Interquarter comparison of markers of subclinical mastitis: somatic cell count, electrical conductivity, N-acetyl-beta-glucosaminidase and antitrypsin.

We have evaluated the usefulness in mastitis diagnostics of new diagnostic measures using the continuous nature of the variation in quarter milk samples of four inflammatory markers: somatic cell count, N-acetyl-beta-glucosaminidase (EC 3.2.1.30), electrical conductivity and antitrypsin. A two-way analysis of variance was performed for each of the inflammatory markers on results from 273 cows sampled six times at 14 d intervals. Adjustment for the individual cow's mean removed approximately half the total variation for all four markers, indicating that comparisons in marker levels between quarters without taking the cows' overall level into account, as is done in diagnostic keys based on fixed threshold, is a rash procedure. After adjusting for the individual cow means, the total variation of the marker observations was partitioned in three sources of variation: between the four quarter levels, between different sampling days, and the experimental error, i.e. the variation within each quarter over the study period. As expected, the variation between quarters accounted for the largest percentage of the variation (44-65%). In contrast the variation between days was only 16-34% and of the same magnitude as the experimental error. These figures can profitably be taken into account when designing experiments in future mastitis research studies. They imply that mastitis diagnostics based on quarter milk samples collected over time will be influenced only slightly by random biological variation and that the same relative differences that exist in marker levels between the four quarters of a cow as a rule will be disclosed irrespective of sampling day. Thus, the udder health state is relatively stable when determined by these inflammatory markers. In addition to simple correlations between markers measured on the same day, correlations of the logarithm of the interquarter ratios and of the experimental errors were computed. Stronger associations between the markers emerged and these correlations reflect more precisely than simple correlations the real diagnostic relations of the markers.