52 • Alternatives to Animal Use in Research, Testing, and Education 
limitations include various information deficien- 
cies, such as: 
• inadequacies of information on most of the 
survey and data collection methodologies, 
• difficulties with definitions, 
• problems with categorizing animals under 
areas of use, 
• reporting requirements of different data 
sources, and 
• an inability to verify completeness of data 
sources. 
For example, there is often a discrepancy in the 
definition of the term “use.” In some cases, the term 
reflects the number of animals acquired; in other 
cases, it corresponds only to those used in labora- 
tory experiments. This distinction is frequently 
obscured in the data sources, and only after care- 
ful reading of the documents (and, sometimes, per- 
sonal inquiry) was the definition used in each case 
clarified. This leads to large differences in num- 
bers, since not all animals acquired are used in 
experiments. It also makes any comparative anal- 
ysis between surveys very unsound. 
In addition to this problem of the difference be- 
tween production and use, the extrinsic problem 
of the number of animals not used in a procedure 
because they do not fit the proper criteria comes 
into any extrapolation of animal use from labora- 
tory-animal market share data. A substantial pro- 
portion of the animals bred for research die or 
must be discarded because they do not meet pro- 
tocol specifications (age, sex, weight, general 
health). The number has been estimated as be- 
tween a few percent of those acquired to almost 
50 percent. In general, the unused proportion of 
a species is inversely related to the cost of the ani- 
mals. In other words, the more expensive the ani- 
mal, the less likely it will be unused, once bred 
or purchased. Thus, nonhuman primates are much 
less likely to go unused than are mice or rats; in 
some cases 50 percent of a rodent species may go 
unused. Using only one sex of a rodent species in 
a given experiment, for example, would account 
for 50 percent of the animals going unused. This 
information must be borne in mind when compar- 
ing “production” with "use” and when estimating 
animal use. 
Overall, these limitations reflect on the accuracy 
of the data and any projections based on them. 
The limitations are such that the only reasonably 
credible source for current use and projections 
is APHIS, particularly its institutional data sheets 
(the Annual Report of Research Facility discussed 
earlier). Only the detailed APHIS institutional data 
sheets for fiscal years 1982 and 1983 were used 
in this assessment, though those for earlier years 
were also available (although they would not have 
had any data on mice or rats, which were not even 
listed on the form until 1982). Consequently, the 
APHIS data are less reliable for the years before 
1982 inasmuch as these are based on reports to 
Congress that did not contain late -reporting insti- 
tutions. (The Animal Welfare Enforcement Reports 
to Congress underestimate use of the mandated 
species by 10 to 20 percent due to the cutoff date 
and do not treat data from Federal institutions con- 
sistently (17).) For some species, such as fish and 
birds, only rough estimates of use could be ob- 
tained, due to the diffuse nature of use and the 
fact that they are not included in the APHIS data. 
Critical Evaluation of 
Animal-Use Estimates 
In evaluating the reliability of various data 
sources, the following parameters were con- 
sidered: 
• ability to trace the methodology used in pro- 
ducing the numbers, including the survey 
technique; 
• ability to extrapolate to nonreporting institu- 
tions, which implies that there is a clear state- 
ment as to which institutions did or did not 
report data; 
• method of data collection, whether some for- 
mal manner or through a few interviews, re- 
sulting in broad estimates; and 
• ability to determine the fraction of animals 
reported as being actually used in lab experi- 
mentation, as contrasted to, for example, ani- 
mal husbandry. 
These parameters were chosen because meet- 
ing these criteria permits extrapolation of the 
limited data to the entire population of institutions. 
