Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 237
4
DETAI LED DESCRIPTION OF THE
OPERATION OF AN ILLUSTRATIVE SYSTEM
INTRODUCTION
An illustrative priority-setting system has been developed to
demonstrate the application of principles described in Chapter 2 and to
demonstrate the feasibility of the approach to designing priority-setting
systems.
The system responds to the approach taken in Part 1, in that it uses
the same classes of intended use and incorporates two of the toxicity
tests used by the Committee on Sampling Strategies and the Committee on
Toxicity Data Elements. The approaches differ in that the illustrative
system is structured to be specific for a human health effect, whereas
those two committees take a more general approach.
ELEMENTS OF DESIGN
Authors of priority-setting systems have several choices to make:
goal of the system, universe of chemicals considered, structure of the
system, and assessment.
GOAL
The goal chosen for this illustrative priority-setting and testing
system is to assess accurately the public-health concern about all
chemicals to which there is human exposure. This goal was chosen to
reflect the wide range of chemicals of interest to NTP. A goal might be
chosen to be oriented more toward regulation, such as minimizing the
impact on human health caused by chemicals to which there is exposure.
The Committee on Priority Mechanisms did not choose this goal, because
NTP does not have regulatory authority and because NTP does testing for
purposes other than regulation.
UNIVERSE
The universe of chemicals to be considered is defined to include all
chemicals to which there is potential human exposure. The universe
defined by the Committee on Sampling Strategies has five categories of
intended use: food chemicals, drug ingredients, pesticide chemicals,
cosmetic ingredients, and general industrial chemicals (TSCA Inventory)
There are two additional categories: "other," such as pollutants; and
"unknown."
237
OCR for page 238
ASSESSMENT
The degree of public-health concern is a combination of degree of
human exposure and degree of toxicity to humans. For purposes of
designing this illustrative system, toxicity is limited to
carcinogenesis. This health effect was chosen because of its severity
and the high degree of public interest in it. Also, more effort has been
devoted to discovering carcinogens; therefore, more data are available
for designing the system.
Both toxicity and exposure are expressed as high, medium, or low,
because available information seems inadequate to define more than three
levels. Toxicity, in this case carcinogenic potency, is defined in terms
of TD50 (Ames et al., 1982~: high is a TD50 less than 102 ~g/kg-d,
medium between 10~-and 106 ~g/kg-d, and low greater than 106
~g/kg-d. His definition assumes a linear dose-response curve. The
category of "low" toxicity also includes noncarcinogens. For exposure,
high is greater than 107 person-grams per year (p-g/yr), medium is
between 107 and 105 p-g/yr, and low is less than 105 p-g/yr.
Note that, in this population-exposure formulation, large exposure of
a small number of people could be equivalent to small exposure of a large
number of people. Which situation is more important is a matter of
social judgment.
STRUCTURE
A multistage structure was chosen, so that the system could handle
many thousands of chemicals and fit the current institutional structure
of the NTP selection system. A multistage structure provides for using
small amounts of data to assess large numbers of chemicals in the early
stages and examining fewer chemicals in depth in later stages.
RULES FOR SELECTION
The result of each stage is an assessment of the toxicity and
exposure of each chemical considered in that stage and an estimate of
confidence in each assessment. Having these assessments from each stage,
it is still necessary to decide whether a chemical should receive further
consideration or be removed from consideration. A mathematical model
chooses rules for selection such that the errors in assessment are
minimized, for a chosen amount of resources for selection and testing
(Appendix B).
That optimization model not only guided the Committee in designing
the illustrative priority-setting scheme, but also allowed it to examine
the values of and interactions among various indicators of exposure and
238
OCR for page 239
toxicity. Even the simplified model used for choosing the rules of the
illustrative system validated some intuitive judgments about testing
priorities and provided valuable additional insights; further development
of the model may be justified to sharpen the rules for chemical selection
in an operational priority-setting system and to modify them as
additional information is gained by operating the system and conducting
the tests it selects.
DESIGN PARAMETERS
Several characteristics of the universe were estimated and serve as
data for the model used to develop the rules for the illustrative
system. These estimates were based on fragmentary data and are intended
to be replaced with better estimates based on data obtained from
operating the system. The universe was estimated to consist of 70,000
discrete chemicals. The Committee on Priority Mechanisms intends that
the definition of the universe be considered flexible and subject to
change as needed. The estimate of 70,000 is not intended to become
either a maximal or a minimal number of chemicals to be considered. On
the basis of the above definitions of degree of exposure, the overall
distribution of low, medium, and high exposure is 0.7, 0.2 and 0.1. On
the basis of the definition of carcinogenic potencies, the estimated
distribution of potency is 0.95 low, 0.04 medium, and 0.01 high. These
estimates are only for purposes of designing the illustrative system and
are highly speculative. Estimates of the proportion of carcinogens in a
group of chemicals may vary widely (U.S. Congress, 1981), because of the
definition of "carcinogen" and the group of chemicals considered.
