Kronfeld D S, Donoghue S, Copp R L, Stearns F M, Engle R H
J Dairy Sci. 1982 Oct;65(10):1925-33. doi: 10.3168/jds.S0022-0302(82)82440-5.
A study of 395 cows in 21 herds tested the suitability of applying the normal distribution to 23 blood variables in setting reference ranges and the potential of multiple regression equations for predicting energy and nutrient intakes, milk and fat production, and reproductive variables from blood measurements. After adjustment for herd and lactational group within herd, deviations from the normal distribution were significant for nine blood variables: glucose, cholesterol, glycerol, triglyceride, acetoacetate, urea, uric acid, creatinine, and chloride. Multiple regressions were calculated for group-averages of ration variables on group-medians of blood variables, and for production and reproductive variables on blood variables in individual cows. Coefficients of determination were .06 to .14 for milk production and reproductive variables and .55 to .79 for ration variables. Best serum predictors of ration variables were glutamic-oxaloacetic and glutamicpyruvic transaminases and cholesterol, phosphorus, triglyceride, and globulin. These results suggest that the normal distribution should not be assumed for blood variables: instead it should be tested. Multiple regressions on blood variables may be useful in predicting energy and nutrient intakes, and regression analysis may help in selection of blood measurements for inclusion in profiles.
