Literature was searched for studies performed in adult dairy cattle that simultaneously measured starch degradability in the rumen (RSDeg) and starch digestion in the total tract to compute postruminal starch digestion (PRSDig). Forty-one studies with 161 dietary treatments were used to form the data set. Of these diets, the major starch source was corn for 83 diets, small grain for 58 diets, and sorghum for 8 diets. Corn RSDeg was more variable than other sources. As measured in vivo across all starch sources, the percent RSDeg was influenced only by the amount of starch consumed, with the amount of degradation being approximately 75% at low starch intakes and decreasing to about 60% when 4kg or more of starch were consumed. Small grain starch had greater RSDeg than corn or sorghum starch, which were approximately equal. The PRSDig of corn and small grain starches were approximately equal, but sorghum was about 15% less. Across all diets, models derived from the Cornell Net Carbohydrate Protein System predicted percentage of total-tract digestibility of starch very accurately, but overpredicted RSDeg and, as a result, underpredicted percent PRSDig. Calculation of RSDeg using a French model predicted the mean RSDeg with greater accuracy but less precisely. The relative differences in RSDeg percent among starch sources was correctly predicted by these models. A model using a revised rate of digestion as a way of combining effects of starch type and processing was developed, which predicted corn starch RSDeg and PRSDig with greater accuracy than nutrition models but only slightly better than using the mean observed degradation or the French calculation. Inaccuracies in prediction of RSDeg may be due mainly to processing effects and particle sizes, but these were not well reported in literature studies and were difficult to estimate. More accurate assessment of RSDeg and PRSDig will require better and more consistent reporting of grain processing. Based on this study, the French calculation is the most accurate of the models examined, although adjustments will be required to improve accuracy.
Read full abstract