Bayesian estimation of sensitivity and specificity of Coxiella burnetii antibody ELISA tests in bovine blood and milk

Serological tests for Coxiella burnetii (the causative agent of Q fever) antibodies are usually based on enzyme linked immunosorbent assay (ELISA) although this method is not thoroughly evaluated. The objective of this study was to determine the sensitivity and specificity of an ELISA for detection of C. burnetii antibodies in milk and blood samples, using latent class models in a Bayesian analysis. Blood and milk samples of 568 lactating cows from 17 Danish dairy cattle herds collected in 2008 were used.The best combination of sensitivity and specificity estimates was revealed at a sample to positive (S/P) cut-off of 40 for both blood and milk ELISAs. At this cut-off, sensitivity of milk ELISA was 0.86 (95% posterior credibility interval [PCI] [0.76; 0.96]). This was slightly but insignificantly higher than sensitivity of blood ELISA (0.84; 95% PCI [0.75; 0.93]). The specificity estimates of the ELISA methods on milk and blood were equal at 0.99. No conditional dependence was observed between the specificity estimates of the two test methods. However, the sensitivity estimates of both tests were significantly reduced when conditional covariances ≥40 were used. Collection of milk samples from lactating cows is relatively easy, non-invasive and inexpensive and hence milk ELISA may be a better option for screening lactating cows. But, blood ELISA is an option for screening non-lactating cattle.