Estimation of the Diagnostic Accuracy of a Multiplix Real-Time PGR Assay and Bacteriological Culture for Four Bovine Intramammary Pathogens

摘要

Bacteriological culture is an intramammary infection (IMI) diagnosis tool in widespread use, but it may lack diagnostic sensitivity and there is a significant time delay for diagnostic results. There is a growing interest in employing multiplex real-time polymerase chain reaction (mr-PCR) assays, which enable simultaneous testing for the presence of several different bacterial species in milk samples. However, routinely using mr-PCR technology in the field before assays' error rates and pitfalls arewell understood may lead to erroneous farm-level decision making. Directly estimating sensitivity (Se) and specificity (Sp) values for a test assay presumes perfect knowledge of the true results of all samples tested; bacteriological culture is unsuitable as a gold-standard. Estimates of Se or Sp of mr-PCR assays evaluated against bacteriological culture of field samples or against bacteria-spiked milk samples are unsuitable for informing field use of the mr-PCR assay. Bayesian modelling is pertinent asit can use the results of multiple tests applied in parallel to the same samples in order to estimate simultaneous distributions of credible error rates for all tests. The aim of this research was to estimate, through Bayesian modelling, the bacterial species-specific Se and Sp values of both bacteriological culture and a mr-PCR test as applied to mammary quarter milk samples from clinical mastitis cases and to samples from apparently normal mammary quarters.