Standardization, specificity, and diagnostic sensitivity of four immunoassays for carcinoembryonic antigen

Four "sandwich"-type immunoassays for carcinoembryonic antigen (CEA) based on monoclonal antibodies (Abbott CEA-RIA Monoclonal, Pharmacia/Wallac Delfia CEA kit, Roche CEA EIA Duomab 60, and our in-house immunoradiometric assay) were compared for 357 samples from colorectal cancer patients and samples from 48 patients with chronic liver disease or with acute, irrelevant diseases. Relative to a common 5 micrograms/L reference limit, all four assays agreed regarding classification in 92% to 94% of the samples in the colorectal cancer samples, with the Abbott and Roche assays giving slightly more "normal" values. Cross-testing of CEA standards and the 1st International Reference Preparation 73/601 showed that calibration differences could not be eliminated by the use of a common standard. It has been earlier demonstrated that the monoclonal antibodies in the Roche assay have an epitope group specificity slightly different from those of the other three assays. This can explain why the correlations involving the Roche assay were weaker (r = 0.84-0.87) than those obtained with the other assays (r = 0.94-0.97). I conclude that agreement between the assays studied is as good as can be expected when different antibodies are used; nevertheless, consistent discrepancies between assays in several patients still necessitate the use of the same assay during follow-up.