Search

We conducted a continuous trial with 32 steers (Bos indicus × Bos taurus crosses) weighing from 100 to 450 kg to measure the aNDFom pool size in the ruminoreticulum Q N D F at slaughter, and two simultaneous 4 × 4 Latin Squares designs with some of the same animals of the continuous trial to measure the total digestibility of aNDFom T D N D F in the gastrointestinal tract. The diets were based on corn silage (CS) and soybean meal (SBM), and on CS and concentrate mixtures of SBM and ground yellow corn (GC). In the Latin square trial, the dry matter intake rates were set for an average daily gain equal to 0.6 kg/d. The treatments were dietary aNDFom intakes scaled to body weight B W at 3, 6, 9, and 12 g / kg · d combined with markers chromium (Cr) and europium (Eu) that labelled CS fibrous particles, and yttrium (Y) that labelled fibrous particles of the concentrate mixtures. The variables Q N D F and T D N D F were predicted by novel derivations of the models most often used to describe fecal excretion profiles of the markers Cr, Eu, and Y. We investigated the following compartmental models: the multicompartmental (MC), the single compartment (GN), and the two-compartment (GNG1) models. We used nonlinear mixed effects and generalized linear mixed-effects models to fit the compartmental models to and make inferences from the fecal marker profiles. In vitro digestion kinetics yielded the parameters that represent the potentially digestible, indigestible and digestion rates of feed aNDFom. All solutions of the GNG1 model violated a critical assumption for some treatment combinations and were deemed unacceptable and removed from the comparative analyses. The best-suited model to interpret fecal marker profiles, according to the information-theoretic (I-T) approach, was the MC model. The solutions for the GN model had less support for describing the fecal marker profiles than the MC model. In addition, several measures of model adequacy were used to empirically evaluate the model predictions against observed variables and revealed that, despite the choice based on I-T, both models MC and GN were equally capable of predicting the Q N D F and T D N D F variables with both accuracy and precision.