222 • Alternatives to Animal Use in Research, Testing, and Education 
Under these laws, the public also has access to 
collections of published and unpublished nonpro- 
prietary data gathered to support administrative 
actions such as rulemaking. This "public docket’’ 
contains all reports, literature, memos, letters, and 
other information considered in taking the action. 
Once information has been obtained by the Fed- 
eral Government, it may be shared within and 
among Government agencies. Often such sharing 
is very informal, and with informality comes un- 
predictability and oversights. Various committees 
have been set up to facilitate intragovernmental 
networking, such as the Interagency Regulatory 
Liaison Group of the late 1970s, the Interagency 
Risk Management Council, and the Interagency 
Toxic Substances Data Committee. These efforts 
increase the amount of information available to 
solve particular problems. They also reduce du- 
plicative information requests made of industry. 
In 1980, an interagency Toxic Substances Strat- 
egy Committee examined the sharing of informa- 
tion, focusing principally on the data held by Fed- 
eral agencies (20). The Committee noted there were 
then more than 200 independent data systems, 
mostly incompatible. Barriers to sharing informa- 
tion included diverse methods of identifying chem- 
icals and differing reliability and review of the data- 
bases. The Committee noted that coordination of 
Federal agencies’ chemical data systems could re- 
duce duplication of information gathering, mini- 
mize delay, and, to some extent, decrease uncer- 
tainties in decisionmaking. The benefits of such 
coordination would likely extend beyond the Fed- 
eral Government to State and local governments, 
industry, labor, public interest groups, academic 
institutions, international organizations, and for- 
eign governments. 
BARRIERS TO USING AVAILABLE INFORMATION 
Data Quality and Comparability 
Before data are to be used, the user must be con- 
fident of their quality. This judgment is based on 
a variety of facts and inferences. People will fre- 
quently take into account the professional repu- 
tation of the investigator or the investigator’s in- 
dustrial, academic, or professional affiliation or 
organization. If the person has no reputation, good 
or bad, many scientists will not rely on that inves- 
tigator’s data. This phenomenon is most acute with 
investigations carried out in foreign countries and 
published overseas (14). Further, many scientists 
will not (and perhaps should not) trust results that 
have not been peer-reviewed. Lastly, some orga- 
nizations tend not to trust any data that they have 
not generated. 
It is important to assess the quality of data. Thus, 
even though numerical databases are convenient 
because they contain data in summary form, often 
there is no way to determine from the informa- 
tion contained there how reliable the data are (un- 
less they were peer-reviewed before being put into 
the system). This problem has been addressed by 
the National Bureau of Standards (NBS), the Chem- 
ical Manufacturers’ Association, and others. A 
workshop held in 1982 (16) recommended that 
computerized databases (discussed at length later 
in this chapter) include the following “data qual- 
ity indicators” that would allow the user to deter- 
mine reliability for specific needs: 
• the method(s) used to obtain the data, 
• the extent to which the data have been eval- 
uated, 
• the source of the data, and 
• some indication of the accuracy of the data. 
An important part of evaluating data is compar- 
ing them with data obtained using similar meth- 
ods— that is, validating the data. In deciding, for 
example, to rely on a particular test protocol, it 
is necessary to be confident not only that the test 
is a useful model of the effect of interest, but also 
that the results can be trusted, even though they 
are unexpected. For many investigators, valida- 
tion involves repeating at least a portion of an ex- 
perimental protocol in their own laboratories. 
They might also compare the results with those 
generated by other procedures with which they 
are more familiar. 
