Predicting visual assessment of allergic patch test reactions by non-invasive measurements

The quantification of allergic patch test reactions to several compounds was evaluated on patients using laser Doppler velocimetry (LDV), transepidermal water loss and colorimetry. The correlation between visual scores and instrumental measurements was investigated by the implementation of a modern statistical method: a classification tree. The LDV showed the highest correlation with the severity of the reaction. The best prediction model was obtained when the data of the three instruments were considered together. Using the classification tree methodology, a model of classification, independent of human observation, was developed. The technique can not only predict an allergic reaction for a new patient, but can also give a measure of how reliable its predictions are. This new method, when entirely standardised and approved, could help in the harmonisation of patch test reactions in different laboratories, thus allowing a more homogenous interpretation of these reactions.