Using milk components to estimate the risk of energy imbalance in Holstein cows by means of receiver operating characteristic (ROC) analysis

Negative energy balance (NEB) in dairy cows during early lactation significantly contributes to metabolic and infectious diseases, traditionally diagnosed via costly and time-consuming serum non-esterified fatty acids (NEFA) analysis. This study aimed to develop a practical and cost-effective diagnostic test for NEB based on milk components analysed routinely. Data from 692 Holstein cows (5-35 days in milk) located at five Czech dairy farms were analysed using multiple logistic regression and receiver operating characteristic (ROC) analysis. Results showed that 99 cows (14.3%) were classified as NEB+ (NEFA > 0.6 mmol/l). Cows in the NEB+ group exhibited a significantly higher milk fat content (P < 0.001) and milk fat-to-protein ratio (P < 0.001), and lower lactose concentrations (P < 0.001) compared to NEB- cows. Key indicators of lipomobilisation, such as C18:1, C18:0, and monounsaturated fatty acids (FA), were significantly higher (P < 0.001) in NEB+ cows, while saturated, short-chain, and medium-chain FA were lower (P < 0.001). The developed prediction models, incorporating milk fat and specific FA (e.g. C18:1, C18:0, C14:0), demonstrated high diagnostic efficacy. The area under the ROC curve (AUC) values ranged from 0.84 to 0.92 for individual farms and reached 0.83 for the combined dataset. Using the Index of Union method, optimal cut-off points yielded sensitivities between 0.72 and 0.86, and specificities between 0.72 and 0.85. For the overall model, both sensitivity and specificity were 0.76. In conclusion, the proposed diagnostic test, leveraging milk components, offers a reliable and practical tool for early NEB detection in dairy cows. This facilitates timely intervention, thereby mitigating adverse health and economic impacts. Further validation with larger and more diverse datasets is recommended.