Improving estimates of weight gain and residual feed intake by adjusting for the amount of feed eaten before weighing

In Australia, a trait under consideration for genetic selection to improve feed efficiency is residual feed intake (RFI), which is defined as the amount of feed eaten by an animal less what would be expected from the animal's growth rate and body weight. Accurate estimates of RFI therefore require accurate estimates of weight gain. Results presented here on steers finished in a feedlot to liveweights of 540 or 600 kg show that, when feed intake is being measured, weight gain can be estimated more accurately using the amount of feed eaten in the previous 3-5 days (as an adjustment for gut fill) than if feed eaten in the 80 h before weighing is ignored. This is demonstrated by a much lower residual mean square from modelling the weight of each animal as a quadratic growth curve over time if terms are included for feed eaten on the current and previous 3-5 days. An analysis of measurement errors associated with fitting the equation used to calculate RFI: Feed intake = constant + beta (w) x mean metabolic weight + beta (g) x weight gain + error (i.e. RFI) (1) indicates that the relatively high measurement errors associated with weight gain but comparatively low measurement errors associated with metabolic weight will result in upward biases in the partial regression coefficient beta (w) and downward biases in beta (g). For example, in a 105-day feed intake test of 44 steers (mean start/end weights 440/600 kg), the estimate of beta (g) was 1.26 based on weight gain estimated by a simple linear regression of each animal's weight over time (LIN), compared with 2.20 using weight gain estimated from the difference between first and last weight of each animal adjusted for the amount of feed eaten on the current and previous 5 days (DIFFadj). From a shorter test, based on weight gains from day 15 to 50 in the automatic feeder pens, the estimate of beta (g) was 0.40 using LIN and 1.67 using DIFFadj. These results illustrate the potential magnitude of the downward bias in beta (g) if inaccurate estimates of weight gain are used to fit equation 1. The higher estimates for beta (g) obtained using DIFFadj may still have some downward bias but are closer to the theoretical values published by SCA (1990) for the amount of metabolisable energy required for weight gain. Adjusting for the amount of feed eaten before weighing therefore increased the accuracy of estimated weight gain and reduced the biases in beta (g) and beta (w), so providing better and more stable estimates of residual feed intake.