Fossomatic Method of Somatic Cell Counting in Milk: Collaborative Study

Fossomatic, a fully automatic, fluoro-optoelectronic somatic cell counter, was evaluated in 6 laboratories. One set of 6 duplicate milk samples, fresh and preserved, was read with instruments as routinely calibrated to each laboratory’s direct microscopic somatic cell counts (DMSCCs), i.e., local standard. A second set was read after standardizing the instrument to a 3-sample common standard (DMSCC) shipped to each collaborator. Cell levels of the unknowns ranged from 371 to 2301 thousand (th)/ml by DMSCC, with an s per sample of 14.8% and s of a level mean of 1.8%. Fossomatic values averaged 1014 compared to 1031 th cells/ml by DMSCC, had homogeneous variance for duplicates within laboratories (s = 4.0%), and had an error per observation of 6.2%. The linear coefficient of 0.9966±0.0039 (for fresh milk and common standard) for the Fossomatic values regressed on DMSCC gave an excellent and unbiased estimate of the cell concentration in a milk sample. A small bias was evident in only 5 of 24 linear regressions (fresh or preserved within each standard) where slopes differed significantly from 1. All 5 were on values derived from a local standard with maximum biases of +18% and —28% (at a 1 million level). Preserved milk had slightly lower (P < 0.01) cell concentrations than fresh milk but only by 3.6% at 1 million. Although condition X level was significant (P < 0.01), the 6 cell levels differed by a magnitude of only 27— 45 th cells/ml. Between-laboratory variation (the component was 4.6%) and laboratory X level interaction were both significant, but their contributions to error were small. The spread in laboratories at any cell level was about 12% and the error of estimate for a sample sent to 2 laboratories was 6.8%. Although interactions were significant with this precise method, they were not of such a magnitude as to invalidate cell count estimates in practice. The Fossomatic method has been adopted as official first action.