262
A Proficiency Test System to Improve Labratory and Method Performance and Produce Reference Values for Component Calibration Samples for Infrared Milk Analysis

Tuesday, July 22, 2014: 3:15 PM
3501D (Kansas City Convention Center)
David M Barbano , Cornell University, Ithaca, NY
Karen L Wojciechowski , Cornell University, Ithaca, NY
Caterina Melilli , Northeast Dairy Foods Research Center, Ithaca, NY
Abstract Text: Our objectives were: 1) utilize data from 10 to 12 labortories running duplicate chemical analyses of 14 milk samples with an orthogonal matrix of fat, protein, and lactose concentration  to calculate all-lab mean reference values for fat, protein, lactose, and total solids for each sample, 2) monitor the analytical performance of the reference chemical analysis methods, and 3) evaluate and improve the analytical performance of individual laboratories conducting chemical reference methods. The chemical reference methods used were Mojonnier either extraction for fat, Kjeldahl for true protein, spectrophotometric enzymatic assay for anhydrous lactose, and forced air oven drying for total solids. Statistical outliers were removed and all-lab mean reference values and within and between laboratory variation (i.e., Sr and SR) were calculated for each sample for each component. The set of 14 milk samples with all lab mean reference values was used to run diagnositic performance evaluation and calibrate infrared milk analyzers. The proficiency of each lab was evaluated utilizing Z-scores, Pareto diagrams, and Euclidian distance (ED) plots.  Performance of the ether extraction, Kjeldahl true protein, and oven drying total solids methods were improved (P < 0.05), due to the improved ability to identify and trouble shoot between laboratory differences in results when using the modified milks with the orthogonal design of variation in composition.  Residual plots of an individual lab’s results minus the all-lab mean for each sample provides an evaluation of individual lab bias and proportional deviations from the all-lab mean over a wide range of concentration of each component.  Residual and ED plots allow an experienced evaluator to identify the sources of poor performance by an individual lab and recommend corrective measures.  Over a period of years, the feed back and method performance trouble shooting based on residual and ED plots has improved within (Sr) and between (SR) laboratory and method performance.  

Keywords: Proficiency Testing, Infrared Milk Analysis, Method Performance