proportional to toxic effect?
10. Potent1at1on/antagon1sm on repeated dosing.
Dr. WHhey pointed out that 1n his three original examples where he made
fairly realistic assumptions (100% uptake, uptake the same for all routes,
absorption rate equal to 10 times the elimination rate), the acceptability
of route extrapolation would be poor, questionable, and excellent, respec-
tively (Table 2-8). Much would depend, however, on the specifics of partic-
ular exposure patterns, physlcochemlcal properties, etc. He stated that
"route extrapolation may thus be reduced from a generalized to a case-by-
case consideration, particularly where mechanisms, species differences, and
other parameters, like those considered above, are Involved. It would, at
this stage, be unfair and Inappropriate to put precisely defined limits on
the application of pharmacoklnetic principles to the question of route
extrapolation."
1871A -41- 04/11/84
-------
TABLE 2-8
Magnitude of Rate Coefficients (t-|/2)
Absorption
minutes
Elimination
(t]/2) minutes
Extrapolation
Ability
Case 1
Case 2
Case 3
0.9
6.9
144.0
9.0
69.0
1440.0
Poor
Questionable
Excellent
1871A
-42-
03/21/84
-------
Dr. WHhey displayed the equation he uses to calculate blood levels 1n
his one-compartment model and pointed out that the critical factor 1s know-
Ing the absorption efficiency, which In turn requires a great deal of data
to determine. He said, "It has been suggested that, where the site of
uptake, or portal of entry, 1s the primary target tissue (as H Is 1n the
case of pulmonary damage arising from the Inhalation of manganese or chro-
mium), the application of either pharmacoklnetlcs or route extrapolation 1s
precluded. In the case where the target organ 1s the same for both Inhala-
tion and systemic [gastrointestinal] uptake, as 1t 1s 1n the case of
paraquat, a special consideration of the relative Insult arising from the
administration of equivalent doses would be a rewarding test of the strength
and the value of pharmacoklnetlcs as a predictive tool." Finally, Dr.
WHhey pointed out that the choice of the pharmacoklnetlc model 1s also
Important; the addition of compartments to the model or an Increase 1n
elimination rate can greatly alter the blood profiles. Another complicating
factor 1s that the pharmacoklnetlcs may be changed by the test substance;
for example, the steady state blood levels Increase or decrease considerably
1n the simple model when the elimination rate 1s Increased or decreased,
respectively, during the administration.
2.4.2. Discussion. Dr. Stara pointed out that the assessment of exposure
to partlculates 1s complicated by the observation that Inhaled materials are
often cleared from the lungs and swallowed, resulting 1n gastrointestinal
absorption, but that the Stoklnger-Woodward approach should work for Inhaled
solvents. Dr. WHhey noted that this approach may not apply to highly vola-
tile solvents and that the Important factor 1s the absorption efficiency.
1871A -43- 03/21/84
-------
Or. Plscator stated that the Stoklnger-Woodward approach should be used with
caution 1n determining ADIs because H 1s based on threshold limit values
(TLVs) and not directly on the data from which the TLVs are derived. Dr.
Stara agreed that some values have limited documentation. Dr. Hartung made
the point that although 1t 1s Important to know 1f the substance 1s having
an effect at the portal of entry, these data are usually not collected. Dr.
O'Flaherty noted that Dr. WHhey's analysis of the variation 1n blood levels
from different routes also Indicated that different levels result from
Intermittent dosing of different frequency within the same route.
2.4.3. Consensus. No consensus was sought at this point. Additional
discussion and an attempt to achieve consensus occurred 1n the workshop on
Multiple Route Exposure, which 1s summarized 1n Section 4.2.
2.5. MULTIPLE CHEMICAL ASSESSMENT
2.5.1. Presentation (R. Hertzberg, P. Durkln). The goal of this session
was to choose an approach to evaluate health risks from exposure to a
mixture of chemicals 1n the absence of Information on their possible Inter-
action. Two options were presented for discussion. In the first, the risk
assessment 1s based on a single chemical, the marker chemical, which 1s
believed to present the greatest hazard as determined by Us toxldty and by
the ratio of the actual exposure to the current regulatory criterion value.
When no criterion has been established, the lifetime threshold concentration
(the level considered acceptable for lifetime exposure) or the 10~s risk-
specific concentration 1s used. In this option, the assessment 1s a rough
estimate of risk and 1s unlikely to be overlnterpreted; however, only a
small part of the available data 1s actually used 1n the assessment.
1871A -44- 03/21/84
-------
The second option and Us background was presented by Or. Durkln, who
stated at the outset that this approach 1s a "default position," I.e., one
that he proposes 1n the absence of data on how chemicals Interact at low
doses. He pointed out that there are two general approaches to assessing
the effects of mixtures of chemicals: dose addition and response addition.
In dose addition, the assumption 1s, 1f the dose-response relationship 1s
known for the substances, then one can be treated as a dilution or concen-
tration of the other depending on Its relative potency, which 1s a ratio of
equltoxlc doses. Figure 2-10 shows how the doses of two chemicals (y and
y ) are combined Into a single expression utilizing potency. In response
addition, the expected response 1s assumed to depend on the response caused
by toxicants 1 and 2 alone and on the degree of correlation of the toler-
ances to the two chemicals within the exposed population. Hence, 1f the
animals that are sensitive to 1 are also sensitive to 2 [I.e., the correla-
tion r 1s 1 and the proportion of the animals responding to 1 alone 1s
greater than the response to 2 (P. > PJ], then the response to the
combined exposure 1s equal to the response of 1 (Pg = P^ Equation 1,
Figure 2-11). If the correlation 1s zero, equation 3 1s used, and 1f there
1s a perfect negative correlation of equltoxlc doses of tolerances, the
proportion responding 1s equal to the sum of the Individual responses
(Equation 4, Figure 2-11). These relationships can be displayed graphically
on an Isobole diagram (Figure 2-12).
As an example of how the different assumptions of dose or response addl-
tlvlty can produce different expected responses, Dr. Durkln gave the example
1n Figure 2-13, 1n which the proportion of the population responding to a
mixture of equltoxlc doses of toxicants 1 and 2 ranged from 20-53X. He also
1871A -45- 04/11/84
-------
y1 = an + B log z1
V2 " a2 + B '°9 *2
Y = a1 + B log (z1 + pz2)
p = POTENCY : RATIO OF EQUITOXIC DOSES
logp = (a2 - an) + B
FIGURE 2-10
Dose Addition
1871A -46- 03/21/84
-------
1) P3 = P1 IFr = 1 AND?! > P2
2) P3 = P2 IFr = 1 ANDP2 > P,
3) P3 = P1 + P2(1 - P^ IFr = 0
4) P3 = P1 + P2 < 1 IFr = -1
FIGURE 2-11
Response Addition
1871A -47- 03/21/84
-------
CD
DOSE OF TOXICANT A
o p p
8 § 8
I . I , I j L
co
o
c
CO
I
CD
fu
Q.
V)
O
cr
o
(D
OJ
3
C
ft)
TO
(0
I/)
(0
Cu
r+
fu
CT>
I
PO
RESPONSE ADDITION (r = 1!
\
\
«
\
\
k
\»
O
o
o
CO
PO
ro
O
O
00
-------
YI = 0.3 + 3 log 19.3
y2 = 1.2 + 3 log 9.7
EXPECTED RESPONSE
DOSE ADDITION 53%
RESPONSE ADDITION 20% (r = 1)
36% (r = 0)
40% (r = -1)
FIGURE 2-13
Example of Dose and Response Addition
1871A -49- 03/21/84
-------
pointed out that as the expected responses get smaller, the differences
between the predictions also become less with the exception that, as long as
equltoxlc doses are used, a completely negative correlation of responses
will always produce a combined response two times the combined response of a
completely positive correlation.
Dr. Durkln also reviewed models of Interactions he had presented at an
earlier meeting (Figure 2-14). Equation 1 1s a rearrangement of the model
for dose addition and does not show any Interaction. In Equation 2, the
Interaction of the two chemicals 1s accounted for by the K coefficient which
1s a constant, p 1s the relative potency, and ir1 and »_ are the
proportions of toxicants 1 and 2 1n the mixture. Equation 3 was developed
by Or. Durkln to model Interactions 1n which the Interaction may change with
dose and therefore has two Interaction coefficients. Dr. Durkln also
pointed out that the use of these models 1s limited by the lack of data to
determine the Individual toxicant's dose-response relations and the Inter-
action of the toxicants. He stated that his survey of the literature on
chronic effects found very little data on Interactions and that this was
clearly not the approach to use.
The approach recommended by Dr. Durkln (Option 2) consists of separate
procedures for noncardnogens and carcinogens. For noncarclnogens, a
"hazard Index" (HI) 1s estimated for each chemical by the sum of the
exposure levels (E) for each route (0 = oral, I = Inhalation, D = dermal)
weighted by the Inverse of the respective lifetime threshold concentrations
(Th):
HI . E0/ThQ + Ej/Thj + ED/ThQ
1871A -50- 03/21/84
-------
1) y = a1 + BloglTr., + p7T2 ) + B log z
2) y = a1 + B log [TTI + pn2 + K (prr., 7r2)°-5 ] + B log z
3) y = a1 + B log [TTI
K2 7r2 (p?r1 7r2)°-5]+ B logX
FIGURE 2-14
Models for Interaction
1871A -51- 03/21/84
-------
The total hazard Index (HIT) 1s estimated by the sum of the Individual
chemical Indices and, when greater than one, 1s considered to be a cause for
concern:
HI,. = HI, + ... + HI for n chemicals
For carcinogens, the procedure 1s a generalized additive response model
assuming a zero correlation of tolerances, I.e., that the risks are Indepen-
dent. For each chemical, the proportion of responders 1n a population 1s
determined by adding the risks from each route of exposure. Then, the total
risk for a mixture of n chemicals Is as follows:
n
R T= 1 - * (1 - Ri)
1=1
where R 1s the risk by all routes for 1. The merits of this option for
both carcinogens and noncardnogens are that all routes and chemicals con-
tribute to the risk estimate and hence more data are used than 1n Option 1.
The assumptions of Independence of action and addltlvHy of effect, however,
probably oversimplify the actual situation. In addition, Or. Durkln noted
that 1t 1s not certain that the types of Interaction mechanisms observed at
high doses are applicable to low doses, and thus, much more Information 1s
needed to verify this approach.
