United States
Environmental Protection
Agency
Office of Health and
Environmental Assessment
Washington DC 20460
Research and Development
EPA/600/S8-85/009 Sept. 1985
&ERA Project Summary
Methodology for
Characterization of
Uncertainty in Exposure
Assessments
Roy W. Whitmore
Virtually all exposure assessments
except those based upon measured
exposure levels for a probability sample
of population members rely upon a
mathematical model to predict expo-
sure. Whenever a model that has not
been validated is used, an uncertainty
associated with the exposure assess-
ment may be present. The primary
characterization of uncertainty is partly
qualitative, i.e., it includes assumptions
inherent in the model. Sensitivity of the
exposure assessment to model formu-
lation can be investigated by replicating
the assessment for plausible alternative
models.
When an exposure assessment is
based upon directly measured exposure
levels for a probability sample of popu-
lation members, uncertainty can be
greatly reduced and described quantita-
tively. In this case, the primary sources
of uncertainty are measurement errors
and sampling errors. A thorough quality
assurance program should be designed
into the study to ensure that measure-
ment errors can be estimated. The
effects of all sources of random error
should be measured quantitatively by
confidence interval estimates of param-
eters of interest, e.g., percentiles of the
exposure distribution. Moreover, the
effect of random errors can be reduced
by taking a larger sample.
Whenever the latter is not feasible, it
is sometimes possible to obtain at least
some data for exposure and model input
variables. This substantially reduces the
amount of quantitative uncertainty for
estimation of the distribution of expo-
sure.
This Project Summary was developed
by EPA's Office of Health and Environ-
mental Assessment, Washington, DC.
to announce key findings of the research
project that is fully documented in a
separate report of the same title (see
Project Report ordering information at
back).
Introduction
This document is written for the profes-
sional who is actively involved in the per-
formance of exposure assessments. It is
intended to facilitate use of the latest sta-
tistical methods for characterizing uncer-
tainty for exposure assessments.
The introductory chapter of the full
report discusses the nature of an expo-
sure assessment which serves as a point
of reference for the remainder of the
report. An overview of the characteriza-
tion of uncertainty for exposure assess-
ments at various stages of refinement is
also presented.
Exposure Assessment
An exposure assessment quantitatively
estimates the contact of an affected popu-
lation with a substance under investiga-
tion. Typically, the magnitude, duration,
and/or frequency of contact are estimat-
ed. The U.S. Environmental Protection
Agency has prepared a draft handbook to
provide guidance pertinent to performing
an exposure assessment. In addition, the
-------
Office of Science and Technology and the
National Research Council have also
developed valuable sources of informa-
tion with regard to the methodology of
exposure assessments.
An integrated exposure assessment
quantifies the contact of population mem-
bers with the substance under investiga-
tion via all routes of exposure (inhalation,
ingestion, and dermal) and all pathways
from sources to exposed individuals (e.g.,
from manufacturing plants to plant work-
ers and/or consumers of manufactured
products). It also estimates the size of
population affected by each exposure
route. Since an integrated exposure as-
sessment is often needed to enable regu-
laboratory decisions, characterization of
uncertainty is addressed in this context.
The overall uncertainty associated with
both the exposure assessment and the
dose-response relationships must be
addressed by risk assessments. Charac-
terization of uncertainty for exposure
assessments is addressed in the full
report; other comparable documents ad-
dress characterization of uncertainty for
risk assessments per se. The primary
purpose of an exposure assessment is to
enable risk assessment and possible
selection of a regulatory option. Regula-
tory decisions must be capable of with-
standing close scrutiny (perhaps even in a
court of law). Therefore, the characteriza-
tion of uncertainty for an exposure as-
sessment should be very thorough and
based upon state-of-the-art quantitative
methods to the extent possible.
The type of exposure (internal dose,
external dose, etc.) investigated by an
integrated exposure assessment depends
upon the input needed to enable risk
assessment. In most cases an exposure
assessment should ideally quantify the
dose of the substance under investigation
that is absorbed by individual body organs
in order to enable the most reliable esti-
mates of health effects. Presently it is
practical to quantify absorbed organ-
specific doses only in certain circumstanc-
es—e.g., when the affected organ is the
skin (dermal route) or when exposure is
due to radiation. Moreover, the absorbed
whole-body dose often cannot be quanti-
fied. Instead, indicators of the absorbed
dose are often used by necessity in expo-
sure assessments. These indicators of
absorbed doses may be either environ-
mental exposure or body burden. Envi-
ronmental exposure refers to the levels of
the substance under investigation con-
tacted through the air, water, food, soil,
etc. Body burden refers to levels of the
substance in body fluids or tissues, e.g.,
blood, urine, breath, fatty tissues, hair,
nails, etc.
An integrated exposure assessment
generally partitions the affected popula-
tion into subpopulations such that all
members of a subpopulation are affected
by the same sources of exposure. For
each source, there can be one or more
pathways by which the substance under
study travels from the source to the
affected population. For each source and
pathway, a population member can be
exposed via one or more of three absorp-
tion routes: inhalation, ingestion, and
dermal. An integrated exposure assess-
ment often begins by estimating expo-
sure for a subpopulation via a specific
combination of source, pathway-,^ and
route. Exposure is then aggregated across
pathways and sources to estimate expo-
sure via a particular route for subpopula-
tion members.
An exposure assessment addresses the
exposures experienced by the subjects
(human or nonhuman) of a specified pop-
ulation. The most complete characteriza-
tion of these exposures is the population
distribution of individual exposures. If the
population distribution of exposure was
known, then all parameters of the distri-
bution would also be known.
