Evaluation of a Commercial Model for Predicting Growth Performance of Pigs with Varying Diet Composition and Stocking Density

The objective of this study was to validate the growth performance predictions of a commercial model (CAMERA®; INTL FCStone, New York, NY) by comparing the optimally-solved estimates with observed growth performance from published studies. Three studies were selected to create 3 feeding scenarios: 1) variation in dietary energy concentrations and fiber sources (Nitikanchana et al., 2015); 2) variation in dietary lysine level (Menegat et al., 2017); and 3) variation in space allowance (Carpenter et al., 2018). For each validation scenario, the growth performance of pigs from the best-performing treatment group was first estimated, calibrated using the observed performance, and used as the baseline for the prediction of other treatment groups. The model estimates were then compared with the observed growth performance to determine the prediction accuracy: deviation, % = (estimated value – observed value)÷ observed value × 100. Results from scenario 1 indicated that the model-estimated final body weight (BW), average daily gain (ADG), average daily feed intake (ADFI), and feed-to-gain ratio (F/G) were reasonably close to observed performance for pigs fed medium energy with 8.7% neutral detergent fiber (NDF) as well as those fed low energy with 22.1% NDF. For pigs fed medium energy with 16.1% NDF, the model accurately estimated final BW and ADG (0.5 and 1.1% deviation, respectively), but overestimated ADFI and F/G (3.3 and 2.1% deviation, respectively). The model underestimated the ADG and ADFI (-2.6 and -3.8% deviation, respectively) and overestimated the F/G (5.0% deviation) of pigs fed low energy with 16.4% NDF. Carcass yield differences were not accurately captured by the model among pigs fed various NDF levels. For validation scenario 2, model-estimated growth responses were generally underestimated (-5.5% deviation) and were not sensitive to changing dietary lysine levels. For validation scenario 3, the model accurately predicted final BW and ADG (\u3c 0.9% deviation) but overestimated ADFI and F/G (3.6% deviation) of pigs allowed restricted space. Model- estimated growth responses were generally accurate for pigs that received increasing space by gate adjustment or pig removal, except for an overestimation of ADFI (3.0% deviation) for the pig removal treatment. In summary, the commercial model was able to capture changes in growth performance of pigs that received various dietary energy and fiber concentrations as well as pigs with changing space allowance. However, the model was not able to predict carcass yield in response to changes in dietary fiber nor the differences in growth performance due to variation in dietary lysine. For model improvements, the consistency of prediction accuracy and ease of user operability should be enhanced