Bayesian accuracy estimates of environmental sampling for determining herd paratuberculosis infection status and its association with the within-herd individual fecal culture prevalence in Québec dairies

Abstract

The objectives of this retrospective analysis were to: 1) estimate the diagnostic sensitivity (Se) and specificity (Sp) of bacterial culture of environmental samples for determining Mycobacterium avium subsp. paratuberculosis (MAP) infection status in Québec dairies, using a Bayesian Latent Class Model (BLCM); and 2) determine the association between the number of positive environmental samples and the individual fecal culture (IFC) apparent and true MAP within-herd prevalence.Environmental and individual fecal samples were collected from 87 commercial dairy herds participating in previous research projects. Environmental samples included two composite samples of 20 g collected from different locations within each of the following sites: an area where manure from the majority of adult cattle accumulates, a manure storage area and another site of manure accumulation chosen by the veterinarian. Samples were cultured using the MGIT Para TB culture liquid media and the BACTEC MGIT 960 system. The Se and Sp of environmental sampling were estimated using a one-test-one-population BLCM. Herds were considered positive for environmental sampling if at least one out of the six samples collected was positive. The apparent and true IFC within-herd MAP prevalence estimates for each herd were obtained using a two-stage cluster BLCM, then merged in a single dataset with the environmental sample results. The association between the within-herd MAP prevalence results (apparent and true), and the number of positive environmental samples was assessed using a zero-inflated negative binomial (ZINB) model. In all BLCMs, median posterior estimates and 95 % Bayesian credible intervals (BCI) were obtained with OpenBUGS statistical freeware.Se and Sp of environmental sampling were 43.7 % (95 % BCI: 32.5−55.5) and 96.2 % (95 % BCI: 84.2−99.8), respectively. Overall, the number of positive environmental samples increased with the apparent and true MAP within-herd prevalence. The true prevalence was higher than the apparent prevalence for a given number of positive environmental samples. The probability of not observing a positive environmental sample decreased with the prevalence.Despite its imperfect accuracy, environmental sampling is an inexpensive and non-invasive sampling method to determine MAP infection status in tie-stall herds that can be used as a proxy to estimate the true within-herd prevalence. The absence of positive environmental samples in a single sampling visit is likely an indicator of a very low within-herd prevalence rather than being MAP exempt.