Risk analysis of toxicants

An aspect of chemistry, uncertainties, is a current hotspot in the risk analysis of toxicants, writes Alberto Mantovani, Research Director at Istituto Superiore di Sanità – Roma, Italy

Uncertainties can be defined as the gaps of knowledge and/or of data sets and/or of methodologies that can exert an unwanted influence on the outcome of a risk assessment. In principle, uncertainties are unavoidable; hence, a transparent description and weighing of relevant uncertainties should be a necessary component of risk assessment. It is often difficult to accurately assess the impact of each specific uncertainty, but the combined effect of identified uncertainties may – and should be- characterised.

Important examples of uncertainty analysis are provided in recent (2018) opinions by the European Food Safety Authority on widespread contaminants, with endocrine disrupting potential, such as dioxins or the perfluoroalkyl PFOS and PFOA. Differently from substances like pesticides, for which further data can be requested to applicants, for contaminants, well-investigated areas may co-exist with major data gaps.

In the case of PFOS and PFOA, the data on humans allowed to define a tolerable weekly intake (TWI), further supported by the identification of toxicological mechanisms. On the other hand, the exposure assessment was weakened by insufficient data on the presence of the substances in European foods, as well as by the frequent use of analytical methods with insufficient sensitivity.

The inherent presence of uncertainties in science-based risk assessment and the need for robust and transparent approaches for dealing with uncertainties have been the core topics of the International Conference on Uncertainty in Risk Analysis (Berlin, 20-22 February 2019), jointly organised by EFSA and the German Federal Institute for Risk Assessment (BfR).

Many interventions pivoted on case studies characterised by data-rich scenarios, allowing for a refined analysis and management of uncertainties (for instance, the risk assessment of the endocrine-disrupting plasticizer bisphenol A). However, this article wishes to point out the problems arising from data-poor scenarios, which may represent the harsh reality in too many instances.

This topic was discussed in an ad hoc satellite workshop within the Conference: “Accounting for uncertainty in data-poor scenarios: Cases studies on risk analysis in food safety”.

Representative instances of data-poor situations include: emergencies when risk assessors are requested to provide fast advice with limited information available (for instance, contamination of food chains by substances on which limited information is available); emerging issues stirring the need for timely decisions by risk managers (examples discussed: contamination of a staple food by poisonous weeds related to climate change, assessment of chemical mixtures, detection of rare allergens in common edible vegetables); countries where data collection still presents gaps, yet risk analysis of food safety issues is needed, beyond the simple fulfilment of legal limits (one example presented was the risk assessment of veterinary drug residues in the Republic of Georgia).

In data-poor scenarios, scientists acting as risk assessors should resist to the will (and the push by risk managers) for giving simple and straightforward answers, by ignoring uncertainties; actually, these should be identified and weighed especially when data sets are not rich. While in some data-poor scenarios (e.g., emergencies), close interaction and even overlapping between risk assessors and risk managers is unavoidable (and maybe useful), the final decision must remain firmly in the hands of risk managers. The risk assessor task is to provide an outcome that risk managers can elaborate into a straight message, i.e., the decision whether to act and how.

Therefore, a critical issue is the communication of the risk assessment outcome. Especially in data-poor scenarios the final form of the risk assessor opinion should present as clearly and completely as achievable all the elements that are required to make decisions and/or to improve the data set in order to reduce the uncertainty burden.

In data-poor scenarios, the boundaries between uncertainty and variability, which is inherent to the investigated population, may become blurred. Mixing-up the two issues may result in conceptual as well as factual hindrances for risk analysis. For instance, the population variability in the consumption of certain food items can be hidden by the lack of reliable food consumption databases. As a consequence, it will be unfeasible to make but rough estimates and assumptions on consumption of food items, without any information on the distribution of high consumers of certain foods, and differences related to age or gender.

In the meanwhile, risk assessors should not fall into a pessimistic attitude. For instance, the uncertainties related to population variability may be reduced by focusing on population subgroups which are more exposed and/or more susceptible (e.g., children). Such subgroups can be identified by “diagnostic risk assessment”: we may name in this way deterministic or semi-probabilistic approaches used to screen (“diagnose”) whether there is a potential problem, by using transparent elements of conservativeness.

Indeed, values, such as precaution, may be introduced in risk assessment when a time-effective response is needed to protect public health: for instance, data on a food commodity known to be highly contaminated taken as representative of the whole food category for which there are no data. In each case, the selection of conservative or precautionary assumptions must be transparent and understandable, as well as be based on the available data, however limited; ideally, these options should be discussed and agreed with the risk manager.

Expert Knowledge Elicitation (EKE) is a robust and standardised procedure implemented by EFSA, which can be of special use in data-poor scenarios to describe uncertainties and their distribution and range. EKE may also facilitate the reduction of the range of uncertainties, by promoting consensus through discussion. Obviously, EKE requires the availability of a group of experts, independent and with different backgrounds. If well-managed, EKE can be both effective and efficient.

Finally, the discussion on uncertainties must not be confined to risk assessors. Risk managers are all-important, starting from problem formulation and ending with the task of developing the assessment outcome, with its uncertainties, into options. If clear and complete, the evaluation of uncertainties should be viewed as an additional strength of a risk assessment opinion, as it increases the transparency. Thus, especially considering data-poor scenarios, it is pivotal that risk managers are trained to understand and cope with uncertainties.


Please note: This is a commercial profile
Alberto Mantovani
Research Director
Istituto Superiore di Sanità – Roma, Italy
Tel: +39 06 4990 2815


Please enter your comment!
Please enter your name here