Comparison of Individual Food Analyses to Composites for Pyrethroid Pesticides

Abstract

PP-30-098 Background/Aims: Dietary exposure can be measured through analysis of foods that comprise the intake for an individual over a specified period. Collection and analysis of many individual foods can be too expensive. To minimize cost and participant burden while maximizing the quality of information obtained from the data, experiments were conducted to facilitate comparison of pyrethroid pesticides concentrations in composite and individual food samples. This would determine whether composite food samples would give comparable results, and thus allow for a more cost-effective analytical scheme. Methods: Using foods commonly consumed and collected from 9 participants from a well-defined community, assessments of various compositing schemes were made using extant food consumption and pesticide residue data from What We Eat in America, National Health and Nutrition Examination Survey, and Pesticide Data Program databases. Simulation experiments used the expected distribution of pesticides in foods to design an effective scheme to yield 20 composites from 67 individual samples. The composite scheme was largely driven by 4 food types that, according to the Pesticide Data Program database, were expected to have high levels of permethrin: tomatoes, apples, broccoli, and lettuce. Using this information, the expected amount of permethrin in each individual collected sample was estimated, and then proportioned into the 20 composites. Results: The analysis of the 67 individual and 20 composite samples indicated that piperonyl butoxide was detected in 16 of 20 composite samples, bifenthrin was detected in 11 composites, and permethrin and esfenvalerate were detected in 10 composites. These pyrethroids were also detected in the individual samples at about the same percentages. Conclusion: The a priori determination of permethrin using extant databases was useful in creating the composites, but was unable to ensure measurable levels detected in every sample. Therefore, using extant databases alone should not only replace actual measurements, but should also prove to be valuable in developing a compositing scheme that yields similar numbers of detectable levels.