STAGE 1
me purpose of Stage 1 is to provide a mechanism for scanning the
entire population of chemicals to be considered by the system. It is
intended that this scanning could be performed in many ways: by category
or combination of categories of intended use or by chemical structure.
The scanning need not be performed in one year, but might be performed
over a period of a few years.
Stage 1 is limited to data in machine-readable data bases, so that
human intervention is minimized. The Committee on Priority Mechanisms
chose to use the data bases and searching capability of the Chemical
Information System (CIS) developed by the National Library of Medicine
(NLM) and EPA. This system was chosen because many chemicals of interest
are already included in the data bases. In addition, the Structure and
Nomenclature Search System of CIS provides a searching capability for
chemical structures and thus makes possible a crude analysis of
structure-activity relationships.
239
OCR for page 240
EXPOSURE DATA ELEMENTS
The purpose of exposure data elements in all stages is to contribute
to an attempt at estimating exposure on the basis of surrogate data--data
that are related to exposure only incompletely or inaccurately.
"Exposure" is a concept that embraces the quantities of material that
reach people, the number of people exposed, the rates and patterns of
exposure, the characteristics of the exposed population, and other
factors. A fully developed system for priority-setting not only would
take into account as many of these dimensions as feasible, but also would
ensure that they interacted properly with the dimensions of toxicity.
However, complexity may not be justified for screening decisions at early
stages of the system, and it also obscures the logic of the illustrative
system. Therefore, the committee conducted its analysis with a single
exposure variable to which degree of hazard would be proportional for a
constant degree of toxicity. If a linear dose-response relationship is
assumed for toxicity as a first approximation, then exposure can be
measured by the total mass of a substance that is ingested, inhaled, or
otherwise taken in by all members of the population, in units of
person-grams per year. This quantity can be estimated by summing the
products of the per capita annual intakes and the number of persons for
all exposed groups with different intakes.
Two surrogates of exposure--class of intended use and production
volume--are used to define the possible classifications of a Stage 1
exposure data element (Table 3~; each classification consists of a pair
of subelement classes--use and production volume. The first is derived
from the definitions of intended use devised for the select universe, and
the second from the TSCA Inventory or other automated sources.
We assume that, on the average, a higher fraction of food-chemical
and drug production than of cosmetic, general TSCA chemical, or other (as
yet undefined) chemical production eventually results in human exposure.
New classes would need to be examined to test this assumption. Prime
candidates for new classes are general environmental chemicals in air and
water (including degradation products and "natural" pollutants) and food
constituents and their products. If one accepts the exposure assumption,
then the production volume required for a high probability of exposure
should be lower for food chemicals and drugs than for chemicals in the
other use classes. Table 3 lists an illustrative exposure
classification, in which there is a decreasing probability that exposure
is high; where the probabilities are identical, the list is in decreasing
order of confidence in the estimates. The ability of the classifications
to identify high-exposure chemicals cannot be determined without
estimating the numbers of chemicals in the universe of concern that fit
these classifications. Table 4 is a hypothetical classification and
would need to be revised as data and experience accumulate.
240
OCR for page 241
TABLE 3 Illustrative Stage 1 Estimates of Probability of Exposure
in Relation to Use and Production
Probability
That Exposure Is:
UseaProduction, lb/yr Low Medium High
F,D>104 0.40 0.40 0.20
P,C>105 0.40 0.40 0.20
G,O> 108 0.40 0.40 0.20
U> 106 0.40 0.40 0~20
G,O1o6_1o8 0.50 0.35 0.15
P,C104-105 0.50 0.35 0.15
U1o4_1o6 0.68 0.20 0.12
F,DU 0.68 0.20 0.12
F,D,U<104 0.73 0.18 0.09
P,C,G,O,UU 0.73 0.18 0.09
P,C< 104 0.75 0.17 0.08
G,O104-106 0.75 0.17 0.08
G,O<104 0.76 0.17 0.07
a F = food chemical.
D = drug.
P = pesticide.
C = cosmetic.
=
general commerce (TSCA).
Other (known, not previously classified).
unclassified.
241
OCR for page 242
1
TABLE 4 Illustrative Estimates of Distr ibution of Production Volumes
in Relation to Use Categories
Frac
tion
Fr action by Production Volume, lb/yr
of All
Chem
Use Class icals > 108 1o6 _ 1o8 105 _ 1o6 104 - 105 <104 ua
b -- 0.001 0.003
Food 0. 12
Drug 0.10
Pesticide 0.05
Cosmetic 0.05
0.003 0.003 0.11
c 0.001 0.009 0.020
0.03 0.06
0.001 0.002
c 0.001 0.004 0.015 0.025 0.055
0.004 0.093
0.001 0.002 0.002 0.045
c 0.0002 0.006 0.005 0.013 0.026
b -- 0.001 0.001
c 0.001 0.002 0.003
__
0.002 0.002 0~044
0.014 0.03
__
General 0.066 0.02 0.06 0.06 0.09 0.28 0.3
Other 0.01d 0
o
o
Unknown 0.01d 0
o
o
o
o
a In inventory, but without stated production volume.
b Estimated distribution of production volumes listed in CIS.
c Estimated distribution of production volumes if all were known.
d At present--these categories will grow.
