In vitro estimations of the rate and extent of ruminal digestion of starch-rich feed fractions compared to in vivo data

An experiment was conducted to study the rumen digestion characteristics of whole feeds (WF) and the neutral detergent fibre (aNDF) and neutral detergent soluble (NDS) fractions of a range of starch-rich feeds using an automated in vitro gas production (GP) technique. In addition, the ruminal digestibility values predicted from the GP data were compared to previously acquired in vivo data. Nine feeds with starch concentrations ranging from 389 to 712 g/kg dry matter and with known in vivo digestibilities were subjected to neutral detergent extraction. The GP for each WF and the corresponding aNDF fractions were measured in duplicate in buffered rumen fluid during 72 h on two occasions. The fermentation residues were collected and analyzed for aNDF concentration to estimate their true organic matter (OM) and NDF digestibility. The GP from the NDS fraction was calculated by subtracting the GP from the aNDF fraction from the GP of the WF. A three-pool Gompertz model was fitted to the GP profiles (R² = 0.99) and a two compartment, mechanistic and dynamic rumen model was used to predict the digestibility of the potentially digestible feed fraction and the effective digestion rate (k_d). The true OM and NDF digestibility determined for the WF ranged from 0.804 to 1.011 and from 0.362 to 1.107, respectively. The NDF digestibility determined for the aNDF fraction ranged from 0.410 to 0.985. The effective k_d values estimated using GP data varied from 0.118 to 0.282/h for the WF and from 0.123 to 0.301/h for the NDS fraction, and were less (P<0.05) for maize compared to small grains (SG) but did not differ between barley and wheat (P>0.05). The effective k_d values for the aNDF fraction ranged from 0.039 to 0.082/h and did not differ (P>0.05) either between maize and SG or between barley and wheat. The predicted ruminal NDS digestibility determined using GP data closely matched the in vivo data describing starch digestion (R² = 0.81). The effective k_d values for the WF were strongly related (R² = 0.94) to those for the NDS fractions. The results indicate that when measured with the GP technique, the differences in the digestion characteristics of maize and small grains are less than those previously reported in studies using the in situ method. It is concluded that the predicted NDS digestibility determined using GP data corresponded well to the in vivo starch digestibility. Our results also suggest that the first order digestion rates of NDS (starch) in starch-rich feeds can be accurately determined by incubating WF samples in the GP system and using the GP kinetic data in a dynamic, mechanistic rumen model.