Predictive equations for dietary energy are improved when independently developed for dry and wet food which could benefit both the pet and the environment.

Abstract

Measuring energy availability through metabolizable energy feeding studies is the "gold standard" for establishing metabolizable energy concentration. However, predictive equations are often used to estimate metabolizable energy in dog and cat pet foods. The goal of this work was to evaluate the prediction of energy density and compare those predictions to each other and the energy needs of the individual pets. Feeding studies used 397 adult dogs and 527 adult cats on 1,028 canine and 847 feline foods. Individual pet results for the estimate of metabolizable energy density were used as outcome variables. Prediction equations were generated from the new data and compared to previously published equations. On average the dogs consumed 747 kilocalories (kcals) per day (SD = 198.7) while cats consumed 234 kcals per day (SD = 53.6). The difference between the average prediction of energy density and the measured metabolizable energy varied from the modified Atwater prediction 4.5%, 3.4% (NRC equations), 1.2% (Hall equations) to the new equations calculated from these data at 0.5%. The average absolute values of the differences between measured and predicted estimates in pet foods (dry and canned, dog and cat) are: 6.7% (modified Atwater), 5.1% (NRC equations), 3.5% (Hall equations) and 3.2% (new equations). All of these estimates resulted in significantly less variation in the estimate of the food expected to be consumed than the observed variation associated with actual pet consumption to maintain body weight. When expressed as a ratio of energy consumed to metabolic body weight (weight in kilograms3/4) the within species variation in energy consumed to maintain weight was still high as compared to the energy density estimates variance from measured metabolizable energy. The amount of food offered as the central point in a feeding guide, based on the prediction equations, would on average result in an average variance between 8.2% error in the worst case estimate (feline dry using modified Atwater estimates) and approximately 2.7% (the new equation for dry dog food). All predictions had relatively small differences in calculating food consumed when compared to the differences associated with the variation in normal energy demand.