New mid-FTIR herd management tools (HMT)

Introduction of new mid-FTIR herd management tools (HMT) for early warnings of nutritional and health issues in high producing dairy cows

Wopke Beukema (1) Ruth Santamaria (1) Tom Gates (2) Heather Dann (3) Rick Grant (3) and Dave Barbano (4)

(1) Delta Instruments B.V., Kelvinlaan 3, 9207 JB, The Netherlands (
(2) St. Albans Cooperative Creamery, 140 Federal St, St Albans City, VT 05478 
(3) Miner Institute, 586 Ridge Rd, Chazy, NY 12921 (,
(4) Cornell University, 289 Stocking Hall, Ithaca, NY 14853 (


Delta Instruments, St. Albans Cooperative Creamery, Cornell University and Miner Institute research collaboration on milk production efficiency has enabled the development of the first generation mid-FTIR herd management tools (HMT). A set of predictive fatty acids (FA) models has been developed to provide farmers with early warnings on nutritional and health problems of dairy cows before these become severe. The first tool is a set of milk FA models grouped according to their chain length and relation to biosynthetic origin within the cow. FA from C4 (i.e. 4 carbon) to C14, known as de novo FA, are synthesized in the udder from volatile FA produced in the rumen. C18 and higher are pre-formed FA mobilized from the adipose tissue or transferred directly from the diet. The C16:0 and C16:1 are mixed origin FA that can be synthesized de novo or may be pre-formed FA. Farmers will be provided with information they can use to effectively monitor and improve milk composition (i.e. fat and protein percentage) and production volume of the herd (bulk tank milk testing), as well as with early warnings on gastrointestinal diseases  of the early transition period (i.e. lactation weeks 1 to 6) cows (individual cow milk testing). The second tool concerns a milk-based blood non-esterified fatty acids (NEFA) prediction. Negative energy balance stimulates the mobilization of energy from fat stores and causes blood NEFA levels to rise. High NEFA levels in the blood signal that the body is in a severe negative energy status and there is an increased risk of metabolic diseases (e.g. ketosis, displaced abomasum, etc.) during the transition period. The milk-based blood NEFA prediction tool will provide farmers with information about the severity of the negative energy balance status of a transition cow and metabolic (clinical and subclinical ketosis) diseases. These early warnings of the health issues of a transition cow could enable preventive measures to avoid these conditions from becoming severe, while improving the transition process from pregnancy to lactation.


Delta Instruments, St. Albans Cooperative Creamery, Cornell University and Miner Institute research collaboration on milk production efficiency has enabled the development of the first generation mid-FTIR herd management tools (HMT). A set of partial least squares (PLS) chemometric predictive models for mid-FTIR spectra of milk have been developed to measure groups of fatty acids (FA) as they relate to the biosynthetic origin of the milk FA. In addition, a (PLS) prediction model to predict the level of non-esterified fatty acids (NEFA) in cow’s blood based on the mid-FTIR milk spectra of individual cow milk samples has been developed.

The first tool consists of a set of milk FA models grouped according to their chain length and relationship to biosynthetic origin of milk FA within the cow, and have good potential to be correlated with biology, metabolism and feeding of dairy cows. FA from C4 (i.e. 4 carbon) to C14, known as de novo FA, are synthesized in the udder from volatile FA produced in the rumen. The C18 and higher are pre-formed FA mobilized from the adipose tissue or transferred directly from the diet (Woolpert, et al., 2016). The C16:0 and C16:1 are mixed origin FA that can be synthesized de novo or may be pre-formed FA.

The concentration of NEFA in the blood of lactating cows is used as an index of how much fat is being mobilized by a dairy cow from adipose tissue at the beginning of the lactation. When blood NEFA and blood beta hydroxybutryate (BHB) are too high, cows are susceptible to a range of metabolic health issues, such as displaced abomasum, ketosis, retained placenta, and others (McArt, et al., 2012, Ospina, et al., 2010). These metabolic health issues are highly responsible for substantial financial losses due to decreased milk yield, poor reproduction, increased susceptibility to immunosuppression and culling rate (Ingvartsen, 2006, McArt, et al., 2012).

