Prediction of the nutritional value of European compound feeds for rabbits by chemical components and in vitro analysis

Abstract

Chemical composition and in vitro analyses were used to predict the nutritional value of 164 experimental rabbit diets evaluated in six European Laboratories under standardised conditions. The equations were mainly developed by stepwise regression analysis with over two third of the samples (111) used as calibration set. The other third (53) was used as validation set, and a study of the residues was undertaken to calculate the error of validation. Twenty three different equations have been proposed to predict the nutritional value (mainly gross energy digestibility, GEd; and digestible energy, DE) of rabbit diets, as a function of the available variables. Acid detergent fibre (ADFom) was the chemical variable most closely related to GEd and DE ( R 2 = 0.49 and 0.43, respectively, RSD = 0.033 and 0.62, for GEd and DE, respectively), but the in vitro DM digestibility (DMd inv) predicted the energy value with greater accuracy ( R 2 = 0.7, 0.52, for GEd and DE, respectively) and lower standard error (RSD = 0.025, 0.58 for GEd and DE, respectively). The latter equations were improved ( R 2 = 0.81, 0.74 for GEd and DE, respectively) when ether extract (EE) and Lignin (sa) were included. The use of additive equations that predict the DE from the main constituents that supply energy (protein, ether extract and carbohydrates) did not increase the precision, nor decrease the validation error respect to the simplest ones. Digestible Energy was predicted with a similar accuracy and validation errors than GEd. Crude protein digestibility (CPd) was better predicted from chemical analysis (Lignin (sa), R 2 = 0.49) than for DMd inv. The further inclusion of CP slightly increased its coefficient of determination (0.53). The error of validation was relatively low (0.050 as average) and of the same magnitude than the RSD of the equations.