2.5.2. Discussion. The primary points of discussion were 1) whether all
effects could be considered equivalent, since values of HI based on differ-
ent effects are summed to derive HI (Sllbergeld, Lappenbusch), and
1871A -52- 04/11/84
-------
2) whether subthreshold effects of different chemicals are additive
(Hartung, Plscator). Dr. Plscator made the point that 1f sufficient data
were available to predict Interactions, then enough data would be available
to do a chem1cal-by-chem1cal assessment and a possibly misleading total 1s
not needed. Other participants noted that, 1n general terms, 1t 1s likely
that the total effect of a mult1chem1cal exposure will be the sum of
Individual effects (Schnelderman) and that Individuals 1n a population will
display different susceptibilities to chemicals (Nelson). Or. Andelman
expressed concern that for the noncardnogens 1n the additive model, a total
of 1 or more was cause for concern while lower values were not and suggested
that the approach be used as a relative ranking of risks. Dr. Hartung stat-
ed that the approach should be termed "the best speculation" at this time.
In their postmeetlng memoranda, several participants expressed support
for the add1t1v1ty approach as the "least unreasonable" or "most defensible"
method, especially 1f H 1s used to establish a relative ranking of risks
for a number of sites (Hattls, O'Flaherty, Clarkson, Durkln). Drs.
O'Flaherty and Durkln pointed out that this approach 1s the most conserva-
tive 1n that 1t yields the largest estimated risk. Dr. Schnelderman accept-
ed the add1t1v1ty approach as an Interim solution and suggested that for
cancer effects the summation be limited to cancers at the same site or 1n
tissues that have a similar embryonic origin. Other participants cited
Inadequacies of using an additive method Including Its failure to account
for possibly large Interactive effects (WHhey) and multiple tumors result-
Ing from exposures to carcinogens (Blngham, Suter). Dr. Hartung advocated
more research into Interactions between chemicals at low doses and chronic
exposures, and Dr. Sllbergeld suggested that the Agency may be better off in
some Instances by basing the assessment on a single chemical, I.e., dioxin.
1871A -53- 03/21/84
-------
2.5.3. Consensus. No consensus was clearly articulated on this topic,
although a number of participants felt that the addltlvlty method was an
adequate Interim approach. Two recommendations were made to the Agency:
1. Collect more data on chemical Interactions, especially at low
doses and chronic exposures, to better define the addltlvlty
model.
2. Consider that the application of the hazard Index approach be
limited only to chemicals or classes of chemicals causing a
similar effect.
1871A -54- 03/21/84
-------
3. PRESENTATION OF NEW TOPICS (Session II)
3.1. APPROACHES USING STRUCTURE-ACTIVITY RELATIONSHIPS
3.1.1. Presentation (R. Bruins, K. Ensleln, S. BrodeMus). Current U.S.
EPA methods for estimating toxicant effect levels 1n humans depend upon the
availability of data, however limited, on mammalian toxldty. The principle
of analogy, where untested members of a chemical class are treated Identi-
cally to tested members, has also been used but has been criticized as over-
simplified. In this session, the use of structure-activity relationships
(SARs) to estimate effect levels (I.e., NOAELs) and endpolnts (I.e., cancer)
for untested chemicals was discussed. Presentations described the use of
equations based on substructural fragment descriptors 1n the prediction of
various effects and the use of physlocochemlcal properties to predict aquat-
ic toxldty. Participants were asked to consider the potential applicabil-
ity of these approaches to the overall risk assessment methodology. Aspects
of particular Importance are the amount of Information needed to Implement
such an approach, and the reliability of the result 1n terms of accuracy
(how well 1t works) and biological rationale (why 1t works).
Dr. Ensleln presented an overview of the development and application of
SARs to risk assessment. He defined the approach as an attempt through
statistical modeling to explain or model biological responses with equations
using parameters that describe known structures and their biologic activity.
Historically, SAR has been used to optimize desired characteristics of
Pharmaceuticals and pesticides or to elucidate mechanisms of action of
closely related chemicals. Two general approaches are used 1n SAR. The
1871A -55- 04/11/84
-------
first uses molecular orbital calculations to precisely describe the struc-
ture of the chemicals; 1n the second, empirical equations are derived using
parameters that generally describe a substance's chemical and physical prop-
erties. In the second approach, the usual equation has the following form:
Biological response = Constant + (Coefficient x Parameter ) * (Coeffi-
cient,, x Parameter-) ... + (Coefficient x Parameter )
The parameters are octanol/water partition coefficients, electronic con-
stants, Indicator variables, substructures, and other physical parameters.
Ideally, one would wish to develop equations for a series of related chemi-
cals; however, few series have sufficient data, one of which Is the poly-
aromatic hydrocarbons. Dr. Ensleln also cited some data on the performance
of the equations he has developed for Identifying mutagens, carcinogens, and
teratogens (Table 3-1). He pointed out that the performance tests were not
conducted with Independent data; I.e., some or all of the data were used 1n
the development of the equations. He also noted that the effect of the size
of the database on equation performance was tested for the rat oral LO
equation; Increasing the database from 1500 to 2000 chemicals did not
Improve the equation's performance. Finally, Dr. Ensleln reviewed disadvan-
tages 1n the SAR approach. The toxlcologlcal activity for only 60-90% of
compounds can be estimated because the remainder have unique nondescrlbable
structures or do not have sufficient data on the parameters used 1n the
equations. In addition, all discriminant analyses have a gray area In which
no conclusion can be made from the discriminant value (Dr. Ensleln
arbitrarily Identifies values of 0.3-0.7 as Indeterminate). Also, SAR 1s
not currently applicable to mixtures and needs more verification to Increase
Us credibility.
1871A -56- 03/21/84
-------
TABLE 3-1
Performance of SAR Equations
Number of False False
loxlc Endpolnt Compounds Tested Positive Negatives Accuracy
(X) (X) (X)
Mutagenldty 472 5 5 90
Cardnogenlcity 334 6 4 90
Teratogenlclty 430 5 7 89
1871A -57- 03/21/84
-------
The second presentation was made by Dr. Broderlus, who discussed the
development of SAR at the EPA Environmental Research Laboratory (Ouluth) and
his research Into the joint action of multiple organic toxicants. The
Initial SAR work Involved the calculation of bloconcentratlon factors 1n
aquatic animals for a variety of organic chemicals from one parameter, log P
(the octanol/water partition coefficient). The results, from -120 compounds
tested 1n trout, showed a linear relationship between log P and bloconcen-
tratlon; however, predictions made for compounds not used 1n the Initial
modeling were accurate only within 1 order of magnitude.
A second phase of the work was to develop models to predict acute
toxldty of Industrial compounds to aquatic organisms based primarily on two
parameters, log P and water solubility. Approximately 200-300 compounds
Including organic solvents, ethers, and chlorinated benzenes were used to
construct and verify the models. The results were that these compounds
could be grouped Into three classes that could be modeled with varying
success. Compounds of low toxldty, which were also termed "nonspecific"
toxicants because their primary effect was narcosis ("a reversible state of
arrested activity of various protoplasmic structures"), fit the model well,
showing good correlation between toxldty and log P and solubility.
Compounds of moderate toxldty showed some correlation, and highly toxic and
specific compounds had toxldtles that showed little relationship to the
modeling parameters.
Dr. Broderlus 'also presented his research Into the joint action of
multiple toxicants 1n aquatic species, primarily the fathead minnow. He
stated he has limited his study to strictly additive joint action as
1871A -58- 03/21/84
-------
Initially described by Bliss and others. Using various binary combinations
of ketones, such as 2-octanone and acetophenone, and alkyl alcohols, he
found that these compounds displayed concentration addition in their
effects. Other compounds, such as 2,4-pentanedione, 2-chloroethanol, and
benzaldehyde, did not show additive effects when tested in combination with
previously tested chemicals, and thus 1t was proposed they had different
modes of acutely toxic action.
3.1.3. Discussion. Dr. Hartung noted that although SAR can be useful,
some compounds are not suitable for the SAR approach and cited as an example
the fluoro acids, which interfere with intermediate metabolism. In general,
he felt that SAR 1s not applicable to compounds that are similar to bio-
logically active materials, I.e., compounds that play a role In cellular
physiology. Dr. Enslein responded that once a mechanism of action is known
for a toxicant, an SAR evaluation can be developed. Dr. Stara stated that,
if a mechanism of action were known, SAR would probably not be needed to
conduct a risk assessment.
Three participants commented on the presentations of Enslein and
Broderius in their postmeeting memoranda. Drs. Durkin and Sllbergeld noted
that the research presented by Dr. Broderius indicated that his SAR models
worked best for compounds of low toxiclty and that these compounds were of
least interest for regulation and human risk assessment. Dr. Schnelderman
1871A -59- 03/21/84
-------
stated that SAR holds promise and raised the following questions and sugges-
tions 1n regard to the Ensleln SAR approach:
1. In the discriminant equation, what parameters contribute most
to separating materials later labeled as carcinogens from
those labeled noncardnogens? Are these parameters different
for different classes of chemicals? If so, how are classes of
chemicals defined, or determined 1n advance?
2. What kind of self-learning process 1s built Into the Ensleln
procedure? How have the parameters changed with additional
data? How are missing data handled?
3. If comparisons of Ensleln equations with other SAR approaches
and with such testing procedures as short-term mutagenesls
tests were made, these comparisons would {or should) Include a
proposal for stepwlse testing (evaluation) of potential
carcinogens. Can the Ensleln process be set up with cutoff
points to Identify compounds that we can be sure will not be
carcinogens? Dr. Schnelderman noted that, 1f this could be
done, a great deal of testing might be eliminated.
4. Is there some way of adding electrophlle Information to the
existing Ensleln equations? If done, does 1t Improve the
discrimination?
3.2. USE OF REPRODUCTIVE EFFECTS AS ENDPOINTS IN RISK ASSESSMENT
3.2.1. Presentation (J. Stara, J. Hanson). ECAO Is faced with the ques-
tion of how appropriately to use fetotoxlc, teratogenlc, and, In general,
reproductive effect data 1n human health risk assessment. Although such
data are limited, they need to be evaluated and considered 1n the estimation
of potential hazard. Much additional basic and applied research must be
done to provide a satisfactory data base for these purposes. Furthermore,
the extrapolation process using reproductive effects 1s complicated because
of species differences that may Include different placentae, placental
transfer rates, neonatal developmental stages, and lengths of pregnancy
among others.
