Statistical Aspects of Planning Clinical Experiments in Cancer Therapy

An important criterion for assessing the value of any new technique of cancer therapy is the proportion of long-term symptom-free survivors achieved by its use.This is not the only possible criterion, and in those sites where permanent cures are still exceedingly rare it may be more appropriate to find some way of measuring the palliative value of the treatment.There are, however, undoubtedly many sites in which the proportion of long-term survivors is a good criterion of success.I t is therefore important to find ways of planning and evaluating a clinical test of any new treatment method which shall determine this proportion of long-term survivors as accurately and rapidly as possible.Since it is impossible to say in any individual case whether an initially good response will be maintained, it is necessary to study the mortality experience of large groups of patients treated a long time ago in order to discover and utilise any regularities presented by such data.Some ten years ago it was shown (1, 2) that the survival expectation of the unsuccessfully treated cases among cancer patients treated by radiotherapy or by surgery followed by radiotherapy (i.e. those who subsequently died with a persistence or recurrence of the disease) did, in fact, show a striking regularity.Although