Exploring the Potential of a Novel Biomarker Panel in Selective Dry Cow Therapy

Objectives: The objective of the study was to determine whether a panel of protein biomarkers can enhance the diagnostic accuracy of selective dry cow therapy (DCT) when compared to traditional methods, such as somatic cell count (SCC) and California mastitis test (CMT). Materials and methods: The study recruited two commercial dairy farms, one with 900 and the other with 600 milking cows. Aseptic quarter milk samples were collected from cows at the time of dry-off, based on the results of the California Mastitis Test (CMT) to enable a case-control study design. Each sample was tested for somatic cell count (SCC), standard bacteriology plus matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-ToF MS), and protein concentrations of six potential biomarkers, Haptoglobin (Hp), Lactoalbumin (LA), mammary SAA (MAA), Lactoferrin (LF), C-reactive protein (CRP) and cathelicidin (CL). Standard bacteriology was performed by culturing milk samples on blood agar and examined after 24 hours. Samples that yielded no growth were considered negative for mastitis-associated pathogens, while samples with three or more morphotypes were considered contaminated and excluded from further analysis. Samples that yielded one or two morphotypes were considered indicative of infection and were further examined using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-ToF MS) for species identification. The results from the MALDI-ToF MS were considered definitive for identifying the presence of bacteria. For data analysis, a classification tree model based on results of the six biomarkers was fitted, 10-fold cross-validated and pruned. The performance of this tree model was then compared to that of CMT (>0) and SCC (≥ 200 000 cells/ml) calculating Sensitivity (Se), specificity (Sp) and accuracy in distinguishing between bacteria presence vs bacteria absent samples based on standard bacteriology plus MALDI-ToF MS results (used as gold standard). Results: A total of 237 targeted quarter milk samples were collected from the two farms. After exclusion, 191 quarter milk samples (81%) were used for biomarker analysis, of which 104 (54%) had bacterial growth and 87 (46%) had no growth. The multiple biomarker classification tree included splits on all six biomarkers. However, the cross-validation misclassification error was minimised at a tree with only one split, on cathelicidin (CL): if CL >= 0.85, predict bacterial growth; otherwise, predict no bacterial growth. Of the 72 milk samples with CL >= 0.85, 54 (75%) had bacterial growth. Of the 112 milk samples with CL < 0.85, 69 (61%) had no bacterial growth. The final biomarker classification tree model (including just CL) had Se 56% (95% confidence interval: 46-65%), Sp 79% (70-87%) and accuracy 67% (60-73%), the SCC model had Se 79% (70-86%), Sp 39% (30-50%) and accuracy of 60% (53-67%), and the CMT model had Se 92% (85-96%), Sp 25% (17-35%) and accuracy 60% (53-67%). Conclusions: The goal of selective DCT is to minimize antimicrobial usage by targeting only those cows that have a high probability of infection. Traditionally, DCT selection has been based on indicators such as SCC or past cases of clinical mastitis. However, there have been few attempts to investigate if measuring inflammatory biomarkers in milk can provide a more precise method of identifying infected cows for DCT. The study found that the biomarker model had a higher specificity, which could lead to a reduction in the number of unnecessary treatments given to cows. However, it also had lower sensitivity compared to the other methods, raising concerns about the welfare of cows that may be missed. Further research is needed to determine if the inclusion of additional inflammatory biomarkers, in conjunction with other data such as SCC, could enhance decision-making at the time of dry-off and ultimately reduce antimicrobial usage in DCT