1871A -60- 03/21/84
-------
Dr. Hanson Introduced her presentation by stating that, at present,
there 1s little experimentally based guidance for the use of reproductive or
prenatal toxldty endpolnts 1n risk assessment. In addition, she said a
wide range of endpolnts 1s possible as a result of a single exposure to a
teratogen during development. To Illustrate her points, Dr. Manson present-
ed data on two teratogens that act through very different mechanisms. The
first, methyl-N-n1tro-N-n1troguan1d1ne, 1s a direct-acting alkylatlng agent
that specifically kills S-phase, or rapidly proliferating cells and, as a
result, Induces 11mb malformation. Her research found that this agent has a
very narrow range of doses that will produce a doubling or tripling of
effect and hence manifests a sharp demarcation between doses causing malfor-
mation and those causing fetal death. In addition, the unique pattern of
11mb deformation caused by the chemical was found to be a consequence of
slightly different rates of cell proliferation 1n the limbs of one side of
the fetus and that there was a very small "window" (~3 hours) during which
exposure would cause the pattern of 11mb malformation. Furthermore, the
distribution of the compound 1n the fetus had relatively little to do with
the ultimate malformation or Initial cellular toxldty. Dr. Manson conclud-
ed that for this type of direct-acting chemical, which 1s typical of almost
all research drugs that have been studied 1n teratology, the final pattern
of malformation depends more on the specific location of susceptible sites,
I.e., the sites of the cellular replication, than on the chemical Itself.
The second teratogen presented as an example was nltrofen, a herbicide
that has little toxldty 1n the adult animal. Dr. Hanson's research found
that, unlike methyl-N-n1tro-N-n1troguan1d1ne, nltrofen has a unique mecha-
nism of action 1n that H acts as a thyroid hormone analog. In addition,
1871A -61- 03/21/84
-------
she pointed out that although the kidneys of the exposed offspring were
grossly malformed, kidney function was normal up to 100 days of age, thus
raising the question of how to choose an endpolnt 1n risk assessment for
teratogens. Her conclusion was that 1t 1s unlikely that a general approach
to the assessment of risk for teratogens could be developed and that each
compound should be assessed Individually. In addition, she said her work
Indicated the Importance of performing thorough qualitative studies and
Identifying an appropriate toxic endpolnt before attempting quantitative
extrapolation to humans.
3.2.2. Discussion. Dr. Manson began the discussion by stating that 1n
her opinion teratogens are a very rare type of toxicant. Dr. Sllbergeld
suggested that 1t was possible that more chemicals may be considered terato-
gens 1f one studied subtle detrimental effects on the central nervous system
(CNS) resulting from prenatal exposure. Dr. Manson said that there Is prob-
ably no chemical that acts only on the CNS. Dr. Legator proposed that 1t 1s
possible that the wide divergence 1n gross morphologic effects resulting
from teratogens 1n different species may actually stem from a common mode of
action that occurs 1n the very early stages of development. Dr. Nelson
agreed with this suggestion and asked Dr. Manson what teratogenlc effects
may occur at very low exposure levels. She said that there 1s no easy way
to Identify such effects.
Dr. Calabrese raised the question of using rodents for test species,
since studies have Indicated that they are more sensitive to teratogens than
are humans. Dr. Clarkson stated that for methyl mercury, humans were the
most sensitive. Dr. Calabrese said the difference between methyl mercury
1871A -62- 03/21/84
-------
and other teratogens may be due to difference 1n b1oact1vat1on and then
asked Or. Hanson 1f teratogens, 1n general, had thresholds below which no
effect occurred. She replied that there were Insufficient data for a
conclusion but that 1t was her belief that teratogens had thresholds. Or.
WUhey noted that his studies of the pharmacoklnetlcs of teratogens led him
to question the term "dose" 1n teratology because he had found that some
I1p1d-soluble compounds, which one would suspect would cross the placenta
easily, did not. Dr. Hartung also noted that the exposure patterns used 1n
teratology studies rarely duplicate human exposure.
In their postmeetlng memoranda, Drs. Schnelderman and Sllbergeld noted
the need for the consideration of a broader range of effects than terato-
genldty 1n the development of a policy 1n this much-neglected area. Or.
Sllbergeld cited fetotoxlns and gonadal toxins as candidates for regulation,
and Dr. Schnelderman mentioned pregnancy wastage and reduced fertility as
possible endpolnts. He also cited three steps that are currently possible
1n the development of a policy:
1. Establishment of a minimum number of endpolnts to be studied.
2. Creation of a database on spontaneous abortions 1n humans.
3. Promotion of studies of effects of some toxicants on germ
cells 1n whole animals.
3.3. USE AND BIOLOGICAL JUSTIFICATION OF MATHEMATICAL MODELS
3.3.1. Presentation (R. Hertzberg, E. O'Flaherty). Several models and
calculation schemes have been adopted or proposed for use 1n predictive
health risk assessment. Among these are add1t1v1ty of risk and dose,
1871A -63- 03/21/84
-------
species conversion by dose per body weight, time-weighted averaging for any
dose rate, and route equivalence using the Stoklnger-Woodward formula.
These calculation methods must be evaluated for possible errors based on
known biological mechanisms, behavior, or numerical methods.
In his presentation, Dr. Hertzberg discussed the general principles
underlying the development of a mathematical model of a biological process
and pointed out that most models used 1n toxlcologlc risk assessment may
benefit from a reexamlnatlon of their assumptions and validity. He divided
the model building process Into three steps: 1) development of a mathemati-
cal and theoretical description of a given biological system or process,
2) calibration of the model with a broad range of applicable data, and
3) verification and validation of the model with a second set of data. He
also stated that 1t 1s Important to carry through the model's assumptions to
the final calculation to prevent overlnterpretatlon of the model's predic-
tions. As an example of this process, Dr. Hertzberg proposed a simple model
to justify the surface area approach to Interspedes dose conversion. He
outlined three assumptions: the dose at the target tissue determines the
toxic effects; the function and anatomy of organs are similar for mammalian
species; and, 1n this case, metabolism detoxifies the toxicant. He also
discussed the Implications of these assumptions for the final form of the
model. As a second example, Dr. Hertzberg described an approach for esti-
mating a cleanup level based on known prior exposures. He pointed out that
the Initial curve used by the ECAO, an ADI line based on the results of
laboratory experiments utilizing continuous exposures, does not represent
the exposure pattern of humans 1n which exposures decrease after a site Is
1871A -64- 03/21/84
-------
cleaned up. Hence, the question 1s whether the total human exposure consti-
tutes an acceptable level: how can one choose a post-cleanup level that, 1n
light of previous exposures, will be safe? One approach 1s to require that
the time-weighted average for any time after cleanup be acceptable, that Is,
assure that the average human exposure duplicates the laboratory situation.
Dr. O'Flaherty focused her talk on one aspect of route-to-route extrapo-
lation, first-pass effects. The text of her talk follows.
"The purpose of this presentation was to Illustrate several Important
features of first-pass metabolism that must be taken Into account by pharma-
coklnetlc or tox1cok1net1c models. These features have direct relevance to
Issues surrounding route-to-route extrapolation.
"The first point made was that the lung should not be too readily dis-
missed as a metabolizing organ. Although 1t 1s true that the total metabo-
lizing capacity of the lung 1s much less than that of the liver, there are
circumstances under which metabolism 1n the lung may be a significant frac-
tion of total metabolism. An example 1s shown 1n the Table 3-2 taken from a
paper by Roth and Wlersma (1979). It shows the predicted clearance of
benzo(a)pyrene (BP) In the lung and liver of control ("basal") rats and of
rats pretreated with 3-methylcholanthrene [3-MC] to Induce BP metabolism.
The predictions were made by Incorporating experimentally measured metabolic
capacities and physiologically reasonable blood flow rates Into a simple and
generally accepted model of organ clearance.
"It 1s apparent from the table that although pulmonary metabolism of BP
1n non-pretreated rats 1s only 0.4 percent of hepatic metabolism, 1n 3-MC
treated rats the predicted BP metabolizing activity of the lung 1s 60
percent that of the liver. The reason 1s not differential Induction, since
metabolizing capacity was Induced to the same extent 1n both organs. The
explanation for the Increased relative Importance of the lung 1n metabolism
1871A -65- 03/21/84
-------
TABLE 3-2
Predicted Clearance of Benzo(a)pyrene In Rats*
Predicted Clearance (mi/minutes)
Organ
Lungs
Liver
Lungs/liver
Blood Flow
(ml/minutes)
44.8
11.2
4.0
Basal
0.027
6.08
0.004
3-MC Induced
6.64
11.0
0.60
*Source: Roth and Wlersma, 1979
T871A
-66-
03/21/84
-------
of BP 1n 3-MC treated rats 1s that liver metabolism has become perfuslon-
rate-Hmlted while lung metabolism remains capacity-limited, so that the
full effect of Induction 1s observable 1n the lung but not the liver.
"This distinction between perfus1on-!1m1ted and capacity-limited metabo-
lism 1s fundamental to an understanding of first-pass metabolism. The total
metabolic capacity of an organ 1s related to a useful parameter, Intrinsic
clearance or CL, ., which 1s expressed 1n units of ml/m1n and, thus, 1s
the counterpart of organ perfuslon rate, 0- The first-pass metabolic
behavior of a compound presented to the liver 1s determined by the relative
magnitudes of Cl. . . and Q. where the subscript L designates the liver.
Either perfuslon rate Q or Intrinsic clearance CL can be rate-
L I n 191_
limiting.
"If CL 1s much less than Q , metabolism 1s rate-limiting at all
I M L f L L
or nearly all doses. Saturation of hepatic metabolizing enzymes will occur
at low doses. First-pass metabolism will be Insignificant, and dose depen-
dence of first-pass metabolism undetectable. But because of the small value
of CL. . . , systemic half-life t,/2 will be dose dependent. Dlphenyl-
hydantoln 1s an example of a compound 1n this class. It 1s at least 98 per-
cent bloavallable at all doses (Jusko et al., 1976). However, Its systemic
t,/? 1s dose dependent even at doses within the therapeutic range.
