A Bayesian evaluation of six diagnostic test for foot-and-mouth disease for vaccinated and non-vaccinated cattle

The sensitivity and specificity of six ELISA tests for foot-and-mouth disease (FMD) to discriminate between sero-converted (for non-structural FMD virus proteins) and non-sero-converted cattle were evaluated for vaccinated and unvaccinated cattle. Since none of the tests could be considered as a proper reference test and for about half of the tested sera the true status (sero-converted or not for non-structural proteins, i.e. presence of antibodies) of the animals was unknown, a Bayesian analysis employing a latent class model was used that did not rely on the use of a reference test or gold standard. Prior information about prevalence for subsets of the data and specificity of the tests was incorporated into the analysis. The specificity of the six tests for vaccinated and non-vaccinated cattle ranged from 96 to 99%. For vaccinated cattle, one test stood out with an estimated sensitivity of 94% (95% CI from 89.8 to 98.1%). Second best for vaccinated cattle were two tests with estimated sensitivities of 85% (95% CI from 78.9 to 89.7%) and 92% (95% CI from 86.2 to 95.6%). For non-vaccinated cattle, the sensitivities of these three tests were around 97%. The remaining three tests showed lower estimated sensitivity for vaccinated cattle, ranging from 57 to 79%