background image
24/27
It is worth to notice than the current approach can be iteratively extended to the study of the
combined effects of s
D
, s
Bi
and s
K
by replacing the conditional distribution
(
)
*
,
D
v
Fo s
K
f
in the current
approach by
(
)
*
,
,
D
Bi
v
Fo s
s
K
f
, where the latter is calculated from Equation (35).
6 Conclusions

The prediction of the contamination of food by packaging substances is encouraged by the
current regulation in Europe as an alternative to costly and time consuming experimental testing.
Recent crises (e.g. isopropylthioxanthone in baby milk in Europe) and new developments
(recycled materials, active packaging) have besides heightened interest in the issue of scientific
uncertainty, and on how it should be handled in risk assessment and decision-making. As
acknowledged in the Working Principles for Risk Analysis recently adopted by the Codex
Alimentarius Commission "Constraints, uncertainties and assumptions having an impact on the
risk assessment should be explicitly considered at each step in the risk assessment and
documented in a transparent manner. Expression of uncertainty or variability in risk estimates
may be qualitative or quantitative, but should be quantified to the extent that is scientifically
achievable". (Codex, 2005).
This work proposes a general framework to quantify and analyze the effects of uncertainty in
mass transport models used to predict the contamination of food products. The applications were
focused on the design of packaging materials, but the proposed approach is enough general to be
applied to sanitary surveys or controls by enforcement laboratories. Two typical users of these
approaches were particularly envisioned i) the producer of finished packaging material, who
formulates the material (e.g. a film) and the ii) the packaging filler (food industry), who puts the
packaged food product on the market. The first user has a good knowledge on the formulation of
its material but an imprecise description of the final use of the material. Its objective is to
demonstrate that is product is compliant for most of conditions (contact time, temperature, and
type of food). By contrast, the second user is ignorant of the formulation of the material but has a
good knowledge on the final use of the material and on possible dependencies between factors
that enhance the desorption of packaging substances (temperature, fatty contact...). Its objective
is to put on the market on food product, which is at minimum compliant, but also to demonstrate
that its product is safer than the comparable product from its competitor. Besides, both users are
subjected to scientific uncertainty inherent to insufficiently tabulated transport properties,
imprecise extrapolation from semi-empirical models of transport properties.
Physical transport models were first extensively reviewed for both monolayer and multilayer
materials. They were adimensioned to minimize the number of input parameters and to highlight
the physical quantities, whose uncertainty may have similar effect on the prediction. From a
phenomenological description of the desorption of packaging substances, it was illustrated how
an association of mass transport resistance could be replaced by a simpler association, for which
calculations are achievable and whose desorption rates overestimate the "true" desorption rate.
The complete approach was described as multi-attribute value trees and applied to both
monolayer and multilayer materials. Since this approach is progressive and generate a rapid
overview of the parameters and related uncertainty, which affects the predictions, it is particularly
appropriate i) for food industry users to audit its providers, ii) for the packaging industry to
simulate mass transport in very complex multilayer structures. Besides, since this worst-case