Negative energy balance is a known risk factor for decreased fertility in dairy cows. This study evaluated the accuracy of plasma concentrations of nonesterified fatty acids (NEFA), β-hydroxybutyrate (BHBA), and insulin-like growth factor 1 (IGF-1)—factors related to negative energy balance—in predicting decreased fertility. One plasma sample per cow was collected from 480 cows in 12 herds during the period from d 4 to 21 in milk and analyzed for NEFA, BHBA, and IGF-1. For each cow, data on breed, parity, calving date, gynecological examinations, and insemination dates were obtained. Milk samples from 241 cows in 7 of the participating herds were analyzed for progesterone concentration to define the first day of luteal activity. The diagnostic sensitivity (Se) and specificity (Sp) at different cut-off concentrations of NEFA, BHBA, or IGF-1 were calculated and related to individual cow fertility status, measured as anestrus (ANEST), delayed first artificial insemination (DFAI), or delayed conception. Positive and negative predictive values (PV+; PV−) were calculated considering different levels of (within-herd) prevalence. Strata (i.e., subgroup)-specific Se and Sp and associations between test results and fertility parameters were investigated using logistic regression. The NEFA and BHBA tests for ANEST and DFAI had the highest combined Se and Sp and were thus evaluated further. Cut-off values with Sp around 80% were used in this step to provide a reasonable number of test-positive cows, representing a practical situation. This corresponded to a cut-off value for the NEFA test of 400 µEq/L (Se 0.27–0.45) and for the BHBA test of 1.8 mM (Se 0.15–0.30) across all cows included in the study. The estimated Sp was generally higher than the original 80%, but the corresponding Se was further decreased when the test was used in heifers compared with older cows. The true prevalence of ANEST in the study population was 27%, which gave a PV+ of 0.36 to 0.45 and a PV− of 0.76 to 0.79. With 35% true prevalence of DFAI, PV+ was 0.29 to 0.38 and PV− was 0.64 to 0.66. Thus, overall test performance was low when metabolic indicators measured as single values in early lactation were used to predict fertility in dairy cows, but accuracy was influenced by cow-level factors such as parity. The prevalence of the target condition (in this case, decreased fertility) also influences test usefulness and should be considered when planning test systems and interpreting test results.
Journal of Dairy Science, 2012, Vol 95, Issue 12
sensitivity; specificity; negative energy balance; two-graph receiver operator characteristic