Source of uncertainty	Consensus judgement	Consensus rationale	Information notes
U1 (omitted commodities)	–/●	a) Inclusion of the omitted commodities in the assessment would be expected to reduce the MOET possibly by more than 20% in the Italian adult population, but by less than a factor of 2. The key points, which in combination, support this judgement are (i) the mean contribution (and the associated standard deviation) of the 35 plant commodities to the overall diet of plant origin (see Table E.1, Note 1) and (ii) the contribution of the 35 selected commodities to the total long‐term exposure, which is weak to some substances (e.g. carbofuran, diazinon, methomyl, oxamyl), as best reflected by the lower bound calculations reported in Table E.3, Note 3). No increase of the MOET is anticipated, i.e. the multiplicative factor can only be below 1. b) The multiplicative factor can be lower for other populations, mainly adult populations as children and toddlers are closer to the Italian adult population regarding the contribution of the 35 plant commodities to the overall plant diet. Nevertheless, consensus judgement still remains within the same range (–/●).	Notes 1, 2 and 3.
U2 (ambiguity in consumption data)	●/●	a) Perfect information can change the MOET in both directions, but the change would be small and would not exceed 20% because (i) the model combines average consumptions and average concentrations for each commodity, and (ii) the pesticide/commodity combinations driving the risk are unaffected by this source of uncertainty. b) No differences are expected between populations.	Note 4
U3 (accuracy of consumption data	●/●	a) Perfect information can change the MOET in both directions, but the change would be small and would not exceed 20%. No significant methodological limitation was identified in the consumption surveys. The effect of under reporting was considered as limited because it is unlikely to affect fruit and vegetables, which are the commodities of importance for exposure to pesticide residues. The effect of over‐reporting was also considered limited based on detailed records of consumption data for subjects with exposures exceeding the 99th percentile (Annex C, Table C.03). b) No differences are expected between populations.	Note 5 and 6
U4 (sampling variability of consumption data)	–/●	a) Perfect number of consumers in the survey (i.e. number high enough to ensure reliability at the 99.9th percentile of exposure) would result in a decrease of the MOET due to the high probability that consumers in the upper tail of the exposure distribution will not be sampled (highly exposed consumers). b) As the number of subjects varies considerably between populations, differences are expected between the different populations, especially the ones with lower number of subjects for which the underestimation of the exposure at P99.9 would be expected to be larger (e.g. 322 subjects for the Dutch survey).	Note 7
U5 (Representativeness of the consumption data)	●/●	a) Perfect information can change the MOET in both directions. A decrease is more plausible than an increase due to the recent increase in fruit and vegetable consumption by a few percents, not captured yet by recent surveys. The change would however be small and would not exceed 20%. Additionally, apart from oranges, being one of the risk drivers, the positive trend in fruit and vegetable consumption does not concern commodities identified as risk drivers. Despite the lack of information on whether ethnical differences are well accounted in surveys, the 35 commodities selected are basic commodities which are most likely consumed by all populations. b) No differences are expected between populations.	Note 8
U6 (use of invariable recipes and conversion factors by the RPC model)	●/●	a) Perfect information can change the MOET in both directions. The change would be small and within 20% as over‐ and under‐estimations resulting from variations in recipes tend to cancel out in the long term. b) No differences are expected between populations.	Note 9
U7 (pesticide/commodity combinations without occurrence data)	●/●	a) Solving this uncertainty can only decrease the MOET. However, the impact would be extremely low because only one substance/commodity combination is affected by this issue. b) No differences are expected between populations.	Note 10
U8 (metabolites not accounted)	–/●	a) Solving this uncertainty can only decrease the MOET. The MOET could be decreased possibly by more than 20% but less than a factor of 2, considering the cases where a residue definition for risk assessment differs from the residue definition for monitoring and cases where information on the residue definition for risk assessment is missing. b) No differences are expected between populations.	Note 11
U9 (ambiguity of occurrence data)	●/●	a) Perfect information can change the MOET in both directions, but the change would be small and would not exceed 20% because the model combines average consumptions and average concentrations for each commodity. b) No differences are expected between populations.	Note 4
U10 (analytical uncertainty for occurrence data)	●/●	a) Perfect information would have a very limited impact on the MOET since the total number of measurements is high and analytical uncertainties are expected to average out. b) No differences are expected between populations.	Note 12
U11 (sampling variability of occurrence data)	●/+	a) Perfect information on sampling variability regarding occurrence data will only be expected to increase the MOET by a factor up to 2. The key points driving this consensus are the following: (i) the median estimate of the MOET at the 99.9th percentile resulting from the Tier II calculations is unstable and the value used as model output for the uncertainty analysis is likely to underestimate the real one; (ii) based on the monitoring data from 2014 to 2018, the probability to encounter a concentration of 4.9 mg/kg (which was observed for the sum of omethoate and dimethoate in one olive sample only) is expected to be smaller than 1 out of 79 (total number of determinations), and the sensitivity analysis E (i.e. excluding the residue concentration of 4.9 mg/kg) suggests that the 99.9th percentile of the exposure distribution may be overestimated by a factor up to 2. b) The overestimation may be smaller for populations that have a smaller contribution of omethoate and dimethoate in olives for oil production to the cumulative exposure (German adults, United Kingdom toddlers, Bulgarian children, Dutch toddlers, Dutch children, Danish toddlers).	Note 7
