Probabilistic risk assessment model for allergens in food: sensitivity analysis of the minimum eliciting dose and food consumption

https://doi.org/10.1016/j.fct.2007.09.109Get rights and content

Abstract

Previously, TNO developed a probabilistic model to predict the likelihood of an allergic reaction, resulting in a quantitative assessment of the risk associated with unintended exposure to food allergens. The likelihood is estimated by including in the model the proportion of the population who is allergic, the proportion consuming the food and the amount consumed, the likelihood of the food containing an adventitious allergen and its concentration, and the minimum eliciting dose (MED) distribution for the allergen. In the present work a sensitivity analysis was performed to identify which parts of the model most influence the output.

A shift in the distribution of the MED reflecting a more potent allergen, and an increase in the proportion of the population consuming a food, increased the number of estimated allergic reactions considerably. In contrast, the number of estimated allergic reactions hardly changed when the MEDs were based on a more severe response, or when the amount of food consumed was increased.

Development of this work will help to generate a more accurate picture of the potential public health impact of allergens. It highlights areas where research is best focused, specifically the determination of minimum eliciting doses and understanding of the food choices of allergic individuals.

Introduction

To assess the risk arising from unintended exposure to food allergens a deterministic approach may be applied, where point estimates are used to calculate the outcome (Spanjersberg et al., 2007). In worst case scenarios usually the observed values of high percentiles, for example, the 97.5th percentiles or maximum are used. In less worst case scenarios mean values may be used. This approach often leads to conclusions such as “an allergic reaction cannot be excluded”, which is a poor basis for decision making in risk management. Furthermore, in this approach neither variability (for example, heterogeneity over different members of a population) nor uncertainty (for example, lack of knowledge about the true value) can be taken into account. In probabilistic modelling, distributions are used to represent the input parameters, making it possible to account for both variability and uncertainty and to quantify the risk.

A risk assessment model for allergens in food based on probabilistic techniques resulting in a quantitative assessment of the risk and detailed information on the predicted number of allergic reactions has been developed at TNO (Spanjersberg et al., 2007). This approach is now recognized as a way forward in allergen risk assessment (Health Council of The Netherlands, 2007, Europrevall & FSA, 2007). In summary, in this model several input parameters can be identified, such as the proportion of the population who is allergic to a specific allergen, the likelihood that an allergic person consumes a certain product and if so in what amount, the likelihood that the food accidentally contains the allergen and if so the concentration of the allergen. These input parameters determine the distribution of intake of the allergen. By combining the distribution of allergen intake with the distribution of minimum eliciting doses (MED) among the population, the number of allergic reactions can be estimated more realistically.

In most probabilistic assessments, the majority of the variance in the output distribution is attributable to variability and/or uncertainty in a selected number of inputs. The identification of significant contributors to output variance has useful implications for future research. It makes it possible to target resources to a small number of important inputs, rather than spread the resources across the entire set of inputs (Cullen and Frey, 1999). Sensitivity analysis of a model is used to identify these key contributors to variability and uncertainty in model outputs by describing the change in output as a function of the change in the input variable. By comparing the magnitude of changes it is possible to identify those inputs which have the most influence on the output.

Validation of the model by a comparison of the predictions with observational data is not possible at this moment due to the lack of observational (incidence) data. Nothing it known about how the incidence of allergic reactions in a specific population relates to the consumption of a food containing an adventitious allergen.

This model has been developed to assess the risk associated with a given allergen concentration in a food. This means that an established allergen concentration (distribution) will be the starting input parameter. With these sensitivity analyses we wanted to investigate the major sources of variance and uncertainty affecting the risk associated with such given concentration (distribution). Therefore, we performed sensitivity analyses, involving different scenarios in which the input parameters of the MED part and the food consumption part of the model were varied.

Section snippets

Methods

As the basis for the sensitivity analyses, a case study was used for the probabilistic risk assessment model. The probabilistic model and case study have been described in more detail by Spanjersberg et al. (2007). The model consists of two elements: the distribution of allergen intake and the distribution of MEDs within a population. Within the model the values of these two factors relative to each other, and the shape of their distributions determine the probability of an allergic reaction.

Sensitivity analyses

The input parameters that were investigated both for the food consumption part and the MED elements are presented in Table 1b. For each scenario the input variable that is changed is presented together with the change in the distribution of the input variable.

Probabilistic modelling

The modelling and the sensitivity analyses were performed using the software package @-RISK (Palisade Corporation, 1997). For this purpose, the parameters that were kept identical throughout all scenarios, as well as the experimentally-derived distributions described in the earlier sections that were subject to the sensitivity analyses, were used as input parameter.

Each model of one scenario was run for 20,000 iterations and 25 simulations, indicating that the model calculated the probability

Results

All scenarios were performed for two eating occasions (breakfast and lunch). However, the predicted risk at lunch was similar or lower than the predicted risk at breakfast and led to the same conclusions regarding the sensitivity analyses. Therefore, only the predicted risks at breakfast are shown in this paper.

Discussion

To determine the risk arising from unintended exposure to food allergens using a probabilistic model is a step forward from the deterministic approach, since a probabilistic model gives an estimate of the proportion of the population at risk of having an allergic reaction. Particularly in scenarios where a zero risk will not be attainable, the most accurate estimation of the (residual) risk and remaining uncertainties is crucial, and probabilistic quantitative risk assessment is a tool

Acknowlegdements

The authors would like to thank the members of Expert Group on the Determination of Eliciting Doses of the Food Allergy Task Force of ILSI Europe for their input. Furthermore, ILSI Europe and the Ministry of Health, Welfare and Sports of The Netherlands are acknowledged for financially supporting this research.

This work was commissioned jointly by the Food Allergy Task Force of the European branch of the International Life Sciences Institute (ILSI Europe) and the Ministry of Health, Welfare and

References (17)

There are more references available in the full text version of this article.

Cited by (0)

View full text