The determination of the "optimal" dose is an essential step in the development of a molecule. In the case of anti-thrombotic agents, the search for this "optimal" dose is based on dose-effect relationships on biological criteria in phase I, and, often, radiological criteria in phase II trials. The main objective of these dose studies is not to directly evaluate the benefit-risk ratio of the molecule under development, but to find the dose which will be tested in phase II to estimate the benefit-risk ratio. Errors of choice of dosage observed at the end of phase III trials may be due to problems of extrapolability of the results of the dose studies due to too strict a selection of subjects included and therefore not representative of the target population of the new treatment or to the use of intermediary criteria for the evaluation of the antithrombotic effect. However, these dosage errors are still mainly due to an inadequate search for the "optimal" dose despite the fact that the ethnical and economic consequences are not negligible.