U12 (representativeness of the occurrence data)	●/●	a) Perfect information can change the MOET in both directions, but the change would be small and would not exceed 20%. Sensitivity analysis G did not confirm the theoretical expectation that samples coded as ‘selective sampling’ lead to an underestimation of the MOET. The impact of this source of uncertainty is however difficult to evaluate due to inconsistencies in the interpretation of the term ‘selective sampling’ at member‐state level. b) No differences are expected between populations.	Note 13
U13 (extrapolation of occurrence data)	●/●	The impact of this source of uncertainty is nil because this type of extrapolation was not needed in the present assessment.	Note 14
U14 (pooling of occurrence data from all Member States)	●/●	a) Perfect information could change the MOET in both directions, depending on the country and pesticide/commodity combination. Considering the large number of pesticide/commodity combinations, the overall impact of this source of uncertainty on the MOET is expected to be low due to an averaging effect. Moreover, populations usually consume a mixture of imported and locally grown commodities. b) No differences are expected between populations.	Note 15
U15 (unspecific residue definitions)	●/+	a) This source of uncertainty affects the risk drivers omethoate and dimethoate on olives for oil production. Omethoate and dimethoate differ in potency by a factor 4. Perfect information on the exact ratio between the two compounds is expected to increase the MOET by a factor up to 2. This is based on information about the evolution of the omethoate/dimethoate ratio on olives after harvest (high ratios are unlikely to be associated to high levels of the sum of the two compounds) and the results from sensitivity analysis K where omethoate was assumed to not be authorised as this is the case. b) This uncertainty would be expected to have smaller impact in populations consuming less olive oil	Note 16
U16 (left‐censored data: assumption of the authorisation status of pesticide/commodity combinations)	●/●	a) There was a general agreement that this type of uncertainty would not impact the MOET by more than 20% towards both directions, because it is subject to two factors acting in opposite directions. This is supported by the sensitivity analysis H in which all pesticide/commodity combinations with detection rates exceeding 1% were considered as authorised. b) No differences are expected between populations.	Note 17
U17 (left‐censored data: assumption about the use frequency)	●/●	a) Perfect information would tend to increase the MOET because the assumption that all samples were treated with at least one active substance (assigned an authorised use) is considered conservative (e.g. organic farming is not considered, information on percentage of quantifiable measurements in Annex C, Table A.09). However, sensitivity analysis B shows that the magnitude of the impact is below 20%. b) No differences are expected between populations.	Note 18
U18 (left‐censored data: assumption on the residue level)	●/●	a) Perfect information would have a limited impact on the MOET based on the results of sensitivity analysis B, and assuming a similar finding in the opposite direction if left censored data were imputed to the level of LOQ instead of 1/2 LOQ. b) No differences are expected between populations	Note 18
U19 (assumption about pesticides in drinking water)	●/●	a) Perfect information would increase the MOET, based on the fact that the five most potent substances of the CAG assumed to be present at 0.05 μg/L in drinking water are not approved in the EU and on quantitative information about the contribution of drinking water in Annex C, Table C.02. The impact is however lower than 20%, as suggested by sensitivity analysis I. b) Differences between populations are suggested by sensitivity analysis I (to be discussed/addressed under EKE Q3).	Note 19
U20 (missing processing factors)	++/+++	a) Perfect information would increase the MOET based on indicative information about PFs related to pesticide/commodity combinations driving the risk, and sensitivity analysis C. b) This sensitivity analysis suggests smaller impacts in all the other populations (in some populations, the judgement would be ++/++) which can be considered in EKE Q3.	Note 20
U21 (Use of processing factors in the EFSA food classification and description system (FoodEx))	●/●	a) Perfect information could change the MOET in both directions, depending on the food consumed and the actual recipe/processing. Considering the large number of recipes and processing types in a long‐term assessment, contrasting effects are expected and would tend to average out across foods. In addition, as this source of uncertainty does not concern most risk drivers the overall impact is expected to be limited. b) No differences are expected between populations.	Note 21
U22 (analytical uncertainty for processing factors)	●/●	a) Perfect information would have a very limited impact on the MOET considering the number of measurements and the fact that analytical uncertainties are expected to average out. b) No differences are expected between populations.	‐
U23 (accuracy of processing factors)	●/●	a) Perfect information on the actual levels in processed commodities would only further decrease the processing factor, and therefore the multiplicative factor can only be above 1. The magnitude of the impact is however very low considering the small number of processing factors concerned by this source of uncertainty and the low values of these PFs. b) No differences are expected between populations	Note 22