242
0 0.01
0 0.01
OCR for page 243
TOXICITY DATA ELEMENTS
Two surrogates of toxicity, RTECS status and chemical class, are
used in Stage 1. Each is used to estimate the likelihood that a
substance may have toxic effects. These surrogates indicate that the
likelihood that a given chemical or class of chemicals will have health
effects is different from the likelihood that any substance in the
universe of chemicals will have those health effects.
Given the present state of knowledge, it is only slightly possible to
assess the potential for an adverse response on the basis of chemical
structure. But it would be desirable to use knowledge of
structure-activity relationships (SARs) to identify potentially toxic
substances by analyzing substructures that have been associated with
adverse effects in humans or animals. Such an assessment is ordinarily
based on expert knowledge, experience, and intuition and is used in
considering the type of testing that may be required.
A number of systems have been created to place SAR analysis on a more
formal footing. They range from simple classifications of key types of
substances to sophisticated statistical treatments that involve weighting
of subgroups and from detailed treatments of specific health effects to
general considerations of toxicity. These approaches have merit as
research efforts, but none has evolved enough to provide a practical or
accurate method for identifying potentially toxic chemicals. A more
detailed review of the possible contribution of SAR analysis is given in
Appendix D.
To be usable in Stage 1, any data base used for SAR analysis must
contain a large proportion of the universe of chemicals being
considered. The data must also be susceptible to a search for chemical
subgroups that can be associated with specific types of toxicity. The
SANSS data base may provide a potentially useful data base. It can be
searched for any of 271 functional groups containing at least two
nonhydrogen atoms, any of 137 specific cyclic nuclei, and a number of
hydrocarbon radicals. Each structural group or feature is described by a
code that can be modified to allow for other structural features, such as
the attachment of phenyl nuclei and aromaticity in ring structures. At
the simplest level, the system provides a machine-readable way of
identifying all the compounds in the system's collections that contain a
specified structural group. The system may also be programed to identify
structures with specified subgroups; thus, the code for phenols can be
modified to produce separately monocyclic, dicyclic, and tricyclic
phenols. And the system allows identification of compounds that have
more than one specified functional group; this permits the
identification, in lists of compounds that contain a given structure, of
substances in which that structure is accompanied by another structure
that might modify its biologic activity.
243
OCR for page 244
About 80 of the SANSS specific functional-group codes are associated
with one or more human health effects, as shown in Table 5. The groups
are assumed to produce the ascribed effect either by direct action or
after metabolic activation. The table is illustrative and should not be
considered comprehensive or definitive. The estimated degree of
association between a given chemical structure and a specific form of
toxicity is indicated in the table as low (L), medium (M), or high tH).
An association between a chemical structure and an effect and the degree
of that association imply that chemicals that contain the structure are
more likely to cause the particular health effect in question~than are
randomly selected chemicals in the universe being considered.
RTECS, published by the National Institute for Occupational Safety
and Health, is the most comprehensive (but not the only) summary of
information on the toxicity of chemicals that is available in easily
accessible, computerized form. In the 1980 edition, positive results of
toxicity testing of some 45,000 chemicals are reported. In general,
negative results of toxicity testing are not reported, so RTECS is biased
toward toxicity. But RTECS does report negative results of lifetime
bioassays for carcinogenicity by NTP (and earlier by the National Cancer
Institute).
Consequently, mere listing in RTECS slightly raises the probability
of high or moderate toxicity at the expense of the probability of low
toxicity. RTECS also reports specific positive results, such as
carcinogenicity, mutagenicity, teratogenicity, reproductive effects, and
a variety of acute and other chronic toxicities. Nevertheless, in
practice, even listing as a carcinogen carries with it a false-positive
rate, because of the imperfect correlation in cancer hazards among
different species, experimental uncertainties, faulty experimental
procedures, or mere error. The absence of a CAR code implies that the
substance may have proved negative in a bioassay, but the false-negative
rate for such an assumption is high. Similar problems beset other
toxicity codes.
In a fully developed priority-setting system, each toxic effect of
concern would be related to at least one RTECS code. Matrices of
probability for a given effect would be constructed to show how the
underlying prevalence rates are modified for appearance or nonappearance
of particular codes, which are retrievable by computer search. In this
report, we discuss only carcinogenicity, both because time was too
limited to develop other "performance characteristics" and because the
theory and data base for carcinogenicity are better developed than those
for other kinds of toxicity.
Six codes are thought to be related to carcinogenicity: NTP POS,*
NTP NEG, CAR, NEO, ETA, and MUT (MTDS in the latest RTECS). NTP POS
*Many RTECS codes refer to NCI-positive or -negative results, rather than
to quantitative results of tests conducted according to NTP-approved
protocols.
