600Z92002
Friday
June 5, 1992
Part V
Environmental
Protection Agency
Draft Report: A Cross-Species Scaling
Factor for Carcinogen Risk Assessment
Based on Equivalence of mg/kg3/VDay;
Notice
-------
24152
Federal Register / Vol. 57, No. 109 / Friday. June 5. 1992 / Notices
ENVIRONMENTAL PROTECTION
AGENCY
[FRL-4139-7]
Draft Report: A Cross-Species Scaling
Factor for Carcinogen Risk
Assessment Based on Equivalence of
mg/kg3/4/Day
AGENCY: U.S. Environmental Protection
Agency.
ACTION: Request for comments on the
draft report: A Cross-Species Scaling
Factor for Carcinogen Risk Assessment
Based on Equivalence of mg/kg3'4/day.
SUMMARY: Three Federal regulatory
agencies, the Environmental Protection
Agency, the Food and Drug
Administration, and the Consumer
Product Safety Commission, are today
asking for public comments on the draft
report: A Cross-Species Scaling Factor
for. Carcinogen Risk Assessment Based
on Equivalence of mg/kg 3'4/day.
The report is intended to serve as the
basis for a common and unified science
policy among these three agencies on a
default methodology for determining
equivalence of doses—to be used when
existing agent-specific data are
insufficient for a case-by-case
determination—when extrapolating
results of rodent carcinogen bioassays
to humans.
The public is invited to comment, and
public comments will be considered in
final revision of the report and in the
final adoption of science policies by the
participating agencies on cross-species
extrapolation of equivalent doses in
assessing potential human risks from
putative chemical carcinogens.
Commenters are asked to focus on the
report's discussion of several issues: (1)
The bearing of empirical data on
carcinogenic potencies in experimental
animals and in humans to the
appropriate choice of a dose-scaling
methodology; (2) the use of allometric
scaling as a means for suggesting
appropriate dose scaling methods; (3)
the appropriate use of pharmacokinetic
and other data in defining a default
methodology and particularly in
supplanting such default assumptions
with case-specific, data-based analysis
of dose equivalence; (4) distinguishing
the contributions of pharmacokinetic
and pharmacodynamic factors to
species differences in a carcinogen's
potency; and (5) the advisability of
adopting the proposed dose-scaling
methodology as a common default
methodology for the participating
agencies.
The complete text of the draft report
is published as the last section of this
notice.
DATES: The draft document is being
made available for public review and
comment until August 4,1992.
Comments must be in writing and must
be postmarked by August 4,1992.
INSPECTION AND COPYING: This notice,
references, supporting documents, and
other relevant materials are available
for inspection and copying from the
ORD Public Information Shelf at the
EPA Headquarters Library, 401 M Street,
SW., Washington, DC, Telephone: (202)
260-5926 or FTS: 260-5926. The Library
is open daily between the hours of 8
a.m. and 5:30 p.m., except weekends and
holidays.
ADDRESSES: Comments may be mailed
or delivered to: Project Officer for Cross-
Species Scaling Factor Report, c/o
Technical Information Staff, Office of
Health and Environmental Assessment,
U.S. EPA (RD-689), 401 M Street, SW.
(room 3703), Washington, DC 20460.
FOR FURTHER INFORMATION CONTACT:
Dr. Lorenz Rhomberg, Human Health
Assessment Group, Office of Health and
Environmental Assessment, U.S. EPA
(RD-689), Washington, DC 20460,
Telephone: (202) 260-5723 or FTS: 260-
5723.
SUPPLEMENTARY INFORMATION: This
document reports a consensus reached
by representatives of the U.S.
Environmental Protection Agency (EPA),
the Food and Drug Administration
(FDA), and the Consumer Product Safety
Commission (CPSC) in discussions
conducted under the auspices of the
Interagency Pharmacokinetics Group, a
workgroup of Federal scientists dealing
with issues of common interest arising
in the application of pharmacokinetics
to chemical health risk assessment. The
report is a product of the Interagency
Pharmacokinetics Group. It comprises
an analysis of empirical and theoretical
aspects of the cross-species dose-scaling
question, together with an argument for
adopting the method of scaling daily
administered doses by body mass raised
to the % power to achieve presumed
equivalence in lifetime carcinogenic risk
in different mammalian species. These
recommendations have been reviewed
and endorsed by the EPA, the FDA, and
the CPSC.
If such a policy is adopted, it would
replace the current practices in
carcinogenic risk assessment of scaling
daily administered amounts by body
mass (as at FDA) or by body surface
area (as at EPA and CPSC). The
consensus recognizes that there is
considerable scientific uncertainty
around any scaling method; it does not
claim to have overturned these previous
methods with one of superior scientific
validity or reduced uncertainty. Rather,
in view of the benefits of having the
major practitioners of carcinogen risk
assessment in the Federal government
adhere to a single, consistent
methodology, the proposal provides a
common default procedure to encourage
consistent analyses in cases where
agent-specific information is insufficient
to suggest appropriate dose-
equivalencies on a case-by-case basis.
Such case-specific information is always
to be preferred to the default
methodology proposed herein, and its
development and appropriate use are
encouraged. Since the scaling
methodologies in current use by the
agencies participating in this proposal
are within the span of scientific
uncertainty surrounding the cross-
species scaling question, it is not
proposed to retroactively change or
adjust any risk assessments completed
under current policies.
This document has undergone a
preliminary interagency review under
the auspices of the Ad Hoc Working
Group on Risk Assessment of the
Federal Coordinating Council for
Science, Engineering, and Technology
(FCCSET). This request for public
comment and a concurrent external
scientific peer review will contribute to
the development of a final report on this
topic. This final report of the
Interagency Pharmacokinetics Group
will provide the basis for a
recommendation of a uniform, default
science policy on interspecies scaling for
carcinogen risk assessment, to be
endorsed by the FCCSET Working
Group and used by a broad segment of
Federal agencies.
Dated: May 22,1992.
F. Henry Habicht II,
Deputy Administrator.
Contents
I. Introduction
II. Approaches to Choosing a Cross-Species
Scaling Factor
A. Empirical Approach
B. Allometric Approach
1. Species Differences in Pharmacokinetics
2. Species Differences in
Pharmacodynamics
3. Toxicological Equivalence
4. A Physiological Time Approach to
Toxicological Equivalence
III. Discussion
IV. Conclusions
V. References
A Cross-Species Scaling Factor for
Carcinogen Risk Assessment Based on
Equivalence of mg/kga/4/Day
I. Introduction
As a matter of necessity, the potential
for a chemical agent to cause toxic
reactions in humans is often
-------
Federal Register / Vol. 57. No. 109 / Friday, June 5. 1992 / Notices
24153
investigated by exposing and observing
the reactions of experimental animals,
usually rats and mice. This practice
rests on the high degree of physiological,
biochemical, and anatomical similarity
among mammalian species; the
biological reactions in the experimental
animals may be taken as evidence that
humans might show similar responses to
the same chemical exposures. When the
objective is to use the animal data to
predict the degree or probability of
response in humans—that is, when the
aim is quantitative extrapolation—one
must define the dose levels for humans
and animals that are expected to
produce the same degree of effect. For
this, it is necessary to take into account
the pronounced difference in scale
between the tested model organisms
and humans. That is, even if
fundamental similarity is presumed, one
must allow for the fact that humans are
much larger than experimental rodents
and will experience chronic exposure to
a toxicant for a longer lifetime.
Defining such "lexicologically
equivalent" doses has been problematic.
Alternatives that have found use include
scaling daily administered amounts by
body weight or by body surface area;
scaling cumulative lifetime intake by
body weight; equating exposures to
contaminated air, food, or water
according to the concentration of toxic
agent; and others. Despite considerable
study and debate (Pinkel, 1958; Freireich
et al., 1966; Mantel and Schneiderman,
1975; Rail, 1977; Hoel, 1977; Hogan and
Hoel, 1982; Calabrese, 1983,1987; Crump
et al., 1985; Davidson et al., 1986;
Gillette, 1987; Vocci and Farber, 1988;
Hill et al., 1986), no alternative has
emerged as clearly preferable, either on
empirical or theoretical grounds. The
various Federal agencies conducting
chemical risk assessments have
developed their own preferences and
precedents for cross-species scaling
methodology. This variation stands
among the chief causes of variation
among estimates of a chemical's
potential human risk, even when
assessments are based on the same
data.
The variety of cross-species scaling
methods in use correctly reflects the
uncertainty about the best procedure,
but the resulting disagreement in risk
estimates results in some awkwardness
in the regulatory arena. Increasingly,
regulatory procedures are being
mandated that establish decision points
contingent on whether a certain human
risk level is to be expected according to
"generally accepted" risk assessment
procedures. Variation in methodology
frequently leads to ambiguity as to
whether regulatory action should take
place. It has therefore become important
to resolve differences in cross-species
scaling assumptions.
A second impetus for reexamining the
scaling question comes from the
increasing availability of comparative
pharmacokinetic information on toxic
agents. Pharmacokinetic analysis uses
data on absorption of agents into the
body, distribution among the tissues,
metabolic activation or detoxification,
and elimination to develop a picture of
the disposition of a dose by the body
and consequent exposure of the actual
target tissues of toxic action.
Pharmacokinetic differences among
species clearly contribute to the
magnitude of equipotent doses.
However, the appropriate use of such
information for the dose equivalency
question hinges on resolving the role of
pharmacokinetics compared to that of
species differences in the magnitude of
toxic reaction to a given degree of
target-tissue exposure (i.e.,
"pharmacodynamics"). Distinguishing
the roles of these two aspects of potency
scaling has been hampered by
imprecisely articulated rationales for the
various methods.
In view of the above considerations,
the Federal agencies with primary
responsibility for conducting chemical
risk assessments have endeavored to
define a uniform cross-species scaling
methodology and rationale for use when
extrapolating results of rodent
carcinogen bioassays to humans.
Discussions and debate on the issues
have been held under the auspices of the
Interagency Pharmacokinetics Group
(IPG), an ongoing workgroup of Federal
scientists that deals with issues of
common interest arising in the
application of pharmacokinetics to risk
assessment. The present report is a
product of the Interagency
Pharmacokinetics Group, and represents
a statement of the consensus
recommendation resulting from these
discussions.
The consensus is that, in the absence
of adequate information on
pharmacokinetic and sensitivity
differences among species, doses of
carcinogens should be expressed in
terms of daily amount administered per
unit of body mass raised to the %
power. Equal doses in these units (i.e., in
mg/kg3' 4/day), when experienced daily
for a full lifetime, are presumed to
produce equal lifetime cancer risks
across mammalian species. This
proposed scaling method has the
advantage of being intermediate
between the two currently used methods
(scaling daily amount by body mass or
by body surface area). It is not merely a
compromise; it is as well supported by
the empirical data on carcinogen
potencies in animals and humans as the
methods it would replace. It also has an
explicit rationale (the concept of
species-independent "physiological
time") that may be derived from
principles of interspecific allometric
variation in anatomy, physiology, and
pharmacokinetics. That is, it can be
interpreted as a correction for readily
observable scale differen ;es among
species as their essentially similar
biology varies in a regular quantitative
way as a function of size.
The consensus does not pretend to
have solved the underlying scientific
issues. Former methodologies have not
been shown to be in error; the
consensus should not be construed as
overturning previous assumptions and
replacing them with one of superior
scientific validity. Rather, the consensus
achieves the benefits of having all
Federal risk assessments adhere to a
single, consistent methodology that is in
accord with current scientific knowledge
on the scaling question. Moreover, the
method corresponds to a fully
articulated rationale with explicitly
stated assumptions about the roles and
interactions of various underlying
determinants of carcinogenic potency.
This aids in consistent and scientifically
appropriate application. Furthermore, as
information is gained on how the
biology of carcinogenesis varies among
species, it will be clearer how the
arguments and previous presumptions
should be modified to accommodate
these new insights.
The balance of this document reviews
the evidence and arguments that may be
adduced to address the question of
cross-species scaling of equally
carcinogenic doses, and outlines the
support for the recommended position of
equipotent doses in terms of mg/kg3'4/
day.
