Analysis of production responses to changing crude protein levels in lactating dairy cow diets when evaluated in continuous or change-over experimental designs1

The objective of this study was to evaluate by meta-analysis the effect of experimental design on the production response functions obtained when changing crude protein (CP) levels in lactating dairy cow diets. The final database of studies meeting the selection criteria contained 55 publications with 23 classified as using a continuous (80 diets) and 34 classified as using a change-over (173 diets) experimental design (2 publications reported results from trials using both designs). Mixed model, weighted analysis of covariance was conducted on production measures in response to CP including the continuous covariates year of publication and average days in milk (DIM) and the discrete classification covariate of experimental design. The model was evaluated for curvilinearity in the response to CP, variance heterogeneity, and coincidence in the response between experimental designs, with P < 0.05 indicating significant effects and P < 0.15 indicating trends. On average, cows in experiments using continuous designs had a lower initial DIM, were on treatment longer, and produced a greater yield of milk and milk protein than cows in experiments using change-over designs. Production responses to increasing CP were increased dry matter intake (DMI), milk and component yield, and feed efficiency (milk Mcal/kg of DMI), and decreased milk N:intake N. Response in milk yield and feed efficiency to increased CP interacted with experimental design where continuous experiments had greater milk yield and feed efficiency response at higher levels of CP. Interaction between CP and design effects on yield of milk protein and fat or milk N:intake N did not approach significance. The database is limited by the lower number of continuous studies and the differences in average DIM between designs; nevertheless, it is concluded that DMI, milk protein yield, milk fat yield, and milk N:intake N responses to CP did not depend on experimental design. Response of milk yield and feed efficiency to CP interacted with experimental design; however, prospective research on the influence of experimental design is required to test these results.