Prediction of cow pregnancy status using conventional and novel mid-infrared predicted milk traits

The objective of this study was to determine the ability of conventional milk cow characteristics and novel traits predicted by mid infrared (MIR) obtained from milk recording to predict the pregnancy status once the cow was inseminated. Conventional milk recording, spectral, and reproductive data collected in Luxembourg Hoslteins between 2008 and 2010 were used. Cows were defined as pregnant if they were positively checked and calved between 267 and 295 d later after the last AI or if they had calved between the later intervals when no checks were recorded. Pregnant or not within 3 intervals after last AI (<=35 d, 45-60 d, and 60-90 d) was modeled using logistic regression models firstly as a function of conventional cow milk characteristics and extended to fatty acids as novel traits predicted by MIR in a second step. The lactation curve characteristics for milk, fat, protein, and lactose yields were estimated using modified best prediction method. Test-day fatty acid contents were estimated from collected MIR spectra using an appropriate calibration equation. Two third proportion and one third of the whole data set were randomly selected for calibration and validation models respectively. The relation between the predicted and observed probabilities of cow pregnancy was approximately linear for calibration and validation models. The sensitivity-specificity combination for cow pregnancy increased when fatty acids were added to conventional milk characteristics as inputs to the different models (from 78 to 85% for sensitivity and from 40 to 52% for specificity). Results based on those models showed that it would be possible to help breeders to manage cow fertility using such tool implemented in the milk recording organizations.