Evaluation of two herd-level diagnostic tests for Streptococcus agalactiae using a latent class approach

Abstract Streptococcus agalactiae mastitis persists as a significant economic problem for the dairy industry in many countries. In Denmark, the annual surveillance programme for this mastitis pathogen initially based only on bacteriological culture of bulk tank milk (BTM) samples, has recently incorporated the use of the real-time PathoProof Mastitis PCR assay with the goal of improving detection of infected herds. The objective of our study was to estimate the herd sensitivity (Se) and specificity (Sp) of both tests of BTM samples using latent class models in a Bayesian analysis while evaluating the effect of herd-level covariates on the Se and Sp of the tests. BTM samples were collected from all 4258 Danish dairy herds in 2009 and screened for the presence of S. agalactiae using both tests. The highest Se of PCR was realized at a cycle threshold (Ct) cut-off value of 40. At this cut-off, the Se of the PCR was significantly higher (95.2; 95% posterior credibility interval [PCI] [88.2; 99.8]) than that of bacteriological culture (68.0; 95% PCI [55.1; 90.0]). However, culture had higher Sp (99.7; 95% PCI [99.3; 100.0]) compared to PCR (98.8; 95% PCI [97.2; 99.9]). The accuracy of the tests was unaffected by the herd-level covariates. We propose that screenings of BTM samples for S. agalactiae be based on the PCR assay with Ct readings of S. agalactiae is low.