United States
Environmental Protection
Agency
Office of Health and
Environmental Assessment
Washington, DC 20460
Research and Development
EPA/600/S8-90/065 Sept. 1990
&EPA Project Summary
Statistical Methods for
Estimating Risk for
Exposure Above the
Reference Dose
A statistical method has been
developed that provides a risk
estimate for noncarcinogenic effects
at a given dose. The method uses a
categorical regression procedure to
model severity of effect as it relates
to experimental dose. Toxicity data
are analyzed from multiple animal
experiments that span different
species, target organs, toxic effects,
and exposure conditions. The data
are screened for homogeneity with
respect to experiment duration and
route of exposure. The resulting
dose-response curve provides an
estimate of the risk of adverse effects
that may be useful in estimating risk
for exposures above the reference
dose (RfD).
This Project Summary was developed
by EPA's Environmental Criteria and
Assessment Office, Cincinnati, OH, 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
The U.S. Environmental Protection
Agency (U.S. EPA) is charged with the
responsibility of protecting public health
from environmental pollutants. In this
capacity, the U.S. EPA establishes a
reference dose (RfD) for noncancer
toxicity of individual chemicals (U.S. EPA,
1988; Barnes and Dourson, 1988). The
U.S. EPA's formal definition of the RfD is:
"An estimate (with uncertainty
spanning perhaps an order of
magnitude) of a daily exposure to the
human population -(including
sensitive subgroups) that is likely to
be without appreciable risk of
deleterious effect during a lifetime."
The determination for the RfD for a
given chemical involves several
judgmental steps. First, the literature on
its toxic effects is evaluated. The most
scientifically sound study with the most
appropriate NOAEL (no-observed-
adverse-effect level) of the critical effect
is then generally chosen, and that
NOAEL is divided by uncertainty factors
to arrive at the RfD. To date, there has
been much concern regarding the risk of
adverse effects for exposures above the
RfD, but no reliable method for
estimating this risk has been developed.
The evaluation of toxicity data for
noncarcinogens is complicated by the
multiplicity of possible, endpoints, and the
variation both in the severity of effect and
in the response rate. Standard dose-
response models most often assume a
fixed severity (e.g., lethal) and endpoint
(e.g., cancer), and would then need to be
generalized into a multivariate form to be
applicable to noncarcinogenic effects.
Since response rates for noncarcinogenic
effects are rarely reported, multivariate
dose-response models are seldom
developed, and "dose-response" analysis
usually relates dose only to the severity
of the observed effects. The approach
presented here assigns the severity
descriptions to ordered categories and
-------
models the "dose-category" relationship.
Modeling the risk of adverse effects using
categorical regression was proposed
previously (Hertzberg and Miller, 1985;
Hertzberg, 1987), but the statistical
algorithm then used was limited.
Results and Conclusions
Several different computer programs
and groupings of the data have been
investigated in order to find the best
approach for applying categorical
rogression to this dose-severity
modeling. Much progress has been
made: available data were put in a form
conducive to such an analysis, a model
was selected, a statistical algorithm was
found to perform the analysis, ways of
presenting the results were explored,
some goodness-of-fit measures were
evaluated, and a mainframe programming
package was written to perform the
analysis. For this document, a logistic
transform was used to regress the
severity of effects on the covariate, dose.
Graphs displaying s-shaped dose-
response curves with 95% confidence
bands were then generated. The curves
are statistically derived and use all of the
available toxicity data in determining the
risk of adverse effects at given doses.
The proposed regression procedure
should be useful in risk-based decisions
that can directly use animal data. For
example, the Margin of Exposure (MOE)
method compares the existing exposure
with the NOAEL for the critical effect.
The MOE method could then be
augmented by considering the estimated
risk at the existing exposure and not just
the MOE ratio.
The progress described above does
not, however, provide a final solution to
the risk problem. Some difficulties arise
because of the data available for analysis.
Almost all of these data are from animal
studies, so the risk estimates produced
by the regression are animal risk
estimates that must be a manipulated
further to produce human risk estimates.
Thus, more research is needed to
produce human risk estimates for
exposures above the RfD. Also, each
record in the data set represents
information from an entire dose group,
not from an individual animal. The
interpretation of the animal risk estimate
depends, then, upon the fact that the unit
of input to the regression is the dose
group.
Recommendations
Additional research is necessary to
ensure that this categorical regression
procedure is successful in estimating risk
levels at specific doses. Examples of
areas requiring further research are as
follows:
1. Methods need to be developed that
will validate the assumptions made
by the model whenever the
regression is performed.
2. The goodness-of-fit measures
provided by the regression
procedure need to be evaluated
relative to their usefulness for dose-
response modeling. New goodness-
of-fit measures need to be
developed.
3. Data should be found and analyzed
that allow characterization of
individual animal effects, instead of
dose group effects.
4. An investigation should be made of
other models that may be superior in
their predictive abilities to the logistic
model.
5. The enigma of the extrapolation of
animal risk to human risk is thus far
unsolved by this process. Ways to
develop a human risk estimate need
to be found.
References
Barnes, D.G. and M.L. Dourson. 1988,
Reference Dose (RfD): Description
and Use in Health Risk Assessment.
Reg. Toxicol. Pharm-acol. 8: 471-
486.
Hertzberg, R. 1987. Fitting a model to
categorical response data with
application to species extrapolation
of toxicity. In: Proceedings of the
26th Hanford Life Sciences
Symposium, Modeling for .Scaling to
Man. October 20-23, 1987. Battelle
Pacific Northwest Laboratories,
Richland, WA. (In press) .
Hertzberg, R. and M. Miller. 1985. A
statistical model for species extrap-
olation using categorical response
data. Toxicol. Ind. Health. 1(4): 43-
57.
U.S. EPA. 1988. Reference Dose (RfD):
Description and Use in Health Risk
Assessments. Integrated Risk
Information System (IRIS). Online.
Intra-agency Re-ference Dose (RfD)
Work Group, Office of Health and
Environmental Assessment, Environ-
mental Criteria and Assessment
Office, Cincinnati, OH.
•fr U.S. GOVERNMENT PRINTING OFFICE: 1990/748-012/20096
-------
-------
Rick Hertzberg is the EPA Project Officer (see below).
The complete report, entitled "Statistical Methods for Estimating for Exposure
Above the Reference Dose," (Order No. PB 90-261 504/AS; Cost: $.8,00
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:
Environmental Criteria Assessment Office
Cincinnati, OH 45268
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use S300
EPA/600/S8-90/065
------- |