Comparison of fifteen immunoassays for the measurement of serum MUC-1/CA 15-3 in breast cancer patients.

BACKGROUND Quality control results for serum MUC-1/CA 15-3 assays have always shown large discrepancies. METHODS This multicentre study of 15 methods (labelled M1-M15) measured coded sera from 35 patients with breast cancer without recurrence (group 1), 46 patients at 1st metastasis (group 2), and 39 patients with advanced metastases (group 3). Results were compared using parametric statistics, ANOVA, principal component analysis, and receiver operating characteristic (ROC) curves. RESULTS Mean MUC-1/CA 15-3 concentrations varied widely (75.1-303.0 U/mL, 24.8%) among methods. The false positive (FP) rate for group 1 was 8/521 (1.5%); for group 2 and group 3 false negative (FN) results were 21/680 (3.1%) and 11/583 (1.9%), respectively. Using the ROC cut-offs, we found no FPs for group 1 and no FNs for group 3. However, group 2 showed 16 FNs. All p-values for Pearson's correlation were <0.0001 between methods, except for M11. When comparing methods using different antibodies, discordance rates reached a maximum of 15.2%. Principal component analysis revealed a grouping of methods using: CanAg monoclonal antibodies (mAbs) (M2, M7 and M12); Centocor/Fujirebio mAbs (M3-M6, M8-M10, M14-M15) and Biomira mAbs (M1 and M13); and Centocor/Fujirebio mAbs (M11). CONCLUSIONS Results were more consistent among methods using the same antibody type. Principal component analysis showed that antibody type was the strongest determinant of immunoassay results.