II. Approaches to Choosing a Cross-
Species Scaling Factor
There are two broad and
complementary approaches to choosing
a cross-species scaling factor. The first
is empirical; one may seek cases in
which human epidemiologic data allow
a direct estimate of an agent's potency,
and then investigate the success of
various scaling methods in predicting
that potency from animal data. The
second approach is theoretical, and is
grounded in the principles of allometry,
which is the study of the regular
variation in features of anatomy and
physiology as a function of overall body
size. The strategy for this second
-------
24154
Federal Register / Vol. 57, No. 109 / Friday, June 5, 1992 / Notices
approach is to develop a scientific
rationale for a particular scaling factor
by investigating the allometric variation
of the biological features and processes
that influence and underlie carcinogenic
potency.
Clearly, in many cases there will be
agent-specific ways in which humans
and experimental animals differ in a
nonsystematic fashion. These may
include metabolic activation or
detoxification, interaction with key
receptors or target molecules, and
others. Such factors create
unpredictable deviation from the general
pattern of scaling, and must be
discovered and accounted for on a case-
by-case basis. The factor proposed here
is a default scaling factor, by which is
meant one that is to be applied in the
absence of adequate case-specific
information. Lacking such information,
one provisionally assumes that the agent
in question is an example of a "typical"
or "average" chemical that follows a
general pattern of cross-species potency
differences. This presumption may be
modified as information becomes
available, but the default assumptions
still serve as the benchmark against
which the new information is evaluated.
A. Empirical Approach
This approach attempts to find a
factor value that is empirically
successful in producing good estimates
of potency in humans from data on
potencies in other species. The
underlying reason why such a factor
works is a secondary consideration. The
advantage of an empirical approach is
that, by directly examining carcinogenic
potencies (rather than influences on
potency, such as pharmacokinetics), all
relevant factors are included. The
disadvantage is that the data are few
and of low resolution. One must hope
that the agent-specific factors,
mentioned above, average out to give a
good estimate of the general
relationship.
A number of studies have sought
general scaling factors empirically.
Freireich et al. (1966), testing and
extending the suggestion of Pinkel
(1958), examined maximum tolerated
doses (MTDs) of 18 antineoplastic drugs
in mice, rats, hamsters, dogs, monkeys,
and humans. LDios were used for
rodents, and were presumed to be an
equivalent level of toxicity to an MTD.
Doses from experiments of different
length were reexpressed in terms of an
exposure regimen of 5 consecutive days,
on the assumption that cumulative dose
is proportional to effect. The authors
concluded that, when doses were
expressed as mg/m2 body surface area/
day, good predictions of human MTDs
were obtained from all animal species,
but that body weight scaling of doses
overpredicted human MTDs (i.e.,
underpredicted potency in humans) by a
margin that increased as one
extrapolates from smaller and smaller
species. Since an MTD is intended to be
a dose causing no lethality, while an
LDio causes 10% lethality, the
equivalence of these two end points can
be questioned. Antineoplastic drugs
typically have very steep dose-response
curves, however, and survival near the
MTD is maintained by close monitoring
and intervention, which the rodent LDio
determination lack.
Collins et al. (1986,1990) have found
that the human MTD for 16
antineoplastic drugs is well predicted on
average by the mouse LDio when doses
are expressed as mg/m2 of body surface
area. (If the MTD is considered to be a
less severe end point, in such
comparisons potencies in the larger
species are overestimated vis-a-vis
those in rodents; a bias would then be
created that would increase the
apparent success of surface area scaling
compared to scaling by body weight.)
That is, if these endpoints of acute
toxicity are taken as equivalent, scaling
doses in proportion to surface area
tends to equalize toxicity across species.
Moreover, Collins et al. (1990) compared
the blood levels (in terms of the areas-
under-the-curve of concentration in
plasma as it declines over time, or "C x
T") that correspond to equally toxic
administered doses and found that these
were an even better predictor, in that
they displayed less case-by-case
variation. These results illustrate three
points that are returned to in Section B,
below: (1) Scaling administered doses in
this way tends to equalize blood levels
across species; (2) areas-under-the-curve
of blood concentration can serve as a
predictive measure of the toxic response
to a dose, even across species; and (3)
obtaining pharmacokinetic data on
internal dose measures can increase the
precision of the cross-species prediction
of equivalently toxic doses by
accounting for case-by-case variation.
Travis and White (1988) reanalyzed
the Freireich et al. (1966) data set and
nearly doubled the number of drugs by
adding a similar data set of Schein et al.
(1979). Instead of simply examining the
success of prevously proposed scaling
methods, they used regression
techniques empirically to determine the
optimal power of body weight to
achieve the best fitting allometric
relationship of MTDs across species. For
both data sets individually and for the
combined data set, a power of 0.72 to
0.74 led to the best cross-species
predictions. In the analysis of the
combined data, a power of unity (body
weight scaling) was clearly rejected at
the 95% level of significance, and a
power of 2/3 (surface area scaling) was
barely rejected. The authors discuss the
history of empirical studies of allometric
variation in a number of physiological
features, primarily basal metabolism,
and arque that their result is part of a
general empirical support for scaling by
the 3/4 power of body weight.
The difficulty with applying these
studies to the present question is that
they address acute systemic toxicity of a
rather narrowly defined type rather than
carcinogenesis. Although dose-scaling
for different toxic end points should
have some features in common (notably
pharmacokinetics), it is not altogether
clear how lifelong risks that accumulate
over time (such as cancer risk) should
relate to short-term toxicity dependent
only on immediate insults to target
tissues.
Some empirical studies of
comparative potencies of carcinogens in
different species have been done. Such
studies face the difficulty of precisely
determining potencies in humans based
on epidemiologic data. There is also
some ambiguity in defining potencies in
animals, owing to the variations in rout
of exposure, sex and strain differences,
varying experimental designs, and so on.
Nonetheless, such studies represent the
direct investigation of the question at
hand.
The National Academy of Sciences
(NAS, 1975) examined the potencies of
six carcinogenic agents in bioassays
using mice and rats and from human
epidemiologic studies. They
recommended as a dose measure
cumulative lifetime amount of agent
administered (in mg) per kg body
weight. Such scaling is more
"conservative" (i.e., predictive of higher
human risk from animal results) than
either surface area scaling or body
weight scaling (from which it differs by
a factor of 35, owing to the lack of
adjustment for differences in length of
lifetime). The NAS conclusion was not
based on formal quantitative
comparison with surface area scaling
(mg/kg2/3/day) or body weight scaling.
The paucity of carcinogen potencies in
humans known directly from
epidemiologic data limits the precision
of such comparisons. Crouch and
Wilson (1979) instead investigated dose
scaling between rats and mice in about
70 ingestion cancer bioassays from the
National Cancer Institute testing
program. They measured potency by the
parameter of a fitted one-hit dose-
response model (in units of risk per mg/
-------
Federal Register / Vol. 57. No. 109 / Friday, June 5. 1992 / Notices
24155
kg/day), focusing on the tumor site/type
producing the greatest potency
(excluding testicular tumors in Fisher
344 rats, and skipping cases in which
potency was less than twice sensitivity
in either species). A geometric mean of
potencies in each sex (which were
highly correlated) was used.
Interspecies comparisons were based on
the best-fitting line of unit slope on a
plot of the logarithm of potency in rats
against the logarithm of potency in mice.
The intercept of such a line gives the
geometric mean of the factor by which
the rat potency must be divided to give
the mouse potency. Body weight scaling
predicts a factor of one (i.e., equal risk
per mg/kg/day in both species) while
surface area scaling predicts a factor of
about 2.1 to 2.3, depending on the exact
body weights. (For comparison, the
scaling by mg/kg3' 4/day, as advocated
herein, predicts a ratio of about 1.8 or
1.9.) The results depend on the strain of
rat used. In the 17 cases of comparison
between Osborne-Mendel rats and
B6C3F1 mice the mean ratio of potencies
was 0.40; these rats were somewhat less
sensitive than mice, contrary to the
expectations of both scaling
methodologies. When Fischer 344 rats
were compared to the same mouse
strain (18 cases) a mean ratio of 4.5 was
obtained, indicating that rats were even
more sensitive than surface area scaling
would expect. (A geometric mean of
these two ratios is 1.3. To attempt
definition of a general mammalian
cross-species allometric relationship
using only two species is fraught with
pitfalls, especially when they are as
close in size as are rats and mice.
Nonetheless, for the purposes of this
discussion one may note that, using
typical body weights—70 kg for a
human, 40 g for a mouse, 467 g for a rat
of unspecified strain, 500 g for an
Osborne-Mendel rat, and 360 g for a
Fischer rat—the ratio of 1.3 implies
scaling by body weight to the 0.89
power.)
Crouch and Wilson (1979) also
examined ratios of rodent potency to
epidemiologically derived human
potency, comparing "insofar as
possible" studies with the same route of
exposure and duration in fraction of a
lifetime. Owing to imprecision in the
epidemiologically based human
estimates, no precise curve fitting was
attempted, but the authors state that
humans appear to be more sensitive to a
mg/kg/day dose by about a factor of 5
compared to either rats or mice. (Using
the typical body weights listed
previously, a factor of 5 corresponds to
scaling doses by a power of body weight
of 0.7 and 0.8 based on the rat and
mouse results, respectively.)
A similar comparison of rats and
mice, based on an expanded base of 187
NCI bioassays, was conducted by
Crouch (1983). (Despite the larger
original database, there were only a few
more chemicals in the final analysis,
apparently owing to more stringent
requirements for significance of
portency estimates.) Again, the rat
strain influenced the results: for
Osborne-Mendel rats the mean ratio
was 0.63 while for Fischer 344 rats it
was 2.29. (A geometric mean of these
two ratios is 1.20.) Separate analysis of
males and females changed these ratios
only slightly. An analysis irrespective of
rat strain yielded a ratio of 1.62. (Using
the typical body weights listed
previously, rations of 1.20 and 1.62 imply
scaling by body weight to the 0.92 and
0.80 power, respectively.)
Gaylor and Chen (1986) examined
data on rats, mice, and hamsters in the
extensive database of Cold et al. (1984)
on TDsoS, the dose (in mg/kg/day)
leading to a halving of the actuarially
adusted percentage of tumor-free
animals at the end of a standard
lifespan. The tumor site/type showing
highest potency (i.e., lowest TD5o) was
chosen to represent the species, and
only agents with responses in both
species were included. For 190
compounds administered in the diet, the
geometric mean ratio of TD5os in rats
and mice was 0.455=1/2.20. That is, rats
were on average about 2.2-fold more
sensitive. (Using the typical body
weights listed previously, this
corresponds to scaling by body weight
to the 0.68 power.) Ratios for other
routes of exposure varied somewhat,
although based on much lower sample
sizes than the ingestion results cited
above. By gavage, 32 compounds had a
mean ratio 1/1.32, in drinking water 10
compounds had a mean ratio of 1.45 (i.e.,
rats were less sensitive), and by
inhalation 7 compounds had a mean
ratio of 1/11.2 (i.e., rats were much more
sensitive).
Chen and Gaylor (1987) investigated
NCI/NTP cancer bioassays of
compounds administered orally to rats
and mice. They compared "virtually safe
doses" (VSDs), defined as doses
associated with a lifetime cancer risk of
one in a million. These were determined
by the method of Gaylor and Kodell
(1980), i.e., a linear extrapolation was
conducted from an upper bound on a
fitted multistage model dose-response
curve. Thus, both the rat and mouse
VSDs are in some sense "upper
bounds." Chemicals were included if
judged by the NTP to be positive in at
least one species, and when in only one,
if there was at least a positive trend in
the other species for the same tumor
site/type. Unlike the studies mentioned
above, Chen and Gaylor (1987) focused
on Correspondence of VSDs at the same
site and sex across species. VSDs were
expressed in terms of concentration
(parts per million [ppm]); as discussed
further in the following section on
allometry, since intakes of contaminated
media (air, food, water) tend to be
proportional to body surface area, the
expectation from surface area scaling is
that VSDs expressed in ppm would be
about equal across species, while body
weight scaling would expect a ratio of
rat to mouse VSDs to be slightly greater
than 2. Again, the results depend on the
strain of rat used: For Fischer 344 rats
the mean ratio is 1.15, for Osborne-
Mendel rats it is 1.68, and for Sprague-
Dawley rats it is 1.78. Ignoring rat strain
gives a mean ratio of 1.27. These results
are intermediate between the
expectations of surface area and body
weight scaling. For ease of comparison
with other studies, one may convert
these ratios from a ppm basis to a mg/
kg/day basis using empirically based
daily food and water consumption
patterns in rats and mice (for food, 5%
and 13% of body weight for rats and
mice, respectively, and for water, 7.8%
and 17% [U.S. EPA, 1984]). On a mg/kg/
day basis, the ratimouse VSD ratios are
0.44-0.53 for Fischer rats, 0.647-0.771 for
Osborne-Mendel rats, and 0.69-0.82 for
Sprague-Dawley rats. (The range
reflects using rat:mouse ratios of water
and food consumption, respectively,
which differ slightly.) Using the typical
body weights listed previously, and
assuming a weight of 540g for Sprague-
Dawley rats, these ratios correspond to
scaling doses by body weight to the
0.63-0.71 power (when based on Fischer
rats, which constituted most of the
cases), 0.83-0.90 power (when based on
Osborn-Mendel rats), and 0.86-0.92
(when based on Sprague-Dawley rats).
