Validation of prediction equations to estimate the nutritive value of broiler chicken diets based on their chemical composition

Abstract

An experiment was conducted to validate the accuracy of previously published prediction equations developed to estimate the coefficient of apparent ileal digestibility (CAID) and ileal digestible content (IDC) of nitrogen (N), crude fat, starch, calcium (Ca), phosphorus (P), energy, and dry matter (DM) in broilers using the chemical composition of diets. Twenty new diets were formulated to have a wide range of chemical characteristics relevant to commercial diets. The CAID of N, crude fat, starch, Ca, P, energy, and DM of the diets were determined in broiler growers fed ad libitum from 15 to 22 days post-hatch. The chemical composition and in vivo digestibility values were used to validate the prediction equations developed from a previous study. Comparison between the determined values and predicted values was used to assess the accuracy of prediction equations using the coefficient of determination (R²), root mean square error of prediction, concordance correlation coefficient (CCC), and mean bias (MB). The most accurate prediction was achieved in terms of R² and CCC for CAID of energy and DM (R² = 0.57 and 0.66, CCC = 0.45 and 0.47, respectively) as well as for IDC of N, starch, energy, and DM (R² = 0.90, 1.00, 0.65, and 0.66, CCC = 0.48, 0.97, 0.51, and 0.47, respectively). The R² and CCC values obtained for CAID of N, crude fat, starch, Ca, and P and IDC of Ca and P were not consistent with the expectation of predictive performance. The R² for IDC of crude fat was high (0.94), however, CCC was moderate (0.43). The determined MB values showed that some equations underpredicted (CAID and IDC of N, crude fat, starch, energy, and DM) and some overpredicted (CAID of Ca and P and IDC of P) the observed values of in vivo study. In conclusion, the equations obtained for CAID of energy and DM as well as IDC of N, starch, energy, and DM could be considered the best fit according to R² and CCC. Moreover, this study highlights the importance of validation with external data before applying each prediction equation to practical situations.