Farm data
The ability to quickly assess the impact of an intervention is sometimes complicated when the number of cows is low. For example, if a 100-cow herd with a 12% incidence of retained placenta starts implementing a program that is expected to help decrease this rate by 20% (on average), the herd’s data should show 2.4 fewer cases per year (on average). Assuming that herd is operating on a flat calving pattern, it would take about five months (on average) before just one less case was detected. It reads ‘on average’ because the biological systems that govern milk production and animal health follow laws of probability which means production changes and behavioural patterns don’t always occur as expected on paper.
Irrespective of herd size, other factors complicate the task of carrying out a meaningful product assessment on your own farm. For example, underlying deviations in productivity and animal health caused by varying year-on-year weather patterns and limitations in the accuracy of past and current records will limit a single on-farm evaluation of a product’s worth.
Assessing the beneficial impact of a specific nutritional product becomes even harder to determine where those expected benefits are of a similar magnitude to the day-to-day variability of production and behavioural data such as milk yield and feed intake. It isn’t necessarily true that the benefits provided by the product are negligible, rather that the real-world day-to-day variability in the data can be substantial and may hide the benefits. For example, corn silage dry matter measurements are easy, yet results have been shown to vary typically 5 to 10 percentage units from one day to the next on the same farm (Weiss and St-Pierre, 2012).
Research papers
Sometimes quick farm ‘trials’ deliver low-quality or inconclusive data because of all the above reasons. Therefore, we will often resort to published research to help us evaluating the merit of a product or concept of interest, and how likely it is to benefit our dairy.
When evaluating published research, we need to make sure the published data we are seeing really applies to our current situation: is the product in that literature piece the same as the product we have available to use?, was it fed at the same rate and comparable conditions?, how many studies report the same type of finding?
Beyond these basic questions, assessing research literature is a specialized task, and there are frequent misconceptions around research, particularly with regards to interpreting research results. These are some of the most common general misconceptions:
- “Only local research matters”: the validity of this statement depends on the topic being studied. For example, if we are evaluating the nutritional value of a type of crop, this will probably depend on characteristics that are intrinsic to each region: soil, climate, agronomic practices, base ration, etc. and in order to obtain accurate estimations of the effects of that crop it is therefore necessary to consider regional scale evaluations. On the other hand, factors that are intrinsic to basic animal physiology or nutrition are less likely to benefit from repeated evaluations at a regional or national level: examples include efficacy of a ruminal buffer, or an anionic product.
- “Trials conducted on research farms are of better quality than field trials”: again, the validity of this specific assumption depends on a number of factors. For example, it is not always possible to obtain meaningful data from the highly controlled but typically small-scale conditions seen on research farms. Traits that require large number of animals to study to an acceptable degree of accuracy (such as reproductive traits) can only be studied in real world conditions and on commercial farms. Indeed, well-planned and monitored multisite trials enrolling commercial herds provide a solid research model that is underused by the animal nutrition industry today.
- “P must be lower than 0.05”: in terms of statistical analysis, ‘P’ tells us how likely a difference between results of two or more groups is random. P=05 means there is 5% chance the difference is random. The closer P is to 0, the less likely the difference is arbitrary and the more likely the difference between the groups is due to the treatment applied. How dissimilar is then P=0.04 from P=0.06 in terms of coming to a reliable conclusion about the efficacy of a product? P<0.05 is a conventional threshold, there is nothing magic about it. Indeed, for ruminants the European Food Safety Authority adopts P<0.10 as the confidence threshold for efficacy trials submitted for the registration of feed additives.
Assess and appraise
In summary, how do you know if a nutritional product is benefitting your dairy herd?
- Look at your own results in as much detail as possible but be wary of the limitations
- Research the product carefully by studying all available published research papers and evaluate the validity and relevance of the findings of these papers in terms of their quantity and quality
Contact us to learn about the extensive research on the benefits of feeding OmniGen and Animate, and how Phibro can support you on evaluating your own farm data.
References
- Weiss, B., and St-Pierre, N. 2012. Trying to Make Sense of Feed Composition Data: Within Farm Variation. In: Proceedings the Mid-South Ruminant Nutrition Conference, 33-44. Grapevine, Texas, USA.
- EFSA FEEDAP Panel. 2018. Guidance on the assessment of the efficacy of feed additives. EFSA Journal 16: 5274.
