An example of expert systems applied to clinical trials: analysis of serial graded exercise ECG test data.

ABSTRACT

Clinical trials collect large amounts of data over time. The use of statistical methods to compare and interpret these serial data often fall short of complete evaluation because the analysis requires clinical judgment. As an alternative, some trials use individual experts or panels of experts to evaluate data, but this method usually requires the participation of clinicians who must spend large amounts of time performing tedious, repetitive tasks. The authors examined the use of expert systems to analyze serial clinical trial data where the analyses required use of clinical judgment. A prototype expert system was built to assess the data obtained from a pair of serial graded exercise ECG tests and reach a decision that would duplicate the decision reached by a cardiologist. The experiment was successful. Expert systems should be further developed and tested in other areas, such as serial coronary arteriography data.