"Induction of metabolism of compounds for which Cl 1s much less
than 0, will shorten the systemic t but have an Insignificant effect
on b1oava1labH1ty.
"If CL 1s much greater than Q , perfuslon 1s rate-limiting.
B1oava1lab1l1ty from an oral dose will be low, and 1n fact may be 0 at low
doses. But because virtually all of an 1ntest1nally absorbed dose reaches
the liver on the first pass, the rate of presentation of absorbed compound
to liver enzymes may exceed the maximum rate at which the compound can be
metabolized, even at Intermediate doses and certainly at high doses. Thus,
the magnitude of the first-pass effect will be both large and dose depen-
dent. On the other hand, because of the large value of Cl, . ., systemic
t,._ will not be dose dependent. Phenacetln 1s one such compound. Its
1871A -67- 03/21/84
-------
b1oava1lab1lHy from a 250-mg dose was zero 1n six human subjects and from a
1-g dose was 0-49 percent 1n four human subjects. But Us systemic t,/2
was 37-74 minutes 1n all subjects, Independent of dose (Raaflaub and Dubach,
1975).
"Induction of metabolism of compounds for which CL 1s much great-
1 n t , H
er than Q will reduce bloavallabUHy further but will have no effect on
systemic t .
"These examples represent the extremes. There 1s a range of Interme-
diate relationships of QL to CL^nt L 1n which both systemic t1/2 and
b1oava1lab1!1ty may be affected by dose and by the animal's status with
regard to enzyme Induction.
"Five consequences of first-pass metabolism were pointed out. These are:
1. First-pass metabolism alters the relationship between applied
dose and Internal dose. Figure 3-1 Illustrates this point.
2. First-pass metabolism leads to the earlier appearance of
larger amounts of metabolite 1n the circulation, so that both
metabolite peak height and total exposure to metabolite are
Increased. Figure 3-2 Illustrates this point.
3. If the metabolite 1s active, plasma levels of the parent asso-
ciated with a given magnitude of effect will be lower after
oral than after Inhalation exposure.
4. If metabolites are associated with a spectrum of effects
different from the effects associated with the parent, dose
dependence of first-pass metabolism could be responsible for
dose dependence of the nature as well as the magnitude of the
effect.
5. Provided that metabolism 1s a major route of elimination, com-
pounds for which Cl^t^L 1s large will have a short systemic
t]/2 and will undergo* a large first-pass effect. On both
these counts, they are poor candidates for route-to-route
extrapolation procedures. Compounds for which CL^nj- L ^
small will have a long systemic t]/2 and will undergo only a
small first-pass effect. On both these counts, they are good
candidates for route-to-route extrapolation procedures."
1871A -68- 03/21/84
-------
AUC.ng-h/nJ
600
400'
200
20 40 60 80 TOO
Dose mctoprotol, mg
FIGURE 3-1
The relationship between a measure of Internal dose, the area under the
concentration, time curve (AUC), and either oral (•) or Intravenous (0)
metoprolol dose 1n five male human subjects. Note that the metoprolol 1s 0%
bloavallable from low oral doses
Source: Johnsson et al., 1975
1871A
-69-
03/21/84
-------
0 6 12 18 24 30 36 42 48 54 60 66 72
FIGURE 3-2
Mean plasma concentrations of nortrlptyllne (solid
major metabolite, I0-hydroxynortr1ptyl1ne (open symbols),
and Intramuscular (A,A) administration to six patients
Source: Alvan et al., 1977
symbols) and Its
after oral (0,i)
1871A
-70-
03/21/84
-------
3.3.2. Discussion. Several participants commented on Dr. Hertzberg's
presentation, noting that the Initial assumptions 1n the dose conversion
example may not be completely accurate or comprehensive (Nelson, Clarkson).
Dr. Hartung said that at a higher level of sophistication, the model
suggests that organ weight should be considered 1n the conversion; however,
Dr. Sllbergeld noted that the organ with the highest level of toxicant 1s
not necessarily the most affected organ. Dr. WHhey also commented on this
approach by pointing out that Gillette (1976) has developed a procedure 1n
which the rate processes for each metabolite are evaluated to find the con-
centration of the active molecule at the target site. Dr. Andelman comment-
ed on the example for estimating ADIs for people near sites that had been
cleaned up and stated that 1t 1s Important to know the age distribution of
the exposed population to determine 1n what way ending exposures at differ-
ent ages may Influence the estimated ADIs.
In her postmeetlng memorandum, Dr. Sllbergeld noted that the model pre-
sented by Dr. Hertzberg could be Improved by the consideration of M1chael1s-
Menten type kinetics and suggested that a model for procardnogens (Hoel,
1983) be reviewed as a useful example. Dr. Schnelderman noted the need for
developing models of diseases other than cancer. He also wrote that
response models as well as dose models need to be considered and that, 1n
general, the effects of peak doses vs. the effect of cumulative doses,
population variability, and tissue-specif1c1ty problems (I.e., first-pass
effects) need to be considered 1n constructing biologically based models.
1871A -71- 04/11/84
-------
4. WORKSHOPS (Session III)
The participants were assigned to four workshop groups as listed below.
The charge to each workshop was to discuss the topic, develop a statement
describing the state of the art, and recommend a course of action that could
be taken at this time.
WORKSHOP TOPIC
Al Sensitive Subgroups
Blngham (Chairperson) Legator Plscator
Calabrese Hanson Schnelderman
Ensleln Nicholson Sllbergeld
Hattls Nlsbet
Bl Multiple Route Exposure
Albert Crump Nelson (Chairperson)
Andelman Durkln O'Flaherty
Clarkson Hartung WHhey
Cornish Mehlman Wyzga
A2 Ranking Severity of Effects
Albert (Chairperson) Hartung Nelson
Blngham Hattls Nicholson
Calabrese Legator Plscator
Durkln Hanson Schnelderman
B2 Use of Exposure Data 1n Assessing Health Risk
Andelman Ensleln Sllbergeld
Clarkson Hehlman (Chairperson) WUhey
Cornish Nlsbet Wyzga
Crump O'Flaherty
4.1. CONSIDERATION OF HIGH RISK (SENSITIVE) SUBGROUPS IN HEALTH RISK
ASSESSHENT
4.1.1. Presentation (L.S. Erdrelch, D. Hattls). The present U.S. EPA
guidelines for calculating an ADI Include a 10-fold uncertainty factor to
account for 1ntraspec1es variability. This 10-fold factor has traditionally
1871A -72- 03/21/84
-------
been assumed to account for the "sensitive Individual." This theoretically
encompasses both 1nter1nd1v1dual variations of a normal healthy adult popu-
lation and high risk subgroups. Some scientists believe that this factor
will prove Inadequate to protect high risk subgroups 1n the population who,
because of predisposing factors, will be more sensitive to specific chemical
Insults. Predisposing factors Include but are not limited to the following:
developmental processes, age, nutritional deficiencies, disease states,
genetic conditions, behavioral factors, and previous or concomitant
exposures. It was the purpose of this workshop to consider this problem and
to evaluate and devise methods that would allow for better Incorporation of
these uncertainties and differences Into the risk assessment process. Two
topics of discussion were presented: the first was an evaluation of the
data needed to assess the variation 1n a population's response to toxicants,
and the second was a rating scheme that would, at least qualitatively,
weight risk assessments with regard to consideration of sensitive subgroups.
The Initial discussion was centered around the data available for an
assessment of the magnitude of the differences 1n Individual responses to
toxicants. Data from pharmacogenetlc studies were presented and their use
1n such an assessment was discussed. The significant advantages 1n using
such data Include the following: 1) they are from studies of humans,
2) they describe the response to xenoblotlcs that have been given 1n
measured doses to a group of Individuals, and 3) 1n some Instances, they are
collected under conditions 1n which 1nter1nd1v1dual variability 1s con-
trolled (I.e., monozygotlc twins). Disadvantages 1n the use of pharmaco-
genetlc data are that they generally show differences 1n a pharmacoklnetic
parameter Instead of an adverse effect and they are derived primarily from
studies of drugs rather than environmental toxicants.
1871A -73- 03/21/84
-------
Dr. Erdrelch pointed out that there was a fairly large body of litera-
ture on studies that measure 1nter1nd1v1dual differences 1n a wide variety
of parameters ranging from lung plutonlum concentrations to the clearance
and metabolism of a number of drugs. She suggested that collection and
analysis of these data might be relevant for estimating the variance 1n
response. A general consensus on this approach was reached and additional
sources of Information were suggested (Schnelderman, Sllbergeld), such as
registries of special diseases or conditions, the Pels Research Institute
longitudinal study, special exposure groups (Industrial, Love Canal), and
perhaps retirement populations (the residents of Leisure World). The sug-
gestion was also made that chemical-specific lists (cross Indexed by effect
and sensitive group) should also be established.
The second topic of this workshop considered an Interim approach pro-
posed by ECAO to address the problem of sensitive subgroups. This qualita-
tive approach adjusts the number exposed to reflect the proportion of high
risk (HR) Individuals and weights the total hazard Index by the size of the
exposed population. The Interim approach also provides Information for
comparing hazards among exposed populations.
The first steps of this procedure would be Included as a part of the
toxldty assessment (Figure 4-1) which summarizes the overall mult1chem1cal
risk assessment process. The rating scheme Itself 1s part of the Integra-
tion of the toxldty and exposure assessments. The Initial steps for
considering sensitive subgroups are as follows:
1. Identify appropriate HR groups for the chemicals 1n question.
These subgroups are to be selected from the general categories
1n Table 4-1, but are limited to well-defined HR groups of
relatively high prevalence
1871A -74- 03/21/84
-------
oo
Monitoring, physical, demo-
graphic and historical data
SITE
OF
HAZARD
Identity nf
Toxlcltv Information
WORLD
LITERATURE
en
i
EXPOSURE ASSESSMENT
Exposure levels, durations,
and routes for all chemicals.
Size and characteristics of
exposed population
TOXICITY ASSESSMENT
Goal; For each chemical, estimate the
minimum effective dose for endpolnts of
various severities, or the cancer risk,
for all dose rates, durations and
exposure routes.