Metzger et al. (1989) expanded
Crouch's (1983) earlier data set by
including all 264 cases from the Gold et
al. (1984) database in which a significant
TD5o was obtained in an oral study of
rats and mice (of any strain), i.e.,
including studies that were not in the
NCI/NTP database. A best-fitting line of
unit slope showed a TD50 ratio of 1.46
between mice and rats. This is
intermediate between the ratio of 1.0
expected from body weight scaling and
2.5 from suface area scaling (using the
authors' assumptions about body
weights—this implies a power of body
weight of 0.86).
-------
24156
Federal Register / Vol. 57, No. 109 / Friday, June 5, 1992 / Notices
A major study of animal-to-human
extrapolation of cancer potencies was
carried out by Allen et al. (1987), and
reported on by Crump et al. (1987,1989}
and Allen et at. (1988). Twenty-three
chemicals were identified that permitted
quantitative evaluation of potency in
humans and in animals. "Risk-Related
Doses" (RRDs) were calculated, defined
as the average daily dose per kg of body
weight that would be expected to result
in an extra cancer risk of 25% over a
lifetime. Chemicals were included even
if RRD estimates were "infinite" for one
species, as happens when no
carcinogenic effect is observed. Unlike
the studies reviewed above, the Allen et
al. (1987) study considered a large
number of alternative ways of
representing the potency in animals as
well as various methods for
extrapolating the resulting RRDs to
humans. Alternative sets of "risk
assessment assumptions" restricted the
animal database according to various
criteria of experimental design, route of
exposure, and tumor type. Different
levels of averaging results over
experiments, sex, and species were
tried. Finally, different methods for
combining the multiple animal results on
a given chemical into a single measure
of its "potency in animals" were
examined. This complexity allows an
admirably comprehensive look at
animal-to-human extrapolation, but it
also makes manifest a problem that is
latent in the other extrapolation studies:
The performance of a scaling factor
depends on how the animal potency is
characterized. A factor that tends to
overpredict human risk can be
"rescued" by a method for characterzing
animal potency that tends to produce a
low estimate, and vice versa.
When the objective is to examine
alternative dose-scaling factors, it would
seem that the best approach is to
examine analyses that aim at broadly
based and unbiased estimates of the
potency in animals. Risk assessment
practices such as using upper bounds on
dose-response curves and extrapolating
from the most sensitive sex and species
of animal are explicitly conservative;
they may be appropriate science policies
for regulatory purposes, but when the
issue is empirically to choose a best-
performing scaling factor, they introduce
a bias, favoring a less conservative
factor to compensate for their
conservatism and restore a good
prediction of the known human potency.
To compare potencies, Allen et al.
(1987) fit a line of unit slope to the data
of epidemiologically observed log RRD
in humans plotted against the predicted
human log RRD based on the animal
data and the chosen scaling
methodology. The intercept of this line
gives an average ratio of the observed to
predicted potency, with a ratio of unity
indicating unbiased prediction. The
analyses discussed prominently in the
Allen et al. (1987,1988) and Crump et al.
(1987,1989) reports show that body
weight scaling leads to a ratio of
approximately one to somewhat less
than one depending on the particular
suite of risk assessment assumptions
chosen (i.e. slightly underpredicting
human risk), while surface area scaling
overpredicts human risk several-fold.
These results are sometimes cited as
tending to support mg/kg/day scaling,
but such a conclusion should be
tempered. The particular choice of risk
assessment assumptions (among many
examined) in the widely cited analysis
is the one with results least favorable to
surface area scaling; most of the
alternatives discussed by Allen et al.
(1987) show that body weight scaling
underestimates human risks by about
the degree to which surface area
overestimates it. Moreover, these
analyses contain a bias of the sort
outlined above—the animal potency for
a chemical is characterized by the
median of the lower bounds on the
RRDs for the various animal data sets
rather than on best estimates. At present
it is unresolved how much the use of
central estimates of animal risk to
predict central estimates of human
risk—a more appropriate analysis for
resolving the scaling factor—would shift
the results toward favoring surface area
scaling.
Two additional studies of
comparative cancer potencies should
briefly be mentioned, both favoring a
somewhat more conservative scaling
factor. Raabe et al. (1983) compared
bone cancer risks from radium in watch
dial painters (who ingested radium by
tipping brushes on their tongues) and in
beagle dogs exposed to radium by
injection. Doses were measured as dose
to bone of deposited radium, so this
camparison can be seen as lacking the
pharmacokinetic component of cross-
species differences. Potency was
measured by the relative mean degree of
life-shortening as a function of does. The
authors argued that a cumulative
lifetime radiation dose per unit of bone
seemed to give good correspondence
between human and dog. This result
could be related to mg/kg/lifetime
scaling for chemical agents.
Kaldor et al. (1988) examined
carcinogenic potency of five
antineoplastic drugs, using potencies
derived from bioassays in rodents and
from the secondary tumors the drugs
caused in human cancer patients. They
argued that potency seemed to be
related to total cumulative lifetime
exposure per kg of body weight.
The empirical evidence on cross-
species scaling of carcinogen potencies
can be summed up as follows. The
correlation of agents' potencies across
species is clearly and strongly
demonstrated. This correlation extends
to humans, so far as is ascertainable
from the limited number of agents for
which potencies can be estimated
epidemiologically. There is a remarkable
agreement among studies that the dose-
scaling methods in current use span a
range that appears approximately
correct. The resolution of the data
available at present, however, does not
permit a clear choice between surface
area and body weight scaling.
Empirically chosen scaling factors tend
to fall in between these two choices in
most cases, but the specific results
depend on the laboratory strains used,
route of administration, details of the
methods for characterizing the
carcinogenic potency in animals, and
the statistical methods used in curve
fitting. The data seem consistent in
indicating that body weight scaling
somewhat underestimates risks in larger
species. The exception is when
Osborne-Mendel or Sprague-Dawley
rats are compared to B6C3F1 mice, in
which comparison the rats are seen to
be less affected even by doses scaled to
body weight. The preponderance of data
are from Fischer 344 rats, however, and
this is the strain used in most modern
bioassays.
Several points should be borne in
mind while interpreting the empirical
scaling data. First, although several
studies are reviewed, they overlap
considerably in their databases; the
individual studies are not independent
tests. Second, the specific results of a
study depend on details of the
methodology. The Allen et al. (1987)
study showed that whether potencies
were averaged over sexes, whether both
benign and malignant tumors were
counted, whether projections were made
for specific tumor sites or for the most
potent site, and other such factors could
swing the analysis toward favoring one
scaling method or another. It is hard
confidently to identify and isolate the
specific contribution of dose scaling
among the many factors that contribute
to the final predictions of human risk.
Third, the epidemiologically based
human potencies that serve as "targets"
for the animal-based extrapolations are
themselves very uncertain and, as in the
animal data, dependent on the specifics
of the methodology used in their
-------
Federal Register / Vol. 57, No. 109 / Friday, June 5, 1992 / Notices
24157
estimation. As a result of this and of the
previous point, the comparability of
animal- and human-based potencies
may be problematic. (For example,
potencies calculated from human data
are usually based on cancers that were
the cause of death following partial
lifetime exposure, while animal-based
estimates usually reflect incidental as
well as fatal tumors arising after full
lifetime exposure.) A final point to be
borne in mind is that the report
empirically derived factors represent
averages over large numbers of cases.
Although the means vary over a narrow
range, the individual chemicals show
ratios of potencies in different species
that span orders of magnitude. Most of
the rat-to-mouse comparisons were
within an order of magnitude of the
average scaling relationship, but several
agents showed a 100-fold difference.
Variances of rodent-to-human potency
ratios were higher, reflecting the
uncertain determination in humans and
the lack of standardized experimental
design. The existence of this scatter of
cases around the mean helps to define
the limits to the resolution of any scaling
method and emphasizes the importance
of case-to-case variation. Moreover, it
provides some insight into the
distribution of uncertainty in the cross-
species dose extrapolation step of risk
assessment.
Despite these shortcomings, the
empirical data support the general
practice of scaling rodent potencies to
humans, and show that, on average, the
current methods perform satisfactorily.
Certainly, any method that produces
average results an order of magnitude
higher or lower than the range
represented by body weight and surface
area scaling would be in contradiction
to the empirical data. The data suggest
that a scaling factor in between the
surface area and body weight scaling
can be considered to have empirical
support.
B. Allometric Approach
The complement to the empirical
investigation of potency scaling is a
more theoretical approach that seeks to
identify the biological factors whose
variation underlies the variation in a
carcinogen's potency across species,
and then attempts to adjust for their
effect. Clearly, these factors are
numerous and, for the most part, poorly
understood. Fortunately, there are some
rather simple and general quantitative
patterns in the variation of many
features of anatomy and physiology
across differently sized mammalian
species, representing broad trends in the
way the essentially similar mammalian
system operates in large and small
editions. Although specific processes
acting on specific chemicals can (and
do) deviate from these broad trends, it is
argued below that the general patterns
can provide a benchmark that expresses
the expectation about a chemical's
carcinogenic potency in small mammals
such as experimental rodents and larger
ones such as humans. This expectation
.can be refined (or refuted) by case-
specific biological and mechanistic data,
when available, showing how the actual
processes of metabolism and
carcinogenesis differ from the
presumptions of the broad trend
analysis that serves as the default.
The aim of a dose-scaling
methodology is to estimate administered
daily doses to experimental rodents and
humans that result in equal lifetime
cancer risks. That is, the scaled doses
are intended to be "toxicologically
equivalent." It is useful to recognize two
components to this equivalence. The
first, which might be termed
"pharmacokinetic equivalence,"
concerns adjustment of the administered
dose to a rodent or human so that the
corresponding tissues that constitute the
targets of the agent's toxicity receive
similar exposures to the toxin. The
second, or "pharmacodynamic
equivalence," relates to the relative
tissue doses that, when experienced
daily for a lifetime, yield equal lifetime
cancer risks. This latter aspect includes,
but goes beyond the question of
"sensitivity" to address species
differences in the operation of the
carcinogenic processes as they relate to
tissue does. For both the
pharmacokinetic and the
pharmacodynamic component, scaling
questions arise and the problem of
defining "equivalence" must be faced.
By way of illustration, consider a
hypothetical agent with rather simple
pharmacokinetics (first order
elimination from a single compartment)
given by intravenous injection to a
mouse and a human. As shown in Figure
1, such a compound will demonstrate an
almost instantaneous peak in its blood
concentration, followed by exponential
decline. If the administered doses are
equal in terms of mg/kg body weight,
the peak concentrations are the same in
the mouse and the human, but the
mouse rids itself of this body burden
faster, owing to its more rapid
metabolism and elimination compared
to the human. As a result, the area under
the curve (AUC) of blood concentration
as it declines with time is much less in
the mouse. If the amount injected is
properly adjusted, as illustrated in
Figure 2, a concentration profile can be
achieved in which the initial peak blood
Concentration is much less in the human,
and yet is balanced by the compound's
longer persistence to generate an AUC
equal to that of the mouse.
BILLING CODE 6560-50-M
-------
CONCENTRATION
Iff
ZJ $
rr <5. -x
03 CQ •
03 03 CL
CD q o
» O §
c c 2
=3 GO Q
Q. CD ^
CD ^ Pi
0 5' ,
I J,g
CD 03 9
Is'
' if
D^-^ 3
CD =:• 00
^ CD ®
"^ C/)
O
H»
^- CD
CD Q.
