Evaluation of a commercial ELISA for H5 low pathogenic avian influenza virus antibody detection in duck sera using Bayesian methods

Abstract

Following the emergence of highly pathogenic avian influenza (AI), active surveillance of infections due to the H5 and H7 subtypes in poultry has increased and been made compulsory in Europe since 2002, by means of annual serological surveys using the haemagglutination inhibition (HI) test. Domestic anseriforms, particularly ducks and geese, are more frequently infected by H5 low pathogenic AI virus, often subclinically, and represent a threat for other terrestrial poultry. 1783 sera, mainly from ducks, have been used to evaluate and compare a commercial ELISA kit detecting H5 antibodies with the currently recommended HI test. Different approaches to calculating specificity and sensitivity have been used, including the original Bayesian method. Results were similar when data were analyzed at the individual and batch levels, and when using different methods of calculation. However, results showed that H5 ELISA had both a higher sensitivity and a lower specificity than the HI test. Given that sensitivity is the most important factor for a screening test, H5 ELISA could therefore be recommended for AI surveillance, followed in cases of positivity by molecular tests aimed at detecting the virus gene.

Keywords: Bayesian methods; Duck; ELISA; H5 antibodies; HI test; Influenza.