A ranking method of chemical substances in foods for prioritisation of monitoring, based on health risk and knowledge gaps

Chemical contaminants are present in all foods. Data on the occurrence of contaminants in foods that are often consumed or contain high contaminant concentrations are critical for the estimation of exposure and evaluation of potential negative health effects. Due to limited resources for the monitoring of contaminants and other chemical substances in foods, methods for prioritisation are needed. We have developed a straightforward semi-quantitative method to rank chemical substances in foods for monitoring as part of a risk-based food control. The method is based on considerations of toxicity, level of exposure including both occurrence in food and dietary intake, vulnerability of one or more population groups due to high exposure because of special food habits or resulting from specific genetic variants, diseases, drug use or age/life stages, and the adequacy of both toxicity and exposure data. The chemical substances ranked for monitoring were contaminants occurring naturally, unintentionally or incidentally in foods or formed during food processing, and the inclusion criteria were high toxicity, high exposure and/or lack of toxicity or exposure data. In principle, this method can be used for all classes of chemical substances that occur in foods, both unintended contaminants and deliberately added chemical substances. Foods considered relevant for monitoring of the different chemical substances were also identified. The outcomes of ranking exercises using the new method including considerations of vulnerable groups and adequacy of data and a shortened version based on risk considerations only were compared. The results showed that the resolution between the contaminants was notably increased with the extended method, which we considered as advantageous for the ranking of chemical substances for monitoring in foods.


Introduction
Food safety is an important prerequisite for good health. With the constant change in food production, processing and dietary habits, there is a continuous need for up-to-date knowledge on the presence of chemical substances in foods. Such knowledge is a critical part of risk assessments of chemical substances in food to ensure food safety. Therefore, monitoring of chemical substances that have a potential to pose a health risk is important (INFOSAN, 2009;van der Fels-Klerx et al., 2015). Data on the occurrence of chemical substances in highly consumed food items and in less consumed but highly contaminated food items are critical for risk assessments related to dietary exposure. Hence, prioritisation of chemical substances for monitoring in foods has to take into consideration i.a. potential health hazards, occurrences and the adequacy of data.
Health-based guidance values (HBGVs), i.e. tolerable daily intake (TDI) or tolerable weekly intake (TWI), define the amount of a specific contaminant that an individual can consume on a regular basis over a lifetime without any appreciable risk to health (EFSA, 2020). Comparison of the HBGVs to the estimated dietary exposure in a geographical region or a population group may be used to rank chemical substances according to the health risk (van der Fels-Klerx et al., 2015).
Alternatively, the margin of exposure (MOE) may also be suitable to rank chemical substances according to risk. The MOE is calculated under consideration of a reference point such as the no observed adverse effect level (NOAEL) or the benchmark dose lower confidence limit (BMDL) for the critical health effect. NOAEL is the highest dose of a compound, at which no detectable adverse effects occur in experimental animals or in a population (EFSA, 2020). The benchmark dose (BMD) is the minimum dose of a compound that produces a distinct, low-level adverse health effect, i.e. a benchmark response (BMR), usually in the range of a <0.5 to 10% increase in a specific adverse effect (EFSA, 2020). The BMDL is the lower boundary of the 95% confidence interval of the BMD. MOE is the ratio of NOAEL or BMDL for the critical effect and the human exposure (EFSA, 2005a).
Owing to limited resources, there is a need for ranking of chemical substances in foods in accordance with their estimated health risk to enable risk managers to perform a knowledge-based prioritisation of chemical substances for monitoring. Here, we present a straightforward semi-quantitative method for the ranking of chemical substances for monitoring in foods, based on their estimated risk for human health and critical knowledge gaps.

Selection of chemical substances for ranking
Expert judgement was used for the selection of chemical substances included in the ranking and for the identification of food items relevant for their monitoring. The chemical groups included were natural toxins, metals and metalloids, persistent organic pollutants (POPs), process-induced contaminants and food contact materials. Veterinary medicine residues and pesticides were excluded as ongoing monitoring programmes are in place in Norway. For other chemicals, there are no established monitoring programmes, therefore, a ranking serving as basis for prioritisation by the risk managers on which substances to monitor for the limited funds available for this purpose each year is needed. Criteria for the selection of chemical substances were high toxicity, high dietary exposure and lack of data on toxicity or occurrence in foods, as further described in the following section.

