Five commercially available ELISA tests for the detection of antibodies against Mycobacterium avium subsp. paratuberculosis in bovine serum were evaluated at the individual animal level using sera from 286 paratuberculosis-free and 110 paratuberculosis-infected dairy cattle. Sensitivity (Se) and specificity (Sp) of the tests were estimated after determination of the cut-off dtheta by TG-ROC analysis or using the cut-off values recommended by the manufacturers, respectively. When the dtheta cut-off values were applied, the five ELISA tests showed sub-optimal Se and Sp. Adopting the cut-offs recommended by the manufacturers, the Sp of four of the five ELISA increased, two tests reaching Sp > or = 99.0%. Test sensitivity clearly depended on the disease state of the animals examined. Se was significantly higher in clinically diseased than in latently infected dairy cattle. Calculation of the positive and negative predictive values indicated that, depending on the test, a considerable proportion of false positive and false negative results have to be expected. Therefore, the suitability of antibody detection for the diagnosis of paratuberculosis in individual animals is questioned.