Bindings between antibodies and antigens work according to the lockand- key principle. This is exploited in biochemical tests, such as the immunoassay, to determine even very small substance concentrations. ELISAs (Enzyme- Linked Immunosorbent Assays), for example, may detect the presence of an infection, of hormones or drugs. At PTB, a new statistical method has been developed which makes the analysis of these measurements far more reliable.
The home pregnancy test is the most well-known example of a sandwich ELISA. In general, such tests allow the detection of antigens by ‘sandwiching’ them between two antibodies. One of these antibodies is linked to an enzyme to generate a detectable signal. In the case of fl uorescent sandwich ELISAs, which have recently been investigated, the estimation of the concentration is based on fl uorescence measurements of a solution. To determine the relation between the measured fl uorescence and the concentration, a calibration is performed for each ELISA – i. e. the same protocol steps are repeated with a solution of known concentration.
The relation between the fl uorescence intensity and the concentration can be described by a statistical model. The estimation of this model (i. e. the calibration) and its application for evaluating the unknown concentration is a challenging task, as was recently demonstrated by an international comparability study using the example of an interferon: concentration estimates deviated by a factor of 2 for some laboratories. The reported measurement uncertainties do not even come close to covering these considerable differences.
At PTB, a new statistical method for the analysis of ELISAs has now been developed. This method makes use of the Bayesian approach and coherently combines the calibration of the model and the evaluation of the concentration. Moreover, the Bayesian approach enables the consideration of prior knowledge and the independent analysis of each individual ELISA. In doing so, clearly more trustworthy estimates and measurement uncertainties are provided.
As a case study, the measurements of the above-mentioned international comparability study were reanalysed, yielding corrected concentration estimates and more realistic measurement uncertainties. Contrary to the originally proposed data analysis, the consistency of the experimental methods was largely verifi ed. The new statistical procedure now enables the reliable evaluation of measurements from fl uorescent ELISAs, and thus, the reliable estimation of the concentration of interferon.
Katy Klauenberg
Department 8.4 Mathematical Modelling and Data Analysis
phone: +49 (0) 30 3481-7614
e-mail: 
katy.klauenberg(at)ptb.de
Klauenberg, K.; Ebert, B.; Voigt, J.; Walzel, M.; Noble, J.E.; Knight, A.E.; Elster, C.: Bayesian analysis of an international ELISA comparability study. Clinical Chemistry and Laboratory Medicine. ISSN (Online) 1437-4331, ISSN (Print) 1434-6621