Comparison of in vivo and in silico growth performance and variability in pigs when applying a feeding strategy designed by simulation to control the variability of slaughter weight.

Variability in bodyweight (BW) among pigs complicates the management of feeding strategies and slaughter. Including variability among individuals in modelling approaches can help to design feeding strategies to control performance level, but also its variability. The InraPorc model was used to perform simulations on 10 batches of 84 crossbred pigs each to characterise the effect of feeding strategies differing in amino acid supply or feed allowance on the mean and variation in growth rate. Results suggested that a feed restriction reduces the coefficient of variation of BW at first departure for slaughter (BW1) by 34%. Growth performance obtained from an in silico simulation using ad libitum and restricted feeding plans was compared with results obtained in an in vivo experiment on a batch of 168 pigs. Pigs were offered feed ad libitum or were restricted (increase in feed allowance by 27 g/day up to a maximum of 2.4 and 2.7 kg/day for gilts and barrows, respectively). A two-phase feeding strategy was applied, with 0.9 and 0.7 g of digestible lysine per MJ of net energy (NE) in diets provided before or after 65 kg BW, respectively. Actual growth was similar to that obtained by simulation. Coefficient of variation of BW1 was similar in vivo and in silico for the ad libitum feeding strategy but was underestimated by 1 percentage point in silico for the restriction strategy. This study confirms the relevance of using simulations performed to predict the level and variability in performance of group housed pigs. Variability of bodyweight among pigs complicates management decisions related to feeding and slaughter; growth models that account for variability can be used to test and to propose feeding strategies to control this variability. This study demonstrates this concept and confirms its applicability in an in vivo experiment. Decision support tools based on such a modelling approach can be developed to propose feeding strategies allowing reaching target levels of performance in many fields (growth rate, economy and environmental impact). [ABSTRACT FROM AUTHOR]