Skip to main content
Food Safety

Technical Annex

A Harmonised Technical Approach on the Parameters Governing Retrospective Cumulative Exposure Assessment

Background and purpose

The assessment of cumulative and synergistic effects of pesticide residues was introduced in European legislation by Regulation (EC) No 396/2005. Between 2008 and 2013, EFSA has elaborated the methodologies to perform these assessments (see examples at the end of the page). These methodologies are flexible and can be customized to different assessment questions and protection goals. They can also be used in a tiered process, depending on the resources available and on the purpose of the assessment.

For an optimal use of these methodologies, large data sets for each input data and complete information should be available for all pesticide/commodity combination contributing to cumulative risks. Unfortunately, this is not the case and in case of missing data or missing information, certain assumptions need to be made.

Therefore, at the Standing Committee for Plants, Animals, Food and Feed - Section Pharmaceuticals, Pesticide Residues (SCoPAFF) on 18-19 November 2013, the European Commission set up a small expert working group with Member States to discuss risk management questions related to the methodology to assess cumulative effects of combined exposure to pesticide residues under development by the European Food Safety Authority (EFSA). The systematic implementation of these parameters would enable producing consistent and comparable assessments.

Since then, discussions have evolved in regular working group meetings. As a first priority, the methodology for the assessment of cumulative exposure based on monitoring data of previous years was developed, the so-called retrospective scenario. The main application of the retrospective scenario falls under Article 32 of Regulation (EC) No 396/2005 as the assessment will make use of the monitoring data collected by the Member States in the EU annual monitoring report on pesticides residues.

The approach to be followed was gradually agreed by EU Member States. A first set of parameters was agreed in the SCoPAFF of 12 June 2015 (Agenda item A.21) and a final set on 19 September 2018 (Agenda Item A.09).

The purpose of this document is to inform interested parties about the agreed approach that will be used by EFSA for the assessment of cumulative exposure based on previously collected monitoring data.

Agreed parameters for cumulative risk assessment using monitoring data (retrospective scenario)

The first step of any cumulative risk assessment is the identification of a toxicological effect which can plausibly be caused by multiple pesticides. Once such effects (e.g. hypothyroidism) are identified, pesticides causing these effects are included in cumulative assessment groups (CAGs) and characterised for the specific effects by the establishment of reference values.

In a second step, exposure to each CAG is calculated according to a stepwise approach, the so-called tiered approach, which is in line with EFSA’s Guidance. A 2-tiered approach is used. As a first step, in Tier 1, the parameters used are more generic leading to a conservative assessment. In Tier 2, the assessment is refined using input parameters that are more realistic, but still conservative. This approach was selected with a view of saving resources, as in case Tier 1 would not indicate any risk then the assessment would not need to continue with a Tier 2 evaluation, thus saving effort.

It was agreed that both tiers should be based on probabilistic modelling, which signals a departure from the currently used deterministic approaches. Deterministic assessments use single values or point estimates as inputs to the exposure calculations. Probabilistic assessments, however, are based on distributions of values. The methodology combines the distribution of consumption data (collected from consumer dietary surveys) for all food commodities (processed or composite foods are converted to Raw Primary Commodities (RPCs)), and the distribution of pesticide residue concentrations found on the RPCs during the 3-year monitoring cycles of the EU Multi-Annual Control Programmes (EU MACP). The output is the distribution of consumer’s exposures to pesticides residues where individual RPC consumptions have been combined with individual pesticide residue concentrations found on the consumed RPCs. This assessment should be representative for all age classes and countries considered in the Pesticide Residues Intake Model (PRIMo).

The EU MACP was selected as a convenient tool offering monitoring results that reflect realistic pattern changes in residue levels due to its 3-year cycle. The 30 commodities included in the programme rotate per year, 10 commodities per year repeated every 3 years. The EU MACP should be based on random sampling, however it was considered that the samples taken during the Member States’ National Control Programmes (NCPs) should also be included in order to increase the number of samples. In this case, sampling might be selective, which means it is slightly more biased than the EU MACP. However, it was agreed not to include enforcement samples (samples that are targeted to already known problems), since sampling in this case is extremely biased, not offering a representative overview of the market. In some cases, despite strict enforcement action taken by the Member States, some samples exceeding MRLs could potentially reach the consumers since sampling is random. In order to further ensure a more realistic overview of the market, it was decided to include in the assessment those samples exceeding the MRLs.

It was agreed to communicate the result of the risk characterisation following the concept of combined (Total) Margin Of Exposure (MOET)see note at the end of the page. The MOET basically represents a safety margin (relating to a whole group of substances) between the real exposure to humans and exposure levels that would lead to a certain adverse health effect (using substance-specific toxicological information for the substances included in the same group and their relative potencies).

