[Problems raised by interim analysis in therapeutic trials].

Interim analyses are analyses performed before complete collection of information on all patients in the study is obtained. The precocity and the repetiveness of these analyses considerably increase the risk of false conclusions of a significant difference. For instance, if ten statistical tests are carried out during the course of a trial at the alpha = 5 per cent nominal level, the overall significance level increases to 19 per cent. Statistical methods allowing repeated analyses are described here: Classical Sequential Analysis (Armitage) and Group Sequential Analysis (Pocok).