Methodenvergleich zur Messung von Zytokinen im Rahmen großer epidemiologischer Studien

In the course of the prospective multicentre cohortstudy PASTURE/ EFRAIM, factors that influence the development of childhood asthma and allergy were investigated. One part has been the creation of an image of the different T-cell-profiles by measuring cytokines at different time points. The usually used ELISA for cytokine measurement is regarded as the gold standard, but is very time-, money-, personell- and sample-consuming. Therefore, cytokine measurement should be switched to a high-throughput system, the „cytometric bead array assay (CBA)“-system from BD Biosciences. The aim of this paper is to validate the CBA-method, to ínvestigate the quality of measurements in comparison to ELISA and the comparability of both methods. The comparison of methods based on 556 samples of fourty study members which were measured in parallel according to the manufacturer’s manual. The ELISA measurements were performed according to the PASTURE-/ EFRAIM- study protocol. The results concerning sample carryover, calibration stability, analytical and functional sensitivity, imprecision, recovery and linearity show, that CBA generates at least as stable and reliable results as the ELISA and is, in most cases, superiour with regard to sample consumption, cost effectiveness, reproducibility, and usability. With regard to the comparability of identical samples, the measurement results show correlations coefficients (Pearson) between 0,854 (TNF-a) and 0,930 (IL-10) for undiluted samples. The concordance corellation coefficient according to Lin, as well as the graphical illustration according to Bland and Altman shows a fair agreement of measurement results for low concentrations and bad agreement in high concentrations. The present results allow the comparison of the results with certain limitations between 5 pg/ml and 100 pg/ml for IL-5, for IL-10 between 20 pg/ml and 300 pg/ml, for IFN-g and TNF-a between 20 pg/ml and 1000 pg/ml. The reason can most likely be seen in the use of different antibody pairs, the existence of heterophilic antibodys, and dilution effects. A solution to the lack of agreement of the measured values between the two methods can be the division to categories. The membership to a certain category may then be compared between CBA and ELISA.