ABSTRACT
In classical survival analysis, one implicitly assumes that all individuals under study will experience the event of interest if the follow-up is long enough. A second consequence is that quite obviously, in the presence of a cure fraction, this cure rate in itself becomes a parameter of interest. Indeed, if, for example, a new treatment lead to a fraction of cure people would like to be able to quantify it. Going one step further, in a clinical trial comparing two treatments people would be interested by the effect of the new experimental treatment both on the fraction of cure as well as on the time-to-event for the uncured patients. The work on cure models have mainly been developed according to two different axes, leading to two main families of cure models. The first one starts from the idea that the population is actually a mixture of two sub-populations.
