Estimation of measurement uncertainty in pesticide residue analysis.

Proficiency test results from 5 countries involving 61 separate interlaboratory proficiency tests for pesticide residues were examined in this study. A total of 24 different matrixes and 869 relative standard deviations of the mean (or median) pesticide residue concentration were statistically evaluated in relation to the Horwitz function. The aim was to determine whether or not the concentration-dependent relationship described by Horwitz would hold for the much narrower range of chemicals and concentrations covered in routine pesticide residue analysis. Although for fatty (animal-derived) matrixes the variability increased as the concentration decreased in line with the Horwitz equation, the between-laboratories relative standard deviations for nonfatty matrixes (fruit, vegetables, and grain) remained at 25% over the entire concentration range of 1 microg/kg to 10 mg/kg for the pesticides studied. Given these findings, the Horwitz equation remains valid for calculating uncertainties involving pesticide residues in fatty matrixes. However, for pesticide residue analyses involving nonfatty matrixes, a constant relative standard deviation of 25% is more appropriate for calculating uncertainties, particularly when a reported result is assessed against a regulatory limit.