Evaluation of the measurement uncertainty in screening immunoassays in blood establishments: computation of diagnostic accuracy models.

The European Union regulation for blood establishments does not require the evaluation of measurement uncertainty in virology screening tests, which is required by ISO 15189 guideline following GUM principles. GUM modular approaches have been discussed by medical laboratory researchers but no consensus has been achieved regarding practical application. Meanwhile, the application of empirical approaches fulfilling GUM principles has gained support. Blood establishments' screening tests accredited by ISO 15189 need to select an appropriate model even GUM models are intended uniquely for quantitative examination procedures. Alternative (to GUM) models focused on probability have been proposed in medical laboratories' diagnostic tests. This article reviews, discusses and proposes models for diagnostic accuracy in blood establishments' screening tests. The output of these models is an alternative to VIM's measurement uncertainty concept. Example applications are provided for an anti-HCV test where calculations were performed using a commercial spreadsheet. The results show that these models satisfy ISO 15189 principles and that the estimation of clinical sensitivity, clinical specificity, binary results agreement and area under the ROC curve are alternatives to the measurement uncertainty concept.