Characterization of Uncertainty
Since it is uncertainty of the predicted
output that is of interest, the probability
distribution of the output variable and the
variance of this distribution are advo-
cated as primary characterizations of the
uncertainty associated with the predicted
output. Within this context, the estimated
distribution of exposures could be re-
garded as a characterization of the uncer-
tainty with regard to the prediction of an
individual population member's exposure
level. However, estimation of the popula-
tion distribution of exposures is the pri-
mary goal of an exposure assessment,
and it is the uncertainty with regard to
this estimated distribution of exposures
that must be addressed.
The determination of which methods
are appropriate for characterizing the
uncertainty of an estimate depends upon
the underlying parameters being esti-
mated, the type and extent of data avail-
able, and the estimation procedures util-
ized. As a result, a great deal of the full
report actually addresses estimation pro-
cedures for exposure assessments. The
methodology considered most appropriate
for characterizing the uncertainty asso-
ciated with an exposure assessment is
then discussed with regard to each esti-
mation procedure. For example, when the
population distribution of exposure is
being estimated, characterization of un-
certainty addresses the possible differ-
ences between the estimated distribution
of exposures and the true population dis-
tribution of exposures.
An exposure assessment is often based
on one or more exposure scenarios and
associated mathematical models. An
exposure scenario ideally considers all
potential sources of the substance that
could come in contact with population
subjects. Exposure prediction models
based upon transportation and fate of the
substance would then be used to describe
the pathways from sources to contact
with population-members. These models •
may be very complex, using data on levels
of the substance at sources and/or
ambient monitoring sites and the geo-
graphic distribution of the target popula-
tion to estimate the population distribu-
tion of exposures. Alternatively, these
models can be relatively simple models
that predict an individual population
member's personal exposure as a func-
tion of several input variables.
The initial exposure assessment for a
substance is often based upon a very
limited amount of data. Under such condi-
tions, the primary characterization of un-
certainty may be qualitative. Exposure
assessments based upon limited data are
discussed in Chapter 2 of the full report.
Chapter 3 discusses assessments
based upon the estimated joint probabil-
ity distribution of model input variables,
given a model that predicts personal
exposure as a function of the input vari-
ables. Depending upon the extent of data
available to validate the model as well as
the input variable distributions, this type
of assessment has the potential for reduc-
ing uncertainty relative to the methods
discussed in Chapter 2. It also has the
potential for a more quantitative charac-
terization of the uncertainties.
The uncertainties associated with an
exposure assessment can potentially be
reduced further by basing the assess-
ment on measured values of model input
variables for a sample of population
members, given a model that predicts
personal exposure as a function of the
input variables, as discussed in Chapter
4. Chapter 5 discusses exposure assess-
ments based on measured exposure lev-
els for a sample of population members,
which has the potential for minimizing
the uncertainties associated with an
assessment.
-------
When the exposure levels experienced
by individual population subjects are
measured, they generally reflect the total
exposure across all sources and path-
ways, and if possible across all routes as
well. However, when the modeling ap-
proach is employed for an exposure
assessment, a model often predicts an
individual population member's exposure
due to a specific combination of source,
pathway, and route. Methods for combin-
ing estimated exposure distributions over
sources, pathways, and routes are dis-
cussed in Chapter 6. Finally, Chapter 7 of
the full report discusses combining the
estimated exposure distributions over
disjointed subpopulations to produce an
estimate of the exposure distribution for
the total population. Methods for charac-
terizing the uncertainties associated with
these estimated exposure distributions
are also discussed in Chapters 6 and 7.
A hypothetical example of a subpopula-
tion exposure assessment that progress-
es through several stages of refinement
is presented in Appendix A of the full
report. The sections of the example in
Appendix A correspond directly to the
sections of the first five chapters of the
full report and are numbered accordingly.
Summary
Virtually all exposure assessments
except those based upon measured expo-
sure levels for a probability sample of
population members rely upon a model to
predict exposure. The model may be any
mathematical function, simple or com-
plex, that estimates the population distri-
bution of exposure or an individual popu-
lation member's exposure as a function of
one or more input variables. Whenever a
model that has not been validated is used
as the basis for an exposure assessment,
the uncertainty associated with the expo-
sure assessment may be substantial. The
primary characterization of uncertainty is
at least partly qualitative, i.e., it includes a
description of the assumptions inherent
in the model and their justification. Sensi-
tivity of the exposure assessment to
model formulation can be investigated by
replicating the assessment for plausible
alternative models.
When an exposure assessment is based
upon directly measured exposure levels
for a probability sample of population
members, uncertainty can be greatly
reduced and described quantitatively. In
this case, the primary sources of uncer-
tainty are measurement errors and sam-
pling errors. A thorough quality assur-
ance program should be designed into the
study to ensure that the magnitude of
measurement errors can be estimated.
The effects of all sources of random error
should be measured quantitatively by
confidence interval estimates of parame-
ters of interest, e.g., percentiles of the
exposure distribution. Moreover, the ef-
fect of random errors can be reduced by
taking a larger sample.
Whenever the latter is not feasible, it is
sometimes possible to obtain at least
some data for exposure and model input
variables. These data should be used to
assess goodness of fit of the model and/or
presumed distributions of input variables.
This substantially reduces the amount of
quantitative uncertainty for estimation of
the distribution of exposure and is strong-
ly recommended. It is recognized, how-
ever, that it may not always be feasible to
collect such data.
-------
Roy W. Whit more is with Research Triangle Institute, Research Triangle Park, NC
27709.
James W. Falco is the EPA Project Officer (see below).
The complete report, entitled "Methodology for Characterization of Uncertainty in
Exposure Assessments," (Order No. PB 85-240 455/AS; Cost: $ 17.50, subject
to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC 20460
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
BULK RATE
POSTAGE & FEES Pfi
EPA
PERMIT No. G-35
Official Business
Penalty for Private Use $300
EPA/600/S8-85/009
------- |