Estimates of variability for somatic cell count measurements in the Iowa dairy industry

The SCC collected from 15 instruments in 12 laboratories were used to quantify accuracy, repeatability, and reproducibility in the Iowa dairy industry. For each of three trials, milk was sampled at the morning milking from 30 different Holsteins in the Iowa State University herd. Identified samples and unidentified duplicates were provided for each participating instrument. Mean SCC was 418,000 cells/ml, and mean SCC for duplicates ranged from 9000 to 3,966,000. Accuracy for a set of 30 duplicates was lowest for trial 1 (CV = 16.4%) and highest for trial 2 (CV = 7.6%). Intraclass correlations estimated repeatability and were .99 for all but one instrument. Coefficients of variation for repeatability (weighted mean = 11.4%) were similar to estimates for accuracy (weighted mean = 11.0%), but reproducibility was considerably lower (30.0%). Samples were classified by SCC as very low, < 125,000; low, 125,000 to 249,000; medium, 250,000 to 500,000; and high, > 500,000. Repeatability for high samples was higher than repeatability for very low SCC samples; coefficient of variation for high SCC samples was 6.8% but was > 25.0% for samples with < 500,000 cells. Repeatability was within standards set by the industry, but current procedures for quality control may not adequately address reproducibility.