Under EU legislation (Article 32, Regulation (EC) No 396/2005), EFSA provides an annual report which analyses pesticide residue levels in foods on the European market. The analysis is based on data from the official national control activities carried out by EU Member States, Iceland and Norway and includes a subset of data from the EU-coordinated control programme which uses a randomised sampling strategy. For 2018, 95.5% of the overall 91,015 samples analysed fell below the maximum residue level (MRL), 4.5% exceeded this level, of which 2.7% were non-compliant, i.e. samples exceeding the MRL after taking into account the measurement uncertainty. For the subset of 11,679 samples analysed as part of the EU-coordinated control programme, 1.4% exceeded the MRL and 0.9% were non-compliant. Table grapes and sweet peppers/bell peppers were among the food products that most frequently exceeded the MRLs. To assess acute and chronic risk to consumer health, dietary exposure to pesticide residues was estimated and compared with health-based guidance values. The findings suggest that the assessed levels for the food commodities analysed are unlikely to pose concern for consumer health. However, a number of recommendations are proposed to increase the efficiency of European control systems (e.g. optimising traceability), thereby continuing to ensure a high level of consumer protection.

Practical "top-down" approaches appear to be the most suitable for the evaluation of measurement uncertainty in pesticide residue testing laboratories, where analytical procedures are routinely applied to a large number of pesticide/food combinations. The opposite approach, "bottom-up" evaluation of measurement uncertainty, leads to great difficulties in evaluating all of the pesticides in a consistent way. Among the top-down approaches, there are two main ways in which measurement uncertainty can be estimated: One is based on default values, which are based on previous extensive interlaboratory experience and the proven accuracy of the laboratory; these include the Horwitz equation or the fit-for-purpose relative standard deviation (FFP-RSD). The other is based on experimental data from the quality control work of the laboratory: within-laboratory reproducibility, interlaboratory validation, or a combination of results obtained in proficiency tests. The principal existing guidelines from various bodies (Eurachem, Nordtest, and Eurolab) all propose different approaches for calculating measurement uncertainty. In this paper, the main top-down approaches are evaluated and compared using the data from the European Proficiency Test Database for Fruits and Vegetables and the Multiresidue Method validation databases obtained from the National Reference and Official Laboratories in Europe. The main conclusion of the comparative study is that a default expanded measurement uncertainty value of 50% could satisfy all of the requirements for facilitating and harmonizing, worldwide, the intercomparability of the pesticide residue confidence results between laboratories.

The obligation for accredited laboratories to participate in proficiency tests under ISO 17025, performing multiresidue methods (MRMs) for pesticide residues, involves the reporting of a large number of individual z scores making the evaluation of the overall performance of the laboratories difficult. It entails, time and again, the need for ways to summarise the laboratory's overall assessment into a unique combined index. In addition, the need for ways to continually evaluate the performance of the laboratory over the years is equally acknowledged. For these reasons, following 14 years of the European Union Reference Laboratory for Pesticide Residues in Fruits and Vegetables (EUPT-FV), useful formulas have been designed to globally evaluate the assessment of the participating laboratories. The aim is to achieve a formula which is easy to understand, which can be applied and which fits the purposes of long-term evaluation detecting positive and negative trends. Moreover, consideration is needed for a fair compensation of bad results in MRM, taking into account the large number of compounds that are covered. It is therefore important to be aware of the difficulties in getting satisfactory values from a wide range of compounds. This work presents an evaluation of the main well-established combined z score formulas together with those new ones developed here which have been applied to the European proficiency test results (EUPTs) over the years. Previous formulas such as the rescaled sum of z score (RSZ), the sum squared of z score (SSZ) and the relative laboratory performance (RLP) are compared with the newer ones: the sum of weighted z scores (SWZ) and the sum of squared z scores (SZ2). By means of formula comparisons, conclusions on the advantages, drawbacks and the most fit-for-purpose approach are achieved.

No abstract