In an inter-laboratory study, we compared six routinely employed immunoassay sets: Architect i2000 and AxSYM, Abbott Laboratories; Elecsys 2010, Roche Diagnostics; ELSA, CIS-BioInternational; Immulite 1, Dia- gnostic Products Corporation; and IRMA-mat, Byk-Sangtec Diagnostica. Using all analytical systems, we measu- red identical groups of clinical samples completed with selected calibrators and control samples. The reproducibility of measurements was calculated from the results of measurements of 21 duplicates of control materials. The coefficient of variance (CV) values ranged from 2.1% (Elecsys 2010) to 6.7% (ELSA). The parameters of Passing-Bablok regression show significant systematic differences among analytical systems. Data from a Bland-Altman diagram suggest that these differences project onto another, still more significant are individual differences among individual samples. Usually, differences among systems increase with increasing CA 19-9 concentrations. Our aim was to find out whether analytical discrepancies would decrease when we related the results obtained to the cut-off values for individual analytical sets. Though the cut-off values differ among various systems, no similar clinical efficacy appears to be attained. The behaviour of indi vidual systems is quite different for identical control materials and does not necessarily duplicate the calibration for biological samples. We are presenting the results of analyses of control sera Bio-Rad (difference 72 %) and DGKC (difference 89%). The results of determining CA 19-9 cannot be generalized from one analytical technique to another, which is even true in cases where the same monoclonal antibody is used.

        Key words: CA 19-9 immunoassay, intra-assay accuracy, CA 19-9 cut-off value, reliability of results, tumour marker.