Opportunities and limitations of milk mid-infrared spectra-based estimation of acetone and β-hydroxybutyrate for the prediction of metabolic stress and ketosis in dairy cows

Association between hyperketolactia and production in early-lactating dairy cows

Study aims were to investigate associations of hyperketolactia (HYKL) status of Holstein dairy cows between 6 and 60 d in milk (DIM), defined by milk acetone (mACE) and β-hydroxybutyrate (mBHB) content, with daily milk yield and composition. Milk samples (∼5.0 million) were collected over a 5-yr period (2014-2019) within the milk recording system in Poland. Concentrations of mACE and mBHB determined by Fourier-transform infrared spectroscopy were used to categorize samples into 4 ketolactia groups. Based on threshold values of ≥0.15 mmol/L mACE and ≥0.10 mmol/L mBHB, ketolactia groups were normoketolactia (NKL; mACE <0.15 mmol/L and mBHB <0.10 mmol/L), BHB hyperketolactia (HYKLBHB; mACE <0.15 mmol/L and mBHB ≥0.10 mmol/L), ACE hyperketolactia (HYKLACE; mACE ≥0.15 mmol/L and mBHB <0.10 mmol/L), and ACE and BHB hyperketolactia (HYKLACEBHB; mACE ≥0.15 mmol/L and mBHB ≥0.10 mmol/L). To investigate ketolactia association with production outcomes, a linear model was developed, including ketolactia group, DIM, parity, their interactions, year-season as fixed effects, and random effects of herd and cow. Among all milk samples, 31.2% were classified as HYKL, and of these, 52.6%, 39.6%, and 7.8% were HYKLACEBHB, HYKLBHB, and HYKLACE, respectively. Ketolactia groups differed for all traits studied in all parities and DIM. Among HYKL groups, lowest milk yield was found in HYKLACEBHB cows, except for 6 to 30 DIM in first- and second-lactation cows. Milk yield of HYKLBHB cows was higher than that of NKL cows until 20 to 30 DIM, and then it was lower than NKL cows. Milk yield of HYKLACE cows was mostly lower than NKL cows. Energy-corrected milk (ECM) yield of HYKLACEBHB cows was higher than that of NKL cows until 30 to 35 DIM for second lactation and third lactation or greater, and in the whole study period for first lactation. The yield of ECM for HYKLBHB cows was mostly higher than that of NKL cows, whereas HYKLACE cows had higher ECM than NKL cows until 15 to 25 DIM and then was lower for the HYKLACE group. Milk composition differed among HYKL groups. Highest milk fat (MF) and lowest milk lactose (ML) contents were observed in HYKLACEBHB cows. Cows in HYKLACEBHB and HYKLBHB groups had higher MF and lower milk protein (MP; except in 6-8 DIM in first lactation) and ML content than NKL cows. Milk fat content was higher in HYKLACE than NKL cows in first lactation and during the first 30 to 40 DIM in older cows. Lactose content was lower in HYKLACE than in NKL cows within 30 to 40 DIM; afterward it was higher in NKL cows. Lower MP content was found in HYKLACE than in NKL cows, except during 6 to 9 DIM for cows in first lactation and third lactation or greater. In conclusion, HYKL is associated with altered milk production in all parities, but a range of these negative relations depends on ketone status addressing both ACE and BHB contents. Further research is needed to ascertain underpinning biochemical defects of HYKL from elevated ACE, alone or in combination with BHB, during early lactation.

Association of herd hyperketolactia prevalence with transition management practices and herd productivity on Canadian dairy farms-A retrospective cross-sectional study