Methodology Used;
Equivalence Calculations
• Interspecles conversion of dose
and duration
• Route-to-route conversion
Endpoint Interpretation
• Special methods for fetotoxlc
endpolnts
Risk or ADI Calculations
• Ranking severity of effects
• ADIs based on quanta1, contin-
uous or graded data
• Less-than-lifetime exposure
• Use and biological justification
of math models
• Sensitive subgroups in U.S.
population
Alternative Methods Where Information
Is Lacking
• Structure-activity approaches
ADIs or Risk Levels
for single chemicals
INTEGRATION
Coal; Report numbers of people ex-
posed above ADI, LOAEL, FEL, etc.
(for noncarcinogens) or above a
given risk level (for carcinogens)
for the chemical mixture. Report
"safe" or "low-risk" levels which
could serve as remedial goals.
Methodology Used (and exposure data
regul
Mixture assessment (requires
data on relative proportions of
chemicals In mixture)
Multiple route assessment (re-
quires data on exposure routes)
Sensitive subgroups In exposed
population (requires data on
exposed population character-
istics)
o
\
r\>
v>
CO
FIGURE 4-1
Information flow and methodology use 1n conducting site-specific mult1chem1cal health risk assessment,
showing use of each methodology component under development by ECAO-C1ndnnat1
-------
TABLE 4-1
Biological Factors Predisposing Individuals to
Hypersuscept1b1l1ty to Pollutants*
Factors
Rationale
I Developmental Processes
Examples: Pre- and Neonatal
Young children
II Genetic Disorders
Examples: G-6 PD deficiency
Sickle-cell trait
III Nutritional Deficiencies
Examples: Vitamin C deficiency
Protein deficiency
IV Existing Disease
Examples: Heart
Lung
V Behavioral Factors
Examples: Smoking
Alcohol
Dietary patterns
Immature enzyme detoxification
systems
Higher rate of GI absorption
Hemolysls 1n presence of certain
chemicals
Predisposes toward hemolytlc anemia
Potentiates effects of pollutants
Affects metabolism of Insecticides
Increased mortality In pollution
episodes demonstrated 1n epldemlo-
loglc studies
Existing conditions are aggravated
by respiratory Irritants
Increases exposures and therefore
risk: Interferes with normal
cleansing mechanisms of lung
Liver damage, synerglsm with other
chemicals
Increases exposures: deficiency
states Increase risk
VI Previous or concomitant exposures
Examples: Occupational exposures
Drugs
Increases exposure level and,
therefore, risk. Effects may be
synerglstlc or antagonistic with
environmental exposures
Possibility of Interactions
*Source: Adapted from Calabrese, 1978
1871A
-76-
03/21/84
-------
2. Determine the prevalence of the HR groups 1n the general
United States population or approximate subgroups (I.e.,
racial groups for certain genetic factors).
3. Incorporate this Information Into the rating scheme of the
Integrated risk assessment.
This Interim approach was not too well received. Many of the criticisms
leveled at 1t were summarized 1n the second presentation of the workshop,
made by Dr. Hattls. His criticism was grouped Into three areas:
1. The proposed approach assumes a discontinuous population
subgroup model of "hypersuscept1bH1ty"; however much of the
variability between Individuals 1n susceptibility may appear
as continuous because of a host of factors acting as quantita-
tive characters, and not based on obviously discernible
membership of people 1n discrete, countable subgroups.
2. The proposed approach appears to Implicitly assume that the
average hypersusceptlble person 1s only about twice as suscep-
tible as the average nonhypersusceptlble person. (If the
difference was actually this small, 1t wouldn't be significant
amid the other uncertainties of quantitative risk assessment.)
3. The proposed restriction to hypersusceptlble groups constitut-
ing at least 1% of the population, while H flows naturally
from the proposed calculating method, 1s arbitrary and poten-
tially seriously misleading. Consider, for example, a case
where there exists a 0.1% subgroup that 1s ten thousand times
as susceptible to a toxin as the average Individual. Quanti-
fying the size of the population that 1s very much more
susceptible than average 1s the essence of a quantitative risk
assessment for a threshold character with widely varying
sensitivity.
Dr. Hattls suggested that, Instead of the proposed approach that
evaluates sensitive subpopulatlons on a chem1cal-by-chem1cal basis, a better
approach would be to analyze the sources of variation 1n sensitivity, define
acceptable methods for estimating the variability contributed by each source
and then Incorporate these estimates Into a stochastic simulation model to
compute the overall distribution of population sensitivity for the chemical.
1871A -77- 03/21/84
-------
He proposed three sources of variability:
1. Exposure 1s effectively more for a subgroup because of either
biology or behavior.
2. Parameters determining normal/abnormal function are closer to
(or even beyond) the boundary of abnormality for the subgroup
(these groups have less "reserve capacity" so smaller Insults
would be required to exceed their ability to handle Insults).
3. There are more parameter changes per unit of dose for a
particular subgroup because of biological differences either
1n pharmacoklnetlc parameters (I.e., faster rate of absorp-
tion) the fraction of a chemical metabolized by "safe" or
"dangerous" pathways or differences 1n the quality or quantity
of receptors.
He further suggested that these three sources could be analyzed as follows:
1. Have a reasonably standardized analysis of susceptibility for
the exposure-related factors, depending only on the routes of
anticipated exposure of the target population.
2. Do a series of standard analyses of population variability for
different well-characterized physiological functions that are
affected by toxic agents (I.e., oxygen transport and nerve
conduction). For substances that affect unknown physiological
parameters, use a "typical" variability profile borrowed from
one of the well-characterized functions for "best estimate"
calculations, and some more extreme assumption of variability
for "conservative" or "worst case" projections.
3. To the extent possible, analyze pharmacoklnetlc and receptor-
related variability chemical by chemical. Where chemical-
specific data are not available, 1t may often be necessary to
adopt some general rules based on previous experience with
specific chemical classes or chemicals affecting specific
receptors.
The results from such an analysis could be entered Into the stochastic
model and used to estimate the overall susceptibility of a population to
each chemical, with default values being used 1n the absence of actual
Information. The computer would then construct a synthetic population of
1871A -78- 04/11/84
-------
whatever size was desired and, according to the distribution of susceptibil-
ities 1n these three categories, would simulate the size and extent of
susceptibility.
4.1.2. Discussion. The discussion was opened by Dr. Blngham who asked
the question, "Is 1t worth the effort to get something better than the
10-fold factor and, 1f so, then how do we go about 1t?" One response was
that, before a different approach could be derived, the question of the
types of effects to be considered Important had to be addressed because 1t
might not be necessary to have a 10-fold factor for all effects. Some may
require a 100-fold factor, others a 2-fold factor. Dr. Nicholson then
suggested that the utility of a 10-fold factor could be checked by looking
at the range of problt slopes for dose-response curves occurring 1n the
population.
Dr. Dourson then briefly discussed the current agency policy, which 1s
to pick the "critical" effect (the adverse effect which occurs at the lowest
dose), find the NOAEL for 1t, and apply a 10-fold factor. The rationale
behind such an approach 1s that protecting against the "critical" effect
should protect against all adverse effects. Unfortunately, this approach
does not take Into consideration the situation 1n which the log problt
slopes for different effects are not parallel, such that an effect with a
higher dose has a more shallow slope and results 1n a lower ADI than that
based on the critical effect. This approach 1s predicated on threshold
effects; H may not matter which effect 1s used as long as the ADI Is below
the threshold for all adverse effects. Dr. Legator then Indicated that
1871A -79- 03/21/84
-------
thresholds for Individuals have to be distinguished from population thresh-
olds, and the question was raised by Or. Calabrese of how super1mpos1t1on of
effects Influences a 10-fold factor.
Dr. Nlsbet pointed out some clear differences between children and
adults who are exposed to toxic chemicals. Children (under 9) spend 4-8
times more time outdoors; they breathe 3-5 times more air per unit of body
weight (bw); and they take 1n 3-4 times more food per unit of bw, 5 times
more fluid, and 6 times more dairy products; breast-fed Infants receive
10-20 times more I1p1d-soluble chemicals. Also, because children spend more
time In contact with the soil and their skin 1s more permeable, H 1s
estimated that they receive 5- to 10-fold more exposure than their parents.
These differences are 1n addition to those differences that might be shown
by pharmacogenetlc Information. Dr. Ensleln also pointed out that there are
also several large studies on populations with various disease states and
that these might provide Information as to variability.
Dr. Blngham closed the discussion by querying, "Should EPA come up with
chemical specific lists of high risk groups?" Dr. Nlsbet suggested that
these be drawn up by groups of people and types of effects. The general
consensus was that the scheme presented was Inappropriate, and that evalua-
tion of sensitive subgroups should be part of an ADI determination rather
than the calculation of the hazard Index.
1871A -80- 04/11/84
-------
4.1.3. Workshop Statement. Dr. Schnelderman prepared a statement on the
conclusions reached by the workshop participants. He grouped the conclu-
sions Into the following three areas:
1. "One 'safety-factor' number (I.e., division by 10 or 100) 1s
Inadequate to take Into account the variability 1n the human
population. Safety factors developed for adults are Inappro-
priate for use with children. Action: Develop needed data
relating to safety factors for different segments of the
population. In the Interim use current factors, setting a
deadline as to when they should be reconsidered.
2. "For future changes 1n safety factors, base safety factors on
the following data:
a. Slope of the dose-response curves (the more shallow the
slope, the larger the factor required).
b. The specific toxldty and Its severity (the greater the
need to protect against the toxldty, the larger the
safety factor needs to be. A lifetime of Illness or
reduced functional ability 1s usually more Important than
a short-term, reversible Incident).
3. "For noncancer effects, remain with the ADI concept, rather
than a 'hazard Index.1 While the public-health consequences
are greater when more people become 111 (hazard Index), 1t 1s
Individuals who become 111, and even a small number of persons
are to be protected against serious consequences."
4.2. ASSESSMENT OF MULTIPLE ROUTE EXPOSURE
4.2.1. Presentation (W. Pepelko, R. Hartung). Under certain conditions
humans are exposed to environmental pollutants by more than one route of
exposure. The three major portals of entry Include the lungs, GI tract, and
skin. While 1t Is possible to be exposed by other routes such as Intrave-
nous or Intramuscular Injection, the use of these routes 1s generally limit-
ed to treatment with therapeutic drugs or other nonenvlronmental exposures.