C
H
m
/ 2661 'S sunf 'Aepuj / 601 'ON 'Zfi 'PA /
paapaj
-------
ro
i—
CONCENTRATION
§
03
O
0 O
•k • •
D -i 10
p p p p
co £* en b>
i i i
P P P
•**i m
If!
® w S
=- c
£D 1-1-
CD 5'5-
CDCi'D
-Q — o
^^ ^J ^^^
E. ^ o
m
CO
O)
O
00
O
o
o
to
o
1
CD
o>
o
oo
o
ro
o
o
/ Z661 'S aunf 'Xepuj / 601 '«N '/S '!OA /
-------
24160
Federal Register / Vol. 57, No. 109 / Friday, June 5, 1992 / Notices
This example illustrates two points:
that knowledge of a compound's
pharmacokinetics can suggest scaling of
administered doses so as to equalize the
exposure of internal targets of toxicity,
and that "equal" internal exposure
requires further definition. The area
under the concentration curve
encompasses both the amount of a
compound that is present and the
duration of its presence, providing a
measure of the compound's opportunity
to interact with the targets of toxicity.
Moreover, since the AUC is the integral
of concentration X time—that is, the
"sum" of many momentary
concentration levels—dividing the AUC
by the time interval over which it is
measured gives the average
concentration during that interval. As
such, the AUC is more representative of
the target organ's total exposure to the
agent than is the peak concentration.
The AUC provides a measure of the
agent's opportunity to participate in
critical reactions at the target site. For
example, for DNA-reactive compounds,
the AUC is predictive of the rate of
generation of DNA adducts (Hattis,
1990), while for moderate levels of
receptor mediated carcinogens it tends
to be proportional to average receptor
occupancy. For such reasons,
pharmacokinetic equivalence is usually
defined in terms of equality of AUCs.
If this hypothetical chemical is
assumed to be a carcinogen, an added
difficulty in defining pharmacodynamic
equivalence is also readily apparent. It
should be remembered that equally
carcinogenic doses are defined in terms
of exposures repeated every day over a
full lifetime. An adjusted daily dose that
yields pharmacokinetic equivalence for
one day's exposure of the target organ
(as illustrated in Figure 2) is repeated for
2 years in the lifetime of a mouse, but 70
years in a human's. Furthermore, if the
agent's stress on the physiological
system at any given moment is not
proportional to its concentration, the
fact that the pharmacokinetically
"equivalent" equal AUCs are achieved
from different time-patterns of target
organ exposure (as seen in Figure 2)
could affect the carcinogenic
consequences. These and other issues
will be discussed at greater length
further on in this document; they are
raised here to emphasize that
pharmacokinetic equivalence need not
lead to carcinogenic equivalence
without first employing further scaling
considerations.
Clearly, actual pharmacokinetic and
pharmacodynamic processes will be
more complex than the simple
considerations mentioned above would
indicate. Nevertheless, there are some
well recognized general trends in
species differences (e.g., the higher
metabolic rate in small mammals, the
longer tumor latency in humans via-a-
vis experimental rodents) that clearly
influence the appropriate scaling of
doses of carcinogens, and for which we
should attempt to account in our scaling
rationale (Boxenbaum, 1982,1983;
Schmidt-Nielsen, 1970,1975,1984; Travis
et al., 1990; Ings, 1990). An analysis of
the effects of major general trends in
cross-species physiological differences
not only helps guide our choice of
appropriate scaling factors, but it
provides the benchmark against which
increasingly available case-specific data
on the complex details of
pharmacokinetics and carcinogenesis
may be compared. Without such a
framework, the impact of data on a
single component—metabolic activation
of a carcinogen in a target tissue in mice
and humans, for example—is difficult to
guage (U.S. EPA, 1987a,b). The analysis
presented below is not a definitive
solution to the cross-species scaling
problem. Rather, it is presented as an
attempt to accommodate present
knowledge about the major quantitative
trends in comparative anatomy and
physiology into a scaling rationale with
explicity stated assumptions.
The scaling of the myriad
physiological processes that underlie the
processing of carcinogens and their
toxic effects can be drawn together into
a single scheme by referring to the
concept of physiological time. This
concept proposes that quantitative
differences across mammalian species
in physiological processes can be seen
largely as the consequence of
fundamentally similar anatomical and
biochemical machinery operating at
different rates in differently sized
species, smaller species having faster
physiological "clocks." By correcting for
these differences in size and time one
can express dose-response problems in
terms of a single scale-free mammalian
system in which scaled doses should
yield equal responses. (It is this very
similarity, after all, that leads us to use
experimental animals as surrogates for
humans in risk assessment.) In the
sections that follow, the issues of
pharmacokinetic and pharmacodynamic
equivalence are considered in turn.
1. Species Differences in
Pharmacokinetics
The physiological time concept
emerges from the study of the allometry
of key physiological and anatomical
variables that affect pharmacokinetics.
Allometry studies the variation in
features (and the consequences of that
variation) as a function of body size and
some other parameters. Most
quantitative features that vary among
mammals are well described by the so-
called allometric equation,
Y = a PV,
where b is the power of body weight
[W] to which attribute Ymaintains a
constant proportionality, a. A review of
the large literature on this subject is
beyond the scope of the present paper.
The reader is referred to a number of
excellent reviews (Adolph, 1949;
Kleiber, 1932,1961; Lindstedt and
Calder, 1976,1981; Schmidt-Nielsen,
1970, 1975,1984).
The key point for the present
argument is that there is great regularity
in the value of b for certain classes of
attributes relevant to pharmacokinetics
(Travis et al., 1990). Volumes and
capacities (blood volume, volumes of
distribution, organ sizes, lung capacity,
etc.) tend to remain in approximately
constant proportion to body size (i.e.,
b~1.0) in large and small mammals.
Rates, in contrast, tend to maintain
proportionality with body weight to the
3/4 power (i.e., bxO.75). Such rates
include cardiac output, minute volume,
basal metabolic rate and oxygen
consumption, glomerular filtration rate,
and many others. Consumption rates
also tend to scale this way, including
daily intakes of food, air, and water. A
rate that scales in this way becomes
smaller per unit weight (or volume) in
larger animals. For example, a human
has a total cardiac output (mL/min)
about 300 times greater than a mouse,
but in proportion to the human's 2000-
times more massive body, the rate of
blood delivery per gram of tissue is
approximately seven-fold smaller (in
terms of mL/min/g).
Several authors have suggested that
this consistent scaling of rates of
physiological processes leads to a useful
concept of physiological time (Dedrick
et al., 1970; Dedrick, 1973; Boxenbaum,
1982,1983,1984,1986; Lindstedt and
Calder, 1981; Mordenti, 1986; Lindstedt,
1987; Travis et al., 1990). A mouse is
carrying out the same set of
physiological processes as a human, but
each process proceeds at a rate some 7-
times faster. The various processes stay
in proportion to one another, but all of
them are relatively sped up in smaller
species. If one scales the units of time
by dividing them by the fourth root of
body mass (i.e., min»W~1/4, correcting
the physiological time scale) then the
time-course of physiological processes
becomes congruent across species. If
time were measured according to some
internal, physiological standard (such as
-------
Federal Register / Vol. 57, No. 109 / Friday, June 5, 1992 / Notices
24161
heartbeats, breaths, blood circuit times,
clearance half-lives, etc.), rather than in
minutes, then the rates of
pharmacokinetic processes, the time
course of disposition of a dose, and even
life milestones and lifespan would all be
about equal across species. (As
discussed more fully below, humans
tend to be an outlier in the relationship
of lifespan to W l/ 4, living longer than
expected. Some authors have addressed
this by including brain weight as a
second factor in the allometric equation
[Boxenbaum, 1986].)
This concept is illustrated by the
simple example introduced in the
previous section (shown graphically in
Figure 1)—a single intravenous dose of a
compound to a mouse and a human, and
its subsequent blood concentration as it
is removed from a single body
compartment. (The simplicity is for
illustration; the argument can be shown
to hold for more complex
pharmacokinetic models as well, e.g.,
Travis et al., 1990.) If doses are scaled to
body weight (mg/kg) then initial
concentrations are equal, but the blood
level takes much longer to decline in the
human, owing to slower processing of
the compound. The human has a bood
volume (which is proportional to body
weight) some 2000-fold higher than the
mouse, but the compound must be
cleared from this volume by processes
(metabolism and/or excretion) that
operate only 300-fold faster (or seven-
fold slower per unit blood volume). As a
result, the human has an area under the
blood concentration curve (or AUC) that
is 7-fold higher. The AUC has units of
[cone.]'[time], e.g., (mg/L)»min.-
There are two kinds of scaling one
could imagine to accommodate the
species difference in pharmacokinetic
behavior. The first has already been
illustrated in Figure 2; one could give a
smaller initial dose to the human—one
that is seven-fold smaller in terms of
mg/kg but equal in terms of mg/kg31 *.
The initial concentration is lower, but
this is balanced by the slower removal
to give the same AUC as seen in the
mouse.
Alternatively, one could give the same
initial mg/kg dose, but scale the time
axis, expressing time in "physiological
time units" (i.e., minutes divided by
W ll"). This is illustrated in Figure 3.
Such graphs are sometimes called
"Dedrick plots," following the
demonstration of Dedrick et al. (1970)
that scaling time in this way leads to
congruity of methotrexate
pharmacokinetics among several
species. The mouse and human curves
are identical on such a graph, falling to
the same concentration after the same
amount of physiological time has
elapsed. (Of course, it still takes 7-times
more minutes in a human for a given
interval of physiological time to elapse.
The AUC in the usual chronological time
units is still bigger in the human, but in
units of [conc.]»[physiological time] it is
equal.)
BILLING CODE 6560-50-M
-------
CONCENTRATION
P
CD C
00" CO
CD c
00 3
00 =
CD 0)
Q_ =J
5'g
C O.
1-3
00 O
o 5
— K 00
-o a>
3" O
*< c
<2.^
9. CD
O CO
CQ DJ
8 3
~~ CO
CD
-------
Federal Register / Vol. 57, No. 109 / Friday. June 5, 1992 / Notices
24163
It can be shown that these two scaling
approaches—shrinking doses or
stretching the time scale—give
equivalent ways of dealing with scale
differences as long as saturable
pharmacokinetic processes do not figure
prominently (O'Flaherty, 1989). For
example, consider the slightly more
complex case of repeated dosing.
Figures 4 and 5 show blood
concentration versus time curves for
bolus dosing repeated at regular
intervals. If dosing is daily (i.e., inter-
dose intervals are equal for animal and
human in clock time, as in Fig. 4) then
scaling the bolus amount by W31 *
achieves an equal area under the curve
after a given number of days, as well as
an equal average steady-state blood
concentration. Alternatively (Fig. 5), one
can give equal mg/kg doses spaced
according to equal intervals of
physiological time (e.g., daily in the
mouse and every seven days in the
human) to achieve the same end.
BILLING CODE 6560-SO-M
-------
CONCENTRATION
ff
CD
Q.®
5J CD
CD £1
-T
C °-
3 °-
|j:
II
^ C/)
—. C/)
CD §.
V CD
Q.
CO
CD
O
C
W
CD
V)
/ 2661 'S
/ 601 -ON 79 'PA /
(eiapaj
-------
CONCENTRATION
8
o
-^ to co
3-0(0
13 CD C
03 0) -i
3 CD °
30.01
CD Q.
V QJ DO
(D Q.
3 Q-
o °
C ™.
v> ^
(D(Q
<^ O
f ?
33
II
03 W
^ W
0.0
^> Sr
< CD
CD Q.
Q.O
0) Q.
CD
O
>
S91W
/ Z661 'S
/ 601 -ON 'ZS 'PA /
{ejapaj
-------
24166
Federal Register / Vol. 57, No. 109 / Friday, June 5, 1992 / Notices
The foregoing examples are of course
simplified and hypothetical, designed to
illustrate the principles of allometric
variation in physiological rates and
volumes and their impact on the relation
of administered dose to the degree of
"internal" exposure. The same
principles, however, can be shown to
apply to much more complex
pharmacokinetic systems as well,
including multicompartment models,
multiple routes of uptake and
elimination, and multiple metabolic
pathways causing carcinogenic
activation and/or detoxification. The
arguments have been most extensively
developed by Mordenti (1986),
O'Flaherty (1989), and Travis et al.