The objective of this observational study was to assess the relationship between herd-level prevalence of hyperketolactia (HPH) with management practices of the transition period and herd milk production. Dairy herds (n = 71) were selected based on their inclusion in a herd management risk assessment study (August 2014-March 2018) using a Vital 90 (Elanco) Risk Assessment tool (one assessment per farm). Data from multiple milk recording test-days (Dairy Herd Improvement, DHI; Lactanet) were included in the analysis. Tests performed within ±6 mo relative to each farm's risk assessment date were included (10 ± 2 SD tests per farm). The majority of the farms were located in Ontario (83%). For each farm DHI test, the data set included herd average milk yield (kg/cow per day), average milk fat and protein (%), average somatic cell count (cells/mL), average days in milk (DIM), number of cows tested for ketosis, number of ketosis-positive tests (milk β-hydroxybutyrate ≥0.15 mmol/L), and proportion of cows by parity groups. Overall HPH (5-21 DIM) was calculated based on data available per farm (sum of all positive tests within 5-21 DIM/sum of all cows tested within 5-21 DIM). Each farm average was obtained by considering all test-days. A logit-transformation was applied to hyperketolactia prevalence. Linear regression models (PROC GLM and MIXED of SAS, Version 9.4) were used to predict herd HPH (milk β-hydroxybutyrate ≥0.15 mmol/L within 5 to 21 DIM; the outcome of interest). Four initial models (far-off, close-up, and fresh periods, and DHI) were separately built to assess associations between their variables and HPH; a final model considered variables selected in the initial models. Univariable (liberal P < 0.25) followed by multivariable models were used to build specific models for each period of the risk assessment. Herd prevalence of hyperketolactia was 27 ± 14%, with an average herd size of 141 ± 110 cows. The final HPH model (R² = 24.8%) included weighted milk yield, the proportion of primiparous cows, water access in the close-up period, and access to rest areas or stall access in the fresh period. Herd prevalence of hyperketolactia was negatively associated with milk yield [odds ratio, OR = 0.96 (95% confidence interval 0.92-0.99)] and proportion of primiparous cows [OR = 0.98 (0.96-0.99)]. The odds of hyperketolactia were greater with poor water access and quality (<5 cm of linear access per cow; dirty water; only 1 water location in pen) than with ≥10.2 cm of linear access per cow; clean water; >2 water locations in pen [1.23 (1.11-2.39)] in the close-up period. The odds of hyperketolactia were greater in farms providing limited access to rest areas in the fresh period than in farms providing constant access to rest areas, without dead-ends [1.64 (1.03-2.80)]. In Canadian dairy herds, HPH in early lactation was associated with certain transition-period management practices and was negatively associated with herd productivity.

Subclinical Ketosis in Dairy Herds: Impact of Early Diagnosis and Treatment

Clinical and subclinical ketosis (SCK) in dairy cows occurs during the lactation period frequently in many herds, causing a reduction in milk yield and alterations in milk quality with significant economic losses for farmers. SCK is defined as a preclinical stage of ketosis characterized by an elevated ketone body level without clinical signs. Often many cows develop an elevated ketone body level during the first weeks of lactation even though it never goes up to a critical point causing clinical signs. This study aimed to evaluate the prevalence of SCK in Sicily and assess the effect of a treatment with propylene glycol (PG) to control the SCK, thus, reducing the negative effect on milk quality yield. This cross-sectional study was carried out on 22 farms located south-east of Sicily and 1,588 cows in lactation. A total of 3,989 individual milk samples were collected from calving to 80 subsequently days to check the β-hydroxybutyrate (BHB) values in order to establish the SCK status by the Fourier Transform Infrared Spectroscopy. Moreover, the contents of fat, protein, lactose, casein, urea, somatic cell count and acetone were evaluated to identify a correlation between SCK and milk quality. A total of 1,100 cows showed BHB values higher than 0.10 mmol/L. These cows were considered SCK positive, were separated from the rest of the herd, and treated with PG (400 g/head per day), all SCK cows were treated with PG and cows without SCK were not treated. The results showed a prevalence of 41.5% of SCK-positive cows during the first 9 days of lactation. The comparison among the cure rate of treated cows shows that the treatment was most effective in the first 7 days of lactation (76.5% of treated cows) than in the following days. PG positively influenced the milk quality parameters, except for the fat proportion. Moreover, the animals treated with PG showed also an increase in milk yield, supporting the economical sustainability of treatment.

Evaluating the performance of machine learning methods and variable selection methods for predicting difficult-to-measure traits in Holstein dairy cattle using milk infrared spectral data

