Prediction of mastitis using milk somatic cell count, N-acetyl-beta-D-glucosaminidase, and lactose.

The objectives of this work were 1) to examine the responsiveness of SCC, lactose concentration, and NAGase activity in milk to changes in bacteriological status and 2) to develop models for predicting bacteriological status of mammary glands. Data included 550 cows in 10 commercial herds. Natural logarithm NAGase and log cell count were most responsive to changes in bacterial status. The log NAGase was relatively more effective in identifying major from minor pathogen infections, whereas log SCC was better able to differentiate between infected and uninfected classes. Non-transformed NAGase, SCC, and lactose were considerably less responsive to infection status. Logistic regression of bacterial status on herd, lactation number, milk, log SCC, log NAGase, and stage of lactation was performed. The least significant variables were removed in a stepwise process. Final predictors of infection status were herd, log SCC, and log NAGase. The role of log SCC was to discriminate infection from no infection, whereas log NAGase discriminated major from minor pathogens. The log NAGase, alone or in combination with log SCC, added substantially to the detection power of the model. Chi-square goodness of fit tests found no significant differences between observed and predicted infection probabilities. Substitution of herd averages for log SCC and log NAGase for the herd variables resulted in significant differences between predicted and observed herd infection probabilities.