Several diagnostics test are available for the detection of Mycobacterium avium subsp. paratuberculosis (Map), the ELISA test being the most widely used, because it is simple, fast to do and cheap.The objective of the study was to validate the IDEXX ELISA to detect Map shedding dairy cows in management conditions, prevalence and infection states existing in southern Chile with different statistical approaches. The validation was carried out between September 2003 and July 2004 using 1334 animals from 27 different dairy herds in southern Chile. The herds were selected according to the number of animals per herd, type of management and the expected prevalence of Map in the herd (non infected and infected herds). In order to validate the ELISA test, faecal culture was used as the reference test. We used two-by-two tables, Receiver Operating Characteristics Analysis and Logistic regression as statistical methods. Latent class models such as maximum likelihood methods and Bayesian inference, which do not need a gold-standard test, were also used.The results showed a clear difference in diagnostic sensitivity, specificity and the cutoff values exist among the different methods that were used. The correct selection of the cutoff value should be carried out with a combination of the used methods, taking the ROC analysis as the base and complement it with the logistic regression results. Finally, an important conclusion was that the correct selection of the cutoff values was closely related to the prevalence of Map in the herds. Thus, it is important to have information about the prevalence in a herd in order to make a correct interpretation of the ELISA results.

Sponsorship

Grant research supported by FONDO SAG, Ministry of agriculture, Chile