(1990). The complete elaboration of the
allometry of pharmacokinetics is too
complex to detail here, but a few
important points should be made.
First, the ability to predict the
pharmacokinetic consequences of
variation in the dozens of parameters
that affect a chemical's uptake,
distribution, processing, and elimination
rests on the regularity in their cross-
species variation and the congruence of
these patterns for certain classes of
parameters (rates, volumes, etc.). If
physiological features varied
haphazardly across species, or if all
features had independent allometric
patterns unrelated to one another, then
no dose scaling method could be defined
(W31 * or any other) to approximate
pharmacokinetic equivalence without
first knowing the compound's
pharmacokinetics in detail.
Owing to their importance, it is well
briefly to examine the starting
assumptions that form the basis of the
allometric, "physiological time" concept
and its predictions. They are: (a)
Volumes and capacities (organ sizes,
blood volumes) retain proportionality to
W; (b) the absolute rates of
physiological processes are proportional
to W31"; these rates include cardiac
output, minute volume, glomerular
filtration, and the rates of specific
metabolic steps; (c) physicochemical
and thermodynamic properties of
compounds (solubilities in various
tissues) are equal in all species; and (d)
for metabolic pathways with saturable
metabolism, the Michaelis constant (the
substrate concentration at which half
the maximum reaction velocity is
achieved) is invariant, while the
maximum velocity scales as W314. A
corollary to points (a) and (b) is that
when rates are figured relative to body
size (or to a volume, or in terms of
concentration rather than absolute
amount), they scale as W3"/W = W^
as illustrated by the cardiac output
example shown earlier.
Most of the above assumptions are
well supported by data on comparative
anatomy and physiology, as detailed in
the allometry references cited
previously. Collectively, they embody
the concept of a basically similar
mammalian physiological and
anatomical plan that varies primarily in
scale from one species to another. The
most problematic issue is the scaling of
rates of individual metabolic
transformation reactions as W3/ 4. Not
only are there few data on such scaling,
but some individual metabolic enzyme
activities are shown to vary rather
haphazardly across species (e.g.,
Gillette, 1987; Calabrese 1986a,b).
Several points should be made,
however. First, there are data that
support the proposition of H/3'4 scaling
in specific cases (e.g., Reitz et al., 1988).
Second, overall metabolic rate (O2
consumption, resting metabolic rate)
clearly scales as W31 *-, indeed, this is
the issue around which physiological
allometry was developed. Scaling an
individual metabolic step in this way
corresponds to keeping it in proportion
to general metabolism, which seems the
best default. Third, daily intake of
natural toxins (the usual targets of
carcinogen-metabolizing enzymes)
depends on intake of air, water, and
food (which all scale as W314). That is,
scaling detoxification processes in
proportion to their anticipated load also
predicts W3' * scalirtg.
Consideration of these points leads to
the view that W314 scaling of the rates of
individual metabolic transformation
reactions can be viewed as a benchmark
around which different species (and
individuals within a species) vary from
instance to instance. Such variation
does not invalidate the general scaling
argument, nor does it provide evidence
for any different scaling factor. Rather,
the variation simply illustrates that any
single conception of cross-species
scaling can accommodate only the
general trends, not the diversity of
particular instances. Clearly, when data
on metabolic conversion are available in
a particular case, they should be used in
preference to the W314 default. In fact,
instances of chemical-, dose-, and
species-specific variation in metabolic
transformation of a chemical may
constitute the principal reason for
deviation from the allotmetric default
assumptions herein laid out.
Accordingly, empirical determination of
such metabolic variation constitutes the
most important pharmacokinetic data
that can be brought to bear on the
, estimation of target tissue exposures.
A second major point to bear in mind
about the allometric analysis of
pharmacokinetics is that the cross-
species consequences of variation in the
many physiological parameters depend
not on the individual parameters, but on
their interrelation. It is misleading
simply to examine the scaling of one
component (say, metabolic activation) in
isolation. One must remember that the
many quantitative differences across
species are having their influences
simultaneously; it is their interactions
and net results that determine the
consequences for doses to the tissues.
For example, metabolic rates alone are a
less important determinant of the
fraction of. a dose that is metabolically
activated than is the ratio of metabolic
activation rates to rates of other
competing processes (such as renal
clearance) that remove a compound
from the body.
The third major point is that, despite
the variety and diversity of underlying
pharmacokinetic processes that may
obtain from one case to another, the
allometric analysis of pharmacokinetics
makes rather general and simple
predictions about how administered
doses should relate to target tissue
exposures in experimental rodents and
humans. These predictions are:
For a given dosing pattern in which
amounts are scaled to body weight, fhe
tissue exposures (as measured by areas
under the concentration curve) tend to
be bigger in larger species by the ratio of
human to animal body weight to the 1/4
power (which amounts to almost seven-
fold for mouse-to-human scaling and not
quite four-fold for rat-to-human scaling).
If the administered amounts are kept in
proportion to W31 * (rather than to W]
the doses tend to be
"pharmacokinetically equivalent" in the
sense of yielding similar areas under the
curve of concentration over time. Since
daily intakes of air, food, and water tend
to be in proportion to W31 * across
species, calling exposures to
environmental media equivalent on a
ppm basis (i.e., when they are equally
contaminated) produces essentially the
same expectation of pharmacokinetic
equivalence as scaling by W314 (Hattis,
1991).
In fact, all variables containing [time]
in their units will scale in a way that
leads to the human value being bigger
by the ratio of body weights to the 1/4
power. If these variables are
reexpressed in terms of "physiological
time units," i.e., [time] 'H/'1'4, then their
values are equal across species.
The above conclusions apply to
parent compound and to metabolites,
since (in this generalized scheme)
-------
Federal Register / Vol. 57, No. 109 / Friday, June 5. 1992 / Notices
24167
metabolites are also subject to scale-
affected clearance processes. In humans
a metabolite may be formed more
slowly, but the amount that is formed
persists longer, resulting in similar
AUCs as seen in rodents. The
pharmacokinetic equivalence applies
not only to an agent's concentration in
blood, but also to concentrations in any
specified organ or tissue. Thus, the
scaling applies to the AUC of the
ultimate carcinogenic species (be it
parent compound or metabolite) at the
particular site in the body that
constitutes the target of carcinogenesis
(presuming the target site to be the same
across species).
The proportion of the administered
dose that ends up having any particular
ultimate fate (e.g., being excreted
unchanged, being metabolized by a
particular biochemical pathway at a
particular site, being excreted as a
conjugate in the urine, etc.) is predicted
to be the same independent of species.
That is, if a mouse given 10 mg/kg of an
agency ends up metabolizing 4 mg/kg
into a form that has an AUC in the
spleen of 100 (mg/L)«min, then the
allometric prediction for a human given
10 mg/kg is that 4 mg/kg will be-
metabolized, but the AUC in the spleen
will be 700 (mg/L)»min, owing to the
metabolite's slower clearance.
A difficult situation arises when the
active carcinogen is neither the parent
compound nor a stable metabolite, but
rather a very reactive metabolite,
perhaps an intermediate formed
ephemerally during the course of
metabolic transformation. If this
reactive compound is removed by
spontaneous reaction (rather than
further enzymatic processing) and if
such spontaneous reaction is so rapid
that the moiety never leaves the tissue
in which it is formed, then the removal
rate may no longer be species-
dependent; instead, it may hinge only on
physicochemical properties of the
reactant and its milieu. In such a case,
without species differences in
persistence, the AUC of the reactive
moiety in its tissue of formation may be
proportional to the amount formed. Such
AUCs would tend to be equalized when
doses are scaled to body weight, rather
than to W 3/ 4 (Travis, 1990).
It may be well to reiterate at this point
that the reason for constructing these
general allometric arguments is to
predict the AUC of the proximate
carcinogenic agency at its site of action
in those cases (which constitute the
majority of cases at present) for which
no better means exists to determine
relative target tissue doses in rodents
and humans. Clearly, if better means
exist to characterize target tissue
exposures, they should take precedence.
Pharmacokinetic modeling of a
particular compound may demonstrate
that the allometric presumptions are in
error. Two possible causes of such error
are: (a) species differences in metabolic
processing that do not adhere to the rule
of proportionality to W3/ 4, and (b)
saturation of metabolism in one but not
the other species as a result of
comparing markedly different dose
levels or dosing regimens. The
importance of the "reactive metabolite"
scenario outlined in the previous
paragraph is best determined by case-
specific characterization of metabolic
activation and its effects.
Macromolecular adducts may be
particularly useful in this regard since,
under certain circumstances (including
negligible repair), their accumulation in
a tissue would be expected to be
proportional to the AUC of the adduct-
forming moiety in that tissue.
It must be conceded that, in actuality,
mice and rats are not simply scale-
model humans; certain particular
characteristcs (metabolism among them)
do not necessarily vary in a simple way
with body size. However, the long-
standing toxicological practice of using
rodent exposures to toxic agents as
surrogates for the human experience
rests on the belief that, to a first
approximation, the similarities that stem
from a shared mammalian anatomy and
physiology outweigh the differences.
The species differences in size, uptake
rates, basal metabolism, blood flows,
organ sizes, and so on are clearly
important to acknowledge in any
dosimetric scheme. The allometric
arguments adduced here attempt to
construct a logical and consistent
framework for investigating cross-
species dosimetry. This framework
provides a basis for articulating the
expected consequence of those broad
general patterns of cross-species
difference in size scale and time scale
that we understand, while providing
rebuttable default positions for those
aspects, such as chemical-specific
metabolism, that are less well
understood.
2. Species Differences in
Pharmacodynamics
The overall aim of dose scaling is to
achieve toxicological equivalence across
species. The foregoing section discussed
pharmacokinetic equivalence. For such
results to be useful for carcinogen risk
assessment—that is, to complete the
equation of exposure and tumorigenic
response—it remains to determine what
toxicological consequences to expect
from given target tissue exposures in
humans and animals. As argued earlier,
the principles of pharmacodynamic
equivalence are far from self-evident.
The issues about pharmacodynamic
equivalence fall into three categories.
First, the appropriate measures of
"delivered dose" would seem to depend
on details of the mechanism of toxic
action, details that are frequently poorly
understood. In the foregoing section,
scaling of administered doses was
discussed in terms of tendency to
equalize the AUC, an integrated
measure of target tissue concentration.
Although this is a frequent and widely
accepted measure of a target organ's
exposure to a toxin (Voisin, et al., 1990),
its use as a measure of carcinogenic
equivalence of .doses rests on the
presumed proportionality of the rates of
toxicological reactions to the AUC. If
the underlying reactions that comprise
the process of carcinogenicity are
markedly nonlinear with target-tissue
concentration, if they include capacity-
limited steps or magnitudes below
which significant stress on the system is
absent, then proportionality of toxic
response to the AUC (or to any other
easily characterized summary measure
of target-tissue exposure) becomes
problematic. Thus, use of the AUC as an
"equivalent" tissue dose should be
regarded as a default that corresponds
to the presumption that the processes
constituting carcinogenicity operate in
proportion to the concentration of the
carcinogen at the target. In particular
applications, this assumption should be
critically examined, and relevant data
brought to bear, if possible.
The second issue returns to the
question of scale. For corresponding
organs bathed in an equal concentration
of carcinogen, a human will have many
more target cells exposed than a rodent,
only one of which need be transformed
to found a tumorigenic clone. Moreover,
during the course of a full lifetime under
this dosing regime, a human's cells will
be exposed for much longer and undergo
many more cell divisions (NAS, 1975;
U.S. EPA, 1987a). Although this would
seem to suggest a much larger
sensitivity to carcinogens in larger
species, the empirical evidence shows
instead a rough lifetime-to-lifetime
equivalence across species of both the
magnitude of spontaneous cancer risk
and the age pattern of its appearance.
