Combination of milk variables and on-farm data as an improved diagnostic tool for metabolic status evaluation in dairy cattle during the transition period
S Heirbaut,X P Jing,B Stefańska,E Pruszyńska-Oszmałek,B Ampe,C Umstätter,L Vandaele,V Fievez
DOI: https://doi.org/10.3168/jds.2023-23693
Abstract:Milk composition, particularly milk fatty acids, has been extensively studied as an indicator of the metabolic status of dairy cows during early lactation. In addition to milk biomarkers, on-farm sensor data also hold potential in providing insights into the metabolic health status of cows. While numerous studies have explored the collection of a wide range of sensor data from cows, the combination of milk biomarkers and on-farm sensor data remains relatively underexplored. Therefore, this study aims to identify associations between metabolic blood variables, milk variables, and various on-farm sensor data. Second, it seeks to examine the supplementary or substitutive potential of these data sources. Therefore, data from 85 lactations on metabolic status and on-farm data were collected during 3 wk before calving up to 5 wk after calving. Blood samples were taken on d 3, 6, 9, and 21 after calving for determination of β-hydroxybutyrate (BHB), nonesterified fatty acids (NEFA), glucose, insulin-like growth factor-1 (IGF-1), insulin, and fructosamine. Milk samples were taken during the first 3 wk in lactation and analyzed by mid-infrared for fat, protein, lactose, urea, milk fatty acids, and BHB. Walking activity, feed intake, and body condition score (BCS) were monitored throughout the study. Linear mixed effect models were used to study the association between blood variables and (1) milk variables (i.e., milk models); (2) on-farm data (i.e., on-farm models) consisting of activity and dry matter intake analyzed during the dry period ([D]) and lactation ([L]) and BCS only analyzed during the dry period ([D]); and (3) the combination of both. In addition, to assess whether milk variables can clarify unexplained variation from the on-farm model and vice versa, Pearson marginal residuals from the milk and on-farm models were extracted and related to the on-farm and milk variables, respectively. The milk models had higher coefficient of determination (R2) than the on-farm models, except for IGF-1 and fructosamine. The highest marginal R2 values were found for BHB, glucose, and NEFA (0.508, 0.427, and 0.303 vs. 0.468, 0.358, and 0.225 for the milk models and on-farm models, respectively). Combining milk and on-farm data particularly increased R2 values of models assessing blood BHB, glucose, and NEFA concentrations with the fixed effects of the milk and on-farm variables mutually having marginal R2 values of 0.608, 0.566, and 0.327, respectively. Milk C18:1 was confirmed as an important milk variable in all models, but particularly for blood NEFA prediction. On-farm data were considerably more capable of describing the IGF-1 concentration than milk data (marginal R2 of 0.192 vs. 0.086), mainly due to dry matter intake before calving. The BCS [D] was the most important on-farm variable in relation to blood BHB and NEFA and could explain additional variation in blood BHB concentration compared with models solely based on milk variables. This study has shown that on-farm data combined with milk data can provide additional information concerning the metabolic health status of dairy cows. On-farm data are of interest to be further studied in predictive modeling, particularly because early warning predictions using milk data are highly challenging or even missing.