On-farm strategies for the prevention and detection of Gram-specific clinical mastitis in dairy cows
MetadataShow full item record
Controlling mastitis in dairy herds relies on good prevention and detection methods. This dissertation describes two areas of research relating to mastitis control. In the first objective, the efficacy of 2 vaccines against Escherichia coli mastitis in mid-lactation dairy cows was evaluated. Secondly, in a series of 3 studies, milk and activity sensor data were used to derive models for clinical mastitis (CM) detection, and models were tested for their ability to indicate the causative pathogen type. Primiparous and multiparous animals were vaccinated with 1 of 2 commercially available J5 vaccines (V1 or V2) or served as unvaccinated controls (CTL). Intramammary challenge with E. coli approximately 84 d later resulted in few treatment differences in the clinical and behavioral responses, except that vaccinated cows exhibited fever (≥ 39.4 °C) 3 h earlier and laid down for longer periods than CTL. Although vaccinated cows had similar severity and duration of CM, V1 cows produced more serum IgG1 and IgG2 than V2 cows. Our results indicated that the effects of vaccination were diminished in mid-lactation, and that antibodies are not the limiting factor in defending against induced E. coli mastitis. Multiple regression models, incorporating the slope changes in relevant milk and activity sensor data, were developed to indicate all CM cases (ACM), or specifically, CM due to Gram-negative (GN) or Gram-positive (GP) bacteria. Gram-specific models had greater detection accuracy (> 80%) than the ACM model (75%) when evaluated using the model training dataset, but independent evaluation demonstrated reduced sensitivity (Se) of detecting CM by all models (GN, 62%, ACM, 56%, and GP, 32% Se). Data in the 3 d prior to CM were more important in detecting GN pathogens, whereas the best GP models incorporated changes more than 1 week prior to CM detection. Still, model performance was imperfect. Next, models were rederived from a dataset that better reflected the infection distribution of the herds its use was intended for. However, the Se of detecting CM in real-time, across 2 farms, was < 21% for all models, and categorization by Gram-status had no benefit. An insufficient number of CM cases was considered to contribute to the poor detection performance of models and limited repeatability across farms. Consequently, models derived in this study were inadequate for implementation as mastitis detection tools. In the future, development of new sensors and application of more sophisticated algorithms to the field of mastitis detection may improve the accuracy of models using sensor data.
General Audience Abstract
Mastitis is an important disease of dairy cattle that adversely affects animal welfare, productivity, and milk quality. Controlling mastitis in dairy herds relies on good prevention and detection methods. In this dissertation, we investigated two elements of mastitis control: 1) the effects of vaccination in protecting against mastitis, and 2) the ability of on-farm sensor data to detect clinical mastitis (CM) and indicate the causative pathogen type. Coliform bacteria commonly cause CM, and vaccination against these bacteria can reduce the severity of the disease. We evaluated the effect of 2 different vaccines on the clinical, behavioral, and immune response in cows with experimental mastitis caused by Escherichia coli. Our findings indicated that the effects of vaccination had diminished at the time of experimental mastitis, as vaccinated cows had no improvement in clinical recovery compared with unvaccinated controls. Although no clinical or behavioral differences were observed between the 2 different vaccines, the antibody response differed, suggesting that antibodies are not the key player underpinning the mechanisms of vaccination against induced coliform mastitis in mid-lactation. Rapid detection and diagnosis of mastitis is important to reduce effects on the cow, and to support decision making for the appropriate intervention. We aimed to develop and test mastitis detection models that utilized data collected by on-farm sensor technologies. Milk and activity parameters, which may be differentially affected by mastitis depending on the pathogen causing infection, were used in multiple regression models for detecting any CM case, or specifically CM caused by Gram-positive or Gram-negative bacteria. Gram-specific models were initially estimated to have > 80% accuracy in classifying cows with and without mastitis, but further validation demonstrated that the models were not repeatable when tested independently. Subsequently, models that were more suited to the farms they were to be implemented on were developed, and tested, revealing limited performance in detecting any case of CM, or CM due to the Gram-specific pathogens. Model derivation was limited by an insufficient number of CM cases to represent the variation in different cases of CM within the Gram-positive and Gram-positive classifications. Although our models did not show promise as a mastitis detection tool, milk and activity data may be incorporated with other sensor data for improved detection and diagnosis of mastitis.
- Doctoral Dissertations