Materials and Methods

De novo fatty acid predictive model set

To develop the specific milk-based mid-FTIR prediction models, spectra of modified milk calibration samples, bulk tank milks from various regions of the US and individual cow milks were used in combination with chemical reference chemistry (Kaylegian, et al., 2006).

A variety of FA and groups of FA were measured producing results from infrared in grams of FA per 100 g of milk: C16:0; C18:0; C18:0 cis9, cis12; C18:1 trans 11 total FA; de novo (C4:0 to C14:0), mixed origin (C16:0, C16:1, C17:0), pre-formed (C18:0 and longer); total FA, total unsaturated FA, total cis FA; total trans FA; monounsaturated FA; and polyunsaturated FA.

Two fat concentration independent measures of FA structure were also done on each sample: mean FA chain length (expressed as mean carbon number per FA) and mean FA unsaturation (expressed as double bonds per FA). The measure of total FA (not fat) in g/100 g of milk is used as a new basis for a more accurate measurement of total fat content in the milk (Wojciechowski & Barbano, 2016). This approach eliminates most of the weakness of traditional measures of FA by IR using the fat A (C=O stretch) and fat B (C-H stretch) because it compensates sample by sample for differences in FA compositions when trying to estimate the total fat content of the milk in comparison to ether extraction (Kaylegian, Dwyer, et al., 2009, Kaylegian, Lynch, et al., 2009). The relative proportion of the total FA in milk represented by an individual or group of FA can be expressed on a relative basis as a percentage of total FA in the sample. Thus, it is possible to produce a simulated gas chromatograph FA analysis of milk fat (using PLS models for individual FA) directly from the same (IR spectra) of milk on the IR for fat, protein and lactose concentration.

The validation of IR FA results was done with split sample analysis to compare IR and gas liquid chromatography (GLC) for FA estimates on samples throughout the study. LactoScope FTIR (Delta Instruments, The Netherlands) provided measurements of milk components and FA composition, while the reference testing for FA composition was done with gas liquid chromatography using a Varian CP-SIL88 capillary column (Variant, Inc., Lake Forest, CA), installed in a Hewlett Packard 6890 GC System equipped with an automatic liquid sampler and a flame ionization detector (Hewlett Packard Co., Wilmington, DE) (Wojciechowski & Barbano, 2016, Woolpert, 2016). The calibration adjustment of the fat, true protein, anhydrous lactose and all FA measured on the LactoScope FTIR was done once a month using a set of 14 modified milks described by Kaylegian, et al. (2006) that had reference values in (g FA per 100 g of milk) for each of the individual or groups of FA measured. The set of calibration samples was produced monthly at Cornell University, and was used to check the calibration during the month (Wojciechowski, et al., 2016).

Blood NEFA predicted values in milk model 

There are no NEFA in milk, so a model to predict blood NEFA from a milk sample uses differences in the milk spectra that are correlated with changes in blood NEFA.

A group of 60 individual Holstein cows from the dairy herd at Miner Institute was monitored for the first three weeks of lactation. Cows were milked 3 times per day. Within more or less one milking of the time of blood collection, a milk sample was analyzed using the LactoScope FTIR. A WAKO NEFA HR test kit was used as an in vitro enzymatic colorimetric method for the quantification of NEFA in blood serum, and these values were used as a reference value for development of a PLS regression model to predict blood NEFA from the mid-IR milk spectra.

The final PLS model had 9 factors, using the following ranges (300 to 2800, 1800 to 1700, 1585 to 1000 cm-1) with a standard cross validation of 172 µEq/L. Validation milk and blood sample pairs (n=53) were collected from Holstein cows from a different herd. The mean value for the blood reference test was 713 µEq/L of serum and the mean value for the milk-based blood NEFA prediction was 703 µEq/L of serum with a standard deviation of the difference (SDD) of 218 µEq/L for the 53 validation samples (Barbano, et al., 2015).

