1. ACUTEX Software

ACUTEX software is designed by INERIS to allow model fitting and predictions of foreseeable health impact of accidental chemical release. This can be used as a supportive tool to derive EU AETLs that can be adapted to the various national situations as land use planning or emergency situations, which require some assessment of the risk of resulting effects.

Assessing the risk of an adverse effect after exposure to a chemical substance can be done with the help of mathematical modelling. Depending on the data, several models are available. For now, this software can only handle categorical data. This applies to binary response data (e.g., mortality: dead versus alive; irreversible toxicity: presence or absence of congenital abnormalities, infertility, blindness, severe burns, necrosis, etc.; reversible effects: apparition or not of a transient unconsciousness, etc.). It also applies to graded responses (e.g., for instance, four levels of severity: no effect, mild effect, severe effects and death).

This software is divided in two parts: BinReg (for binary data) and CatReg (for categorical data when the number of categories is strictly greater than 2).

Data requirements

Some models require more data than others do. However, to run any of them, the following is the minimum requirement:

• N: the number of groups of individuals (usually animals)
• C_i: the exposure concentration of group i
• n_i: the number of individuals of group i exposed at concentration C_i.
• y_i: the number of individuals (among n_i) with an adverse response (usually death) after exposure.
• τ_i : the exposure duration of group i.

Heuristics behind the models

For any of these models, it is assumed that the individual responses to exposures are independent to each other. Therefore, the Binomial model is appropriate to model the actual number of individuals that have an adverse response. That is: where pi is the probability of observing an adverse response in group i.

On the other hand, one can observe that the biological response to an exposure depends on the dose received. Hence the probability p_i is a function of the dose D_i. We can write: p_i =F(D_i) where F is a cumulative distribution function. The neatest discriminations between the different models are about the cumulative distribution function F used and/or the definition of dose adopted.