Development of a model to predict dietary metabolizable energy from digestible energy in beef cattle

Understanding the utilization of feed energy is essential for precision feeding in beef cattle production. We aimed to assess whether predicting the metabolizable energy (ME) to digestible energy (DE) ratio (MDR), rather than a prediction of ME with DE, is feasible and to develop a model equation to predict MDR in beef cattle. We constructed a literature database based on published data. A meta-analysis was conducted with 306 means from 69 studies containing both dietary DE and ME concentrations measured by calorimetry to test whether exclusion of the y-intercept is adequate in the linear relationship between DE and ME. A random coefficient model with study as the random variable was used to develop equations to predict MDR in growing and finishing beef cattle. Routinely measured or calculated variables in the field (body weight, age, daily gain, intake, and dietary nutrient components) were chosen as explanatory variables. The developed equations were evaluated with other published equations. The no-intercept linear equation was found to represent the relationship between DE and ME more appropriately than the equation with a y-intercept. The y-intercept (-0.025 ± 0.0525) was not different from 0 (P = 0.638), and Akaike and Bayesian information criteria of the no-intercept model were smaller than those with the y-intercept. Within our growing and finishing cattle data, the animal's physiological stage was not a significant variable affecting MDR after accounting for the study effect (P = 0.213). The mean (±SE) of MDR was 0.849 (±0.0063). The best equation for predicting MDR (n = 106 from 28 studies) was 0.9410 ( ± 0.02160) +0.0042 ( ± 0.00186) × DMI (kg) - 0.0017 ( ± 0.00024) × NDF(% DM) - 0.0022 ( ± 0.00084) × CP(% DM). We also presented a model with a positive coefficient for the ether extract (n = 80 from 22 studies). When using these equations, the observed ME was predicted with high precision (R2 = 0.92). The model accuracy was also high, as shown by the high concordance correlation coefficient (>0.95) and small root mean square error of prediction (RMSEP), <5% of the observed mean. Moreover, a significant portion of the RMSEP was due to random bias (> 93%), without mean or slope bias (P > 0.05). We concluded that dietary ME in beef cattle could be accurately estimated from dietary DE and its conversion factor, MDR, predicted by the dry matter intake and concentration of several dietary nutrients, using the 2 equations developed in this study.