When arguments from first principles
lead to answers that are clearly off
track, it indicates that key factors have
not been brought into consideration. In
this case, the role of species differences
in repair processes may enter. Also, the
number of cells (or cell divisions) at risk
may be less different among species
-------
24168
Federal Register / Vol. 57, No. 109 / Friday, June 5, 1992 / Notices
than presumed, owing to slower
turnover, stem cell populations that are
not proportional to tissue volume, or
other factors. The point is raised here
simply to emphasize that size and
timespan differences across species may
have key roles in comparative
pharmacodynamics just as they do in
comparative pharmacokinetics, although
the particulars are not clear at present.
In the face of this difficulty, it has been
the ususal practice to assume lifetime
equivalence when projecting
carcinogenesis patterns across species,
an assumption that has held up well in
experience. This point will be returned
to below.
The third issue in pharmacodynamic
equivalence also parallels one in
pharmacokinetics—that of the
uniqueness and species-specificity of
carcinogenic responses that tends to
obscure overall trends and patterns. The
pharmacodynamic reasons for
differences in sensitivity of potential
target organs among species are perhaps
more obscure than the pharmacokinetic
reasons, but they surely exist. As with
the case-by-case particulars of
pharmacokinetic processes, the
idiosyncratic and species-specific
variations in responsiveness to
carcinogenic stimuli create an
unavoidable envelope of uncertainty
around the predictions of a scaling
methodology that can only characterize
the average behavior of carcinogens
overall. When data are available that
enable the investigator to incorporate
knowledge of species differences in the
carcinogenic reactions to a given level
of target-tissue dose, they should be
considered in the analysis and
incorporated when appropriate.
Although certain pieces of the puzzle
of cellular and molecular biology that
underlie carcinogenesis are known, and
despite rapid progress, it not yet
possible to undertake a detailed
analysis of the magnitudes and causes
of species differences in the
carcinogenic process. At present, there
can be no empirical and allometric
characterizations of general cross-
species trends, as has been done in this
report for the pharmacokinetic part of
the equation. One can, however, make
use of the observation of general
lifetime-equivalence, noted above, to
suggest how the insights of cross-species
patterns in pharmacokinetics might be
applied to the question of toxicological
equivalence.
3. Toxicological Equivalence
When experimental animals and
humans are exposed to a chemical in
such a way that they experience equal
areas-under-the-curve of the proximate
carcinogenic agent (be it the parent
compound, a metabolite, or a reactive
intermediate of metabolism) at the
target of toxic action, then they will
have their susceptible tissues exposed
to equal average concentrations of the
carcinogen over the exposure period.
Over the course of a full lifetime of
exposure, the lifetime average target-
tissue concentrations are equal
(although the total accumulated AUC is
larger in humans, by virtue of their
longer lives). The earlier discussion of
pharmacokinetics argued that, if daily
administered doses are scaled in
proportion to W3/ * (or if exposures of
equal duration are equated on a ppm
basis), such equality of resulting AUCs
tends to result across mammalian
species.
If the empirical principle of lifetime-
to-lifetime equivalence is applied, then a
possible presumption is that such
pharmacokinetically equivalent lifetime
exposures (in terms of equal average
concentrations of the carcinogen at its
target) should be equivalent in the
degree of lifetime cancer risk they
engender (although other interpretations
of the consequences of pharmacokinetic
equivalence are possible). That is, it
may be assumed that equal carcinogen
concentrations at the target lead to
equal degrees of impact at the cellular
level which, if continued for a lifetime,
yield equal lifetime probabilities that a
tumor will be caused in that target
organ.
The reasons for approximate lifetime
equivalence in the carcinogenic process
among species of different body size and
lifespan are not clear. One can,
however, rationalize this observation by
extending the concept of physiological
time from pharmacokinetic processes to
cover pharmacodynamic processes as
well. The following section explores this
approach.
4. A Physiological Time Approach to
Toxicological Equivalence
It is helpful to begin by considering
the case of "zero" dose, i.e., by
examining background or spontaneous
carcinogenesis. Although the common
cancer types differ somewhat, humans
and experimental animals have roughly
similar lifetime cancer rates. Moreover,
the latency periods are greatly different
in animals and humans, but in a way
that is roughly proportional to lifetime.
Age-specific incidences are also roughly
parallel when time is measured not in
years, but on a lifetime scale (Cutler and
Semsei, 1989). If these equivalencies
were not so, we would either never see
tumors in experimental animals (since
they would die of other causes before
the 20-to-40 year latency was
completed), or we would find humans to
be overwhelmed with spontaneously
arising tumors during childhood. These
results from spontaneous carcinogenesis
appear to be paralleled by chemically
induced cancers, in that such cancers
also arise and progress on a "lifetime"
time scale in experimental animals and
humans.
The above results suggest that
carcinogenesis proceeds more slowly in
larger animals, in a way that makes its
progress roughly constant per lifetime,
rather than per unit of clock time. This is
in accord with the current risk
assessment practice of equating lifetime
cancer incidences in humans and
rodents. It would seem that the concept
of physiological time—that large
animals carry on their life processes at
an overall slower pace than smaller
ones—proves as useful in examining
pharmacodynamics as it does for
pharmacokinetics. As argued in the
previous section, the rates of the
underlying pharmacokinetic processes
tend to operate in proportion to a size-
dependent physiological time "clock,"
which allows appropriate scaling to
explain and correct for species
differences in pharmacokinetic end
points." In the case of carcinogenesis,
the component physiological features
and processes are less easily observed,
but the "pharmacodynamic end point"
can be seen in the above-mentioned
cross-species patterns of spontaneous
carcinogenesis. In sum, not only may
"pharmacokinetic time" vary among
species in a regular way,
"pharmacodynamic time" may do so as
well. Total lifespans of different species
generally scale in rough proportion to
W ll 4 (Sacher, 1959; Lindstedt and
Calder, 1976,1981). (In terms of the
physiological time concept, the
"processes of living" that proceed at a
rate proportional to W 3l *—or on a per
kg basis, to W~lli—go slower in a
larger animal, and so take chronological
time in proportion to W114 to go "to
completion.") Hence, the two
physiological time scales are quite
similar. However, humans live longer
than their allometric prediction by about
a factor of five.
The above discussion of
pharmacodynamics suggests that
carcinogenesis (in common with other
physiological processes) proceeds more
slowly in humans than in rodents, in a
way that tends to be equivalent on a
lifetime basis. Together with the
pharmacokinetic results outlined
earlier—namely, that scaling daily
administered doses in proportion to
W3l 4 tends to result in
"pharmacokinetically equivalent"
-------
Federal Register / Vol. 57, No. 109 / Friday, June 5. 1992 / Notices
24169
exposures to corresponding organs and
equal steady-state concentrations of
agents and their metabolites—this
suggests that administered doses of
carcinogens be considered equal in
lifetime risk when expressed in units of
mg/kg 3/ 4/day. One possible
interpretation of this line of reasoning is
that tissues experiencing equal average
concentrations of the carcinogenic
moiety over a full lifetime should be
presumed to have equal lifetime cancer
risk. Under the arguments on
pharmacokinetic allometry set out
earlier, such equality of average
concentrations would tend to be
produced by daily administered doses
scaled in proportion to W314. However,
if the pharmacokinetically equivalent
doses can be obtained by experimental
means, under this line of reasoning, such
results could replace the allometric
presumptions, and equal risks would be
expected when average daily AUCs are
equal (or equivalently, when average
concentrations are equal). If the default
allometrically based assumptions about
pharmacokinetics are adhered to by a
particular compound, the introduction of
data in place of assumptions will leave
the answer unchanged. Other
interpretations of the question of the
cross-species toxicological equivalence
of delivered doses are possible, and the
issue remains one on which further
insight would be helpful.
If we use a scale of pharmacodynamic
time based on the equivalence of
lifetimes, then the 35-times larger
exposure of human tissues to
carcinogens that results from a lifetime
of doses scaled by mg/jy3/ 4/day
results in an equal lifetime cancer risk
because the affected physiological
processes of carcinogenesis themselves
are operating more slowly (by
assumption, 35-times more slowly). A
given span of clock time that a tissue
spends under a given concentration
regime yields less risk in a human (since
the tissue has spent less
"pharmacodynamic time" exposed).
It should be clear that not every
empirical measure of "internal dose" is
equally informative about species
differences. As noted earlier, the amount
of a dose metabolically activated, for
example, may be equal in a mouse and a
human, but the human's AUC of
metabolite at the target may be much
larger. If an empirical measurement or
modeled result is to be used as a
surrogate for "internal dose" in a cross-
species extrapolation, its value in
animals and humans should be
compared to the predictions of the
default assumptions of allometrically
scaled pharmacokinetics (which should
be aided by a full analysis of the
uncertainties in the available data and
of reasonably likely alternative
pharmacokinetic modeling approaches).
With this kind of analysis, it is possible
to judge whether those default
assumptions have actually been
contradicted by data for the case at
hand.
Once again it should be stressed that
the arguments set out here are intended
as defaults. They attempt to gauge the
expected effect of known major cross-
species trends in the rates and
magnitudes of the underlying
physiological processes, both in the
internal disposition of a dose and its
subsequent carcinogenic effect. Just as
the pharmacokinetic presumptions may
be able to be replaced with sufficiently
validated case-specific modeling, the
pharmacodynamic presumptions may be
replaced with suitable biologically
based dose-response models. The true
pharmacodynamic situation is clearly
more complex than represented here. In
particular, there may be dose-rate
effects, in which higher concentrations
have more-than-proportionally stronger
effect (Hattis, 1990). The effect of one
moment's exposure may also depend on
age or on the degree of exposure earlier
in life. Such effects have no
generalizable patterns, however, and
cannot serve as a basis for default
scaling of effects. Again, we seek a
simple default principle to guide our
expectations, while allowing for the use
of case-specific experimental or
epidemiologic insights (when available)
to improve the estimate based on the
simplifying assumptions.
It should also be pointed out that this
scheme, with its explicit treatment of
time, pharmacokinetics, and
pharmacodynamics, provides a
conceptual framework for examining
such crucial emerging issues as risks
from partial lifetime exposures,
potencies in children vis-a-vis adults,
and other similar questions. Failing to
provide such an explicit argument from
stated assumptions dooms a scaling
factor to be inapplicable to such
questions and provides no means for
incorporating biological insights, such as
data on pharmacokinetics and
mechanism of action, when they are
available.
III. Discussion
This proposal aims at arriving at a
very broad generalization about
carcinogen exposures that can be
considered of equal risk in experimental
animals and humans—one that can be
applied to potentially carcinogenic
chemicals lacking adequate information
on pharmacokinetics and mechanisms of
action. It attempts to provide a rational
basis for a prima facie characterization
of potential risks in humans, consistent
with our empirical knowledge of
carcinogen potencies in animals and
humans and with the known general
consequences of species variation in
body size and the rates of physiological
processes.
To achieve this wide applicability and
generality, it is necessary to rely on
simplified, broad patterns and trends of
biological variation, while bypassing
many details and causes of case-by-case
variation. This is not to deny the
importance of these details, nor to
denigrate the value of case-specific data
that show species- or dose-related
differences in uptake, metabolism, or
physiological actions of putative
carcinogenic agents. To the contrary, the
intention is to provide a framework for
the use of such data, allowing (and
indeed, encouraging) one to go beyond
the prima facie case based on overall
trends to address the impact of specific
knowledge about the chemical and its
actions.
The empirical data on carcinogen
potencies estimated in various animal
species and in humans demonstrate the
large variability involved. Although
scaling doses by W3l 4, as proposed
herein, characterizes the trend fairly
well, individual chemicals may deviate
from this overall pattern by two orders
of magnitude or more in either direction.
In the case of the allometric arguments,
there are dozens of points in the chain of
inference where one could raise
counterexamples to simplifying
assumptions, arguing that the
generalized W 3l 4 scaling method
thereby would over- or underestimate
human risks for that case. For example,
Gillette (1985) lists a number of
physiological factors with high
variability that would influence the
accuracy of extrapolation of a dose's
toxicity to an exposed human, not the
least of which is the 20-to-50-fold
variation among individual humans in
their ability to take up and metabolize
an agent and to repair any resulting
damage.
The existence of such underlying
variation means that the extrapolation
of chemically induced risks observed in
one circumstance (say, in a mouse
lifetime cancer bioassay) to another
(say, to people exposed to
environmental pollutants) needs to be
carefully and properly interpreted.
