September 12, 2016


To receive meaningful information from an experiment, you first must know the question you want to answer, then design the experiment to answer it. As AFB International’s principal biostatistician, that’s an important part of my job. I help design experiments to determine whether two pet food palatants are equivalent.

This equivalence testing research is not intended to measure whether the two palatants are exactly the same, but rather whether they’re statistically similar in their performance within a small, tolerable range.


At AFB, we use equivalence testing to ensure our palatants produced around the world are consistent, regardless of where they are produced—and that they remain consistent in quality over time in various conditions. For example, we compare palatants that are:

  • Produced at two different AFB production facilities using the same ingredients and instructions
  • Produced at two different AFB production facilities, where one uses an ingredient translation—a substitution that is more accessible or appropriate to that particular market
  • Produced over time at the same AFB production facility and shipped to various regions


When testing for differences between rations, the best interpretation of preference usually comes from a two-bowl test in which the animals are offered both rations simultaneously. However, a two-bowl test isn’t always the best choice for determining equivalence. Some dogs may demonstrate the similarity by switching between and eating from both bowls. Some dogs may recognize the similarity and, realizing there is no difference, stick with the left or right bowl for most of the meal.

This may create a high amount of variation in responses, which makes finding equivalence less likely, especially when the number of dogs in the trial is small. In this case, monadic (one bowl at a time) feeding trials are preferred to reduce the potential for misinterpreting indifference toward one food from preference for another food. Ultimately, monadic feeding allows us to declare equivalence in less time using fewer dogs.


Recently we conducted an experiment to answer the question: Is there a distinguishable difference between palatants made at two different locations using the same formula? Our data verified the similarities in preference for the palatants produced in these two different locations. The trial was conducted with 40 dogs using monadic feeding over a four-day span. The detailed results, represented in Figure 1, illustrate key concepts of equivalence testing.


[Adapted from graph featured in Equivalence Testing whitepaper download]

The blue and red dots represent the average diet amounts from each location eaten by the dogs over a four-day trial. Equivalence is indicated in two ways:

  • Many dogs have two points that overlap (shown in purple) indicating similar average consumption of both diets.
  • Points that don’t overlap, with vertical black bars connecting the points, indicate dogs with at least 18-gram differences in average consumption of the two diets. In five of these cases, the dogs preferred the palatant from Location 1, but these were balanced almost equally by three dogs that ate more from Location 2.

The points are arranged from left to right in order of increasing consumption by dog. This shows that equivalence was present in smaller dogs offered smaller amounts of food, as well as in larger dogs offered more food.


Because we at AFB are committed to research excellence, we can assure our customers of the global consistency of our products. For more information, see “Equivalence Testing: Insights from the Experts” in the Downloads area at or email me at


Tags: , , , , ,

Download FREE White Paper