U24 (use of fixed values of processing factors)	●/●	a) Only one risk driver is concerned (chlorpyrifos/orange), for which two PFs were used (peeling and juicing) and, in this case, they were median values of four and six independent trials. b) No differences are expected between populations.	Note 23
U25 (processing factors not considered (peeling of commodities with edible peel and washing)	●/●	a) Perfect information on peeling and washing can only result in multiplicative factors above 1. However, the impact is minor because none of the risk drivers is concerned by this source of uncertainty. b) No differences are expected between populations.	Note 24
U26 (OPs and NMCs not included in the CAG)	–/● or ●/●	a) Perfect information on the omitted substances and their inclusion in the CAG can only decrease the MOET. The overall quantification rate of the substances not included in the CAG (16 non‐approved OPs and NMCs) is about 1% of the quantification rate of substances included in the CAG. Results from deterministic exposure calculations in PRIMo for the 7 substances with known ADIs (especially lower bound estimates which are better indicators of risk drivers than upper bound estimates) suggest that their contribution to the cumulative exposure would be minor. On the other hand, regarding the 9 substances for which ADIs are missing, despite the very low quantification rate, account is taken of the fact that one single sample may significantly alter the MOET at the percentile of interest (see sensitivity analysis E in Table 16 in Section 3.1.3), depending on the residue level and the potency of the substance. The experts did not agree on a consensus range for the impact of this source of uncertainty and 2 consensus judgements were retained. However, those experts supporting the judgement (−/.) agreed that the decrease of the MOET was unlikely to be much larger than 20%. b) No differences are expected between populations.	Note 25
U27 (contribution of substances acting through oxidative stress)		Not assessed (see Section 3.2.1)
U28 (substances included in the CAG not causing the effect)	●/●	a) There is a high level of certainty that all substances in the CAG contribute to chronic AChE inhibition. b) No differences are expected between populations.	Note 26
U29 (Uncertainties related to original studies/data quality)	●/+	a) Perfect data quality could either increase or decrease the MOET. The former is judged more plausible because the NOAELs of 2 risk drivers with low data quality (monocrotophos and dichlorvos) used for the CRA calculations were particularly low (0.005 mg/kg bw per d and 0.008 mg/kg bw per d, respectively) compared to all other substances of the CAG and would likely increase if information from studies of perfect quality were available. In general, in case of low‐quality studies and/or lack of statistical analysis, the assessors tend to derive lower NOAELs. b) No differences are expected between populations.	Note 27
U30 (Uncertainties related to the data collection methodology)	●/●	a) Perfect information could either increase or decrease the MOET. The impact would be low (less than 20%) considering that the information in source documents has been checked during the EFSA peer review process or JMPR evaluations, and that the working documents (excel spreadsheets) were checked against EFSA conclusions. b) No differences are expected between populations.	Note 28
U31 (Uncertainty related to the NOAEL‐setting principles)	●/●	a) Using perfect NOAELs for erythrocyte AChE inhibition would most probably decrease the MOET because a NOAEL of 0.04 mg/kg bw per d for dimethoate (a risk driver) would be used instead of the NOAEL of 0.1 mg/kg bw per d. The change would, however, not exceed 20%, considering the contribution of dimethoate to the risk. b) No differences are expected between populations.	Note 29
U32 (Uncertainty related to the study design of the critical study)	–/+	a) Perfect study design such that the NOAELs would precisely reflect BMDL20s could change the MOET by a factor ranging from 1/2 to 2. There are several factors in interplay and with opposing effects, the main ones being the dose spacing and the methods/assays used to measure AChE activity, which are old and differ between substances. This is a major source of uncertainty, as suggested by the differences observed by the EFSA Scientific Committee between NOAELs and BMDs (EFSA Scientific Committee, 2017a,b). b) No differences are expected between populations.	Note 30
U33 (adequacy of the dose‐addition model)	●/●	a) Perfect information would be expected to have minor impact on the MOET as OPs and NMCs have a similar mode of action b) No differences are expected between populations	Note 31
U34 (adequacy of the OIM model)	●/+	a) Perfect information could increase the MOET due to the inherent limitations of the OIM model in the tails of the exposure distribution. The increase could be larger than 20%, although unlikely much larger. The key elements which support this judgement are the following: (i) Most of the commodities driving the risk (i.e. olive oil, wheat and drinking water) are consumed with relatively moderate daily fluctuations and for such commodities the OIM model is expected to produce better long‐term exposure estimates at extreme percentiles (ii) The ratio between the MOET estimates at P50 and P99.9 of the exposure distribution for the Italian adult population is 3, excluding an impact exceeding a factor of 2. b) Differences are expected between populations depending on the number of days in the respective consumption surveys.	Note 32
U35 (adequacy of the UF for intraspecies variability)	●/●	a) Source of uncertainty of marginal relevance as the Italian adult population includes individuals from 18 to 65 years old. b) No differences are expected between populations as none of the 10 populations under consideration includes infants of less than 16 weeks or elderly people.	Note 33