Clearly, the projection of an equivalent
dose is not merely a conversion of units,
with the resulting human dose achieving
an equal factual standing to the original
animal observation. The projection is an
-------
24170
Federal Register / Vol. 57, No. 109 / Friday, June 5, 1992 / Notices
hypothesis, formulated in the face of
uncertainty. In the most basic case—
when there is little additional
information that may be brought to
bear—this hypothesis is framed in terms
of the general features of anatomical
and physiological differences among
species that should affect all chemicals.
It represents a best guess based on
general principles and the recognition of
overall trends. This best guess is
surrounded by an envelope of
considerable uncertainty, owing to the
dozens of particulars that make each
chemical's disposition and toxic effects
in various species unique, despite the
overall trends. When applicable
pharmacokinetic and mechanistic
insights into the particular chemical and
its actions are available, they can (and
should) be used to refine the projections
by identifying and accounting for these
chemical-specific factors.
Every projection of human equivalent
dose, no matter how sophisticated, will
have associated with it both uncertainty
and variability. The uncertainty
concerns whether the scaling method
employed has correctly embodied and
utilized the information at hand (be it
general cross-species trends over all
chemicals or case-specific insights from
pharmacokinetics and mechanistic
studies). The variability arises because
even a sophisticated projection, when
applied to a population of cases, will at
best predict the mean of an array of
actual values that reflect the myriad
individual factors that no analysis can
completely take into account. The "true"
dose of equivalent risk will vary among
exposed humans according to how each
individual deviates from the overall
human norm, owing to genetic factors,
environmental influences, age, sex,
lifestyle, and countless details of
personal history.
The goal of a cross-species scaling
methodology, then, is not to arrive at
"true" values of equivalent does under
all circumstances (for this is impossible,
even in principle). Rather, it is to
embody correctly and without bias the
impact of the information at hand,
providing rational estimates that take
into account what is known, recognizing
that true values will vary around this
estimate as a result of case-by-case
particulars, many of which are either
unknown to vary among the individuals
for whom the projections are being
made.
The proposed scaling of daily
administered doses of putative
carcinogens by W 3l 4 is intended to be
such an unbiased projection; i.e., it is to
be thought of as a "best" estimate rather
than one with some conservatism built
in to assure that any error is on the side
of being overly protective. It should not
be interpreted as a "safety factor" or
other intentional bias designed to "err
on the side of safety." Thus, it is to be
expected that some individual
compounds will have their human
potencies overestimated by this
procedure, while others will have them
underestimated.
This having been said, it must be said,
it must be acknowledged that there is
considerable uncertainty about the best
scaling method to achieve this unbiased
projection. In particular, the empirical
data on comparative carcinogen
potencies are also compatible with both
body weight and surface area scaling,
the methodologies that we propose to
abandon in favor of W3l * scaling. The
W3/ 4 scaling is chosen both to achieve
unity of default methods and because it
can be related to an explicit rationale
based on allometric variation of the
underlying anatomy and physiology.
Former methodologies have not been
shown to be false, however, and it is
considered that risk assessments
conducted under these methodologies
are not in need of revision on account of
any agreement to utilize a common
methodology in the future.
The utility of the "physiological time"
concept for understanding the patterns
of cross-species differences in a
carcinogen's action lies in its simplicity
and generality. Because organ volumes
tend to share a common pattern of
allometric variation, while rates of
physiological processes share another,
the general predictions of cross-species
differences is independent of specific
hypotheses about target organs or
mechanisms of action. One could, for
instance, envisage an alternative
allometric formulation that, rather than
relying on overall patterns for
unspecified organs in all mammals,
focuses instead on the details of specific
organs (common target organs or sites of
metabolic transformation, say) in
specific laboratory animal strains and in
humans. For example, instead of relying
on the approximation that breathing
rates vary as W3l 4, one could make
precise measurements of rates in
B6C3F1 mice and in the humans whose
risks are being evaluated. The utility of
such an approach for a default scaling
factor is doubtful, however, since the
generality of the argument is lost, and
the analysis becomes contingent on the
details of the specific physiological
hypothesis being elaborated. If such
specificity is possible in an individual
instance, it should become part of the
case-specific pharmacokinetic and
pharmacodynamic analysis that
overrides the default methodology.
It is sometimes suggested that there
should be more than one "default"
scaling methodology, with different
generalized procedures to be applied to
different classes of chemical
carcinogens. At present, it is not clear
how such division of cases would be
made, however, nor what the
consequences on a generalized method
should be. For example, tissue area-
under-the-curve of the toxic moiety
would seem to be the best prima facie
dosimeter for the effects of both
genotoxic and non-genotoxic
carcinogens on their target organs.
Similarly, the general allometric
arguments for how AUCs are expected
to vary across species apply both to
agents active as the parent compound
and to those requiring metabolic
activation.
A possible exception to this pattern
has been mentioned earlier. The
generalized allometric pattern assumes
that the rate of clearance of a metabolite
from the target site of toxic action, like
other rates, scales in proportion to W3l 4.
If a compound acts through a very
reactive metabolite that is
spontaneously and fully deactivated by
purely physical-chemical processes
within the target tissue itself, then the
rate of detoxification may be species-
independent, and the AUC may be more
related to the amount metabolized,
which by default is expected to retain
proportionality to body mass (Travis,
1990). Such a situation is not only
plausible, it may be frequent. There is
no particular indication from the
empirical data, however, that different
rules apply to metabolically activated
compounds. Moreover, since the
reactive intermediate scenario breaks
the symmetry of the physiological time
argument, it is difficult to know exactly
what the carcinogenic consequences
should be. This remains an important
problematical area that requires future
attention. For the present, however,
there do not seem to be grounds for
specifying when and how one should
alter the default proposal.
The analysis presented herein is
oriented around scaling doses so as to
yield equal areas under the carcinogen's
concentration curve at the target site.
This definition of equivalence of target
"doses" is in line with common practice.
The AUC provides a measure of the
agent's opportunity to interact with the
target. Equal AUCs over a fixed time
interval correspond to equal average
concentrations of the agent during that
interval. It should be borne in mind,
however, that other measures of target
-------
Federal Register / Vol. 57. No. 109 / Friday. June 5. 1992 / Notices
24171
tissue dose might be more appropriate
for specific mechanisms of
carcinogenicity. For example, if a critical
concentration must be reached or if
there is a nonlinear dependence of toxic
stress on concentration of the agent.
Such alternative have no generalizable
consequences or patterns, however, and
there is no evident way to bring them
into a default methodology. When case-
specific pharmacokinetic analysis is
undertaken, careful attention should
also be paid to the measure of target
tissue dose that is being considered to
yield equivalent lifetime carcinogenic
effect, and alternatives should be
examined.
When AUCs from daily exposures are
equal, then average concentrations of
the agent at the target sites are equal.
And when dosing producing equal daily
average concentrations is continued for
a lifetime, then average lifetime
concentrations are equal. If one
presumes that such average lifetime
concentrations yield equal cancer risk,
then the argument follows common
practice and is in accord with the
general finding that age-specific tumor
incidence patterns tend to be congruent
across species when expressed on a
lifetime scale. (Other presumptions
about the impact of such equal
concentrations can be held, however.)
The underlying biological basis for
lifetime equivalence, and the conditions
under which it might be violated, are not
clear at present. This is an area in need
of further investigation, and increased
understanding will be key to
determining how to scale the results of
cell-kinetically based models of
carcinogenesis from animal models to
humans.
It should be borne in mind that the
arguments for scaling doses by W3/ 4
have been cast in very general terms to
reflect constant, low-level, lifetime
dosing and consequent lifetime cancer
risks. Care should be taken when
applying the methodology to specific
exposure scenarios that deviate from
this pattern. For example, the allometric
arguments are adduced for variation
among mammals. Other groups of
animals have their own characteristic
allometric patterns, but they are
different than the mammalian ones. To
extrapolate across classes of
vertebrates with the proposed
methodology, for example, would violate
the basic presumption of the variation in
a basically similar anatomical and
physiological plan among differently
sized mammals.
The allometric patterns relied on by
the present argument represent variation
among species for adult organisms.
Allometric patterns among variously
sized individuals of the same species
can (and generally do) differ from the
pattern seen from one species to
another. The metabolic and lifespan
patterns across species do not really
describe variation among differently
sized humans, for example. In other
words, the scaling arguments presented
here do not necessarily apply for the
adjustment of doses to larger and
smaller humans. In such cases, it is
probably preferable to use mg/kg
scaling (although the difference between
this and W 3l * scaling is minor).
Similarly, the allometric patterns
describing the changes within an
individual as he or she grows and
matures from child to adult generally
differ from both the cross-species
pattern and from the variation among
differently sized adults. Compared to
adults, children do have faster
metabolic rates and greater intakes of
food, water and air per unit of body
weight, but these relations are not well
described by proportionality W3l 4, as
they are across species. Moreover,
children also have proportionally faster
rates of cell division (i.e., both
pharmacokinetic and pharmacodynamic
time are accelerated compared to
adults). This a complex and problematic
issue that is beyond the scope of the
present document. It is deserving of
further study. At present, it seems most
reasonable to follow current practice,
i.e., to scale doses for adults and
children (and for differently sized
adults) on a mg/kg basis. For similar
reasons, the present scaling arguments
provide no special insight into the
problem of partial lifetime exposures.
Finally, it should be borne in mind
that the scaling arguments are made for
similar levels and patterns of exposure
in animals and humans. When
experimental animals are exposed to
much higher levels than humans (as is
common in carcinogenicity bioassays)
there is the possibility of saturation of
metabolism in animals that is not shared
with human exposures. Such effects will
obscure the usual pattern of equivalence
of internal doses projected on the
assumption of similar exposure regimes.
In other words, dose scaling cannot
solve the high-to-low-dose extrapolation
problem, which must be addressed by
other means. Case-specific
pharmacokinetic analysis can, however,
provide very valuable insight into
differences in target tissue doses
between rodents at high bioassay
exposures and humans at much lower
exposures.
IV. Conclusions
This notice is an announcement of a
consensus reached by the
Environmental Protection Agency, the
Food and Drug Administration, and the
Consumer Product Safety Commission
to consider that lifetime cancer risks
will be presumed to be equal when daily
amounts administered are in proportion
to body weight raised to the 3/4 power.
It should be reiterated that former
methodologies have not been shown to
be in error, and this agreement should
not be construed as overturning those
practices with one of superior scientific
validity.
The empirical data on comparative
carcinogenic potencies in different
species support the general practice of
scaling rodent potencies to humans, and
show that, on average, current methods
perform rather well. The data are not of
sufficient resolution, however, to
distinguish between surface area and
body weight dose scaling. The data are
fully consistent with the proposal
contained herein for scaling by body
weight to the 3/4 power.
Theoretical support for scaling
carcinogen doses by the 3/4 power of
body weight is available from analysis
of the allometric variation of key
physiological parameters across
mammalian species. Such an analysis
has the benefit of providing an
articulated rationale for the scaling
methodology and of setting out the
underlying assumptions explicitly.
V. References
Adolph, E.F. 1949. Quantitative relations in
the physiological constitution of mammals.
Science 109:579-85.
Allen, B.C., A.M. Shipp, K.S. Crump, B. Kilian,
M.L. Hogg, J. Tudor, and B. Keller. 1987.
Investigation of cancer risk assessment
methods. (3 volumes plus summary report).
Prepared for U.S. Environmental Protection
Agency under contract to Research
Triangle Institute, U.S. EPA Contract N° 68-
01-6807. National Technical Information
Service N° PB88-127113.
Allen, B.C., K.S. Crump, and A.M. Shipp. 1988.
Correlation between carcinogenic potency
of chemicals in animals and humans. Risk
Anal. 8:531-61.
Boxenbaum, H. 1982. Interspecies scaling,
allometry, physiological time, and the
ground plan of pharmacokinetics. /.
Pharmacol. Biopharm. 10:201-27.
Boxenbaum, H. 1983. Evolutionary biology,
animal behavior, fourth-dimensional space,
and the raison d'etre of drug metabolism
and pharmacokinetics. Drug Metab. Rev.
14:1057-97.
Boxenbaum, H. 1984. Interspecies
pharmacokinetic scaling and the
evolutionary-comparative paradigm. Drug.
Metab. Rev. 15:1071-121.