1871A -81- 04/11/84
-------
Such routes were therefore not emphasized 1n the discussion. The skin has a
much smaller surface area than the lungs and 1s generally less permeable to
environmental pollutants. Under normal conditions of exposure to environ-
mental pollutants, 1n which the concentrations are very low, entry through
the skin will add very little to body burdens. Thus, 1n most cases, H
should be necessary to consider only two routes of exposure, Inhalation and
IntragastMc.
The major Issue to be resolved was whether exposures by more than one
route should be summed, and 1f so, how? It was proposed that summation 1s
not valid under the following conditions:
1. A significant amount of pollutant 1s absorbed by only one
route.
2. The pollutant chemical 1s completely removed or Inactivated
before reaching the target organ (I.e., first-pass effect for
liver), except for one route.
3. The portal of entry 1s the primary target organ. This brings
up two questions to be considered. If the target organ 1s
different for each route of exposure, should the situation be
treated as two different chemicals with two separate AOIs? In
such cases will exposure by one route (at above threshold
levels) affect the threshold for exposure by another route?
If none of the conditions listed above 1s present, then 1t 1s proposed
that summation of the exposures by more than one route can be done. At the
previous meeting, a mathematical approach was proposed for this summation:
dn = Vl *d2r2 * "• Vl
where d = total dose, d. = the various exposure routes, and r. = the
n 1 I
absorption fraction for each exposure route; d 1s then compared with the
ADI.
1871A -82- 04/11/84
-------
The previous equation 1s simple, straightforward, and easy to apply. It
should be possible to apply this equation successfully 1n cases where the
target tissue 1s the same for each route of exposure, where the elimination
half-time 1s sufficiently long so that the route of exposure will not affect
temporal target organ concentration relationships, and where the absorption
coefficient 1s known for each route of exposure.
The more difficult cases and the ones that were to be emphasized at this
workshop are those 1n which a straightforward summation of d r 1s not
possible. These Include the following examples:
1. The target organs are generally different for two exposure
routes but not exclusively so.
2. Partial removal of an absorbed dose occurs following exposure
by one route but not another.
3. As a result of short elimination, the half-time target organ
concentrations fluctuate during Intermittent oral dosing as
compared with constant levels during Inhalation exposure.
Under these conditions, can absorbed doses by different routes be summed
with any degree of accuracy, and 1f so, how?
Dr. Hartung summarized his presentation 1n his postmeetlng memorandum.
The portion revelant to this workshop 1s as follows:
"Multiple-Route Exposure: The evaluation of multiple route exposures
was analyzed as a special case of route:route extrapolation, and 1n the case
when exposures by different routes produced different effects as a special
case of mult1chem1cal exposure. The simplistic additive approach suggested
earlier, and as summarized 1n the premeetlng Information, was considered to
be Inadequate.
1871A -83- 04/11/84
-------
"The approach taken to the problem was to consider first those condi-
tions under which a multlroute exposure evaluation could be made with confi-
dence. In order to keep the problem tractable, the problem analysis was
considered for a single species only. An evaluation with confidence can be
made 1f and only 1f all of the following conditions are met:
"1. The dose-duration of exposure patterns are Identical for all
routes of exposure;
"2. The systemic effects are Identical for all routes of exposure;
"3. No route-specific effects are found at the portal of entry;
"4. There are no first-pass effects; and
"5. The dose-duration of exposure-response surface 1s known In
terms of absorbed dose (or, the relationships between applied
dose and absorbed dose are known for each route).
"Points 1 and 4 are likely to become unimportant 1f the chemical 1s
cleared very slowly from the body.
"If all of these conditions are met, then the total absorbed dose could
be estimated, and subsequently the systemic response 1n terms of absorbed
dose could be evaluated.
"It was recognized that these preconditions would only rarely, 1f ever,
be met for a specific compound. It was also pointed out, that as these
various preconditions are violated, then the ability to predict the conse-
quences of exposures by various routes deteriorates In ways that are not
easily predictable.
"If there were route-specific effects, then a simple summation of
absorbed doses would probably be misleading, since Interactions between
1871A -84- 03/21/84
-------
effects would not be taken Into account appropriately. Under such condi-
tions (1f dose/response relationships were known for each route of expo-
sure), the problem could be treated as a special case of the simple joint
action add1t1v1ty model as postulated by Bliss (1939). Such a model could
readily be formulated 1n terms of external dose, rather than absorbed dose.
"Since exposures by multiple routes appear to be common, and since dose/
response Information 1s often Imperfectly known for the various routes of
exposure, H 1s tempting to attempt estimations of the effects of multlroute
exposures on the basis of mixtures of chronic and acute data or on the basis
of chronic data alone. Such estimates may be very unreliable and mis-
leading."
4.2.2. Discussion. Several participants commented on the limitations to
the general approach outlined by Dr. Hartung, Including the likelihood that
some chemicals produce overlapping effects through different routes
(Albert), that environmental exposures are at low doses and effects are
likely to be subtle (Nelson), and that the largest database for preparing
dose-response curves 1s for acute exposures (Durkln). Dr. Nelson stated
that, 1n addition to the proposed approach, the risk assessor should be
aware of the limitations In the data and proposed a four-tier system, with
the highest tier corresponding to the highest level of confidence 1n the
assessment and the lowest tier, to the lowest level of confidence. The four
tiers are as follows:
1. All of the necessary Information 1s available, Including dose/
response/time curves for all three routes.
2. Information 1s available on two routes, and the remaining
Information 1s available through experimentation to determine
absorption coefficients, Internal doses, and body burden data.
1871A -85- 03/21/84
-------
3. Acute data are available for all three routes, and chronic
data are available for one.
4. Acute (and chronic) data are available for only one route.
As a practical approach, for a single situation one could Identify which
of Dr. Hartung's preconditions were met or were not pertinent and then gauge
the level of confidence 1n the assessment (Nelson). Dr. Albert pointed out
that the assessment of multiple exposures 1s sufficiently complex that no
one simplistic approach 1s adequate, and recommended a case-by-case approach
using the available data. Dr. Durkln noted that specifying a minimum amount
of data without which an assessment would not be performed 1s a necessity.
In her postmeetlng memorandum, Dr. O'Flaherty noted that the character-
istics of few compounds will be known 1n sufficient detail to meet Dr.
Hartung's five criteria for an Ideal candidate and to allow summation across
routes. Therefore, she proposed the following generalizations on the extent
to which some compounds may meet the criteria:
"1. Internal dose-duration patterns are likely not to differ
significantly across routes of exposure 1f the chemical has a
long t-|/2- Concentration cycling around the mean steady-
state blood concentration 1s damped 1f the compound has a long
t]/2» so *nat f°r such compounds the difference between
dose-duration patterns for Intermittent-exposure routes and
for continuous exposure routes 1s minimized. Alternatively,
even for compounds with short t^/2» ^ patterns of exposure
are roughly comparable the Internal dose-duration patterns may
also be roughly comparable. For example, exposure through
water, although discontinuous, would fall somewhere between
continuous exposure by Inhalation and administration of the
same total dally dose to an animal by single gavage.
"2. Although situations 1n which systemic effects are different
for different routes of administration can readily be
Imagined, we were unable to Identify any compounds for which
this 1s true. It may be possible to Infer that few chemicals
do not meet this criterion.
1871A -86- 04/11/84
-------
"3. Highly chemically reactive compounds, or compounds with long
residence times 1n the tissue of entry, should be suspect with
regard to their potential for causing portal-of-entry effects.
However, 1t cannot be assumed that compounds that are not
highly reactive or have short residence times 1n the tissue of
entry will not cause portal-of-entry effects.
"4. Compounds with long t]/2are 1ess likely to undergo signifi-
cant first-pass effects than are compounds with short t]/2-
Thus, as a result of both considerations 1 and 4, 1t appears
that compounds with long t1/2 "^Y be suitable candidates for
summation across exposure routes."
4.2.3. Workshop Statement. The essential points of the workshop were
agreed upon by Drs. Nelson, Clarkson, and Pepelko. Or. Harlung's five
conditions for conducting an assessment of a multlroute exposure were
considered a valid starting point against which other approaches should be
evaluated. The limitations 1n the available data are likely to preclude
meaningful assessments at the present time, although some gauge of the level
of confidence 1n an assessment 1s possible and desirable.
4.3. RANKING THE SEVERITY OF EFFECTS
4.3.1. Presentation (M. Dourson, P. Durkln). In this workshop, a scheme
was presented for ranking potential adverse effects from chronic exposures,
which 1s to be used for Identifying reportable quantities (RQs) as required
by CERCLA. This act stipulates that once a spill occurs resulting 1n the
release of a hazardous material In a quantity greater than the RQ, the spill
must be reported to the Agency. Under Section 102(b) of the act, the RQ of
any hazardous substance 1s 1 pound, unless a different RQ has been estab-
lished; statutory RQs can be set at 1, 10, 100, 1000 or 5000 pounds. The
current regulatory approach uses six primary criteria for setting an RQ:
aquatic toxldty, mammalian toxldty (acute), 1gn1tab1l1ty, reactivity,
1871A -87- 04/11/84
-------
other toxic effects (chronic), and cardnogenlcHy. For each criterion, a
f1ve-t1ered rating scale 1s set up corresponding with the RQ values of 1,
10, 100, 1000 and 5000 pounds. There are also three secondary criteria
(b1odegradab1!1ty, hydrolysis, and photolysis) that may adjust the RQ up by
one level.
In the scheme presented at the workshop, determination of a reportable
quantity based on chronic toxic effects requires a composite score (CS),
which 1s the product of a rating value for dose (RV ) and a rating value
for effect (RV ). Each RV can vary from 1 to 10; hence, the CS varies
from 1 to 100. The higher the CS, the lower the RQ.
The following 1s a brief description of the calculation of an RQ.
First, an RV. 1s derived as follows for each route of exposure for which
there are data:
1. Identify subchronlc or chronic NOAELs, LOAELs or FELs.
2. Convert above dose to units of mg/kg/day.
3. If above dose 1s based on subchronlc exposure, divide by 10.
4. Estimate a minimum effect dose (MED) for humans by converting
above dose using surface area conversion.
