Review of the regulatory management of food allergens

Allergen thresholds

An allergen threshold is defined in practical terms as the amount of a specific food that would elicit mild, objective symptoms in highly sensitive individuals. The amount of food capable of eliciting a reaction is variable, possibly over an order of magnitude or more between different individuals, with the same type of food allergy. Many factors contribute to this variability. Intra-individual variability may also occur as a result of extrinsic factors, such as exercise, alcohol and concurrent infection. Also the thresholds for different allergenic foods are likely to be different due to differences in the inherent potency of allergens (Taylor et al., 2002).

Where allergenic ingredients are not deliberately added, the reality of food production and processing means that cross-contamination may occur in the supply chain or during food manufacturing. Information on thresholds is critical for developing and maintaining effective allergen control strategies.

7.1 Clinical data

Individual clinical thresholds lie between the No Observed Adverse Effect Level (NOAEL), the highest dose observed not to produce any adverse effect and the Lowest Observed Adverse Effect Level (LOAEL), the lowest dose that is observed to produce an adverse effect (Crevel et al.,2008).

However, the exact point is difficult to determine due to the limitations of dose selection and the sensitivity of clinical measuring techniques. There is also clinical data suggesting that individual thresholds may vary over time. These variables, together with the limitations of human studies, make it unlikely that absolute experimental thresholds for food allergens can be obtained. Because low doses are sometimes used in the diagnosis of food allergies, clinical data could be used to determine the LOAELs for a number of food allergens. Analysis of such data indicates that a wide dose range exists among patients allergic to specific foods. However, the estimation of thresholds from this data was not possible due to the use of different procedure for performing DBPCFC and the lack of NOAEL data (Taylor et al., 2002).
In addition to differences in clinical testing protocols, the data is often based on a few individuals in any one study. There is also the question of whether the patients selected for these studies are representative of the entire population of individuals with allergies to that specific food since most clinics exclude the seriously affected patients (i.e. with a history of anaphylactic shock). Also, a number of factors may affect threshold levels, including exercise, disease, concurrent seasonal allergy and pharmaceutical treatments (Taylor et al., 2009).
A prerequisite to setting thresholds is the systematic collection of clinical food challenge data, using consistent protocols from a representative sample of the full range of food allergic individuals including those at greatest risk. To achieve this, a consensus protocol for clinical studies was developed to facilitate the comparative analysis of data from various sources (Bindslev-Jensen et al., 2004; Taylor et al., 2004). A common clinical testing procedure using low doses of various food allergens is critical to facilitating the combined use of data from multiple sources (Crevel et al., 2008). Work is underway in allergy clinics, mainly in Europe, to generate and analyse the data needed to establish individual and public thresholds.
At the population level, the distribution of thresholds from a range of allergic individuals representing the allergic population, can be generated from individual threshold data. Emerging information from statistical modelling studies, using data sets from published older studies and clinical records, on peanut allergy provides an insight into the feasibility of using threshold distribution to establish population thresholds (Taylor et al., 2009; Taylor et al., 2010). The latter study analysed a large clinical dataset (obtained from University Hospital, Nancy, France) where diagnostic peanut challenges had been conducted on all prospective peanut-allergic patients at that clinic using a consistent challenge protocol over a period of more than 10 years. The study confirmed the usefulness of the approach to predict the eliciting doses .This is an important first step in establishing the evidence base to underpin allergen risk assessment and allergen control measures in the food industry.

Probabilistic modelling is considered to be the most promising approach to allergen risk assessment (Madsen et al., 2009). Probablistic modelling can be used to estimate the possible impact of inadvertent allergen residues in food products (Spanjersberg, 2010). However, the methodology requires quantitative data on consumption patterns within allergic consumer groups and levels of allergens in food as well as population thresholds. Inevitably, where solid data are not available, assumptions are made which may influence the outcome to various degrees (Spanjersberg et al., 2007; Kruizinga et al., 2008).

While significant progress has been made in the past few years in generating threshold data and in developing allergen risk assessment methodologies, there are still a number of areas that require further investigation. For example, the potential impact of food processing and food matrix on thresholds and on allergen detectability, has been recognised. There is also a need for accurate analytical data on levels of allergens in food that is supported by quality control and reference materials.
Another issue is how to assign numerical values for symptom frequency and severity during clinical testing, and how to incorporate this information into risk modelling (Mills et al., 2010). Ultimately, it needs to be acknowledged that, even through the most rigorous allergen control systems, zero risk is not achievable and, in this context, a community consensus on the acceptable level of risk is needed to effectively minimise precautionary labelling (Madsen et al., 2009; Madsen et al., 2010).

7.1.1 Conclusion

Significant advances have been made in the area of thresholds in the last decade including improved methodologies for gathering and analysing clinical data. Emerging evidence indicates that statistical modelling approaches can be used to establish population threshold levels. This is a critical step to underpin allergen risk assessment and guide allergen control measures in food manufacturing.

7.1.2 Recommendation

In collaboration with the Scientific Advisory Group, FSANZ to maintain a watching brief on scientific developments in the area of allergen thresholds.