Fourier-transform infrared (FTIR) spectroscopy is a powerful high-throughput phenotyping tool for predicting traits that are expensive and difficult to measure in dairy cattle. Calibration equations are often developed using standard methods, such as partial least squares (PLS) regression. Methods that employ penalization, rank-reduction, and variable selection, as well as being able to model the nonlinear relations between phenotype and FTIR, might offer improvements in predictive ability and model robustness. This study aimed to compare the predictive ability of 2 machine learning methods, namely random forest (RF) and gradient boosting machine (GBM), and penalized regression against PLS regression for predicting 3 phenotypes differing in terms of biological meaning and relationships with milk composition (i.e., phenotypes measurable directly and not directly in milk, reflecting different biological processes which can be captured using milk spectra) in Holstein-Friesian cattle under 2 cross-validation scenarios. The data set comprised phenotypic information from 471 Holstein-Friesian cows, and 3 target phenotypes were evaluated: (1) body condition score (BCS), (2) blood β-hydroxybutyrate (BHB, mmol/L), and (3) κ-casein expressed as a percentage of nitrogen (κ-CN, % N). The data set was split considering 2 cross-validation scenarios: samples-out random in which the population was randomly split into 10-folds (8-folds for training and 1-fold for validation and testing); and herd/date-out in which the population was randomly assigned to training (70% herd), validation (10%), and testing (20% herd) based on the herd and date in which the samples were collected. The random grid search was performed using the training subset for the hyperparameter optimization and the validation set was used for the generalization of prediction error. The trained model was then used to assess the final prediction in the testing subset. The grid search for penalized regression evidenced that the elastic net (EN) was the best regularization with increase in predictive ability of 5%. The performance of PLS (standard model) was compared against 2 machine learning techniques and penalized regression using 2 cross-validation scenarios. Machine learning methods showed a greater predictive ability for BCS (0.63 for GBM and 0.61 for RF), BHB (0.80 for GBM and 0.79 for RF), and κ-CN (0.81 for GBM and 0.80 for RF) in samples-out cross-validation. Considering a herd/date-out cross-validation these values were 0.58 (GBM and RF) for BCS, 0.73 (GBM and RF) for BHB, and 0.77 (GBM and RF) for κ-CN. The GBM model tended to outperform other methods in predictive ability around 4%, 1%, and 7% for EN, RF, and PLS, respectively. The prediction accuracies of the GBM and RF models were similar, and differed statistically from the PLS model in samples-out random cross-validation. Although, machine learning techniques outperformed PLS in herd/date-out cross-validation, no significant differences were observed in terms of predictive ability due to the large standard deviation observed for predictions. Overall, GBM achieved the highest accuracy of FTIR-based prediction of the different phenotypic traits across the cross-validation scenarios. These results indicate that GBM is a promising method for obtaining more accurate FTIR-based predictions for different phenotypes in dairy cattle.

Effects of a Multicomponent Herbal Extract on the Course of Subclinical Ketosis in Dairy Cows - a Blinded Placebo-controlled Field-study

A blinded placebo-controlled multi-center on-farm trial was conducted in dairy cows with subclinical ketosis to investigate effects of a multicomponent herbal extract. Blood ketone levels were measured weekly in early lactating cows from 16 Swiss herds. Cows were subclassified based on their initial blood-β-hydroxybutyrate levels (≥ 1.0 [KET-low, 84 cows] and > 1.2 mmol/L [KET-high, 39 cows]) and randomly distributed to 3 groups treated orally with herbal extract containing Camellia sinensis, Cichcorium intybus, Gentiana lutea, Glycyrrhiza glabra, Taraxacum officinale, Trigonella foenum-graecum, and Zingiber officinale, sodium propionate, or placebo twice a day for 5 days. Milk yield, milk acetone, blood-β-hydroxybutyrate, glucose, nonesterified fatty acids, gamma-glutamyl transferase, and glutamate dehydrogenase were analyzed over 2 wk. Linear mixed effect models were used for data analysis. No effects were found for nonesterifed fatty acids, gamma-glutamyl transferase, and glucose. Significantly higher glutamate dehydrogenase (29.71 U/L) values were found in herbal extract-treated animals compared to sodium propionate on day 7 (22.33 U/L). By trend, higher blood-β-hydroxybutyrate levels (1.36 mmol/L) were found in the placebo group of KET-high-cows on day 14 compared to the sodium propionate group (0.91 mmol/L). Milk yields of all treatment groups increased. Milking time and treatment showed a significant interaction for milk acetone: sodium propionate led to an immediate decrease, whereas herbal extracts resulted in a milk acetone decrease from day 7 on, reaching significantly lower milk acetone on day 14 (3.17 mg/L) when compared to placebo (4.89 mg/L). In conclusion, herbal extracts and sodium propionate are both likely to improve subclinical ketosis in dairy cows, however, by different modes of action.