QUALITY CRITERIA THAT SUPPORT COMPUTER-ASSISTED RETRIEVAL
AND REVIEW OF SCIENTIFIC DATA
Robert S DeWoskin
Research Triangle Institute, PO Box 12194, Research Triangle Park, NC, 27709-2194, USA
Three basic steps in a quality assurance program are: 1) developing quality criteria that define excellence,
2) evaluating quality by comparing the study to the standards defined in the quality criteria, and
3) implementing quality controls that either reject or remedy studies that fail to meet the standard. Of these 3,
developing quality criteria is the most challenging because quality criteria differ depending on how the study results are used.
Quality evaluations based on different criteria produce different results. In the Regulatory Toxicology Program at Research Triangle Park,
we summarize large amounts of published data on health effects to support risk assessments. The quality of the data incorporated into the
risk assessment is of utmost importance. The first quality criteria that we use is that the study be published in a peer-reviewed journal.
Other quality criteria we like include the US National Research Council criteria on what constitutes a good toxicology or epidemiology study,
the Good Laboratory Practice standards for non-clinical studies, and the Good Clinical Practice standards for clinical studies. Rarely, however,
do authors report compliance with these established criteria, and although the peer-review process focuses on scientific merit, the criteria used
by the peer reviewers is for the most part totally undocumented. This makes it extremely difficult and time consuming for those who review and summarize
the ever increasing amount of published data to evaluate the quality of the data either for scientific merit or for data integrity. This presentation makes
a case for the development of published quality criteria for data integrity (and at least for some level of scientific merit) with associated keywords that
authors could cite in their papers and that would also appear in the abstracted information of electronic databases for published literature. This would facilitate
automated review and retrieval of quality data. Nothing precludes the development of different criteria sets for different end uses, and some examples are given.
Return to Session Information
