Abstract
Enzyme-linked immunosorbent assays (ELISAs) are routinely used for cattle herd diagnosis of Salmonella Dublin infection in many countries. It is also possible to use such tests for individual diagnosis. Passively transferred immunoglobulins may cause false-positive test results in young calves. Also, false-positive test results may be seen in recovered animals several months after infection. False-negative results are seen in acutely infected animals, especially immature animals that are unable to produce a humoral antibody response to infection. To be able to interpret the individual animal test results, it is necessary to take age into account when validating the ELISA. In the present study an age-stratified validation of an indirect Salmonella Dublin serum ELISA as predictor of bacterial excretion was performed using receiver-operating characteristic (ROC) analysis. Three age groups were formed according to the results of an exploratory analysis of the age effect on area under curve (AUC) of ROC curves. The AUC for the youngest age group (0–99 days) was 0.816 (SE = 0.033), which was significantly (z = 4.23, P < 0.0001) smaller than the 0.977 (SE = 0.019) estimated for the next age group (100–300 days). The oldest age group (>300 days) had an AUC of 0.905 (SE = 0.023), which was significantly different from the AUC of both the other age groups (z = 2.21, P = 0.027 when compared with the youngest age groups and z = 2.41, P = 0.016 when compared with the age group of 100–300 days). The results showed that the indirect Salmonella Dublin serum ELISA is most valid for detection of infection in individual cattle from the age of 100 days. Purpose-related test sensitivities and specificities were evaluated at different cutoff values.