Ch. 8— Alternative to Animal Use in Testing • 181 
grams, it is possible to identify likely reactions and 
cascading physiological events in various species, 
techniques first developed for pharmacology (54). 
A similar approach is the use of multitiered clas- 
sification schemes that use large databases to draw 
semiempirical conclusions (36). 
Epidemiologic Data on Humans 
Perhaps the most useful alternative to animal 
testing is epidemiologic studies on humans. Such 
studies were used to detect carcinogenicity in hu- 
mans as early as the 18th century (49,85,87). The 
most well known study detected scrotal cancer 
in chimney sweeps (85). A more recent example 
in which epidemiologic evidence was used to de- 
tect a human carcinogen was the finding that vi- 
nyl chloride causes a rare liver cancer in humans 
(26). A major disadvantage of epidemiologic studies 
is that considerable human exposure can take place 
before a toxic effect is detectable, particularly in 
the case of diseases that take many years to de- 
velop. Another disadvantage is that they can be 
quite expensive to conduct. Privacy must also be 
considered (112), preventing many data that would 
be useful from being collected or analyzed. 
Epidemiologic studies may be divided into three 
general types: experimental, descriptive, and ob- 
servational. Experimental epidemiology is the hu- 
man equivalent of animal testing— providing or 
withholding a substance to determine its toxic or 
beneficial effects . Such studies are greatly limited 
by ethical and legal considerations, as well as the 
difficulties involved in securing the cooperation 
of a large number of people. 
Descriptive epidemiology analyzes data on the 
distribution and extent of health problems or other 
conditions in various populations, trying to find 
correlations among characteristics such as diet, 
air quality, and occupation. Such comparisons are 
frequently done between countries or smaller geo- 
graphic regions, as is the case for cancer statistics 
collected and analyzed by the National Cancer In- 
stitute (9). 
Observational epidemiology uses data derived 
from individuals or small groups. Data would be 
evaluated statistically to determine the strength 
of the association between the variable of interest 
and the disease. In cohort studies, a well -charac- 
terized and homogenous group is studied over 
time. In case-control studies, a control group is 
selected retrospectively based on variables thought 
to be relevant to the effect. Both methods rely on 
an accurate prediction of the variables that are 
important and are subject to various selection 
biases (62,112). 
THE LD 50 TEST 
The LD 50 test is one of the most widely used tox- 
icity tests, and the development of alternatives to 
it is regarded by many as a high priority. As de- 
scribed in chapter 7, this acute toxicity test meas- 
ures the amount of a substance needed to kill half 
the population of the test species. The LD S0 is 
used as a rough indicator of the acute toxicity of 
a chemical. 
The LD S0 is useful for testing biological thera- 
peutics, although there remain few such sub- 
stances for which the LD 50 is the only available 
means of standardization (13,90). Other applica- 
tions, perhaps not so well justified (90), are deter- 
mining doses for other toxicological tests and set- 
ting regulatory priorities. 
J 
There has been political pressure to abolish the 
LD 50 and it has been criticized by many toxicolo- 
gists on scientific grounds. It has poor reproduci- 
bility and the results are difficult to extrapolate 
to humans because there are so many mechanisms 
by which death could occur (70,90,125). 
Despite the many criticisms of the LD 50 , most 
toxicologists agree that acute toxicity information 
has valid uses, and that measurements of lethality 
also are important. Nevertheless, the precision with 
which the LD S0 is measured is often unjustified 
for several reasons. First, most applications of the 
information do not require precision. Second, even 
if the information were precise for a given spe- 
cies, the LD S0 varies so much from species to spe- 
