Factors Contributing to Intra-Individual Variation of Serum Constituents: 3. Use of Randomized Duplicate Serum Specimens to Evaluate Sources of Analytical Error

Abstract

Abstract In study of intra-individual variation of serum constituents, we used analysis of variance to separate biological from analytical variation, the latter being estimated from results for duplicate specimens. However, the replication error term in such a model may grossly overestimate instrumental variability, especially when a significant pre-instrumental error component is present. To separate pre-instrumental and instrumental error, we used the following experimental design: Duplicate serum specimens were obtained from 88 healthy men, the blood being collected at one venipuncture and immediately divided into two portions. The duplicates were randomized and analyzed on one occasion with the "AutoChemist Multi-Channel Analyzer," and pre-instrumental variation was estimated from the results. Instrumental variation was estimated from results for three independent populations with widely different mean analyte concentrations; these samples were analyzed at the same time as the samples just mentioned. Of the 20 serum constituents assayed, sodium, potassium, iron, total lipids, aspartate aminotransferase, alanine aminotransferase, lactate dehydrogenase, and total bilirubin showed significant pre-instrumental error components, as evidenced from significant χ2-tests comparing analytical and "duplicate sample" variation. For total serum protein, albumin, total lipids, cholesterol, and alkaline phosphatase, we found a significant difference (by paired t-test) between values for the first and second specimen.