506 Mid-infrared spectroscopy of milk as a tool to predict subacute ruminal acidosis

Abstract Subacute ruminal acidosis (SARA) has deleterious effects on the health, welfare and production of dairy cows. At present the only way to accurately diagnose SARA is by continuous measurement of rumen pH, which is costly and impractical. Several authors have shown that SARA causes changes to the milk fatty acid profile of lactating dairy cows, and a number of fatty acids have been proposed as potential biomarkers. Mid-infrared spectroscopy (MIR) has been used to predict the concentration of several fatty acids in milk with good accuracy, so we therefore hypothesized that MIR spectra, generated as part of routine milk recording, could be used to predict rumen pH and SARA. Concurrent rumen pH and milk MIR spectral data were collated from two experiments conducted at the Ellinbank dairy research centre (belonging to the State Government of Victoria) between September 2016 and September 2017. A total of 144 rumen pH measurements and milk MIR spectra from 20 fistulated cows were included in the analysis. Several pH metrics were investigated, including mean pH, time below pH 6 and area under the pH curve between milkings, and pH at the time of milk sample collection. Models to predict rumen pH using milk MIR spectra were constructed using partial least squares regression analysis, and prediction accuracy was assessed using a 10-fold cross-validation. The accuracy of all models was moderate, with R2 values between 0.22 and 0.59. SARA was defined as rumen pH less than six for greater than 180 minutes in the period between milkings. Partial least square discriminant analysis models could classify animals as either affected or not affected with SARA with a sensitivity of 81% and a specificity of 72%. Our results indicate that MIR of milk may be a useful tool for monitoring rumen pH in lactating dairy cows.