[Interpretation of the cell concentration of cow's milk for the diagnosis of mammary infection].

The potential utilisation of monthly cell counts per cow to determine their infection status was studied for 30 months on 62 cows from an experimental herd, and by simulation, on different herd models with defined epidemiological characteristics. During this period, the cows of the experimental herd were regularly submitted to bacteriological tests for intramammary infections and individual cell counts. The cows which were not chronically infected by a major pathogen were characterised by cell counts of less than 300 000 cells/ml throughout lactation. The chronically infected cows were characterised by at least two peaks over 300 000 with at least one beyond 800 000, but cell counts could be, at other time intervals, below 300 000 cells/ml. Provided that at least two monthly cell counts were available for each cow, it was possible to discriminate strongly the cows not chronically infected and the cows chronically infected by a major pathogen which are reservoirs of infection. The percentage of correct classifications varied between 74 and 87% in relation with the chosen threshold and the infection status of the herd in which this threshold was used. With the same threshold, there were higher percentages of false negative diagnosis in a herd model where staphylococcal like long-term infections were prevalent and of false positive when there was a high incidence of clinical short-term infections. Monthly cell counts per cow did not correctly reveal short-term clinical infections which are particularly prevalent in herds affected mainly by environmental mastitis. The precision and practical importance of this individual diagnosis was discussed for different herd models.