Survival Time Endpoints in Oncology Trials

Giannarelli D, Bria E, Milella M

Survival times are considered the most adopted endpoints in oncology trials both in phase II and phase III designs. Overall survival is considered as the gold standard for measuring benefit from the regulatory point of view:  It can be easily measured, it is reliable and usually respond to patients’, scientific community and regulatory expectations and requirements. Progression-free survival is also a very attractive composite endpoint but it is subjected to possible biases due to the frequency and accuracy of assessment (which may generate ascertain biases) and to some of the trial design features. Its association (and correlation) with overall survival is not always obvious, although its role in supporting the evaluation of the anticancer effect is unquestionable.

Conventionally, survival analysis is performed by assuming as exponential the survival distribution and as proportional the hazard overtime. With these assumptions, the summarized median survival may be reliably and consistently estimated before all survival times are observed.

The gain (if any) due to the experimental drug is often measured by differences in median survival and hazard ratio, which both well perform if the above specified hypotheses do exist.

New drugs with different mechanism of action, the landscape of many therapeutics alternatives and different clinical scenarios, may undermine such approach. The crossover possibility, mandatory for ethical aspects, confound the ‘real’ effect of the experimental therapy and the adjustment for this opportunity needs to be implemented in the statistical analysis of results. 

Moreover, different metrics can be used, such as restricted mean survival time, cure fraction and survival rates at prespecified timepoints (as well as landmark analyses) can be considered alternative endpoints, especially in immunotherapeutic studies in which a delayed effect is reasonably expected.

From the economic sustainability standpoint, decisions need to be supported by a definition of drug value based on well designed clinical trial and strong measures of magnitude of clinical benefit. Planned analyses and reported results need to be implemented in this view.