Testing cattle with clinical signs of paratuberculosis is an important element of surveillance for paratuberculosis. In many herds, control of paratuberculosis infection is only initiated after detecting clinical paratuberculosis cases. Therefore, the aim of this study was to evaluate the diagnostic-test characteristics of microscopic examination of Ziehl-Neelsen-stained faecal smears for acid-fast Mycobacteria (ZN-test) and serum-ELISA in cattle suspected of clinical paratuberculosis.

Results of all samples submitted for ZN-test and serum-ELISA between April 2003 and April 2006 to our laboratory were retrieved. Results of cattle for which both tests were performed were analysed using three Bayesian models for evaluation of diagnostic tests in two populations without a gold standard, assuming conditional independence of tests (model 1), conditional dependence of tests in both infected and non-infected cattle (model 2) and conditional dependence of tests in infected cattle only (model 3). Sampled cattle were divided into two populations in two different ways using known risk factors for clinical paratuberculosis: region and age. Priors for sensitivity and specificity of tests were based on the literature; uninformative priors were used for prevalence's in the various populations.

For 892 cattle suspected of clinical paratuberculosis, both ZN-test and ELISA results were retrieved: 250 ZN-positive and ELISA-positive, 12 ZN-positive and ELISA-negative, 260 ZN-negative and ELISA-positive, and 370 ZN-negative and ELISA-negative cattle.

Posterior estimates of sensitivity, specificity, and positive and negative predictive values of the ELISA were always higher than those of the ZN-test, irrespective of the population and choice of model. Lower limits of the 95% credibility intervals of the posterior positive predictive values of the ELISA were always ≥99.7%, and of negative predictive values of the ELISA ≥57.0%. Upper limits of the of the 95% credibility intervals of the posterior positive predictive values of the ZN-test in the various models were always ≤99.4%, and of negative predictive values of the ZN-test were always ≤59.8%.

It is concluded that the ELISA is preferred to the ZN-test to confirm the presumptive diagnosis of clinical paratuberculosis. Little diagnostic information can be gained by performing the ZN-test in addition to the ELISA.