Submitted to: Journal of Dairy Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/2/2019
Publication Date: 3/27/2019
Citation: Zanton, G.I. 2019. Effect of experimental design on responses to two levels of metabolizable protein in multiparous dairy cows. Journal of Dairy Science. https://doi.org/10.3168/jds.2018-15730.
Interpretive Summary: For the inferences of experiments to accurately reflect what happens in actual practice, conditions of a study must accurately represent the conditions in which the data will be used in practice. Here, we evaluated whether the inferences made after cows changed between diets of adequate or low metabolizable protein as part of the experimental design were comparable to inferences made when cows were fed these diets continuously (i.e., actual practice). For the overwhelming majority of responses that were evaluated after four weeks on the diets, the inferences were not affected by the experimental design. The most prominent exception to this was the responses in milk fat, which were affected by the design. These results are important because most responses to diets with different levels of protein can be confidently observed over shorter feeding times and applied to conditions of longer-term feeding situations.
Technical Abstract: The objective of this research was to characterize the implications of changing between diets formulated to be adequate (ADMP) or low (LOMP) in metabolizable protein in a Latin Square (LSq) design or when fed the same diets continuously in a randomized complete block experimental design (RCBD). Fifty-four multiparous, early lactation cows (initial average±SD parity 2.8±0.9, 85.8±31 DIM, 715±63 kg body weight, 29.1±2.7 kg/d DMI, and 57.7±5.7 kg/d milk yield) were blocked by parity and DIM and were then randomly assigned to experimental design with 16 cows assigned to LSq and 38 cows assigned to RCBD. Cows within blocks in LSq were randomly assigned to sequence in a 4-sequence, 4-period, 2-treatment LSq balanced for the effects of previous treatment carry-over. Cows within blocks in RCBD were randomly assigned to dietary treatment, which was fed over the same 4, 28-d periods as the cows in LSq. Prior to the initiation of treatment diet feeding, cows were continued on the herd diet for 14 d to collect covariate measurements. Treatment diets were formulated with the intention of producing different animal responses to test the effects of treatment duration and experimental design. To achieve this goal, diets were similar in composition with the exception of exchanging an equal quantity of expeller soybean meal from ADMP (16.5% CP, 28.4% NDFOM, 27.6% starch) for soybean hulls in LOMP (14.6% CP, 31.1% NDFOM, 27.5% starch). Cows were individually fed treatment diets in a tie stall barn once daily for ad libitum consumption, water was available continuously for ad libitum consumption, cows were milked thrice daily, and rBST was administered every 14 d at the beginning and middle of each 28-d period. Milk yield, feed offered, and refused were measured daily, BW was recorded on 2 consecutive d each week, milk composition was measured at 6 consecutive milkings each week, spot samples of feces, urine, and blood were collected during the last week of each period and the covariate period. Experimental designs were analyzed separately using results from wk 4 of each period with mixed effects modeling where P < 0.05 was considered significant. DMI and milk fat yield were not affected by dietary treatment in either design, whereas milk and protein yield was greater for cows fed ADMP in both designs. Milk fat and protein percentage responses and milk energy output responses were different between designs. Milk fat yield and percent responses were affected by previous treatment carry-over in LSq. Metabolic and digestibility measures were very similar between designs with dietary responses generally in the direction and magnitude with literature predictions. Under the conditions of this experiment, inferences on N metabolism and the majority of production measurements were not affected by experimental design, with the principle exceptions of milk fat and protein percent and milk energy output.