Data presented for these variables were back-transformed. Frequency analysis was performed using the FREQ procedure of SAS considering the binary response variable SNAP test using the chi-squared probability. Finally, multivariable logistic mixed models (PROC GLIMMIX) considering the binary outcome variable and histological hepatocyte inflammation score were constructed. Treatments were forced into the models. Cow was considered as a random effect. Cows that developed illness (n = 1 CON) during the measurement phase (d 14–21) were removed from the data set. Statistical significance was declared at P ≤ 0.05 and trends were declared at 0.05 < P ≤ 0.10.