Ranking for monitoring method
The method used to rank chemical substances in foods for monitoring was based on considerations of known toxicity and level of exposure (including the occurrence in food and dietary intake) and vulnerability of one or more defined population groups due to high exposure because of special food habits or resulting from specific genetic variants, diseases, drug use or age/life stages. In addition, the adequacy of both toxicity and exposure data was considered. An overview of the scoring is presented in Table 1. When quantitative data on toxicity and exposure were available, a chemical was scored according to the scoring categories 1, 4, 5 and 6. When either quantitative toxicity or exposure data were unavailable, a chemical was scored according to the categories 2, 3, 4, 5 and 6. The highest possible score was 9 and the lowest possible score was 2, based on the sum of either scoring, i.e. categories 1, 4, 5 and 6, or 2, 3, 4, 5 and 6. The awarding of high, medium or low scores for each category was based on expert judgements. The initial scoring was performed by 15 experts who had extensive knowledge of the toxicity and/or exposure of the various Table 1 Explanation to categories for ranking of chemical substances in foods for monitoring. When quantitative data on toxicity and exposure were available, scores were given in the categories 1, 4, 5 and 6. When either quantitative toxicity or exposure data was unavailable, scores were given in the categories 2, 3, 4, 5 and 6.

Category
Description Score 1. Quantitative toxicity and exposure data available The exposure was above the HBGV or the MOE* was too low 6 The exposure was close to the HBGV or the MOE* was close to an acceptable value 4 The exposure was well below the HBGV or the MOE* was sufficiently high 2 2. Toxicity of the chemical High toxicity 3 Medium toxicity 2 Low toxicity 1 3. Dietary exposure to the chemical** High exposure 3 Medium exposure 2 Low exposure 1 4. Vulnerable groups The exposure was high because of special food habits for one or more groups in the population, or one or more groups in the population were especially vulnerable due to, for example, specific genetic variants, diseases, drug use or age/life stages (<1 year, puberty, pregnant/nursing, elderly) 1 The exposure was somewhat higher because of special food habits for one or more groups in the population, or one or more groups in the population were somewhat more vulnerable due to, for example, specific genetic variants, diseases, drug use or age/life stages (<1 year, puberty, pregnant/nursing, elderly) 0.5 No population group with increased exposure because of special food habits or special vulnerability due to, for example, specific genetic variants, diseases, drug use or age/life stages (<1 year, puberty, pregnant/nursing, elderly) was identified 0 5. Adequacy of toxicity data Toxicity data were insufficient or lacking 1 Some toxicity data were lacking 0.5 Sufficient toxicity data were available 0 6. Adequacy of exposure data (occurrence and/or intake) Exposure data were insufficient or lacking 1 Some exposure data were lacking 0.5 Sufficient exposure data were available 0 BMDL (benchmark dose lower confidence limit); HBGV (health-based guidance value); MOE (margin of exposure); NOAEL (no observed adverse effect level); TDI (tolerable daily intake); TWI (tolerable weekly intake). *MOE was too low/MOE was sufficiently high: o For compounds that were genotoxic and carcinogenic (compounds for which no threshold of toxicity can be identified), a MOE <10,000 based on the BMDL 10 (the lower limit of an one-sided 95% confidence interval on the BMDL, corresponding to a 10% tumour incidence over control), would in general be considered as too low. Considerations with regard to a sufficiently large MOE that would allow to conclude on low risk have to be case-specific and based on the available data. o For non-genotoxic compounds (for which a threshold for adverse effects can be identified), a MOE <100 based on the no observed adverse effect level (NOAEL) or BMDL, would in general be considered as too low. Depending on the available data, the necessary size of the MOE may be judged differently. **Based on occurrence and/or intake, or biomonitoring showing high total exposure, from food as one important source.
groups of chemicals included in the ranking, such as metals, mycotoxins, food contact materials etc., through their own research and long experience in chemical risk assessment work. In addition, after the individual scorings were done by the respective experts for each chemical category, all scorings were discussed in the whole working group in order to achieve a consistent way of scoring by expert judgements. When the database was insufficient for taking an informed decision or the uncertainty was high, the medium score was chosen.

Identification of foods for monitoring of the ranked chemical substances
Food groups considered relevant for monitoring of the included chemical substances were identified by expert judgements based on available occurrence data in foods, preferable from national databases, or if not available, European or international data from databases or scientific papers. The respective food items were selected based on existing data showing considerable prevalence of specific contaminants EFSA (2015c) 4 1 Intake of Ni from food could be a problem for allergic individuals. 5 0 Sufficient data were available. 6 0 Sufficient data were available.
*Explanations of category numbers: 1) Quantitative toxicity and exposure data available; 2) Toxicity of the chemical; 3) Dietary exposure to the chemical; 4) Vulnerable groups; 5) Adequacy of toxicity data; 6) Adequacy of exposure data (occurrence and/or intake).