244
OCR for page 245
:'
l
ye ~ o ~ ~ - ~e
cooux:~a 0
1
V
41 ~ 1
~ V ·44
cq ~ ~ w
~ 0 ~
14 1
0
C Hi--
~ x;a 0
~ U
1 U 15 1
D'- - C ~ ~
~ ~ ~ ~ L'
~ sea 0
C ~
~ 1 1 0
~ ~ ~ a 0
OF
.-,
v
..
~1
a: 1
1
TIC
-
u
. -
4,
:~:
w
¢
Ld
o
u
~4
. -
C)
TIC
4J
. -
:,
1 1 ~ 1
00 ~ 0
a- - ~
:> ~ C: 0
1 ~ 1
-4 C
~ o
aa 0
1 1
..C
~ 0 v
Z ~ ~
1 ~ V
o.
0 .,
1 -~ C a
1 C
1 0 ~
C C
U.
. -
0
¢
C U
a'
~4
U)
in
eC Z
V ~ ~
c
:~:
TIC
4
V
. -
U
0
0 ~ ~
o
v
u
Us ~
. - a,
Cal ~ ~
J ~ O
C7
Ed
:~:
~:
~r
~o
~:! ~ .
,0 1
oo
_I
~ o' o
o,
In
0 0 u ~0 0 ~n
GD ~GD
cr ~ v~
P4
~10
O '0 ~J\zo o o C~
~CJ ~U I Q1 ~ ~Y
ro
0
v
0
V
U
. -
V V ~
~C
_
-I U U
_
·. V ~0
a' c ~ J:
O ~04
.c ~ ~o aC ~
-4 0 ~ U ~ 0 0 :>
0 ~a
~: ¢
245
OCR for page 246
C ~ 0 1 C 1
·~- ~ ~ ~ 0 ~
~ ~ O ~ 4-~ 4
u, 0 ox: Ma 0
1 00
~ 41 1 t.)
Y
U] ~ ~ Ad
4~
TIC
c
3 o 4
4 1
0= 4 0 ~
_ C ~ ~ 4
~ ~ x: a 0
1 ~ ~ 1
=~- c m
~ ~ ~ ~ 4
Hi; 4~ a: a 0
C 14
:, 1 1 do
~ ~ do- - 4
~ u, ~ a 0
:' 4
1 ~ ~ 1 0
4 0 U' 4 U' ~
as- - ~ be- - 4
c' ~ :> ~ a 0
`4 4
1 ~ 1
C ~ ~
O · - 4
o' ea 0
I
~ 0
Z 4
1 ~1
0 4
1 _1 C ~1
4' ~
.,.
00
1 0
C C ~
~ · - ~ . -
C) ~ ~ ~
EN
U)
Z
4
:,
-4
~0 ~
-
C
-
Vc
o
_
US ~
· - o.
O
ED ~ ~
A:
A:
A:
J S X J -1 X ~:
J J J 3:
red 0
or
~:
0= 1
1°
1 1 1 O=
z Z zz 1
1 1 1 `2: 1
r~
r~
_1 1
~o C~
r~
~D
~4
~:
z
c]
z '1 z ~°=z
c> ~
° u 3
~J · - 4 4 111
.C ~U ~C
O go ~C
~o ~o ~e ~c ~
~C ~0
·. · - ~U _ · - 0C)
U) 4 1U _ ~_1 4
~4 C ~>~
e ~4 X
· - ~V 0 · - O OO
~C ~N ~N
246
OCR for page 247
a:
~ o ~ " -
0 of Ha 0
1 0
1 U
dew
to ~ ~ w
c
~ ~ 0 ~
O ~ 1
0 ~ ~ ~ ~
_I ~ of- - 4
a: ~ O
~ ~4
1 U ~ 1
- C
(U dJ o. -
~ ~ ~ a 0
C Ll
:, 1 1 ~
us ~ ~ 0
:,
1 1 U 1
~ 0 ot ~ ~ ~
· -
c, ~ ~ ~ a 0
u
~4
w
tic
~x ~x:
~Ill
1 ~ 1
a,
a" 0
I ~ 3
~ 0
z
1 ~1 ~
u
0 ~ ~
1 _1 ~ W
w
a >~=
1 C
4) ~ . -
~ cn 0
1 0 ~
_
u
I: ~
E.
in
In
:c
~ 8
In
to
ret
up ~
£
rid I ~ ,, - 1 \0
Z
~$ ez ~ ~ ~
~ ~0=1 O-1 Z:
0
0
=0
C e.
