EUROPOEM, A Predictive Occupational Exposure Database for Registration Purposes of Pesticides

Abstract

For registration of agricultural pesticides, the risks for humans, animals, and the environment must be determined. The risk assessment is based on an appraisal of the levels of exposure and the hazards of the active substance(s) in the plant protection product, that is, the agricultural pesticide. Funded by the European Commission (AIR3 CT93-1370), the EUROPOEM database has been developed by a group of experts, representing governments, industry, and academia. The currently available exposure database reflects exposure to operators (mixer/loaders and applicators). The EUROPOEM approach is based on a harmonized protocol for conduct of field studies of operator exposure (presently published as an Organization for Economic Cooperation and Development [OECD] Guidance Document) and a tiered approach to exposure and risk assessment. The database is constructed from exposure data obtained in representative field studies. These field studies are considered according to criteria reflecting the quality of documentation, study design, adequate methodology, number of replicates, and QA/QC elements, for use of the inhalation and dermal exposure data. The resulting exposure data were combined according to comparable use scenarios. From the resulting databases typical surrogate potential exposure values have been obtained, which are determined by their use for either acute or chronic health effects, and the size of the database. For large databases (over 50-100 data points), from many different field studies (10 or more), the 75th percentile is taken if the exposure is considered leading to chronic effects. For smaller databases, a more conservative 90th percentile is taken as surrogate value, or none at all for very small databases (15-20 or less data points from 3 or less different field studies). The choice for the 75th percentile is based on the assumed or observed lognormal distribution of the exposure data, as being the most relevant typical value for long-term effects, since the 75th percentile of log-normal distributions is nominally very similar to a calculated arithmetic mean (AM). The AM, as such however, is irrelevant for log-normal distributions.