Development and validation of a grouping method for pesticides analysed in foodstuffs

The cost effective monitoring of hundreds of pesticide residues in foodstuffs of vegetable origin involves the definition of groups of analytes associated with homogeneous performance parameters of the used analytical method. The performance of the analytical method only needs to be tested on a daily base for some of these group’s compounds. This work proposes a strategy for defining groups of compounds with homogeneous physical–chemical properties based on the evaluation of the similarity of the multivariate pattern of five of these properties namely: molar mass, melting point, vapour pressure at 20 °C, n-octanol–water partition coefficient and solubility in water at 20 or 25 °C. Three independent and competing multivariate analysis tools, namely Principal Component Analysis, Hierarchical Clustering and K-Mean Clustering were used. This strategy was successfully used to group 100 pesticides into nine clusters, with 1–40 pesticides, represented by a compound with within group average properties. The developed grouping method was validated using physical–chemical data from other references or compounds and produced groups consistent with the performance of the studied determination of pesticide residues in dried red bean. The mean analyte recoveries of groups with larger dimension are statistically different for a confidence level of 95%. The within group coefficients of variance of mean analyte recoveries of larger groups ranged from 6.7% to 8.5%. This grouping method can reduce the number of recovery tests performed for the validation of the analytical method or test quality control.