~ U
L. ~
~0~ u o
_I ~ ~N
D4 N ~ ~D
o" ~O~ ~U
247
11
~,
11 0 =
1
~0
u
0
.~4
D
o
OCR for page 274
TABLE 16 Hypothetical Estimates of Probability Distributions of
Carcinogenic Potency as Generated in Stage 3, in Relation to Assigned
Degree of Concern about Carcinogenicity and Degree of Confidence in
Assignment
Degree of
Toxicity
Low
Medium
High
Degree of Confidence in Estimate
(Estimated Probability Distribution
of True Toxicity), %
Low Medium High
96-3-1 97-2-1 98-1 1
75-21-4 57-40-3 40-57-3
40-40-20 30-40-30 25-35-40
m e corresponding distribution of judgments by the expert committee
might be as follows:
Degree of Degree of
Concern,% Confidence, ~
Low Med. High Low Med. High
Stage 3 toxicity dossier 60 30 10 60 25 15
These distributions are consistent with an assumpton that 81% of the
chemicals moving from Stage 2 to Stage 3 are noncarcinogens, 15% are weak
carcinogens, and 4% are strong carcinogens.
Note that the expert committee is not asked to make any judgment
about the value of proposed tests. We propose that systematic and
largely predetermined rules be used to determine the value of a
mutagenicity assay or a lifetime bioassay in clarifying the concern about
carcinogenicity for a specific chemical that has already been assessed by
an expert committee.
The recommendations for testing would then be reviewed by the board
of scientific counselors, or a similar body, to ensure that they are
reasonable, given the status of existing information and the detailed
properties of the substance in question. Unfortunately, this whole
approach to testing decisions is not as well suited to evaluating
reconnaissance testing (e.g., 90-d feeding studies), in which an effect
of concern may be specified only generally, if at all. The performance
characteristics for such tests would have to be either created from the
aggregate of several effects that tests could detect or defined for a
generalized toxicity measure, which would be sought only because exposure
is fairly high and fairly certain.
274
OCR for page 275
SELECTION RULES
As in Stage 2, the Stage 3 rules (Appendix B) operate on the finding
of high, medium, or low toxicity and high, medium, or low exposure;
later, they should be revised to consider degree of confidence. The
choices for Stage 4 are a short-term test, a long-term cancer bioassay,
and no action.
STAGE 4
Because the objective of priority-setting must be the minimization of
misclassification of the potential public-health concern about chemicals,
the system cannot be complete without an analysis of the predictive power
of the tests to which it leads. We discuss here some problems related to
the selection of toxicity tests and the relations between operating
characteristics of tests and selection of chemicals for toxicity testing.
At this stage of the priority-setting process, the goal is to select,
for the chemicals to be evaluated, tests that provide the most useful
toxicity information for the resources invested. Because testing
resources are limited, selection of an optimal test or combination of
tests involves a tradeoff between the total number of chemicals evaluated
and the completeness with which any one chemical can be characterized.
To illustrate how this tradeoff can be made, we constructed a
mathematical model for the allocation of resources for one kind of
toxicity, carcinogenicity (Appendix B). This choice was made both
because methods for carcinogenicity testing are better than methods for
most other toxicity testing and because carcinogenicity testing consumes
a larger fraction of NTP resources than does testing for any other toxic
effect.
Tests of other effects, like tests to assess carcinogenicity, entail
consideration of both cost and accuracy of testing. These tests range in
cost from a few hundred dollars for some short-term tests with bacteria
to more than a half-million dollars for a standard long-term bioassay in
rats and mice (Table 17~. The cost of a given test may vary among
laboratories, depending on the substance being tested, the conditions of
testing, institutional overhead expenses, and other factors (Lave et al.,
1982).
Tests vary in predictive accuracy. None is considered accurate
enough to predict with complete reliability the carcinogenicity of a
substance for humans; that can be established only by epidemiologic
evidence of carcinogenic effects in humans. Nevertheless, the
demonstration that a substance is carcinogenic in appropriately conducted
animal tests is generally considered an adequate basis for classifying it
a presumptive human carcinogen (IRLG, 1979; U.S. Congress, 1981; IARC,
1982~. Interpretation of animal tests is uncertain, however, and
requires expert judgment. Extrapolation from animal test results to
quantitative estimation of human risks entails large uncertainties in
275
OCR for page 276
TABLE 17 Estimates of Costs of Some Carcinogenicity Testsa
.
Testsb TO e of Test
- P
Bacterial cell:
Estimated
Average Cost, No.
$ Responses _
1-4,6-13 Salmonella his-1,200 10
19 Escherichia cold WP2400 2
21 Bacillus subtilis rec-800 1
22-24 Escherichia cold rec-1,500 1
29 Degranulation 2,500 1
37 Yeast cell:
Saccharomyces D7
· ~
cereals 1ae
Mammalian cell:
1,400 1
40-42 Unscheduled DNA synthesis-- 5,200 3
human fibroblasts,
HeLa ceils
43,45 Sister-chromatid exchange-- 3,000 1
CHO cells
44 Chromosomal aberrations-- 7,500 3
CHO cells, rat liver cells
c Transformation--CHO cells 1,400 1
c Transformation--C3H-lOT 1/2 5,400 4
48 TK +/- L5178Y mouse 4,900 1
lymphoblasts
50,51 HGPRT-CHO cells, 6,500 4
V79 cells
Whole Animal:
56-58 Sex-linked recessive lethal-- 10,000 1
Drosophila melanoqaster
(injection)
59 Sister-chromatid exchange--mouse 3,000 1
60-62 Micronucleus--mouse 3,400 2
63 Sperm morphology--mouse 11,400 1
Whole-animal two-species rodent
bioassay
a Modified from Lave et al., 1982.
b See Table 19 for list of tests.
c Test was not conducted in International Collaborative Program (de Serres
and Ashby, 1981) in either CHO or C3H cells, but in BHK-21 cells.
d From Weinstein, 1983.
