|Hall, Mary Beth|
|Rymph, Stuart - UNIVERSITY OF WISCONSIN|
Submitted to: Canadian Journal of Animal Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: July 15, 2005
Publication Date: December 5, 2005
Citation: Hall, M., Rymph, S.J. 2005. Definitions of feed analysis methods as sources of error in nutritional models. Canadian Journal of Animal Science. 85(4):545. Interpretive Summary: There are a variety of nutritional models available to predict requirements, utilization and excretion of nutrients by animals. To allow most accurate use of these models, it is essential that the authors provide users with the specific methods compatible with the model to obtain analytical information used as inputs. This guidance will help to assure that the inputs are correct for the model, and will generate answers correct for that model. Without this guidance, users may use inputs that are incorrect for that model, the outputs will also be in error, and resources devoted to use of the model will be unnecessarily wasted.
Technical Abstract: On the nutrient supply side, equations in models are developed with inputs that describe feed fractions that are assumed to have homogeneous kinetic properties or other nutritionally relevant characteristics. These inputs include quantitative data such as feed composition, or digestion kinetics of specific fractions. It can be a challenge for users to determine local values for inputs if specific, model-associated guidelines regarding the method of assay are not provided. The risk is that a user-selected method does not give the same measure as that to which the model was developed. The assay of "sugar" exemplifies the essentiality of defining method of assay. Three common methods for analysis of feedstuffs for "sugar" using 80% ethanol or water extracts, or total sugars as invert can give three very different answers - they include mono- and oligosaccharides, those carbohydrates plus water-soluble polysaccharides, or free glucose and fructose and that released from sucrose, respectively. The differences in values among the assays will differ among feedstuffs. "Sugar" values that are dissimilar in the carbohydrates they describe will also differ in their digestion characteristics. Although nominally providing values for "sugar", the different assays provide inputs that may or may not be correct for a given model. The potential difference between assays may be large: the difference between values determined with water and 80% ethanol extracts for carbohydrate content of grass stem bases (DM basis) ranged from 100 to 400 g kg-1, depending upon the grass species, with water extracts giving greater values (Smith and Grotelueschen, 1966). Diverse data sets that utilize different assays are often used in model development. The assays used to measure specific inputs in the studies should be noted for use by users to give guidance on methods that are appropriate to provide inputs. If a model is calibrated to data from an assortment of analytical methods, the modelers should recognize potential differences in analytes in specifying the precision of their inputs, and the possibility that the input is actually a relative index rather than an absolute value. Advances in methodology and understanding of nutritional relevance of feed fractions could allow specific method selection with possible improvement in precision after recalibration of the model to the new data. Collaboration among modelers and analysts seems essential to make improvements in this area. Smith, D. and Grotelueschen, R. D. 1966. Carbohydrates in grasses. I. Sugar and fructosan composition of the stem bases of several northern-adapted grasses at seed maturity. Crop Sci. 6:263-266.