-------
24172
Federal Register / Vol. 57. No. 109 / Friday. June 5. 1992 / Notices
Boxenbaum, H. 1986. Time concepts in
physics, biology, and pharmacokinetics. /.
Pharmacol. Sci. 75:1053-62.
Calabrese, E.J. 1983. Principles of Animal
Extrapolation. John Wiley & Sons. New
York.
Calabrese, E.J. 1987. Extrapolation from
animal data. In: Tardiff, R.G., and J.V.
Rodricks (eds.). 1987. Toxic Substances and
Human Risk: Principles of Data
Interpretation. Plenum Press. New York.
Calabrese, E.J. 1986a. Animal extrapolation
and the challenge of human heterogeneity.
/. Pharmacol. Sci. 75:1041-6.
Calabrese, E.J. 1986b. Comparative biology of
test species. In: Hill, T.A., R.C. Wands, and
R.W. Leukroth, jr. (eds.). 1986. Biological
Bases for Interspecies Extrapolation of
Carcinogenicity Data. Prepared for Food
Safety and Applied Nutrition, Food and
Drug Administration, Department of Health
and Human Services, Washington, D.C.
under contract to Federation of American
Societies for Experimental Biology, FDA
contract N° 223-83-2020.
Chen, J.J., and D.W. Gaylor. 1987.
Carcinogenic risk assessment: Comparison
of estimated safe doses for rats and mice.
Environ. Health Perspect. 72:305-9.
Collins, J.M., D.S. Zaharko, R.L. Dedrick, and
B.A. Chabner. 1986. Potential role for
preclinical pharmacology in phase I clinical
trials. Cancer Treat. Rep. 70:73-80.
Collins, J.M., C.K. Grieshaber and B.A.
Chabner. 1990. Pharmacologically guided
phase I clinical trials based upon
preclinical drug development. /. Natl.
Cancer Inst. 82:1321-6.
Crouch, E.A.C. 1983. Uncertainties in
interspecies extrapolations of
carcinogenicity. Environ. Health Perspect.
50:321-7.
Crouch, E., and R. Wilson. 1979. Interspecies
comparison of carcinogenic potency. /.
Toxicol. Environ Health, 5:1095-118.
Crump, K., B. Allen, and A. Shipp. 1989.
Choice of dose measure for extrapolating
carcinogenic risk from animals to humans:
an empirical investigation of 23 chemicals.
Health Physics 57(Sup.l):387-93.
Crump, K.S., A Silvers, P.P. Ricci, and R.
Wyzga. 1985. Interspecies comparison for
carcinogenic potency to humans. In: Ricci,
P.P. (ed.) Principles of Health Risk
Assessment. Prentice-Hall, Englewood
Cliffs, N.J.
Crump, K., B. Allen, and A. Shipp. 1987. An
investigation of how well human
carcinogenic risk from chemical exposures
can be predicted by animal data, with
emphasis upon selection of dose measure
for extrapolation from animals to humans.
Presentation at Twenty-Sixth Hanford Life
Sciences Symposium; Modeling for Scaling
to Man: Biology, Dosimetry, and Response.
Richland, Washington. October 20-23,1987.
Cutler, R.G., and I. Semsei. 1989.
Development, cancer and aging: possible
common mechanisms of action and
regulation. J. Gerontol. 44:25-34.
Davidson, I.W.F., J.C. Parker, and R.P. Beliles.
1986. Biological basis for extrapolation
across mammalian species. Reg. Toxicol.
Pharmacol. 6:211-37.
Dedrick, R.L. 1973. Animal scale-up./.
Pharma. Biopharm. 1:435-61.
Dedrick, R.O., K.B. Bischoff, and D.S.
Zaharko. 1970. Interspecies correlation of
plasma concentration history of
methotrexate (NSC-740). Cancer
Chemother. Rep. Pt. 7,54:95-101.
Freireich, E.J., E.A. Gehan, D.P. Rail, L.H.
Schmidt, and H.E. Skipper. 1966.
Quantitative comparison of toxicity of
anticancer agents in mouse, rat, hamster,
dog, monkey and man. Cancer Chemother.
Rep. 50:219-44.
Gaylor, D.W., and J.J. Chen. 1986. Relative
potency of chemical carcinogens in
rodents. Risk Anal. 6:283-90.
Gaylor, D.W., and R.L. Kodel. 1980. Linear
extrapolation Algorithm for low dose risk
assessment of toxic substances. /. Environ.
Pathol. Toxicol. 4:305-12.
Gillette, J.R. 1985. Biological variation: the
unsolvable problem in quantitative
extrapolation from laboratory animals and
other surrogate systems to human
populations. In: Hoel, D.G., R.A. Merrill,
and P.P. Perera (eds.) Risk Quantitation
and Regulatory Policy, Banbury Report 19.
Cold Spring Harbor Laboratory, Cold
Spring Harbor, L.I., N.Y.
Gillette, J.R. 1987. Dose, species, and route
extrapolation: general aspects. In: National
Research Council. Pharmacokinetics in
Risk Assessment: Drinking Water and
Health, Vol. 8, National Academy Press,
Washington, DC.
Gold, L.S., C.B. Sawyer, R. McGaw, G.M.
Buckman, M. DeVecidna, R. Levinson, N.K.
Hooper, W.R. Hevendor, L. Bernstein, R.
Peto, M.C. Pike, and B.N. Ames. 1984. A
carcinogenic potency database of the
standardized results of animal bioassays.
Environ. Health Perspect. 58:9-319.
Hattis, D. 1990. Pharmacokinetic principles
for dose rate extrapolation of carcinogenic
risk from genetically active agents. Risk
Anal. 10:303-16.
Hattis, D. 1991. Use of biological markers and
pharmacokinetics in human health risk
assessment. Environ. Health Perspect.
89:230-8.
Hill, T.A., R.C. Wands, and R.W. Leukroth, Jr.
(eds.). 1986. Biological Bases for
Interspecies Extrapolation of
Carcinogenicity Data. Prepared for Food
Safety and Applied Nutrition, Food and
Drug Administration, Department of Health
and Humas Services, Washington, DC,
under contract to Federation of American
Societies for Experimental Biology, FDA
contract N° 223-83-2020.
Hoel, D.G. 1977. Some problems in low-dose
extrapolation. In: Hiatt, H.H, J.D. Watson,
and J.A. Winsten (eds.) Origins of Human
Cancer: Book C, Human Risk Assessment.
Cold Spring Harbor Conferences on Cell
Proliferation, Vol. 4. Cold Spring Harbor
Laboratory, Cold Spring Harbor, L.I., N.Y.
Hogan, M., and D.G. Hoel. 1982.
Extrapolation to man. In: Hayes, A.W. (ed.)
Principles of Toxicology. Raven Press, New
York.
Ings, R.J.M. 1990. Interspecies scaling and
comparisons in drug development and
toxicokinetics. Xenobiotica 20:1201-31.
Kaldor, J.M., N.E. Day, and K. Hemminki.
1988. Quantifying the carcinogenicity of
antineoplastic drugs. Eur. J. Can, C. 24:703-
11.
Kleiber, M. 1932. Body size and metabolism.
Hilgardia 6:315-53.
Kleiber, M. 1961. The Fire of Life: An
Introduction to Animal Energetics. Wiley,
New York.
Lindstedt, S.L. 1987. Alloemtry: body size
constraints in animal design. In: National
Research Council. Pharmacokinetics in
Risk Assessment: Drinking Water and
Health, Vol. 8. National Academy Press,
Washington, DC.
Lindstedt, S.L., and W.A. Calder. 1976. Body
size and longevity in birds. Condor 78:91-4.
Lindstedt, S.L., and W.A. Calder. 1981. Body
size, physiological time, and longevity of
homeothermic animals. Quart. Rev. Biol.
56:1-16.
Mantel, N., and M.A. Schneiderman. 1975.
Estimating safe levels, a hazardous
undertaking. Cancer Res. 35:1379-86.
Metzger, B., E. Crouch, and R. Wilson. 1989.
On the relationship between
carcinogenicity and acute toxicity. Risk
Anal. 9:169-77.
Mordenti, J. 1986. Man versus beast:
Pharmacokinetic scaling in mammals. /.
Pharmacol. Sci. 75:1028-40.
National Academy of Sciences (NAS). 1975.
Pest Control Volume 1: An Assessment of
Present and Alternative Technologies.
National Academy Press, Washington, DC.
O'Flaherty, E.L. 1989. Interspecies conversion
of kinetically equivalent doses. Risk Anal.
9:587-98.
Pinkel, D. 1958. The use of body surface area
as a criterion of drug dosage in cancer
chemotherapy. Cancer Res. 18:853-6.
Raabe, O.G., S.A. Book, and N.J. Parks. 1983.
Lifetime bone cancer dose-response
relationships in beagles and people from
skeletal burdens of 226Ra and 90Sr. Health
Phys. 44, Supl.l:33-48.
Rail, D.P. 1977. Species differences in
carcinogenesis testing. In: Hiatt, H.H, J.D.
Watson, and J.A. Winsten (eds.) Origins of
Human Cancer: Book C, Human Risk
Assessment. Cold Spring Harbor
Conferences on Cell Proliferation, Vol. 4.
Cold Spring Harbor Laboratory, Cold
Spring Harbor, L.I., N.Y.
Reitz, R.H., A.L. Mendrala, and P.P.
Guengerich. 1988. In vitro studies of
methylene chloride (MEC) metabolism in
human and animal tissues: Use in
physiologically-based pharmacokinetic
(PB-PK) models. The Toxicologist 621.
Sacher, G.A. 1959. Relation of lifespan to
brain weight and body weight. In:
Wolstenholme, G.E.W., and M. O'Conner
(eds.) The Lifespan of Animals. Little
Brown, Boston.
Schmidt-Nielsen, K. 1970. Energy metabolism,
body size, and problems of scaling. Fed.
Proc. 29:1524-32.
Schmidt-Nielsen, K. 1975. Scaling in biology:
the consequences of size. /. Exp. Zool.
194:287-308.
Schmidt-Nielsen, K. 1984. Scaling: Why is
Animal Size so Important? Cambridge
University Press, Cambridge.
Schein, P.S., R.D. Davis, S. Carter, J. Newman,
D.R. Schein, and D.P. Rail. 1979. The
evaluation of anticancer drugs in dogs, and
monkeys for the prediction of quantitative
-------
Federal Register / Vol. 57, No. 109 / Friday, June 5, 1992 / Notices
24173
toxicities in man. Clin. Pharmacol.
Therapeut. 11:3-40.
Travis, C.C. 1990. Tissue dosimetry for
reactive metabolites. Risk Anal, 10:317-21.
Travis, C.C., and R.K. White. 1988.
Interspecific scaling of toxicity data. Risk
Anal. 8:119-25.
Travis, C.C., R.K. White, and R.C. Ward. 1990.
Interspecies extrapolation of
pharmacokinetics. /. Theor. Biol. 142:285-
304.
U.S. EPA. 1984. Health assessment document
for epichlorohydrin. EPA-600/8-83-03F.
Available from National Technical
Information Service, Springfield, VA. PB85-
132363/AS.
U.S. EPA. 1987a. Technical analysis of new
methods and data regarding
dichloromethane hazard assessment. EPA/
600/8-87/029A (Review Draft, June 1987).
Available from National Technical
Information Service, Springfield, VA. PB87-
228557/AS.
U.S. EPA. 1987b. Update to the health
assessment document and addendum for
dichloromethane (methylene chloride):
pharmacokinetics, mechanism of action,
and epidemiology. EPA/600/8-87/030A
(Review Draft, July 1987). Available from
National Technical Information Service,
Springfield, VA. PB87-228565/AS.
Vocci, F., and T. Farber. 1988. Extrapolation
of animal toxicity data to man. Regul.
Toxicol. Pharmacol. 8:389-98.
Voisin, E.M., M. Ruthsatz, J.M. Collins, and
P.C. Hoyle. 1990. Extrapolation of animal
toxicity to humans: interspecies
comparisons in drug development. Regul.
Toxicol. Pharmacol. 12:107-116.
[FR Doc. 92-13207 Filed 6-4-92; 8:45 am]
BILLING CODE 6560-50-M
*U.S. Government Printing Office : 1992 - 312-014/40149
-------
------- |