-Our objective was to estimate the sensitivity and specificity of the agar-gelimmunodiffusion test (AGID), the ELISA, and the skin test for the detection of Mycobacterium avium subspecies paratuberculosis (MAP) in sheep using Bayesian methods without a gold standard. Fourteen flocks (2 465 sheep) were used. Five flocks (450 sheep) were considered MAP non-infected and 9 flocks (2 015 sheep) had sheep infected with MAP. Sheep were skin tested and blood was collected for AGID and ELISA testing. Results were analyzed using a Bayesian 3-test in 1-population model fitted in WinBUGS. The model allowed for dependence (correlation) between the two serologic tests, but these two tests were assumed to be conditionally independent of the skin test. The estimated specificity was 99.5% (95% PI of 98.9-99.9%) for the AGID; 99.3% (98.4-99.8%) for the ELISA using an optical density measured cutoff of 0.20; 99.2% (98.1-99.8%) using a cutoff of 0.15; 97.5% (95.8-98.7%) using a cutoff of 0.10; and 98.7% (97.3-99.5%) for the skin test. The estimated sensitivities were 8.3% (6.2-10.7%) for the AGID; 8.0% (6.0-10.4%), 10.6% (8.3-13.1%), and 16.3% (13.5-19.4%) for the ELISA using the cutoffs 0.20, 0.15, and 0.10 respectively; and 73.3% (62.3-85.8%) for the skin test. The skin test was specific in non-infected populations and sensitive in infected populations, although in some cases a positive skin test might represent MAP exposure rather than infection. The AGID and ELISA were specific but lacked sensitivity. The AGID and ELISA consistently identified two different populations of infected sheep with only moderate overlap between positive test results.

No abstract