Reduced milk yield, fat production, and fiber digestion are just a few of the symptoms of subacute ruminal acidosis (SARA). This hidden disease can hold back your dairy herd significantly, so Quebec researchers worked to shine a light into the rumen using machine learning.
In addition to production losses, SARA is known to promote inflammation in the cow, resulting in a reduced health status. This condition is technically defined as a rumen pH at or below 5.8 for several minutes per day. The ideal rumen pH is 6.2.
Volatile fatty acids (VFA) are an energy source resulting from the fermentation of feedstuffs in the rumen. However, when these VFAs accumulate in excess, often when high concentrate diets are fed, rumen pH declines, resulting in a shift of rumen microorganisms.
Rumen pH is difficult to measure, so researchers sought to find a proxy measurement to help us better keep a thumb on the status of SARA in a herd.
Rumen changes
Fat is a fantastic source of energy to support cows through lactation. Under normal rumen conditions, unsaturated dietary fats typically enter the biohydrogenation pathway, where they are packaged for easy uptake in the mammary gland. However, when the pH is low, this pathway is disrupted, resulting in a different concentration of fatty acids available for milk production.
The team of scientists initially studied SARA in a research setting, tightly controlling the conditions to see the impact on the milk fatty acid profile. They expected this response might be different in a commercial herd. “Induced SARA isn’t the same as the SARA you’d see in the field,” explained Eric Paquet, a computer scientist from Laval University in Quebec, on the July edition of the “Dairy Science Digest” podcast. He worked in collaboration with Agrinova, Lactanet, agronomist and Ph.D. candidate Felix Huot, and colleague Rachel Gervais to study the possibility of monitoring fatty acid profiles in the milk as a non-invasive way to detect SARA in commercial herds enrolled in Quebec’s monthly milk test of individual cows. Their findings were recently published in the Journal of Dairy Science.
“We’ve learned that rumen pH varies widely,” explained Huot. This heterogeneity in the dataset limited their ability to use fatty acid concentration to identify individual cows with SARA. Sill, Huot said, “We feel confident in applying the tool to find limitations at the herd level.”
Predicted SARA prevalence
Once the machine learning tool became reliable in using Fourier transform infrared (FTIR) spectrometry-estimated milk fatty acid data to predict SARA prevalence, the team applied the computer tool to a larger dataset housed at Lactanet in Quebec. They found an average predicted SARA prevalence of 6.6% over nearly 220,000 cows. Herd prevalence ranged between 0% and 40%. This could certainly be holding your herd back!
Using the larger dataset, the computer tool was able to identify common factors associated with higher predicted herd prevalence of SARA. These included:
- Robotic milking systems
- Herd size
- Seasonal effect
- Higher milk yield
- Depressed milkfat percentage
When compared to a pipeline, cows milked in robots tended to have an increased prevalence of SARA. “When a partial mixed ration is prepared, it separates some of the concentrate away from the fiber,” Huot explained. “This shift in overall diet particle size could decrease rumen pH.”
Smaller herds tend to have higher SARA prevalence, perhaps due to component feeding resulting in “slugs” of grain fed during milking that cause an undesirable decline in rumen pH. Green chop silage is likely the suspect of higher SARA prevalence through the fall.
The team was unable to analyze the link between predicted SARA prevalence and diets of the 3,000 herds, since this information was not available. Therefore, no diet causation can be assumed. It’s suspected that high-producing herds tend to have higher rates of SARA, perhaps due to more concentrate in the ration to support high production. The team also suspects high milkfat is an exemplar of adequate fiber being fed and aligns with rarely logging a positive for SARA in the model.
“It is challenging to predict SARA on an individual cow basis, but this tool has allowed us to provide insight for a herd over time,” said Paquet. “We hope to help herds find SARA so they can work to eliminate it.”
These findings were summarized in a peer-reviewed, open access Journal of Dairy Science article found at www.journalofdairyscience.org. To learn more, listen in to the monthly podcast, “Dairy Science Digest,” on your favorite podcast platform.
