Predicting colostrum quality from performance in the previous lactation and environmental changes

Nine New Hampshire Holstein dairies contributed to a study to investigate if colostrum quality could be predicted by cow performance in the previous lactation and by environmental factors during the 21-d prepartum period. The numbers of days below 5°C (D<), days above 23°C (D>), and days between 5 and 23°C (D) were used in the development of the regression equation. Between 2011 and 2014, 111 colostrum samples were obtained and analyzed for IgG. Producers recorded cow identification number, calf date of birth, sex of the calf, colostrum yield, hours from parturition to colostrum harvest, and weeks on pasture during the dry period (if any). Dairy Herd Improvement data from each cow and weather data were compiled for analysis. Information accessed was predicted transmitting abilities for milk, fat (PTAF), protein (PTAP), and dollars; previous lactation: milk yield, fat yield, fat percent, protein percent, protein yield, somatic cell score, days open, days dry, days in milk, and previous parity (PAR). Colostrum yield was negatively correlated with IgG concentration (r=-0.42) and D (r=-0.2). It was positively correlated with D> (r=0.30), predicted transmitting ability for milk (r=0.26), PTAF (r=0.21), and PTAP (r=0.22). Immunoglobulin G concentration (g/L) was positively correlated with days in milk (r=0.21), milk yield (r=0.30), fat yield (r=0.34), protein yield (r=0.26), days open (r=0.21), PAR (r=0.22), and tended to be positively correlated with DD (r=0.17). Immunoglobulin G concentration (g/L) was negatively correlated with D> (r=-0.24) and PTAF (r=-0.21) and tended to be negatively correlated with PTAP (r=-0.18). To determine the best fit, values >0 were transformed to natural logarithm. All nontransformed variables were also used to develop the model. A variance inflation factor analysis was conducted, followed by a backward elimination procedure. The resulting regression model indicated that changes in Ln fat yield (β=2.29), Ln fat percent (β=2.15), Ln protein yield (β=-2.25), and Ln protein percent (β=2.1) had largest effect on LnIgG. This model was validated using 27 colostrum samples from 9 different farms not used in the model. The difference between means for actual and predicted colostrum quality (IgG, g/L) was 13.6g/L. Previous lactation DHI data and weather data can be used to predict the IgG concentration of colostrum.