Estimation of the diagnostic performance of two ELISAs to detect PCV2 antibodies in pig sera using a Bayesian method

This study was designed to assess the diagnostic characteristics of two PCV2 ELISAs without a gold standard. Four hundred and sixty-five serum samples from finishing pigs (25 herds) not vaccinated against PCV2 were used. Samples were tested by two ELISAs: an in-house ELISA (I-ELISA) and the commercial SERELISA^®PCV2 Ab Mono Blocking kit (S-ELISA). A ROC curve was used to assess the S-ELISA's optimal threshold by taking the I-ELISA as a reference and using the cut-off previously determined by comparison to an cccmonolayer assay (IPMA). This led to an S-ELISA result ≥170 being considered as positive. The sensitivity (Se) and specificity (Sp) of each ELISA were then estimated without a gold standard using a Bayesian approach. The mean Se and Sp values of the I-ELISA were slightly higher than those of the S-ELISA (mean Se I-ELISA=0.90 vs. mean Se S-ELISA=0.86; mean Sp I-ELISA=0.92 vs. mean Sp S-ELISA=0.85). However, the 95% credibility intervals (CI95%) overlapped (Se I-ELISA CI95%=0.85-0.95 vs. Se S-ELISA CI95%=0.82-0.90; Sp I-ELISA CI95%=0.82-0.98 vs. Sp S-ELISA CI95%=0.75-0.94). Both ELISAs appeared to be valuable tools for detecting PCV2 antibodies.