A probabilistic approach to the interpretation of milk antibody results for diagnosis of Johne's disease in dairy cattle

Johne's disease is a serious wasting disease of ruminants that is of high economic importance for the dairy sector in particular. The chronic nature of the disease, the fluctuations in antibody levels and the limited ability of diagnostic tests to identify cows at early stages of infection are huge challenges for the control of the disease. In the United Kingdom, the latter is commonly based on repeated milk ELISA testing of lactating cows, followed by selected culling and improved management practices around calving. In this paper, the dataset built through a large quarterly screening programme conducted in the United Kingdom since 2010 is used to investigate the use of milk ELISA testing for Johne's disease management. Over the study period, 13,509 out of 281,558 cows were identified as high-risk of being infected and shedding mycobacteria in the faeces, based on a case definition of at least two consecutive positive milk ELISA results. Around a third of them were kept in the dairy herd a year or more after being classified as high-risk. However, 16% of these cows did not have any further positive test, suggesting that they might be uninfected animals. The mean specificity and sensitivity of the milk ELISA test were estimated at 99.5% and 61.8%, respectively. The cows in the dataset are categorised in different result groups according to the number of positive test results and whether they are classified as high-risk according to the programme's case definition. The posterior probability of infection is calculated after each test in order to investigate the impact of repeated testing on the belief in a cow's infection status. The interpretation of the results show that most cows classified as high-risk are very likely to be infected, while some other groups that do not match the case definition could reasonably be considered as infected too. Our results show that there is considerable potential for more targeted use of serological testing, including adjusting the testing frequency and implementing the posterior probability approach.