Robust detection of outlying serum protein electrophoretic profiles in External Quality Assessment schemes

External Quality Assessment (EQA) schemes are national programmes to ensure that results obtained on a particular specimen in a given clinical laboratory are compatible with those obtained by other laboratories on the same specimen. EQA schemes are traditionally facing the problem of outlier detection of single analytes, e.g., glucose or cholesterol. Recently, serum protein electrophoresis was tested to be subsequently implemented on a regular basis in EQA programmes. This is a laboratory test which consists of five fractions (albumin, α1, α2, β and γ globulins) and the sum of them is equal to 100% of total proteins. So far, in EQA the five fractions have been analyzed separately as for the univariate analytes, without considering the correlation and the linear relationship between them. The minimum covariance determinant (MCD) estimator (Rousseeuw and van Driessen 1999) provides an interesting robust approach to analyze EQA electrophoresis results from a global standpoint. This talk focuses on comparing the results obtained by the traditional univariate approach and the robust MCD approach (Zhang et al. 2008). As illustrated on electrophoretic data from the Belgian EQA scheme, the MCD approach improves the detection of poor performing laboratories. Simple graphical representation of the outliers will be demonstrated at the end of the talk.