Within-pool prevalence limits for the identification of paratuberculosis infected herds using antibody detection in pooled milk samples

Identification of Mycobacterium avium subsp. paratuberculosis (MAP) infected herds is the first step in paratuberculosis control. Bulk milk or pooled milk sampling is a widely used, cost-effective approach for the surveillance of infectious diseases in dairy cattle. Due to insufficient data about the diagnostic performance, it has rarely been exploited for paratuberculosis. Using a logistic regression model, threshold levels of the apparent within-pool prevalence (WPPapp) were determined for the identification of paratuberculosis positive herds by antibody detection in pools of 50 individual milk samples, which is a usual pool size. A total of 73 milk pools were prepared, 50 from two MAP-positive and 23 from one MAP-non-suspect herds. Four different commercial ELISAs were used. WPPapp was estimated independently based on antibody positive individual milk or serum samples and on positive faecal culture. Antibody detection resulted in lower WPPapp estimates than faecal culture (FWPPapp) and tended to underestimate true prevalence. For pooled milk testing, cut-off values of the four ELISAs were revised to ensure 99% specificity and high sensitivity. For 50% probability of detection, FWPPapp thresholds of 8.9–16.3% were determined, increasing to 20.0–37.8% for 95% probability of detection. The results underline that antibody detection in pools of 50 individual milk samples or bulk tank milk from herds ≤ 50 cows allows only the identification of herds with a very high prevalence of MAP-shedders. In control programs this can be the first step to identify the most affected herds. However, it is ineffective for prevalence investigations, surveillance programs and certification.