Results and Discussion

De novo FA predictive model set

The de novo FA predictive model set was used to test bulk tank milks from 430 farms located in Northeast USA that were sampled and tested 3 to 20 times per month, using mid-FTIR to measure fat, true protein, anhydrous lactose and FA composition. FA data were organized and analyzed by breed: Jersey and Holstein. A wide range of farm management practices and a range of herd sizes were represented in this study. There was an overall breed difference in FA composition as well as a large amount of variation within each breed. FA composition within a farm from day-to-day was fairly consistent, and when a major change in FA composition within a farm was observed, it was usually a major change in feeding that shifted the FA composition.

The most interesting parameters in FA data that were correlated with the concentration of fat and true protein in the bulk milk were the groups of FA – de novo, mixed and pre-formed FA. There is a positive correlation of bulk tank fat and true protein test when de novo FA concentration in milk was higher for both Holstein and Jersey breeds. The correlation for de novo FA is stronger with fat than for true protein (Barbano, et al., 2014).  Increased output per day of de novo FA may reflect better rumen fermentation with higher production of rumen volatile FA and higher production of rumen microbial biomass that provides essential amino acids to support milk protein synthesis.

Given the relationships observed in the data collected from the 430 farms (Barbano, et al., 2015), a sub-population of high de novo and low de novo farms was selected in 2014 and then again in 2015 to determine the differences in feeding and management practices between the high de novo (HDN) and low de novo (LDN) herds, the relationship to bulk tank milk composition and differences in milk payment. In both years, the HDN farms had higher fat and protein test in bulk tank milk than the LDN farms.

The study carried out in 2014 evaluated the relationship between management practices and milk de novo FA content for lower-producing smaller dairy farms containing a variety of breeds. A population of 44 HDN and LDN were selected based on a history of HDN and LDN FA in bulk tank milk. Management practices were assessed during a visit to each farm, and the total mixed ration samples were collected and analyzed for chemical composition. At a farm level, there were no differences in days in milk. This research indicated that overcrowded freestalls, reduced feeding frequency and greater dietary ether extract content were associated with lower de novo FA synthesis and reduced milk fat and true protein yields per cow per day (Woolpert, et al., 2016).

The second study in 2015 was carried out to better understand the relationship between cow comfort indicators and physically effective non-degradable fiber (peNDF) with de novo FA concentration for high-producing Holstein dairy farms. Similar to the previous study, a population of 19 HDN and 20 LDN farms was selected and evaluated based on management and facility design previously shown to affect cow comfort, on physical and chemical characteristics of the diet, forage quality, and on the ration formulation obtained from the farm’s nutritionist. There were no detected differences in farm size, time away from the pen for milking, days in milk, or body condition score between the HDN and LDN. No differences between HDN and LDN farms in milk, fat or true protein yield were detected; however, milk fat and protein content and de novo FA yield per day were higher for HDN farms, as was gross income per unit of milk sold. The results of this study indicated that cow comfort indicators, such as feeding management and stocking density, as well as physical characteristics of the diet, are related to de novo FA concentration in bulk tank milk of Holstein dairy farms and total production of fat and protein per cow per day (Woolpert, et al., 2016).

Blood NEFA predicted values in milk model

Barbano, et al. (2015) were the first to report and validate a blood NEFA prediction model based on the analysis of milk samples from individual cows. There was a considerable cow-to-cow variation in the level and the temporal patterns of change in the relative proportions of the de novo, mixed and pre-formed milk FA that seem to reflect real-time cow-to-cow differences in energy balance and metabolic health status of individual cows. Generally, healthy cows at 7 days in lactation that do not have excessively high blood NEFA will have a relatively high percentage of total FA that are de novo FA (20% or higher), and with increasing days in milk, the de novo value as a proportion of total FA should be in the range of 27–30% of total FA when the cow reaches positive energy balance. Generally, the mixed origin FA as percentage of total FA will increase with days in milk and the pre-formed FA will decrease. Blood NEFA measured on blood is a snapshot of the NEFA concentration at that precise instant in time, while blood NEFA predicted from milk represents an average for the total time between milkings.