500,000d 1
276
OCR for page 277
connection with mechanisms of carcinogenesis and species differences in
carcinogen metabolism (IRLG, 1979; Calkins et al., 1980; IARC, 1982~.
The difficulties in interpreting tests for carcinogenicity have led the
International Agency for Research on Cancer (1982) to publish guidelines
for interpretation.
Long-term bioassays in rats and mice are so expensive (Table 17) and
time-consuming (requiring up to 5 yr for completion) that such tests are
not feasible for more than a small percentage of the many thousands of
chemicals to be tested. Faster and less expensive assays are needed. To
this end, more than 100 short-term tests have been introduced (Hollstein
_ al., 1979~. Emus far, however, such tests have been used only for
screening purposes, and not for definitive predictions of
carcinogenicity, pending further standardization and validation of their
accuracy. The need for standardization of such tests is indicated by the
variability in results of the most widely used short-term screening
test--the Salmonella typhimurium/microsome plate mutagenicity assay.
Table 18 shows the degree of agreement found by the International
Collaborative Program, which had 12 laboratories apply the test to a
series of 19 carcinogens and noncarcinogens (de Serres and Ashby, 1981~.
The discrepancies are attributable at least in part to the use of
differing procedures.
Despite the variations, the Salmonella mutation assay yielded
preponderantly positive results with chemicals known to be
carcinogenic in rodents. Nevertheless, the observed correlation between
mutagenicity and carcinogenicity has varied from one class of chemicals
to another. For some carcinogens (e.g., asbestos and halogenated
hydrocarbons, such as DDT) and for most tumor-promoting agents (such as
phorbol esters and some naturally occurring hormones), the test has given
negative results (Rinkus and Legator, 1979; Ames and McCann, 1981;
Purchase, 1982~. For chemicals of all classes tested to date, its
overall accuracy is 60-80% (see Table 19) (Ames, 1979; Purchase, 1982;
Upton et al., in press). Other short-term tests have received less
systematic evaluation than has the Salmonella test; only limited data are
available on their comparative results (see Table 19~. One naturally
successful combination of the two tests is the Salmonella test with the
in vitro cell-transformation assay.
Combinations of various short-term tests, in batteries or in tiers,
have been observed to yield greater accuracy than any one of the tests
alone (Bridges, 1976; Weisburger and Williams, 1981; Lave et al., 1982~.
Six combinations of tests have been calculated to have predictive
accuracies of 81.6-89.7% (Table 20) for a limited number of animal
carcinogens and noncarcinogens (Table 21~. None of the combinations
appears capable of avoiding a substantial percentage of false-positives
and false-negatives. However, where a false-positive is observed, it may
be suspected that the carcinogenicity of the chemical under consideration
escaped detection because of deficiencies in the animal test used as a
criterion.
277
OCR for page 278
TABLE 18 Correlation among Results of Salmonella/microsome Tests
Performed by 12 Investigatorsa
Investi- 1 2 3 4 6 7 8 9 10 11 12 13
1 1.00 0.81 0.53 0.47
0.47 0.34 1.00 0.62 0.81 0.47 0.34 0.22
1.00 0.65 0.62
0.62 0.53 0.81
0.81 1.00 0.65 0.53 0.15
1.00 0.65 0.65 0.34 0.53 0.53 0.65 0.42 0.34 0.19
4 1.00 0.33 0.53 0.47 0.47 0.62 0.65 0.26 0.15
6 1.00 0.53 0.47 0.47 0.62 0.35 0.26 0.03
7 1.00 0.34 0.34 0.53 0.46 0.30 0.05
8 1.00 0.62 0.81 0.47 0.34 0.22
9 1.00 0.81 0.47 0.34 0.07
10
11
12
13
a From Lave et al., 1982.
1.00 0.65 0.53 0.15
1.00 0.46 0.02
1.00 0.21
1.00
Entries are squared correlation coefficients
computed from test results on 19 chemicals reported to International
Collaborative Program by 12 investigators. 1.00 indicates complete agreement
and O indicates no agreement between investigators.