Table 3
Scoring results for mycotoxins in ranking for monitoring.
Chemical/ chemical group Total score Category scored* Score Rationale for score There was a concern with respect to chronic exposure. The mean dietary exposure in adults was determined at a level 10 6 -times below the highest dose in a mice study that did not indicate genotoxic effects. 4 0.5 The dietary exposure in children was expected to be 2-3 times higher than in adults. For vegetarians, the higher intake of food of plant origin might increase the exposure. 5 1.0 Toxicity data for AOH/AME were limited. A NOAEL had not been determined, and a TDI had not been established. For intake for OTA at 4 ng/kg bw per day, the cancer risk was negligible. European exposure data was below the TDI, but high consumers exceeded the TDI.
EFSA (2006) *Explanations of category numbers: 1) Quantitative toxicity and exposure data available; 2) Toxicity of the chemical; 3) Dietary exposure to the chemical; 4) Vulnerable groups; 5) Adequacy of toxicity data; 6) Adequacy of exposure data (occurrence and/or intake).
G.H. Mathisen, et al. Food Research International 137 (2020) 109499 in the food groups or high consumption of these foods, as both can lead to high contaminant exposure of the consumers, and thus contributing most to the exposure of the population to a certain chemical.

Chemical substances included in the ranking for monitoring
In total, 33 relevant chemical substances or chemical groups were selected by expert judgement as a proof of concept for the development of a ranking method. The inclusion criteria were high toxicity, high exposure and/or lack of toxicity or exposure data. All chemical substances were naturally occurring, unintentionally or incidentally in foods or formed during food processing: • Metals and metalloids including aluminium (Al), inorganic arsenic (iAs), organic arsenic (oAs), cadmium (Cd), chromium (Cr), lead (Pb), methylmercury (MeHg) and nickel (Ni).
Acrylamide 8 1 6 The MOE values across all age groups indicated a health concern. Similar results were found for Norwegian children.
EFSA (2015a), VKM 1 People consuming food products containing increased PAH concentrations such as mussels from contaminated waters, grilled meat, food prepared using open fire etc., are more vulnerable. 5 0.5 Exposure to mixtures of PAHs is usual, and data on carcinogenic effects of mixtures were needed. 6 0.5 Occurrence data for food prepared on fire, grilled food, mussels from contaminated areas etc., were needed.
3-MCPD and its fatty esters 5.5 1 4 The TDI of 2 µg/kg bw per day was not exceeded in the adult population. A slight exceedance of the TDI was observed for high consumers in younger age groups and in particular in scenarios considering infants receiving formula only. *Explanations of category numbers: 1) Quantitative toxicity and exposure data available; 2) Toxicity of the chemical; 3) Dietary exposure to the chemical; 4) Vulnerable groups; 5) Adequacy of toxicity data; 6) Adequacy of exposure data (occurrence and/or intake).
The scoring of selected chemical substances was likewise based on expert judgement. The chemical substances were allocated to experts in the respective fields and evaluated by the established ranking method. Subsequently, the scoring results were discussed among the experts for validation and balancing of individual assessments.
Alternatively to using expert judgement for the scoring, a more systematic approach involving extensive literature searches could have been applied. However, this would have required substantially more effort and time. As the purpose of the present project was to establish a method for the ranking of chemical substances for monitoring and not the performance of risk assessments, the expert-based approach appeared to be appropriate, practical and time-saving.