Both consumers and non-consumers of a RPC are included in both Tiers in order to give the best estimate of exposure.

In line with the safety margin currently used for establishing toxicological reference values, an indicative target of 100 was selected as a threshold for regulatory consideration. A MOET of above 100 would mean that there is a sufficient safety margin and likely no regulatory action would be needed. In case a MOET value is found to be below 100 for a certain percentage of the population then this also does not mean necessarily that consumers are at risk, but it means that risk managers should consider taking action.

pesticides_mrl_cum-risk_tech-details_fig1.jpg

Figure 1:

As illustrated in Figure 1, the output of the evaluation is a distribution of consumer exposures, so a specific point on it is set as a reference to decide whether regulatory action should be considered. Together with the above-mentioned margin of exposure of 100, the 99.9th percentile of the total population was selected as the most protective threshold for regulatory consideration. In distributional statistics, the 99.9th percentile of a set of data is the value below which 99.9% of the data fall. In other words, if there is a MOET of e.g. 115 at the 99.9th percentile of the total population this means that exposure levels for 999 out of 1,000 people would be expected to have a safety margin of 115 (more than 100 and hence regulatory action would likely not be needed.

As already mentioned, progressing in tiers offers a more refined assessment, gradually leading to more realistic evaluations. The main parameters offering room for refinement refer to the samples for which residue concentrations were not quantifiable (NQs) and samples for which data are missing. Every year the rate of samples for which no residue was quantified is approximately 50%. Furthermore, samples are analysed according to different analytical scopes, which do not necessarily comprise all pesticides. Hence, only a part of the samples will be analysed for a given pesticide while for the remaining samples data on that pesticide will be missing. At worst, data may be missing completely for a given pesticide-crop combination.

In order to impute the most realistic values for estimating exposure to those substances for which no residues were quantified in the monitoring data sets, the following main approach was agreed:

In Tier 1, authorisation status of a pesticide is not considered. Either half the value of the limit of quantification (½LOQ) is assigned to the Non-Quantifiable residues (NQs), or the value zero.

Half the value of the limit of quantification (½LOQ) is assigned to pesticide-crop combinations with at least one quantifiable finding in the large datasets of the monitoring exercise. The value zero is assigned to pesticide-crop combinations without any quantifiable findings in the monitoring exercise. In the latter case it is assumed that the pesticide was not used. Using these parameters, in particular the use of ½ LOQ, means a very conservative assessment resulting in overestimation of exposure. It is very unlikely that all substances of a given cumulative assessment group with non quantified residues would all be present in the same sample eaten by a consumer at the level of ½ LOQ.

In Tier 2, use frequencies are taken into account to refine the calculations. However, since reliable and complete data on agricultural uses are currently not yet available, certain assumptions are necessary. The value of ½LOQ is assigned where information on combinations of pesticides observed in the monitoring data (findings on multiple residues) shows co-occurrence of substances. Some of the previously imputed values of ½ LOQ will turn into values of zero where it is unlikely that substances were used in combination, which would also in many cases make no sense from the agricultural point of view. Still the Tier 2 calculation will likely overestimate chronic exposure since it is unlikely that all possible co-occurrences found in the monitoring data on multiple residues would also occur on each and every sample eaten by a consumer over a lifetime.

Regarding Processing Factors (PFs), imputation on values for RACs will overestimate exposure in most cases since many processing operations result in a decrease of pesticides residues and most food is eaten in processed form. It was agreed to use for both Tiers the PF values from all available input sources (e.g. EFSA and national databases like the German BfR or the Dutch RIVM), as this would offer a more realistic approach, compensating for the conservative approach of assigning ½ LOQ to NQs. When PFs are not available, the assumption is that as a worst case all residues are transferred into the processed products, as this is the case in the evaluation of MRL applications.

Drinking water may also contain pesticides according to Directive 98/83/EC, but because the level of pesticides in drinking water is very low, the contribution from this source is low compared to the pesticide residues originating from food. In order to reflect this low occurrence, it was agreed to assign in Tier 1 a level of 0.1μg/L for the 5 most potent substances of the CAG and in Tier 2 a level of 0.05μg/L.

As with current acute exposure assessments, it was concluded that the Variability Factors (VFs) currently used in PRIMo should be included in Tier 1. For Tier 2 it was decided to use the distribution of VF values centered around a more realistic value of 3.6, as this is the average variability factor observed in market samples as reported in an opinion of EFSA’s PPR Panel on the use of the appropriate variability factor(s) for acute dietary intake assessment of pesticide residues.