278
OCR for page 279
TABLE 19 Predictive Accuracy of Various Short-Term Tests
for Carcinogenicitya
Test Accuracy of Test Resultsb
Code
Number Type of Test N Accuracy
Bacterial Mutation Assays
1 S. typhimurium/plate 37 0.70
2 S. typhimurium/plate 37 0.68
3 S. tYphimurium/plate 36 0.59
4 S. typhimurium/plate 38 0.71
5 S. typhimurium/fluctuation 31 0.71
6 S. typhimurium/plate 38 0.63
7 S. typhimurium/plate 33 0.64
8 S. typhimurium/plate 38 0.68
9 S. typhimurium/plate 37 0.57
10 S. typhimurium/plate 37 0.65
11 S. typhimurium/plate 37 0.73
12 S. typhimurium/plate 38 0.68
13 S. typhimurium/plate 38 0.71
14 S. typhimurium/plate 28 0.57
norharman
15 S. typhimurium/ 33 0.55
f luctuation
16 S. tYphimurium/ 38 0. 66
azaguanine res
17 S. typhimurium/E. cold 36 0.67
W2/f luctuation
hepatocytes
18 S. typhimurium/E. cold 36 0.78
WP2/plate
19 E. cold WP/2/plate 34 0.65
20 E. cold 343 18 0.72
Bacterial Repair, Phage Tnduction, Degranulation,
and Nuclear Enlargement Assays
21 B. subtilis, M45 rec- 38 0.79
22 Ee coli, RecA/PolA 34 0.59
23 Ee coli, RecA- 37 Oe62
24 E e coli, RecA-/PolA 36 0.58
25 E. coli, Pol 36 0.58
26 X-induction (gal+) 20 0. 65
27 X-induction (lysis) 36 0.56
279
OCR for page 280
TABLE 19 (continued)
Test Accuracy of Test Resultsb
Code
Number Type of Test _ Accuracy
28 Degranulation of RER 33 0.39
29 Degranulation of RER 30 0.67
ribonuclease post
treatment
30 Nuclear enlargement, 22 0.32
He La cells
31 Nuclear enlargement, 22 0.64
human fibroblasts
Yeast Assays
32 S. cerevisiae XV185-14C 30 0.70
l
33 S. pombe PI 29 0.66
34 S. cerevisiae PG 148 35 0.43
PG-154
PG-155
PG-166
35 S. cerevisiae D4 31 0.42
_ ,
36 _. cerevisiae D6 37 0.65
37 S. cerevisiae D7 35 0.66
38 S. cerevisiae JD1 32 0.72
39 S. cerevisiae red 32 0.72
In Vitro Mammalian Test Systems
40 Unscheduled DNA synthesis, 18 0.39
human fibroblasts
41 Unscheduled DNA synthesis, 21 0.62
human fibroblasts
42 Unscheduled DNA synthesis, 38 0.68
human fibroblasts
43 Sister chromatic exchange, 19 0.53
CHO cells
44 Sister chromatic exchange, 18 0.67
CHO cells
45 Sister chromatic exchange, 33 O.S2
CHO cells
46 Cytogenetic analysis, 21 0.57
micronucleus test
47 Cytogenet~c analysis, 18 0.72
micronucleus test
48 Forward-mutation assay, 19 0.53
mouse lymphoma cells L518Y
280
OCR for page 281
TABLE 19 (continued)
Test
Code
Number
Accuracy of Test Resultsb
Type of Test N
-
AccuracY
49 Gene-mutation assay, CHO 9 0.44
cells, HGPRT gene
50 Gene-mutation assay, CHO 3 0.67
cells, HGPRT gene
51 Gene-mutation assay, 5 0.60
V79 hamster cells
53 Cell transformation 34 0.59
54 Cell transformation, 38 0.82
BKH-21 cells
In Vivo Assays
56 Sex-linked recessive lethal 9 0.44
Droposphila
57 Sex-linked recessive lethal 15 0.53
58 Sex-linked recessive lethal 9 0.44
59 Sister chromatic exchange, 16 0.63
mouse
60 Micronucleus assay, mouse 29 0.66
61 Micronucleus assay, mouse 17 0.47
62 Micronucleus assay, mouse 33 0.45
63 Sperm morphology, mouse 15 0.47
aFrom Lave _ al., 1982.
bFigures tabulated indicate accuracy (number of chemicals correctly
identified divided by number of chemicals tested) for tests 1 through
63. Data subset includes all 42 chemicals listed in Table 21 except
presumptive noncarcinogens 2, 16, 20, 22, and 27, which had high frequencies of
positive results in international study (de Serres and Ashby, 19817.
Diphenylnitrosamine (22) was found to be carcinogenic.
