[Lyell syndrome and its attributability to medications: methodological problems].

Lyell's syndrome, the most serious of the secondary effects of drugs, is a good example of difficult methodology in the establishment of a causal link between the observed accident and a particular drug (association). In each patient, in the absence of a reliable test, in vivo or in vitro, this association rests on a probability step based on comparison of essentially chronological criteria. On a country scale, accumulation of cases attributed to one product only allows a calculation of the risks if the number is known of the patients exposed to the drug. It is possible to make only indirect estimations of this "denominator", calculated from the sales figures. Only prospective epidemiological studies by the method "cases/controls" permits rigorous calculation of the risks of the drugs implicated in the etiology of Lyell's syndrome. They do not control the problem of association in individual cases.