A non-parametric graphical representation of the relationship between survival and the occurrence of an event: application to responder versus non-responder bias.

Reports of cancer clinical trials often attempt to evaluate whether tumour response is associated with prolonged survival. Proper analysis requires accounting for the time-dependent nature of response status. We review a valid and relatively simple method of significance testing for this problem, and develop a corresponding non-parametric method for displaying the association between survival and occurrence of response. The new method applies to many other clinical problems involving representation and the association between survival and the occurrence of an event. We illustrate the method with data from two clinical trials.