Comparison of pre- and post-vaccination ovine Johne's disease prevalence using a Bayesian approach

This study was conducted to evaluate the effectiveness of Gudair™ vaccine in decreasing the prevalence of shedding of Mycobacterium avium subsp. paratuberculosis (MAP) in flocks of varying initial prevalence. Thirty-seven self-replacing Merino flocks from New South Wales and Victoria (Australia) that had been vaccinating lambs with Gudair™ for at least five years were enrolled in the study. These flocks had been tested prior to or at commencement of vaccination using pooled faecal culture, agar gel immunodiffusion or both tests. These pre-vaccination test results were used to estimate pre-vaccination prevalence. Post-vaccination prevalence was estimated from culture of usually 7 pools of 50 sheep collected from the enrolled flocks in 2008–2009, approximately five or more years after commencement of vaccination.A Bayesian model was developed to estimate and compare the pre- and post-vaccination prevalences for the enrolled flocks. Apparent pre- and post-vaccination prevalences for flocks were modelled as functions of the true pre- and post-vaccination prevalences, respectively, and the sensitivities and specificities of the respective diagnostic tests. Logit-normal models were specified on pre- and post-vaccination true prevalences and were then used to make inferences about the median and 90th percentile of the prevalence distributions and their differences. Priors were mostly specified based on published literature or analysis of abattoir surveillance data for this population of flocks.The analysis found a significant decline in ovine Johne's disease prevalence from a pre-vaccination median prevalence of 2.72% [95% probability interval (PI): 1.40; 6.86%] to a post-vaccination median prevalence of 0.72% (0.39; 1.27%). However 30 of the 37 flocks still contained sheep that were shedding MAP in their faeces. The results suggest that vaccination with Gudair™ is usually effective in reducing the prevalence of faecal shedding but the response to vaccination is variable among flocks. The Bayesian approach reported here could be implemented in similar situations to compare prevalences where information from multiple diagnostic tests with varied sensitivities and specificities is available.