The first generation HMT for mid-infrared are useful in supporting farm management decisions that have the potential to improve the economic performance and sustainability of the dairy farm. The FTIR milk analysis is rapid, uses no reagents and can simultaneously perform the measurement of the major components in milk, milk FA composition, milk FA structure, and the other metrics (i.e., estimated blood NEFA) related to the risk for nutritional and metabolic diseases. Farm managers will be enabled with information they can use to improve feed efficiency and the transition from pregnancy to lactation of a dairy cow. The de novo FA model set used has been specifically developed to help the farm managers increase the total fat, true protein and bulk tank milk yield of their herds, increasing their revenue and profitability of their farms. At the individual cow level, the de novo FA set will give early warnings on the high risks of displaced abomasum for lactating cows within 12 days in milk. The blood NEFA predicted values in milk model has been specifically develop for early warning of the metabolic diseases of clinical and subclinical ketosis for lactating cows with up to 60 days in milk.


The authors would like to thank the staff at Cornell University, Miner Institute and the St. Albans Creamery Cooperative for their invaluable contribution that has made the creation of the first generation HMT possible.


Barbano, D. M., Cree, P., Overton, T. R., Dann, H. M., & Grant, R. J. (2015). Prediction of Blood Nonesterified Fatty Acid (NEFA) by FTIR Analysis of Individual ow Milk Samples. Journal of Dairy Science, 98(1):(553).

Barbano, D. M., Melilli, C., & Overton, T. R. (2014). Advance use of FTIR Spectra of Milk for Feeding and Health Management Cornell Animal Nutrition Conference, Cornell University, Ithaca, NY, 105-113

Ingvartsen, K. L. (2006). Feeding and Management-Related diseases in the Transition Cow: Physiological Adaptations Around Calving and Strategies to Reduce Feeding Related Diseases. Animal Feed Science and Technology, 126(3-4):(175-213).

Kaylegian, K. E., Dwyer, D. A., Lynch, J. M., Bauman, D. E., Fleming, J. R., & Barbano, D. M. (2009). Impact of Fatty Acid Composition on the Accurancy of Mid-Infrared Fat Analysis of Farm Milks. Journal of Dairy Science, 92:(2502-2513).

Kaylegian, K. E., Houghton, G. E., Lynch, J. M., Fleming, J. R., & Barbano, D. M. (2006). Calibration of Infrared Milk Analyzers: Modified Milk Versus Producer Milk. Journal of Dairy Science, 89:(2817-2832).

Kaylegian, K. E., Lynch, J. M., Fleming, J. R., & Barbano, D. M. (2009). Influence of Fatty Acid Chain Length and Unsaturation on Mid-Infrared Milk Analysis. Journal of Dairy Science, 92:(2485-2501).

McArt, J. A., Nydam, D. V., & Oetzel, G. R. (2012). Epidemiology of Subclinical Ketosis in Early Lactation Dairy Cattle. Journal of Dairy Science, 95:(5056-5066).

Ospina, P. A., Nydam, D. V., Stokol, T., & Overton, T. R. (2010). Evaluation of Nonesterified Fatty Acids and Beta-hydrozybutyrate in Transition Dairy Cattle in the Northeastern United States: Critical Thresholds for Prediction of Clinical Diseases. Journal of Dairy Science, 93:(546-554).

Wojciechowski, K. L., & Barbano, D. M. (2016). Prediction of Fatty Acid Chain Length and Unsaturation of Milk Fat by Mid-Infrared Milk Analysis. Journal of Dairy Science, 99:(8561-8570).

Wojciechowski, K. L., Melilli, C., & Barbano, D. M. (2016). A proficiency test system to Improve Performance of Milk Analysis Methods and Produce Reference Values for Component Calibration Samples for Infrared Milk Analysis. Journal of Dairy Science, 99:(6808-6827).

Woolpert, M. E. (2016). Management Practices and Communication Stratagies to Improve Milk Fat and Protein Content on Dairy Farm. University of Vermont, Burlington.

Woolpert, M. E., Dann, H. M., Cotanch, K. W., Melilli, C., Chase, L. E., Grant, R. J., & Barbano, D. M. (2016). Management, Nutrition and Lactation Performance are Related to Bulk Tank Milk de Novo Fatty Acid Concentration on Northeastern US Dairy Farms. Journal of Dairy Science, 99:(8486-8497).