167 Analysis of variation in net energy prediction of feed ingredients fed to swine

Precision of NRC and CVB net energy (NE) prediction equations was investigated by taking into account natural and analyzed variation of the chemical components (e.g., moisture, protein, starch, NDF, ADF, sugars, fat) for six feed ingredients used in swine diets. Precision is defined as the variation of the NE formula given the variation of its chemical components. Stochastic variables were substituted for the component values and the variation in nutrient composition was algebraically, not statistically, propagated to the NE value for both equations. This was done both, for the naturally occurring variation as published in CVB, and for a range of analytical variation values associated with wet chemistry analysis based on a range of values obtained by published ring-tests. Consequently, variation of the NE value and the contribution to variation of each chemical component in the NE equation were calculated. The variation of NE prediction using CVB is lower than using NRC. The main contributor for increased variation of the NRC NE is the NDF fraction. Whereas in CVB, this is replaced by the NSP fraction which is computed as a residue from the other chemical component values, forcing the sum of the composition to add to 100%. Furthermore, it was determined that analyzing all nutrients, in particular NDF and ADF, did not always reduce the variation of the NE equations. In conclusion, analytical variation, especially fiber analysis, must be critically examined and, preferably, sum to 100% to increase precision in the prediction of NE in feed ingredients. Otherwise, the use of a residue fraction, although nutritionally difficult to justify, actually increased precision in the NE equations. Note that it is unfair to compare both NE formulas based on precision alone. An interesting follow-up question is to take accuracy also into account when comparing CVB and NRC.