PRINCIPLES OF TOXICOLOGY

13.7 EMPIRICAL MEASURES OF RELIABILITY OF THE EXTRAPOLATION 299
nonresponsive species would appear to be the only way to improve our use (extrapolation) of chronic
cancer bioassay data.
Are Some Test Species Too Sensitive?
A number of strains or species have a significantly higher tumor incidence in a particular tissue than
do humans. The incidence of liver tumors in B6C3F 1 mice was discussed earlier. Another example is
the strain A mouse, a mouse strain sometimes used to test a chemical’s potential to induce lung tumors.
In this particular mouse strain the incidence of lung tumors in the control (unexposed) animals will
reach 100 percent by the time the animals have reached old age. In fact, because all animals will at
some point develop lung tumors, a shortening of the latency (time to tumor) or the number of tumors
at an early age are used, rather than the final tumor incidence measured at the end of the animals’ lives.
The use of positive data from an animal species with a particularly high background tumor incidence
poses several problems. For example, are the mechanisms of cancer initiation or promotion the same
for this chemical in humans? Can the potency of the chemical be estimated or even ranked when it
might not be clear if the enhanced animal response is just a promotional effect of high background rate
or the added effect of a complete carcinogen? Where the biology of the test animal clearly differs from
that of humans is a positive response meaningful without corroboration in another species?
13.7 EMPIRICAL MEASURES OF RELIABILITY OF THE EXTRAPOLATION
What is the Reliability of the Species Extrapolation?
To test the reliability of making interspecies extrapolations, scientists have analyzed the results of a
large number of chronic animal bioassays to ascertain the consistency with which a response in one
species is also observed in another species. In one of the largest analyses performed to date, scientists
analyzed the results for 266 chemicals tested in both sexes of rats and mice. The data forming this
analysis is presented in Table 13.8.
From the findings discussed above, after defining concordance to be species agreement for both
positive and negative results, the authors of this analysis concluded the following:
• The intersex correlations are stronger than the interspecies correlations.
• If only the male rat and female mouse had been tested, positive evidence of carcinogenicity
would have led to the same conclusions regarding carcinogenicity/noncarcinogenicity in 96
percent of the chemicals tested in both sexes of both species (i.e., 255/266 correct responses).
TABLE 13.8 Correlations in Tumor Response in NCI/NTP Carcinogenicity Studies
Observed Outcome %
Concordant
(++ or ––)
Comparison Responses
a
+ +– –+ –– Total
Male rats vs. female rats 74 25 12 181 292 87.3
Male rats vs. male mice 46 43 36 145 270 70.7
Male rats vs. female mice 29 33 36 145 273 74.7
Female rats vs. male mice 46 32 37 156 271 74.5
Female rats vs. female mice 57 23 39 156 275 77.5
Male mice vs. female mice 78 10 23 177 288 88.5
Rats vs. mice 67 32 36 131 266 74.4
Source: Adapted from Haseman and Huff (1987).