Abstract
Bovine respiratory disease (BRD) results from interactions between pathogens, environmental stressors, and host factors. Obtaining a diagnosis of the causal pathogens is challenging but the use of high-throughput real-time PCR (rtPCR) may help target preventive and therapeutic interventions. The aim of this study was to improve the interpretation of rtPCR results by analysing their associations with clinical observations. The objective was to develop and illustrate a field-data driven statistical method to guide the selection of relevant quantification cycle cut-off values for pathogens associated with BRD for the high-throughput rtPCR system "Fluidigm BioMark HD" based on nasal swabs from calves. We used data from 36 herds enrolled in a Danish field study where 340 calves within pre-determined age-groups were subject to clinical examination and nasal swabs up to four times. The samples were analysed with the rtPCR system. Each of the 1,025 observation units were classified as sick with BRD or healthy, based on clinical scores. The optimal rtPCR results to predict BRD were investigated for Pasteurella multocida, Mycoplasma bovis, Histophilus somni, Mannheimia haemolytica, and Trueperella pyogenes by interpreting scatterplots and results of mixed effects logistic regression models. The clinically relevant rtPCR cut-off suggested for P. multocida and M. bovis was ≤ 21.3. For H. somni it was ≤ 17.4, while no cut-off could be determined for M. haemolytica and T. pyogenes. The demonstrated approach can provide objective support in the choice of clinically relevant cut-offs. However, for robust performance of the regression model sufficient amounts of suitable data are required.