5. Assign a rating value to MED as shown 1n Figure 4-2. This 1s
the RVd.
The RV that 1s associated with the MED 1s derived as shown 1n
e
Table 4-2.
The RQ 1s then assigned on the basis of the following relationship
(Table 4-3) using the highest composite score (the product of RV. and
RV for each route of exposure).
1871A -88- 03/21/84
-------
CO
I
oo
VO
I
•o
oc
UJ
CO
o
O
10-
9
8
7
6
5-
4.
3-
2
1 •
0-
T
3
-2
RVd = 10 IF log MED < -3
RVd = -1.5 log MED + 5.5 IF-3 < log MED < 3
RVd = 1 IF log MED > 3
^ \ \ r
-1012
log HUMAN MED (mg/day)
1
4
00
FIGURE 4-2
Rating Values for Doses
-------
TABLE 4-2
Rating Values for NOELs, LOAELs and FELs for the Determination of RVe
RATING EFFECT
1 ENZYME INDUCTION OR OTHER BIOCHEMICAL CHANGE WITH NO PATHOLOGIC CHANGES AND NO
CHANGE IN ORGAN WEIGHTS.
2 ENZYME INDUCTION AND SUBCELLULAR PROLIFERATION OR OTHER CHANGES IN ORGANELLES
BUT NO OTHER APPARENT EFFECTS.
3 HYPERPLASIA, HYPERTROPHY OR ATROPHY BUT NO CHANGE IN ORGAN WEIGHTS.
4 HYPERPLASIA. HYPERTROPHY, OR ATROPHY WITH CHANGES IN ORGAN WEIGHTS.
5 REVERSIBLE CELLULAR CHANGES: CLOUDY SWELLING, HYDROPIC CHANGE, OR FATTY CHANGES.
6 NECROSIS, OR METAPLASIA WITH NO APPARENT DECREMENT OF ORGAN FUNCTION. ANY
NEUROPATHY WITHOUT APPARENT BEHAVIORAL, SENSORY, OR PHYSIOLOGIC CHANGES.
7 NECROSIS, ATROPHY, HYPERTROPHY, OR METAPLASIA WITH A DETECTABLE DECREMENT OF
ORGAN FUNCTIONS. ANY NEUROPATHY WITH A MEASURABLE CHANGE IN BEHAVIORAL,
SENSORY, OR PHYSIOLOGIC ACTIVITY.
8 NECROSIS, ATROPHY, HYPERTROPHY, OR METAPLASIA WITH DEFINITIVE ORGAN DYSFUNCTION,
ANY NEUROPATHY WITH GROSS CHANGES IN BEHAVIOR, SENSORY, OR MOTOR PERFORMANCE.
ANY DECREASE IN REPRODUCTIVE CAPACITY. ANY EVIDENCE OF FETOTOXICITY.
9 PRONOUNCED PATHOLOGIC CHANGES WITH SEVERE ORGAN DYSFUNCTION. ANY NEUROPATHY
WITH LOSS OF BEHAVIORAL OR MOTOR CONTROL OR LOSS OF SENSORY ABILITY. REPRODUCTIVE
DYSFUNCTION. ANY TERATOGENIC EFFECT WITH MATERNAL TOXICITY.
££ 10 DEATH OR PRONOUNCED LIFE SHORTENING. ANY TERATOGENIC EFFECT WITHOUT SIGNS OF
¥ MATERNAL TOXICITY.
oo
-------
TABLE 4-3
Relationship of Composite Score to RQ
Composite Score RQ
81-100 1
41-80 10
21-40 100
6-20 1000
1-5 5000
1871A -91- 03/21/84
-------
4.3.2. Discussion. Several Issues were raised during the discussion and
1n postmeetlng memoranda. Or. Albert defined three questions that he
thought the participants should address relative to the proposed schemes:
1. Do the participants agree with the rating for toxldty 1n
Table 4-2?
2. Do the participants agree with the rating proposed for dose,
and 1f so, should 1t essentially be a log function?
3. Do participants agree with the multiplicative approach for
combining the two rating schemes?
With regard to the first question, Dr. Stara asked 1f the group (or
possibly the National Academy of Sciences) should be asked to develop a
catalog of effects to better define the toxldty rating and to provide Input
as to where FELs, LOAELs, and NOAELs fall 1n the rating scheme. Dr. Hartung
suggested that perhaps criteria should be developed to grade particular
effects. In particular he suggested that the question of reversible or
Irreversible effects be addressed and also that public perception of effects
should be taken Into account. Dr. Schnelderman Indicated a similar concern
with the question of 1rrevers1b1l1ty.
With regard to question 3, several Individuals Indicated that multipli-
cation of what were Implicitly log functions to derive the CS was probably
Inappropriate and that perhaps an additive scheme would be better. They
also noted that this would probably greatly change the outcome of the rank-
Ing procedures presented by Dr. Durkln.
1871A -92- 03/21/84
-------
Two participants made additional comments 1n their postmeetlng
memoranda. Dr. Sllbergeld stated that the ranking of effects presented In
Table 4-2 was biased toward pathological findings (I.e., hyperplasla and
necrosis) and that some statements were so unqualified as to be unlnterpret-
able (I.e., "any decrease 1n reproductive capacity"). She also cited the
lack of distinction between "death" and "pronounced life-shortening," the
lack of a rationale for why teratogenldty without signs of maternal toxlc-
1ty was considered more serious than with signs of maternal toxUHy, and
the absence of consideration of revers1b1Hty/1rrevers1b1lHy. Or. Manson
made similar points and stated that a distinction should be made between
decreases 1n reproductive capacity that occur with general systemic toxldty
and decreases occurring 1n the absence of other types of toxldty. Dr.
Schnelderman also noted the lack of consideration of reversibility, germ
plasm effects, and effects occurring over a lifetime (I.e., reduction 1n
functional ability and reduced IQ), and stated that the use of logarithmic
scales for effect and dose required substantially more justification. He
concluded that the process 1s "far from ready to be used for extrapolation
to NOELs."
4.3.3. Workshop Statement. A workshop statement was written by Dr.
Albert. In addition to summarizing the presentation, the statement Included
the following conclusion.
"The primary objective of the panel was to critique the approach to
rating the toxic effects. Three conclusions were reached by the panel:
"1. There was general endorsement of the principles used 1n the
approach to developing the composite score.
1871A -93- 03/21/84
-------
"2. There were reservations about the scheme presented for ranking
the toxic effects; for example, tentative suggestions were
made that the minimal effects be lumped Into a single category
and other suggestions concerned different approaches to giving
weights to the various kinds of effects than were done by Dr.
Durkln. Other possibilities Included assigning weighting
factors to the quality of evidence used 1n defining the
composite score.
"3. There was concurrence on the fact that the subject 1s highly
complex and one or two days would be required to evaluate all
four aspects of defining a composite score and assigning these
scores to reportable quantities of spilled toxic substances.
The possibility of turning this problem over to the National
Academy was discussed with mixed response."
4.4. USE OF EXPOSURE DATA IN ASSESSING HEALTH RISK
4.4.1. Presentation (R. Bruins, G. Kew). A site-specific health risk
assessment 1s accomplished 1n three phases Including a toxiclty assessment
and an exposure assessment, which may be conducted concurrently, and an
Integration step to combine these two (see Figure 4-1). The goal or output
of a health risk assessment 1s to report the numbers of people who are at
risk of experiencing adverse effects as a result of a given circumstance
(such as uncontrolled waste site or untreated Industrial effluent). For
noncardnogens, the output would be the number of people subject to expo-
sures above an ADI, a LOAEL, a FEL, etc., and the duration or frequency of
exposure. For carcinogens the result can be expressed as numbers exposed to
a given risk level, or as a predicted number of cancer cases. The risk
assessment should also be capable of estimating environmental levels or
exposure concentrations that can serve as goals for remedial measures.
1871A -94- 04/11/84
-------
The risk assessment methodology being developed by ECAO focuses primar-
ily on the toxldty assessment phase (ADI and risk level calculations for
single chemicals) and also on methods used 1n the Integration phase (assess-
ment of chemical mixtures, multiple exposure of routes, and response of
sensitive subgroups). The goal of this workshop was to discuss the Informa-
tion an exposure assessment can provide and how 1t should be used. The
Intent was not to recommend how an exposure assessment should be performed.
Some of the questions with which we were concerned Include the following:
How will the availability or quality of key exposure Information affect the
assessment procedure? What 1s the most appropriate endpolnt (I.e., exposure
concentration or absorbed dose) of an exposure assessment? What special
problems arise when more than one chemical 1s Involved?
4.4.1.1. DATA AVAILABILITY AND QUALITY — An exposure assessment
examines the Identity of chemicals of concern, their movements and concen-
trations 1n various environmental media, and the demographics of potentially
affected populations. The types of Information useful In conducting a risk
assessment are chemical exposure levels, numbers and kinds of people exposed
at each level, and exposure durations.
The exposure level 1s likely to be a critically Important factor, but 1s
very difficult to precisely ascertain. It 1s therefore Important to
consider the effect of exposure uncertainty upon the overall outcome of the
risk assessment. Site assessments could conceivably be requested where the
exposure level 1s unknown. Consider, for example, a situation 1n which the
Identity of chemicals at a site Is known and release 1s verified, but lack
of time, funds, or data has precluded an exposure assessment. Or consider a
1871A -95- 04/11/84
-------
situation where release seems Imminent but has not yet occurred. We may
readily conclude that without the needed data, no risk assessment Is
possible. But consider a similar situation, 1n which an exposure assessment
has been conducted, but limited Information has required that a large number
of assumptions be made to the extent that the uncertainty around the
exposure level 1s extremely large, I.e., several orders of magnitude. This
exposure estimate could 1n effect be as useless as no assessment at all.
Thus, 1n terms of associated variability, the data may fall anywhere on
a continuum from very poor to highly accurate (I.e., where extensive target
monitoring has been conducted). Questions to be considered Include the
following: What level of precision can be expected from a typical exposure
assessment? Should our methodology provide guidance as to what 1s accept-
able? How should exposure uncertainty be treated when exposure estimates
are Integrated with toxldty estimates?
