Abstract
Ethanol was assayed by an average of 200 participants in the UK National External Quality Assessment Scheme, in 26 samples of human serum containing 0·1% fluoride/oxalate and added ethanol from 0·2 to 4·5 g/L. Outliers greater than three standard deviations from the consensus mean for any sample were excluded. Data remaining were grouped by technique and the technique mean and standard deviation calculated. Inter-laboratory variation of 13 technique groups was assessed by the coefficient of variation of measurements and bias from the per cent difference of the technique mean from the target value. Gas chromatography (GC) with packed columns and Sigma alcohol dehydrogenase assay protocols that included a sample deproteinization step, showed better between-laboratory agreement but greater bias. The least variable techniques were headspace analysis with GC-packed columns, Kodak Ektachem, bioMérieux and DuPont aca assays. A significant negative bias was produced by Kodak Ektachem and a positive bias by the Lion alcometer which was the most variable technique.