Organization and analysis of comparative clinical trials entailing prolonged surveillance of patients

For several years the Medical Research Council has encouraged the performance of multicentric clinical trials in cancerology, as well as the publication of their results. Recent methodological progress in the organisation and analysis of these trials are the objective of this article. Edited by a group of British and American statisticians it is, in fact, addressed to non-statisticians. The first part deals with the programme. The second, which is concerned with the analysis, gives the methods for establishing the curves of survival and for comparing them by the logrank test, methods that are illustrated by a numerical example. The way to interpret the results of a therapeutic trial is discussed with reference to 2 concrete examples. This article is applicable to all clinical trials in which a study is made of the survival, that is to say, the time gap up to the death, but it could also apply to trials studying some other event such as: the first relapse, metastasis, recurrence, and, apart from cancer, cerebral vascular accident, thrombosis, rejection of a graft, or death due to some particular cause. The first part collects together ideas, most of which have been published, but only scantily. The second is the first simple publication on the data on survival intended for non-statisticians. The article also indicates how to compare the survival curves by taking into account knowledge of an important prognostic factor, such as the stage or the histology.