Prediction of standardized ileal digestibility and essential amino acid content of ingredients in swine: A meta-analysis

• Predictive models of amino acid content based on crude protein analysis are available.
• Prediction of apparent ileal digestibility from total amino acid of ingredient.
• Conversion of amino acid apparent ileal digestibility to standardized ileal digestibility can be improved.
• Pig weight has a minor effect on the ileal digestibility of amino acids of feeds.

Amino acid (AA) composition and digestibility are major determinants of nutritional value. These factors vary considerably in animal feed ingredients. For pig feed ingredients, standardized ileal digestibility (SID) is commonly used as an estimate of nutritional AA value. In this study, relationships between ingredient proximate composition and AA content, AA content and essential AA apparent ileal digestibility (AID), and between SID and AID were characterized using databases compiled from published research. The effects of ingredients (cereals, cereal by-products, oilseeds, legume seeds and oilseed meals), surgical procedure (T-cannula, re-entrant cannulas, post-valve T-cannulas and ileorectal anastomy) and animal weight (≤ or >25 kg) were modelled. Relationships between independent variables were examined for each AA to avoid co-linearity. The best predictor of AA content was crude protein (significant slope, P < 0.001, high R2, low error) for all ingredients except for faba bean (R2 = 0.143–0.805). For dietary lysine, prediction accuracy was best for peas, lupin and barley, with root mean square errors prediction of 3.84%, 8.08% and 1.89% respectively. Neither ingredients nor surgical procedure (P > 0.1) nor the basal endogenous nitrogen loss estimation method (P > 0.1) appeared to affect the accuracy of essential AA (EAA) AID estimation from tEAA and EAA SID estimation from AID. Apparent ileal digestibility was the best predictor of essential EAA SID (R2 = 0.911–0.987), slope near unity and a significant intercept representing basal endogenous nitrogen loss). Compared to published models, the studied models had improved R2 and offered better accuracy.