Mathematical approaches to detect low concentrations in progesterone profiles.

There is a general need for higher objectivity and accuracy in describing the physiological fertility performance of dairy cows. To develop the alternative meaningful starting points for the selection of genetically superior dairy cows, this study focused on the detection of low progesterone concentrations, which are indicative of estrus events. Three mathematical approaches were used: one based on the exponentially weighted moving average control chart, and two threshold methods, which were developed in-house. Data were collected from one data set that included 97 insemination data of first-lactating Holstein dairy cows, and a second set that included 160 inseminations of primiparous and multiparous Holstein dairy cows. On the basis of these 2 data sets, and using a threshold of 1.2 ng progesterone/ml skimmed milk, the sensitivity of the 3 models was high and ranged between 100% and 93.13%, with an error rate between 4% and 22.17%. The specificity varied between 97.92% and 99.93%. The average concentration levels of true-positive-detected progesterone measures were low and ranged between 0.18 and 0.28 ng progesterone/ml skimmed milk (first data set) and 0.21 to 0.26 ng progesterone/ml skimmed milk (second data set). False-positive-detected low progesterone concentrations during estrus events were closely related to progesterone values around the 1.2 ng progesterone/ml skimmed milk threshold and the detecting rules of the control chart. Thus, we suggest that a threshold of 0.8 ng progesterone/ml skimmed milk is indicative for luteal activity in defatted foremilk. By means of the three methods used, the detection of low progesterone concentrations was possible and it can be assumed that this is a good starting point for further studies (such as interval calculation) in this area.