Rationale for the scoring and ranking of individual chemical substances or chemical groups for monitoring
The chemical substances were evaluated and scored according to the methodology presented in Table 1. The scoring and the rationale for the scores given are shown in Table 2 for metals and metalloids, Table 3 for mycotoxins, Table 4 for POPs, Table 5 for process-induced contaminants, and Table 6 for compounds present in food contact materials. For metals and metalloids, the total scores ranged from 3.0 for nickel and chromium to 7.5 for lead. For mycotoxins, the total scores ranged from 3.5 for PAT and ZEN and modified forms to 8.5 for T-2 toxins, HT-2 toxins and modified forms. For POPs, the total scores ranged from 3.5 for PBDEs to 8.0 for dioxins and DL-PCBs, and PFOS and PFOA. For process-induced contaminants, the total scores ranged from 5.5 for 3-MCPD and its fatty esters to 8.5 for furans. For compounds in food contact materials, the total scores ranged from 3.0 for BPA to 6.5 for BPF, BPS and BPAF.
We have developed the ranking of chemical substances for monitoring in foods as a tool for priority-setting with regard to risk-based food safety control. The method allows the ranking of chemical substances in different chemical classes and is simply based on the scoring of risk and knowledge gaps by expert judgement considering existing data. Knowledge gaps regarding toxicity and exposure are usually not included in risk ranking methods developed by other agencies (NFA et al., 2018;van der Fels-Klerx et al., 2015). We considered the inclusion of such gaps essential for the comprehensive evaluation of the risk potential of dietary contaminants and for their potential inclusion in food monitoring programs. The method developed by us is useful for the ranking of dietary contaminants, such as metals, mycotoxins, persistent organic pollutants, process-induced contaminants and food contact materials, as shown in this paper. However, in principle, the method can be used for all types of chemical substances occurring in foods, including residues of regulated compounds used for a specific purpose in food production, i.e. food additives, flavourings, pesticides, veterinary medicines and packaging materials. The European Food Safety Authority (EFSA) published in 2015 an external scientific report called "Critical review of methodology and application of risk ranking for prioritisation of food and feed related issues, on the basis of the size of the anticipated health impact" (van der Fels-Klerx et al., 2015), which gave an overview of various risk ranking methods. The included methods ranged from rather simple methods such as the Hazard Index (HI), which is the Estimated Daily Intake (EDI) divided by the HBGV, to more complex methods considering the severity of the health hazard, such as Disability Adjusted Life Years (DALY)/Quality Adjusted Life Years (QALY) or Multi Criteria Decision Analysis (MCDA). The simple methods may be used without much prior experience, whereas the more sophisticated methods need specialist training and experience in order to use them in a correct and meaningful way.
The Swedish National Food Agency (NFA) has developed the "Risk Thermometer Tool" in cooperation with EFSA for the risk ranking of chemical substances and for better risk communication. The method, Table 6 Scoring results for chemical substances/chemical groups in food contact materials in ranking for monitoring.
Chemical/ chemical group Total score Category scored* Score Rationale for score References BPF, BPS and BPAF 6.5 2 2 There was not sufficient toxicological data available to assess the toxicity. The TDI is 0.05 mg/kg bw per day, and the exposure was found to be below the TDI. The estimated exposure for Norwegian adults was well below the TDI. *Explanations of category numbers: 1) Quantitative toxicity and exposure data available; 2) Toxicity of the chemical; 3) Dietary exposure to the chemical; 4) Vulnerable groups; 5) Adequacy of toxicity data; 6) Adequacy of exposure data (occurrence and/or intake).
G.H. Mathisen, et al. Food Research International 137 (2020) 109499 which uses MOE, defined as NOAEL/exposure, and integrates severity by adjusting for the severity of the critical health effects, is called severity-adjusted margins of exposure (SAMOE) (NFA et al., 2018). The SAMOE values were divided into five risk classes corresponding to different levels of human health concern (1 -no concern, 2 -no-to-low concern, 3 -low-to-moderate concern, 4 -moderate-to-high concern, 5high concern). The SAMOE method depends, however, on good quantitative data for both exposure and toxicity, which are not always available.
Our method is designed specifically for ranking for monitoring and not for risk communication. It is straightforward as it is based on expert judgement of the existing data and uses a simplified scoring system. Due to the consideration of information on vulnerable groups and missing toxicity and exposure data (categories 4-6, Table 1), the results have a built-in safety factor and uncertainty margin, allowing the ranking of chemical substances in foods for which little data are available. By setting the maximum total score for toxicity and exposure to 6 points and the maximum total score for vulnerable groups and adequacy of toxicity and exposure data to 3 points (Table 1), we have built-in weighing factors that ensure the balancing of existing data and expert evaluation. The method is suitable not only for known contaminants in food, but for all chemical substances that occur in foods, both unintended contaminants and deliberately added chemical substances, even if the knowledge level regarding occurrence and toxic potential might be rather low.

