180 • Alternatives to Animal Use in Research, Testing, and Education 
USE OF NONLIVING SYSTEMS IN TESTING 
Animal use can sometimes be avoided altogether 
with nonliving biochemical or physicochemical sys- 
tems, although most such systems currently re- 
quire animal-derived components. Computer simu- 
lation can also be used when there are sufficient 
data available for substances related to the one 
of interest and when the mechanisms of toxicity 
are at least partially understood. 
Chemical Systems 
Whole animals have been replaced with analyti- 
cal chemistry for tests involving detection of a sub- 
stance or measurement of potency or concentra- 
tion, such as for vaccines, anticancer drugs, and 
vitamins (10). However, toxicity testing in nonliv- 
ing systems is quite limited at this time. 
Recently developed methods of detection or 
measurement are based on the selective binding 
that occurs between a particular substance and 
the antibodies to it. In an assay for botulism toxin 
(which traditionally required up to 200 mice), an- 
tibodies obtained from rabbits are modified so that 
the binding of the toxin can be detected easily . The 
rabbits are initially injected with a small, harm- 
less dose of the botulism toxin. Small amounts of 
blood are then removed from the rabbits at regu- 
lar intervals. In 4 weeks, a rabbit can produce 
enough antibody, with little discomfort, to perform 
tests that would otherwise require thousands of 
mice (32). 
Chemical systems that test for toxicity are based 
on determining whether a substance undergoes 
a specific reaction. For example, it is well known 
that carbon monoxide binds to hemoglobin in the 
blood, thus greatly reducing the blood's ability to 
carry oxygen. The extent to which a substance 
would displace oxygen in hemoglobin can be a 
measure of its ability to produce asphyxiation. Sub- 
stances can also be tested in isolation for their ef- 
fects on enzymes crucial to certain bodily functions. 
An important limit of chemical systems is that 
they do not indicate the extent to which an organ- 
ism can recover from or prevent these reactions. 
For example, a substance that binds strongly to 
hemoglobin may not be a problem because it is 
not absorbed. A substance will not have a signifi- 
cant effect on an enzyme of interest if it is excreted 
before it has an effect. 
Physicochemical systems have some ability to 
determine whether a substance will be absorbed 
and what will happen to it. The tendency of a sub- 
stance to accumulate in a biological system can 
be roughly estimated by the relative proportions 
that dissolve in equal volumes of water and the 
organic solvent octanol (34,55). Artificial skin made 
with filter paper and fats is being tried as a means 
of mimicking absorption of cosmetics and drugs 
(45). Reactivity and other toxicity -related proper- 
ties can be deduced from chemical structure alone 
(109). 
Mathematical and Computer Models 
Advances in computer technology during the 
past 20 years have contributed to the development 
of sophisticated mathematical models of quantita- 
tive structure activity relationships (QSAR). These 
models are used to predict biological responses 
on the basis of physical and chemical properties, 
structure, and available toxicological data. The limi- 
tations of such models are due in part to a lack 
of understanding of the mechanisms by which 
toxic effects occur. 
In applying QSAR, the biological effects of chem- 
icals are expressed in quantitative terms. These 
effects can be correlated with physicochemical 
properties, composition, and/or structure. Fre- 
quently used properties include an affinity for fats 
versus water (octanol/water partition coefficient), 
the presence of certain reactive groups, the size 
and shape of molecules, and the way reactive frag- 
ments are linked together. 
The simplest extrapolation is for a series of 
closely related chemicals. The several character- 
istics they have in common need not be incorpo- 
rated into the model as variables. This type of 
analysis has been performed for several hundred 
families of chemicals and has established that rela- 
tionships within a series are fairly predictable (64). 
Another approach, more broadly applicable, is 
to examine the contributions of various portions 
of a molecule. In more elaborate computer pro- 
