2.2 Use of the NOAEL to quantify risks


Derivation of health-based guidance values, like the ADI or ARfD, using the NOAEL approach has its drawbacks. In determining the NOAEL the data from the entire dose-response curve are not considered but rather only the data from a single dose group showing no response compared to the control group. Another drawback is that the NOAEL is dependent on study design, including number of animals in the different dose groups and the experimental doses selected by the investigators.

Due to these drawbacks, the Benchmark Dose (BMD) approach has been developed. In this approach a dose-response model is fitted to the dose response data as derived from animal studies, and this model is used to estimate the dose belonging to a certain level of response, e.g. 5% increase in body weight. This level of response, the so called critical effect size (CES), corresponds to a level that is associated with a specified low incidence  of risk, generally in the range of 1% to 10%, of a health effect or the dose associated with a specified measure or change of a biological effect, e.g. 5%. In the first case (incidence) the corresponding dose is called the Benchmark Dose (BMD). For continuous effects (e.g. change in body weight, enzyme activity, red blood cell count), the corresponding dose is called the Critical Effect Dose (CED).

In WP3 dose-response analyses were performed using the effect modelling software PROAST, resulting in the derivation of BMDs / CEDs for numerous contaminants present in food. The results are described in Muri et. al (in preparation).

Integrating exposure and effect modelling into quantitative risk modelling

Within current risk assessment practices, the probabilistic approach, when used, is usually used to model either exposure or effect. For example an exposure distribution is typically compared with a point estimate such as a reference dose (RfD) or an acceptable daily intake (ADI). Using this approach only the variation and uncertainty in exposure is acknowledged, while possibly the variation or uncertainty in the other domain (effect) may be more prominent. A consequence of this could be that the attention is focused on only one domain, which may result in spending resources to better quantify variation or reduce uncertainty in that one domain, while the effort should go to the other domain.

To address this the best way is to deal with the variability and uncertainty in both areas together in one single probabilistic analysis. Therefore in WP3 an integrated probabilistic risk assessment (IPRA) model was built where variation and uncertainties in both exposure assessment and hazard characterization are included. In this tool exposure and effect modelling are both included in a probabilistic way. The aim is to specify the probability that a random individual will have an exposure high enough to cause a particular adverse effect. The result of the integration is a distribution of Individual Margins of Exposure (IMoE). Margin of exposure is the ratio between a dose at which a ‘measurable’ low adverse effect occurs (e.g. BMD, CED) and a human exposure level. In analogy, we define the Individual Margin of Exposure (IMoE) as:

IMoE = IBMD / IEXP

In this equation both parameters refer to levels for individual humans. In a population there are differences in both consumption patterns and contamination levels of relevant foods resulting in different exposure levels per individual (IEXP). However, individuale also differ in their sensitivity for adverse effects of compounds (e.g. children, pregnant women and elderly people may be more sensitive compared to a healthy 30-year old person). In IPRA this sensitivity is modelled resulting in individual BMDs or CEDs. The proportion of the IMOE distribution below unity (IBMD = IEXP) is then equal to the proportion of the population that has an exposure level higher than the dose that may cause a harmful effect. For further reading see van der Voet et al. 2007 {see references}.

A distribution of IMoEs fits into the discussion of risk prioritisation of compounds that are both genotoxic and carcinogenic, as discussed by EFSA. The IMoE (or MoE) is however also a promising tool when it comes to weighing risks of exposure to more than one chemical simultaneously in relation to risk management decisions. For example, it may be desirable to reduce the exposure (or the probability of a high exposure) to chemical X with a MoE of 90 because of an undesirable percentage of the population having an intake above e.g. the ADI. In some cases, however, a reduction in risk for chemical X may coincide with the increase of a possible other risk, e.g. elevated levels of chemical Y (for example, increase of mycotoxin levels due to reduced use of fungicides in agriculture). If chemical Y has a smaller MoE (e.g. 50), the risk management decision to reduce the risk of chemical X may be debatable.

The model for integration of exposure and effect modelling into a (distribution of) MOE(s) is available and will be applied in the different cases related to cumulative exposure and risk comparisons.

A semi-quantitative model for risk prioritization: The health impact appreciation system

The MOE in itself is not an adequate parameter for risk comparison or risk prioritization (the CES does not really account for the effect size). For instance one chemical may cause concern because of the large fraction of the population at risk, while another causes a more severe type of effect in a relative small fraction of the population.

Experts within WP3 have developed a semi-quantitative risk severity (or rather, a health impact appreciation) ranking system, which can be used to prioritize risks of different compounds, e.g. in risk-benefit and risk-risk considerations. This system is based on three parameters:

  1. The fraction of the population at risk
  2. The type of effect(s) expected to occur and
  3. The severity of these expected effect(s)

It should be noted that these parameters are not independent. Within the population at risk the individual exposures may vary from just exceeding the ADI or up to several fold above the ADI. But the higher the exposure the more severe the effects are likely to be and more (other) effects might become apparent. A risk manager needs information on all three parameters in order to make the appropriate decisions when it comes to prioritization or risk-benefit analyses. In addition this information will also enable the risk manager to decide on risk reduction measures weighing costs against possible risk reductions.

In the last stage of the SAFE FOODS project the potential of the model will be proven in a risk-risk and/or risk-benefit evaluation.