281
OCR for page 282
TABLE 20 Predictive Accuracy of Several Short-Term Tier
Testing Regimensa
Tier
Number Tests Included in Tierb
No. of
Chemicals
on Which
Results
Predic
tive
are Based Accuracy,
- typhimurium/plate (4)
Differential killing, B. subtilis
M45 Rec~ (21)
Cell transformation, BHK-21 cells (54)
S. typhimurium/plate (4)
Cell transformation, BHK-21 cells (54)
Forward mutation, S. Bombs (33)
S . typhimur ium/plate (4)
-
Forward mutation, S. pombe (33)
Cell transformation, BHK-21 cells (54)
Unscheduled DNA synthesis, HeLa cells (42)
4
typhimurium/plate (4)
Unscheduled DNA synthesis, HeLa cells (42)
Cell transformation, BHK-21 cells (54)
S. typhimurium/plate (4)
Cell transformation, BHK-21 cells (54)
Rabin's test (degranulation)
rat liver cells (29)
_. tyPhimurium/plate (4)
Unscheduled DNA synthesis, HeLa cells {42)
Cell transformation, BHK-21 cells (54)
38
29
29
38
30
38
84.2
86.2
89.7
81.6
83.3
81.6
.
a From Lave et al., 1982.
~ _ _
Numbers in parentheses are code numbers of tests (Table 19~.
282
OCR for page 283
TABLE 21 Chemicals Analyzed by Short-Term Tests (International
Collaborative Study~a
Chemical
Carcinogenicity
Class if icationb
2
3
4
5
6
7
8
9 B-Propiolactone
10 Y-Butyrolactone
11 9,10-Dimethylanthracene
12 Anthracene
13 Chloroform
14 1, 1,1-Trichloroethane
15 2-Acetylaminofluorene
16 4-Acetylaminofluorene
17 Dimethylcarbamoyl chloride
18 Dimethylformamide
19 2-Naphthylamine
20 1-Naphthylamine
21 N-Nitrosomorpholine
22 Diphenyloitrosamine
23 Dinitrosopentamethylene tetramine
24 Urethane
25 Isooroovl N-(3-chloronhenYl)carbamate
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
4-Nitroquinoline-N-oxide
3-Methyl-4-nitroquinoline-N-oxide
Benzidine
3,3',5,5'-Tetramethylbenzidine
4-Dimethylaminoazobenzenetbutter yellow)
4-Dimethylaminoazobenzene-4-sulfonic acid, Na salt
Benzo~alpyrene
Pyrene
_ , _ ,, ~ , ~, _,
Methylazoxymethanol acetate
Azoxybenzene
DL-Ethionine
Methionine
Hydrazine sulfate
Hexamethylphosphoramide (HMPA)
Ethylenethiourea
Diethylstilbestrol
Safrole
Cyclophosphamide
Epichlorhydrin
3-Aminotriazole
4,4'-Methylenebis(2-chloroaniline) (MOCA)
o-Toluidine hydrochloride
Auramine (technical grade)
Sugar (sucrose)
Ascorbic acid
-
+
+
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
-
a From de Serres and Ashby, 1981.
b Based on effects in human populations or experimental animals.
283
OCR for page 284
Other approaches being emphasized in current research efforts include
those for identifying nongenotoxic carcinogens, including cocarcinogens
and tumor-promoting agents (Sivak, 1982; Upton, in press), and the
exploitation of in viva short-term bioassays (Ashby, in press). It may
be envisioned that the test systems of the future will include various
assays in addition to those represented in Table 19. It may also be
expected that the particular combinations of tests used and the sequences
in which they are used will depend on the nature of the chemicals being
tested and their patterns of use. Whether a tier regimen culminates in a
long-term bioassay in rats and mice will depend on the results of the
antecedent short-term tests and their apparent predictive accuracy.
The principles involved in setting priorities for carcinogenicity
testing apply also to the setting of priorities for testing chemicals
suspected of other forms of toxicity. It must be recognized that the
process of predicting the toxic potential of a chemical is extremely
complex. There are no firm rules, or even guidelines, for obtaining
reliable answers. Attention is generally focused on devising animal
models or tests for human health effects with attributable chemical
causes. Thus, in the best circumstances, it may be possible to predict
the biologic activity of a given chemical, identify a useful end point
with an established animal model or test, and draw some tentative
conclusion regarding the chemical's likely toxicity in humans.
Results of toxicity tests on animals have been recorded in RTECS, a
machine-readable file of approximately 15,000 chemicals, although the
universe of chemicals defined by the Committee on Sampling Strategies
contains over 70,000 chemicals. Even information on well-characterized
chemicals is often limited to data on molecular structure and physical
characteristics. A chemical may have several effects on human health
that are individually dose-dependent. To set testing priorities among
different types of toxic effects, two tasks must be accomplished: a
catalog of human health effects must be established, and the various
effects must be ranked according to relative severity.
The relative importance of different types of health effects depends
on their consequences to the affected persons and to society. Ranking of
effects according to severity is implicit in most priority-setting
schemes. It involves not only technical judgments of toxicity, but also
individual perceptions of harm. An approach to determining the relative
severity of toxic effects is described in Chapter 2.
When a chemical is suspected of causing more than one toxic effect,
which is often the case, the combined impacts of all its potential
effects must be taken into account in setting priority for its testing.
Because of the multiplicity, diversity, and complexity of the health
effects of different chemicals, no attempt is made here to develop a
detailed or comprehensive system for this purpose.
284
Representative terms from entire chapter:
data elements
