USE OF ON-FARM DATA AND MILK LEUCOCYTE COMPONENTS FOR MONITORING THE HEALTH OF DAIRY COWS

The aim of this PhD project was monitoring cow health and welfare, using on-farm data and leucocytes components in milk. For this reason, the investigation focused on the two main diseases of dairy cows: mastitis and lameness. Critical aspects are linked with these diseases and some of them were considered and discussed in four different experiments. The first experiment focused on the variability of leucocytes in milk, that affects its use as a detection tool for mastitis. For this reason, the factors that influence neutrophils, lymphocytes and macrophages of milk were analyzed. A total of 848 milk samples from 179 dairy cows bred in 6 farms of Northern Italy were collected and analyzed, with Vetscan DC-Q milk Analyzer (Advanced Animal Diagnostic) and DeLaval Cell Counter (DeLaval), to detect Total Leucocyte Count (TLC), Somatic Cell Count (SCC) and the concentration of neutrophils, macrophages and lymphocyte. The results of the Generalized Mixed Model highlighted that the three fractions of leucocytes were significantly influenced by parity, stage of lactation, and farm. The highest values of neutrophils (>60% of TLC) were found at the beginning and at the end of lactation, the two most critical moments for mastitis onset. In accordance, high level of neutrophils (>63% of TLC) was described as one of the risk factors of high SCC (>100,000 cells/ml), highlighting the possibility to apply neutrophils as indicators of udder issues, after accounting for parity, stage of lactation and management conditions. The second experiment focused on the reduction of antimicrobials usage to treat and prevent the insurgence of intramammary infections, investigating one of the main critical periods for antimicrobial overuse: dry-off. This experiment proposed a Selective Dry Cow Therapy (SDCT) protocol based on SCC (thresholds 100,000 cells/ml for primiparous cows and 200,000 cells/ml for multiparous ones) and Differential Somatic Cell Count (threshold of 69.3%) to identify the cows to treat and not treat at dry-off, with the aim to investigate the effects on udder health (SCC value). 33% of 243 dairy cows were selected to not be treated at dry-off. The analysis of milk sample at the beginning of lactation highlighted similar udder conditions, SCC (P=0.5) and TLC (P=0.7) were not different between cows treated with or without antimicrobials. On the other hand, the abandonment of antimicrobial therapy at dry-off, together with poor hygiene conditions and long dry periods (> 55 days), was identified as a risk factor of high SCC (>100,000 cells/ml) at the beginning of the subsequent lactation. The proposed SDCT protocol permits to select cows to treat and not treat with antimicrobials at dry-off, reducing efficiently antimicrobials usage without compromising udder health, even if adequate hygiene conditions and proper management of dry period is requested for the successful of the procedure. The last two experiments were focused on lameness and on-farm data were explored as early indicators of lameness onset. The aim was to produce a starting point for future studies investigating the possibilities of lameness detection without farmers needed to dedicate more work or incur in additional expenses. In the first experiment, milking (e.g. milk production, milking flows) and behavioral (e.g. lying time, number of steps, lying bouts) data were collected for one year, using only on-farm sensors (Afimilk) in one farm of Northern Italy. From the starting dataset, 3459 daily observations of 67 lame cows were extracted. Generalized Additive Mixed Models were used to model the trends of milking and behavioral parameters in the 30 days before and after the detection of lameness, and derivatives were used to identify the changing points of the function. The results showed a change in most of the parameters before the lameness detection, with different magnitude for the three parity groups of cows considered. The reduction of average milk flow and yield in the first two minutes of milking suggested that the milk ejection is disturbed with lameness onset. Milk production and lying time reported changes, decrease and increase respectively, at least 12 days before lameness detection, emphasizing the possibility, in the future, to detect lameness in early stage using only on-farm data. In the second experiment, Locomotion Score (LS) and Body Condition Score (BCS) were manually evaluated by two observers for 7 months in 3 farms of Northern Italy. The milking data were automatically collected for each udder quarter by sensors implemented in automatic milking systems (AMS) (e.g. milk production, milk conductivity, milk flows, milking duration). At first, the cows were divided into three classes of lameness using LS (not lame, mildly lame and severely lame). A Mixed Model was performed to evaluate the least squares means of milking parameters of the lameness classes (42,569 observations), and the contrast between groups showed differences in almost all the milking parameters, especially number of visits to AMS, milk flows and milk production. For this reason, an eXtreme Gradient Boosting machine learning algorithm was trained and tested to classify cows in the three LS classes, using as predictors only BCS and milking on-farm data. Balance accuracy, sensitivity and specificity higher than 0.9 were found, suggesting good predictive performances of the algorithm. Shapley values were used to identify the contribution in classification of the variables, and BCS, parity, days in milk, milk production, average and maximum milking flows were reported as the main contributors. In conclusion, the use of on-farm data and leucocytes components of milk could be a supportive tool for farmers to detect diseases, to reduce antimicrobials usage and to monitor the health status of cows. This is important to guarantee a reduction of the negative consequences of mastitis and lameness on health and welfare of dairy cows, and consequently to preserve productivity of animals, promoting a sustainable dairy farm.