Expressing the variability in results of metabolizable energy assays

In estimating the metabolizable energy of a food or feedstuff (MEf), the ME of a reference diet (MEr, containing glucose monohydrate) and the ME of a test (MEt, containing the feedstuff substituted for glucose) are determined from a feeding trial: MEf = MEglucose -1/P (MEr - MEt) where P is a proportion of substitution. Traditionally, MEr has been used in calculating the variance of MEf as if it were measured without error, leading to an underestimation of the variance in MEf. Variances of MEf estimated by a method based on the variances MEr and MEt were 4-11% greater for sample data. Underestimation of variances can lead to excess confidence in ME estimates. Too few replicates may be used for the desired confidence. Figures depicting the influence of replication and degree of substitution are included to help select the replication and proportion of substitution necessary to achieve acceptable results.