Other types of data may also be unavailable or of variable quality. If
exposure levels are given but no Information 1s provided on the size or
composition of the exposed population, 1t may still be possible to complete
a health risk assessment. However, the result could be expressed only In
terms of effects and would be lacking 1n the population dimension. This
would allow a qualitative judgment as to whether or not a hazard existed,
but would not permit an estimate of the associated societal cost. The lack
of Information about the characteristics of people exposed raises an Issue
of whether the population should be assumed to be "average" or "sensitive"
with respect to physiological, behavioral, or genetic traits that may affect
esponse. Thus, this topic 1s related to the topic of the "sensitive
subgroups" workshop.
"»871A -96- 03/21/84
-------
As to exposure duration, the assumption of lifetime exposure may be
sufficient to Indicate whether the situation 1s acceptable 1n perpetuity or
whether remedial measures need to be taken. However, more specific Informa-
tion on exposure duration, I.e., duration of past exposure and likely dura-
tion before remedial measures are completed, could be needed 1n decisions to
relocate people, change water source, etc. What are the Impacts 1f this
Information 1s unavailable, and 1s this an Issue that our guidelines need to
address?
4.4.1.2. ENDPOINT OF EXPOSURE ASSESSMENT -- The endpolnt of an expo-
sure assessment may vary according to the purpose, the substance, or the
type of monitoring done. If the substance has been monitored 1n air, food,
soil, or water, exposure may be expressed as the concentration or amount
reaching human exchange membranes. If absorption factors are known or esti-
mated, exposure may be expressed as a whole-body absorbed dose. If human
tissues or excreta are monitored, knowledge of pharmacoklnetlcs 1s needed
for back-calculating to either a body dose or an environmental concentra-
tion. Where multiple routes of exposure may contribute to systemic effects,
whole-body dose may be the best Indicator of effects. On the other hand,
where effects may be route-specific, exchange-membrane concentrations or
amounts could be most useful. Therefore, 1t 1s Important to consider the
significance of the exposure endpolnt 1n determining health risk, and to
evaluate whether our guidelines should address this Issue.
1871A -97- 03/21/84
-------
4.4.1.3. MULTIPLE CHEMICAL EXPOSURE — The preceding questions apply
to single-toxicant exposures. What additional questions are faced by health
risk assessors 1n Interpreting the results of mult1chem1cal exposure assess-
ments? For example, consider a site vicinity where exposure to each of two
chemicals varies over a wide range. If both were transported by the same
route(s), then their respective exposure levels should be directly corre-
lated, and some Individuals might be exposed to the highest observed levels
of both. Alternatively, 1f routes differ, the highest exposures may occur
to different Individuals; this would affect the outcome of the health risk
assessment. Thus, 1t would not be sufficient for an exposure assessment to
simply provide ranges of chemical levels over a given area, but correlations
among chemicals or routes may also be needed. Guidance on this and other
points 1s needed.
Many other questions are likely to arise 1n any attempt to conduct a
risk assessment using this methodology. Hence there 1s a need to conduct
trial assessments 1n order to detect remaining problems. Another goal of
this session 1s to Identify available exposure assessment data that would be
suited to this purpose. Assessments Including data on more than one chemi-
cal and on the exposed population would be most useful. Participants are
requested to provide suggestions and/or exposure data of which they are
aware.
4.4.2. Discussion. Dr. Wyzga of the Electric Power Research Institute
(EPRI) briefly discussed EPRI's current approach to assessing exposure to
atmospheric emissions. From Information on transport and transformation, an
estimate of the atmospheric concentration of a given pollutant 1s made.
1871A -98- 04/11/84
-------
These data are combined with a demographic model of the exposed population
that Includes health status and activity levels as covarlants. Similar
approaches are used for estimating the contribution of the other media
(I.e., water and food) to the total exposure; that 1s, whatever data are
available are used 1n models of the movement of pollutants within and
between the media to determine exposure.
Dr. Nlsbet stated that 1n his experience the modeling approach was of no
use 1n gauging exposures resulting from hazardous waste sites because
exposure conditions are site-specific and therefore do not fit the models.
The monitoring data tend to show wide variability in space and time and
that, although a log-normal distribution is desired, such a distribution is
rare. Or. Silbergeld pointed out that an Important factor missing 1n
exposure assessments is the environment-human Interface, I.e., how people's
behavior influences their exposure to toxic substances. She also expressed
the concern that, because little information is available on the background
rate of various diseases in the population, the health status of a poten-
tially exposed population is not considered 1n exposure assessments. Other
participants also commented that both behavioral and health status data are
needed (Andelmann, Mehlman, Hattis).
4.4.3. Workshop Statement. Dr. Mehlman, chairman of the workshop, sub-
mitted the following summary statement:
"We assume that the type and amount of waste will be reasonably well
characterized (but may be deficient in some cases).
1871A -99- 03/21/84
-------
"Exposure assessments are resource limited and data-limited, and expo-
sures are variable 1n space and time; therefore, exposure cannot be charac-
terized by a single number.
"The following data are desirable for characterizing exposure: monitor-
Ing data from air, water (surface and ground), foodstuffs (especially fish),
and soil must be determined to adequately establish the level of substances
present. Tissue/excreta monitoring 1n humans or animals from the area 1s
extremely valuable 1n assessing exposure.
"Human behavioral observations 1n the area should be made to Identify
exposure routes or sensitive Individuals.
"Attention should be given to proper locations of monitoring wells.
"The output should be explicit as to the variability associated with
exposure, both 1n space and time.
"Monitoring data and exposure assessments come from a wide variety of
sources within and without U.S. EPA. Close communication between those
responsible for each phase must be encouraged, especially 1f a second phase
of monitoring can be done.
"We recommend that the Agency commission a paper discussing the Integra-
tion of environmental monitoring data, exposure data, and health data,
Including the role of epidemiology of exposure."
1871A -100- 03/21/84
-------
REFERENCES
Alvan, G., 0. Borga, M. L1nd, L. Palmer and B. Slwers. 1977. First pass
hydroxylatlon of nortrlptyllne: Concentrations of parent drug and major
metabolites 1n plasma. Eur. J. Cl1n. Pharmacol. 11:219.
Bliss, C.I. 1939. The toxldty of poisons applied jointly. Ann. Appl.
B1ol. 26:585-615.
Boxenbaum, H. 1982. Interspedes scaling, allometry, physiological time,
and the ground plan of pharmacoklnetlcs. J. Pharmacoklnet. Blopharm.
10:201-207.
Calabrese, E.J. 1978. Pollutants and High Risk Groups. John Wiley and
Sons, New York, NY.
Crump, K.S. and R.B. Howe. 1983. The Multistage Model with a Time-Depen-
dent Dose Pattern: Applications to Carcinogenic Risk Assessment. Environ-
mental Criteria and Assessment Office, Cincinnati, OH, U.S. Environmental
Protection Agency. Contract No. 68-03-3111.
Day, N.E. and C.C. Brown. 1980. Multistage models and primary prevention
of cancer. J. Natl. Cancer Inst. 64:977-989.
Federal Register. 1980. 45: 79318-79379.
Food Safety Council, Scientific Committee. 1978. Proposed procedure for
food safety assessment. Food Cosmet. Toxlcol. 16:1-132.
1871A -101- 04/11/84
-------
Glllete, J.R. 1976. Application of pharmacok1net1c principles 1n the
application of animal data to man. Cl1n. Toxlcol. 9:709-722.
Hoel, D. 1983. LINDA ERDREICH WILL PROVIDE THIS.
Oohnsson, G., C-G. Regardh and L. Solve!!. 1975. Combined pharmacokinetic
and pharmacodynamic studies in man of the adrenergic-receptor antagonist
metoprolol. Acta Pharmacol. Toxlcol. 36: Suppl. V, 31.
Jusko, W.J., J.R. Koup and G. Alvan. 1976. Nonlinear assessment of pheny-
toin availability. J. Pharmacokinet. Biopharm. 4:327.
Khera, K.S. 1973. Ethylenethiourea: Teratogenlcity studies in rats and
rabbits. Teratology. 7:243-252.
Khera, K.S. 1974. Teratogenlcity and dominant lethal studies on hexa-
chlorobenzene 1n rats. Food Cosmet. Toxlcol. 12:471-477.
Khera, K.S. and J.A. Ruddlck. 1973. Po!ych!orod1benzo-p-dioxins: Perinatal
effects and the dominant lethal test in Wlstar rats. J.n: Advances in Chem-
istry Series, No. 120, Chlorod1ox1ns--0r1gins and Fate, American Chemical
Society.
1871A -102- 03/21/84
-------
Murray, F.J., F.A. Smith, K.D. NHschke, C.G. Humlston, R.3. Kochlba and
B.A. Schwetz. 1979. Three generation reproduction study of rats given
2,3,7,8-tetrachlorod1benzo-p-d1ox1n (TCDD) 1n the diet. Toxlcol. Appl.
Pharmacol. 50:241-252.
Muska, C.F. and L.J. Weber. 1977. An approach for studying effects of
mixtures of environmental toxicants on whole organism performances, in.:
Recent Advances 1n F1sh Toxicology, A Symposium. Ecological Research
Series, EPA 600/3-77-085.
NCI (National Cancer Institute). 1978. Bloassay of 1,2,-D1bromoethane for
Possible Carc1nogenes1s. Technical Report Series No. 86, DHEW Publ. No.
(NIH) 78-1336.
Raaflaub, J. and U.C. Dubach. 1975. On the pharmacoklnetlcs of phenacetln
1n man. Eur. J. CUn. Pharmacol. 8:261-1975.
Roth, R.A. and D.A. Wlersma. 1979. Role of the lung 1n total body clear-
ance of circulating drugs. CUn. Pharmacoklnet. 4:355.
Stara, J.F., M.L. Dourson and C.T. OeRosa. 1981. Water Quality Criteria:
Methodology and Applications. JJK Conference Proceedings: Environmental
Risk Assessment. How New Regulations Will Affect the Utility Industry,
EA2064, Research Project 1316-6, EPRI Contract No. WA-80-134.
Whlttemore, A.S. and J.B. Keller. 1978. Quantitative theories of cardno-
genesls. Soc. Ind. Appl. Mathemat. Rev. 20:1-30.
1871A -103- 03/21/84
USGPO: 1984-759-102-975
------- |