Estimates of essential amino acid requirements from dose-response studies with rainbow trout and broiler chicken: Effect of mathematical model

A total of 37 dose-response experiments with essential amino acids performe d with rainbow trout and broiler chicken were re-evaluated with different m athematical approaches: an exponential model, a four-parameter logistic fun ction, the saturation kinetics model and the broken line approach. The diff erent approaches were compared both with regard to the goodness of fit (r(2 ) and s(y, x)) and with regard to the allowances which were derived regardi ng the optimal amino acid level in the diet. The experimental design, parti cularly the chosen range in dietary amino acid concentration was found to b e important for the comparison of models. Amongst the non-linear models, th e four-parameter logistic function and the saturation kinetics model appear ed superior to the exponential approach, when the range in dietary amino ac id concentration was very wide and included both a severely deficient basal level and a level that exceeded the needs of the animal by approximately t he factor 2. In these cases, allowances derived from individual experiments were considerably different depending on the model. The allowances based o n the exponential and the saturation kinetics approach were 27.7 and 20.7 g lysine/kg DM and 8.0 and 6.3 g methionine/kg DM, respectively, for rainbow trout. For other amino acids studied in rainbow trout the difference due t o model was less. Consequently, the predicted 'ideal protein' for rainbow t rout was considerably different depending on the model used. The maximum de viation found in different experiments with broiler chicken for the exponen tial vs. the saturation kinetics approach was 13.0 and 9.7 g lysine/kg and 11.4 and 8.2 g sulfur-containing amino acids/kg, respectively. However, the more restricted the range in dietary concentration was, the lesser became the differences between the different non-linear models. No definite recomm endation can therefore be extracted regarding the most suitable, generally applicable mathematical model.