REPCAT was designed to test both plasma and urine samples submitted for query PHEO, and it provides interpretive reports for both sample types.
To obtain statistical data on the utility of REPCAT, the system was subjected to analysis by examining the 4321 urine samples (which included 168 samples from 42 confirmed PHEO patients) collected over the 4 years since EP analysis became routine in our laboratory .
Three data fields were considered critical in examining REPCAT's utility and were those cleared of data in the test datafile before running.
After REPCAT had run and tested each record (a process that took ~40 min for the 4321 records), 384 samples were classified as Q (abnormal--conditionally requiring follow-up) and, of these, 176 were classified as "probable PHEO." The remaining samples were classified either as normal or as having a nonsignificant abnormality.
On comparison of these data with those for the original records, all positive diagnostic PHEO samples had been correctly assigned a "Q" in the PatClas field and each was also rated as "probable PHEO." Thus, no false negatives were generated by REPCAT. This laboratory is part of a major teaching hospital and a tertiary referral center for PHEO diagnosis.
In the group of patients with confirmed PHEO, 20 were assessed by REPCAT to be exclusively NE-secreting, 10 were assessed as exclusively EP-secreting, and the remaining 12 were assessed as "mixed" secretors.
The maximum number of false-positive (possible PHEO) samples classified by REPCAT, calculated by subtracting the number of actual positive PHEO samples from the total number of "Q"s, was 208 (5%), and the maximum number of false positives classified probable PHEO was 86 (2%).
REPCAT has been in full-time use in our laboratory for 5 years and has proven a fast, highly functional tool, providing the performance of an experienced expert in the domain of investigation of PHEO.
In many of these cases, however, the initial interpretive report provided by REPCAT was helpful in speeding the follow-up procedure.
Aside from its accuracy and efficiency as an expert system, we believe that by its incorporation into a database application, REPCAT represents a simple approach to practical expert system development in a relatively narrow domain.