Ranking with or without consideration of vulnerable groups and data adequacy
The scoring of vulnerable groups and adequacy of data (categories 4-6, Table 1) in the ranking may be considered as "uncertainty-based scoring criteria", whereas when they are included together with scoring based on existing knowledge of toxicity and exposure, the method can be considered as "the full ranking for monitoring method". The impact of including the "uncertainty-based scoring criteria" in the ranking was evaluated using the chemical substances in the Tables 2 to 6. The results are shown in Table 7. Whereas ranking by setting scores for all categories, "the full ranking for monitoring method" (Table 1), delivered scores in the range from 2 to 9, the exclusion of the "uncertainty-based scoring" in a shortened version of the method, using only the risk-based categories 1-3 (i.e. toxicity and exposure; Table 1), delivered scores in the range from 2 to 6. Applying "the full ranking for monitoring method", none of the chemical substances received the lowest or the highest possible score, while with the "shortened method" two-thirds of Table 7 Comparison of ranking performed considering risk-and uncertainty-based scoring categories (full method) or only risk-based scoring categories (shortened method).  Mathisen, et al. Food Research International 137 (2020) 109499 G.H. Mathisen, et al. Food Research International 137 (2020) 109499 G.H. Mathisen, et al. Food Research International 137 (2020) 109499 the chemical substances received the lowest or the highest score. The differentiation between the chemical substances was notably increased with the full method as compared to the shortened version. Compounds scored 2 by the shortened method were scored 3, 3.5 or 4 by the full method. Compounds scored 4 by the shortened method received scores of 4.5, 5.5, 6 or 6.5 by the full method (Table 7). With the full method, higher resolution between chemical substances was even possible for those receiving the highest scores, such as 6.5, 7, 7.5, 8 or 8.5. We therefore consider the inclusion of both risk-based and uncertaintybased ranking categories as advantageous for the identification and ranking of chemical substances in foods for monitoring. However, both methods ranked the contaminants largely in the same order.

Identification of foods for monitoring
Food groups considered relevant for monitoring were identified by expert judgements based on available occurrence data in foods. The selected food groups included baby foods, bakery wares, cereal grains and products thereof, coffee/tea, dairy products, drinking water, eggs, fish/seafood, meat, nuts/seeds/pulses, spices, vegetables/fruits and vegetable oils (Table 8). For chemical substances in food contact materials, specific food groups could not be identified as their presence in food depends on the type of packaging material used and the character of the contact. Each food group identified as important contained at least one chemical with the highest scores (7.5 to 8.5) in this study. Three food groups stood out as especially relevant for the monitoring of chemical substances: cereal grains and products thereof, fish/seafood and vegetables/fruits. These foods are included in the dietary recommendations published by the Norwegian Directorate for Health, saying; "Eat whole grain foods every day. Eat fish two to three times a week. Eat at least five portions of vegetables, fruit and berries every day." (Nasjonalt Råd for Ernaering, 2011). The relative importance of each food group varied for the different chemical classes. Whereas metals and metalloids, mycotoxins and process-induced contaminants should be monitored in cereal grains and products thereof, metals and metalloids, POPs and process-induced contaminants should be monitored in fish/seafood, and metals and metalloids, mycotoxins and process-induced contaminants should be monitored in vegetable/fruits. The chemical class identified as relevant for all food groups except drinking water, eggs, spices and nuts/seeds/pulses, was process-induced contaminants. All chemical substances from this class that were evaluated in the present study were scored 0.5 or 1 in category 6 ( Table 5) with rationales for the scoring given as "Little data were available on exposure" or "Some exposure data were lacking", showing that there is an urgent need for monitoring to allow exposure characterisation. As not all chemical substances can be monitored at the same time due to the limited resources available, prioritisation is needed. By restricting the number of the most important food groups for monitoring of each chemical/chemical group, analytical and sampling resources for monitoring could be planned by the risk managers in such a way that several highly ranked chemical substances could be analysed in the same food groups, thus saving resources. It has been suggested that food products consumed in significant quantities and those that may contain elevated contaminant levels should be sampled and tested with high frequency (INFOSAN, 2009). To make sure that foods recommended as healthy are also safe, data on the occurrence of contaminants in these foods should be available.
The actual use of this ranking for prioritisation is to be decided by the risk managers who are in charge of the monitoring. Which substances ultimately to be included in the monitoring will depend on the available funds for monitoring at the time and possibly also other concerns, such as how the various substances can be analysed together, time since last monitoring of the substance, alerts from other countries on health risk from certain substances, media interest etc.  Mathisen, et al. Food Research International 137 (2020) 109499

Conclusions
A method for ranking of chemical substances in foods was developed. Subsequent use of the ranking is meant as a tool for risk managers in their prioritisation for food monitoring programs as part of a riskbased food control. The method is straightforward as it is based on expert judgement by risk assessors of existing risk-and uncertaintybased data and uses a simplified scoring method. The resolution between the chemical substances was notably increased with the full method, which includes vulnerable groups and adequacy of data, as compared with the shortened version. The methodology can be used to compare different classes of chemicals, as well as to compare subtypes of substances within the chemical classes. In principle, this method can be used for all classes of chemical substances that occur in foods, both unintended contaminants and deliberately added chemical substances. At large, the obtained ranking mirrors the need for monitoring and research to obtain new data as have been identified in many risk assessments opinions by EFSA and VKM and highlighted in research papers.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.