Evaluation of nutrition models to estimate performance of young dairy calves: a meta-analytical study under tropical conditions

Mathematical models are important tools to estimate nutritional requirements and animal growth. Very few calf models generated from other countries with different feeding programs, environment and production systems have been evaluated. The objective of this paper is to evaluate two calf models: (i) the National Research Council (NRC) in 2001 and (ii) the updates published by Van Amburgh and Drackley in 2005 and inputted into Agricultural Modeling and Training Systems (AMTS, version 3.5.8). Data from 16 previous studies involving 51 diets for dairy calves under tropical conditions (n=485 calves, initial BW 37.5±4.35 kg and weaning weight of 62.0±10.16 kg) were used. The calves were fed with whole milk, milk replacer or fermented colostrum, plus starter (20.9±1.78% of CP). The accuracy of the average daily gain (ADG) prediction was evaluated by mean bias, mean square prediction error (MSPE), concordance correlation coefficient, bias correction factor (Cb), and regression between the observed and predicted values. The ADG observed from birth to weaning was 0.452±0.121 kg/day. Calves fed with whole milk had greater ADG compared with calves fed milk replacer (0.477 v. 0.379 kg/day) during the milk-feeding period. When all data were pooled (n=51 diets), predictions had a mean bias of ?0.019 and 0.068 kg/day for energy-allowable gain using NRC and AMTS models, respectively. The regression equation between observed and predicted values obtained from energy of diets showed an intercept different from zero (P<0.0001) and slope that differed from unity (P<0.0001). In a second evaluation, when calves were fed only milk replacer, the energy-allowable gain from AMTS showed the lowest mean bias (0.008 kg/day) and 82.1% of the MSPE value originated from random errors. The lowest MSPE, the higher Cb value and no significant slope bias (P>0.05) indicate that the AMTS growth model resulted in accurate predictions for calves fed with milk replacer. However, within these latter two approaches, the goodness of fit (R²) was low, representing low precision. The weight gain estimated by the energy available from the diet was overestimated by 19 g/day when calculated by the NRC and underestimated by 68 g/day when calculated by AMTS. The reasons for this discrepancy need to be understood, for only then new models could be developed and parameterized to estimate animal performance in tropical conditions more accurately and precisely.