The interpretation of serial Johne’s disease milk antibody results is affected by test characteristics, pattern of test results and parallel bovine tuberculosis testing

In the UK, quarterly Johne’s disease milk antibody ELISAs (JD-mELISAs) are commonly used to classify animals which are likely to be infectious, termed “red cows”. “Red cows” are classified following two positive results from the previous four tests (e.g. + - - +). All cattle are also regularly screened for bovine tuberculosis using intradermal avian and bovine tuberculin, and it is advised to maintain a 60 day interval between a tuberculosis test and JD-mELISA. To evaluate the impact of bovine tuberculosis testing on JD-mELISAs, and to quantify the impact of test specificity and “red cow” classification test pattern on the probability of infection. Four years of individual cow milk records with JD-mELISA results were collated from 735 dairy farms and matched to tuberculosis testing records. A two-level multivariable logistic regression model quantified the effect of tuberculosis testing on JD-mELISA result. The specificity and age-dependent sensitivity of a single JD-mELISA were estimated and used to calculate likelihood ratios following each test. Using Bayes’ theorem, the posterior probability of infection with Johne’s disease was calculated for different specificities, ages of cow, and patterns of test results. There were increased odds of a positive JD-mELISA if it was ≤30 days (OR: 2.1) or 31−60 days (OR: 1.2) after a tuberculosis test, compared to >90 days. A larger avian skin reaction at the tuberculosis test was also associated with increased odds of a positive JD-mELISA. The proportion of cows which tested exclusively negative after their first positive JD-mELISA was higher if that JD-mELISA was ≤30 days after a tuberculosis test compared to >90 days. The posterior probability of infection reduced substantially when the test specificity was slightly reduced. In “red cows” classified following two consecutive positive tests, if the test specificity was reduced to 0.95, then the posterior probability of infection was only >95 % if the prior probability was >13 %. If the “red cow” classification was due to two non-consecutive positive tests (+ - - +), the posterior probability of infection was only >95 % if the prior probability was >43 %. Testing for Johne’s disease within 60 days of a tuberculosis test is associated with a higher chance of a positive JD-mELISA and this may reflect a reduction in the ELISA specificity. Relatively small reductions in JD-mELISA specificity can markedly reduce the posterior probability of infection which also depends o