Comparison of four 'second generation' immunoassay systems to determine CA 125 in serum by using a graphical approach
N P Koper1, C M Thomas, L F Massuger
1Department of Obstetrics and Gynaecology, University Hospital Nijmegen, The Netherlands.
Abstract:
Clinical management of ovarian cancer patients is facilitated by CA 125 determinations in serum. Presently, several assay systems based on different concepts and different methodologies are available to measure CA 125. Method comparison analysis of such assay systems is usually performed through (linear) regression analysis, which requires assumptions about the distribution of experimental data and its measurement error. The aim of the present study was to compare four newly developed second generation assay systems for quantitation of CA 125 by utilizing an alternative simple approach to method comparison analysis. This alternative comprises the construction of relative difference plots and mountain plots, previously described by Krouwer et al. (Eur J Clin Chem Clin Biochem 1995; 33:525-7). In addition, the diagnostic value of the assays was illustrated through receiver-operating-characteristic (ROC) curves. Sera obtained from 300 women were assayed for CA 125 using the Abbott IMx CA 125 assay (Abbott), the Centocor CA 125 II RIA assay (Centocor), the Berilux Ov testing kit for CA 125 (Behringwerke), and the CA 125 TR-FIA assay (Wallac Oy). Both the relative difference plots and the mountain plots revealed higher serum concentrations with the Centocor RIA II (Median +33%, P2.5 to P97.5: -25% to 161%) and Berilux (Median +28%, P2.5 to P97.5: -17% to 108%) compared to the Abbott IMx system. The TR-FIA assay system showed lower serum concentrations (Median - 17%, P2.5 to P97.5: -74% to 229%). The combination of relative difference plots and mountain plots demonstrated clearly the wide range of differences between CA 125 assays measuring the same analyte. The relative difference plots provided insight into the distribution of the differences over the range of measurement as well as the identification of outliers. A simple quantitative assessment of the median differences could be made from the overlaying mountain plots. The close correspondence observed between the ROC curves illustrated that assay systems for CA 125 differing in design (type of antibodies used) and format can produce similar results on group level. However, the results of the clinical evaluation underline the importance of the application of assay specific cut-off values.