The following table summarises the agreed parameters as explained above:

Parameter

TIER 1 approach

TIER 2 approach

Expressing Cumulative Exposure when using a probabilistic approach

Combined (Total) Margin of Exposure (MOET)

Combined (Total) Margin of Exposure (MOET)

Threshold for regulatory consideration (MOET, percentile, population)

MOET=100

99.9p

Whole population

MOET=100

99.9p

Whole population

Take into account samples exceeding MRLs?

yes

yes

Period of interest

3-years cycle data sets

3-years cycle data sets

Residue data and Sampling

Use data from the EU-Multi Annual Control Programme (EU MACP) and National Control Programmes (NCPs). Although samples from NCPs are more biased than samples from the EU MACP, they should also be considered. No enforcement samples.

Use data from the EU-Multi Annual Control Programme (EU MACP) and National Control Programmes (NCPs). Although samples from NCPs are more biased than samples from the EU MACP, they should also be considered. No enforcement samples.

Type of dietary surveys

Surveys representative for all age classes and countries considered in PRIMO

Surveys representative for all age classes and countries considered in PRIMO

Type of commodities

All commodities (not only MACP)

All commodities (not only MACP)

Non-quantifiable residues (NQs): pesticide-crop combinations with only NQs

Assign zero value to all NQs

Assign zero value to all NQs

Non-quantifiable residues (NQs): pesticide-crop combinations with 1 or more quantifiable findings

Replace all NQs with ½ LOQ, regardless of the authorisation status.

Assign ½ LOQ based on use frequencies which are estimated from the agricultural use patterns (AUP) identified through the monitoring data. Frequencies of authorised AUPs are adjusted, assuming the frequencies of all AUPs sum up to 100%.

Handling missing data:

missing pesticide-crop combinations

Extrapolation from other commodities based on the EU Extrapolation Guide*.

Extrapolation from other commodities based on the EU Extrapolation Guide*.

Handling missing data:

missing pesticide-sample combinations

Missing measurements for a given pesticide are imputed with concentrations measured in other samples. The highest concentration is imputed to the sample with the highest concentrations for other pesticides.

Missing measurements for a given pesticide are imputed at random with concentrations measured in other samples

Processing factors (PFs)

Use the PF values from all available input sources (e.g. EFSA, BfR, RIVM)

Assume 100% transference of residues to processed commodities when PFs are not available

Use the PF values from all available input sources (e.g. EFSA, BfR, RIVM)

Assume 100% transference of residues to processed commodities when PFs are not available

Pesticide Residues in Water

0.1μg/L for the 5 most potent substances of the CAG

0.05μg/L for the 5 most potent substances of the CAG

Unit-to-unit variability

Use b-distributions with variability factors of 5 and 7 (defined in PRIMo)

Use b-distributions with a variability factor of at 3.6 (average variability factor observed in market samples)

Summary

The approach set out in this document uses the most up-to-date science and data for estimating human exposure from food and drinking water relating to possible effects of simultaneously occurring pesticides. It has been used in the first two EFSA risk assessments relating to groups of substances with effects on the thyroid and the nervous system. The methodology presented here estimates retrospectively dietary exposure using existing monitoring data. This is an important step forward compared to the currently used single substance assessments. By using a tiered assessment approach, refinements can be done gradually. The first tier is always very conservative. If tier 1 calculations result in low exposures, tier 2 assessment is not needed. If the assessment shows that a tier 2 calculation is needed, the tier 2 will deliver a refined and more realistic exposure assessment. It is important to note that even tier 2 is still a conservative assessment and certain conservative assumptions are made where data are missing or incomplete. EFSA also always considers uncertainties in their risk assessment reports and delivers sensitivity analyses to evaluate the impact of using certain parameters, including those presented in this document.

In a second step, work will go on in progressing with a methodology that would also allow the use of cumulative risk assessment in regulatory practice at pre-authorisation stage (prospective scenario), e.g. for making decisions on approval of active substances, authorisation of plant protection products and on the setting of maximum residue levels (MRLs).

Notes

EFSA methodologies

  • Scientific Opinion to evaluate the suitability of existing methodologies and, if appropriate, the identification of new approaches to assess cumulative and synergistic risks from pesticides to human health with a view to set MRLs for those pesticides in the frame of Regulation (EC) 396/2005;
  • Guidance on the Use of Probabilistic Methodology for Modelling Dietary Exposure to Pesticide Residues;
  • Scientific Opinion on the identification of pesticides to be included in cumulative assessment groups on the basis of their toxicological profile (2014 update);

Margin Of Exposure

The Margin of Exposure approach uses a reference point, often taken from an animal study and corresponding to a dose that causes a low but measurable response in animals. This reference point is compared with various dietary intakes in humans. The ratio of the estimated exposure to the relevant toxicological endpoint (before application of any uncertainty factors) is calculated.