Objective

Latent-class models were used to determine the sensitivity (Se), specificity (Sp) and predictive values (PV) of a faecal culture (FC) method and a serum ELISA for paratuberculosis in dairy sheep and goats.

Materials and methods

One hundred sera and faecal samples were collected from each of four dairy sheep and goat flocks with a history of clinical paratuberculosis. Faecal samples were cultured in Herrold's Egg Yolk medium supplemented with mycobactin J and antibiotics. Sera were tested using a commercial serum ELISA. The ELISA results were interpreted at the recommended cut-off and values reduced by 25 and 50%. Assuming independence, Bayesian estimates and posterior 95% credible intervals (PCI's) of the Se's and Sp's were obtained. Estimates of P's and the PV's and 95% PCI's were also estimated for each sub-sample of the data.

Results

In sheep the Se and Sp of the FC were 19% (5%-41%) and 97% (94%-99%), respectively, while in goats they were 11% (4%-21%) and 98% (95%-99%), respectively. The Se and Sp of the serum ELISA in sheep and goats were 39% (17%-67%), 97% (94%-99%), 64% (46%-83%) and 94% (89%-97%) respectively, at the recommended cut-off. In either species, lowering of the cut-off improved Se and positive PV's of serology with minimal loss in Sp's and negative PV's.

Conclusions

The Se of serology differs between sheep and goats. Species-specific cut-off selection in small ruminants is recommended. The performance of the ELISA could be improved by lowering of the cut-off. The sensitivity of the particular FC was quite low; in combination with its laborious and time-consuming process it cannot be suggested as a routine diagnostic procedure.