[Contribution of statistical analytic tools to risk assessment]

Risk assessment is an important aspect of health safety. The main basis of information comes from the identification of adverse effects observed in animal studies, in vitro models, or in silico computer models. These data are used to construct a proof scaffold with an inherent double uncertainty about its clinical pertinence and the probability of occurrence in man. This uncertainty can be reduced with statistical analytical tools, particularly factorial analysis and principal component analysis. Factorial analysis describes a group of variables using a linear combination of underlying common factors associated with a single variable summarizing the specific contribution of each initial variable, thus providing evidence demonstrating the probability of the undesirable effect under consideration. Principal component analysis consists in expressing a group of variables (undesirable effects) by linear combinations of uncorrelated factors. These factors account for a more or less extensive proportion of the data variability thus limiting loss of data pertinence.