Discrepancies in carcinoembryonic antigen measurements: survey and control values vs values for patients.

We analyzed the carcinoembryonic antigen (CEA) test results reported in the College of American Pathologists' (CAP) surveys to determine the relationship between the source of CEA used to manufacture the survey specimens and the discrepancies among analytical methods. With the 1983 survey specimens, which were prepared from metastatic colon carcinoma, laboratories using Roche RIA with Clinetics columns reported results that were only one-half the values reported by laboratories using the Abbott polyclonal enzyme immunoassay. With the 1984 specimens, prepared from a different metastatic colon carcinoma, and the 1985 specimens, prepared from a tissue-culture source of CEA, the Roche results were about one-sixth as large as the Abbott results. These differences are larger than the reported assay differences for patients' specimens. In addition, twofold proportional differences were found when survey and control specimens were tested with different lots of Abbott polyclonal reagent, whereas only random differences were found with 102 patients' specimens. Evidently, assay systems perform differently with proficiency-testing and control specimens than with patients' specimens.