Progression-Free Survival vs Time To Progression (TTP) in Oncology Biostatistics
Survival analysis is one of the most important segments of most Oncology studies. In this blog, i discuss the differences between PFS and TTP and how to address them in Oncology Biostatistics
Darko Medin
7/11/20242 min read
In this article, i will discuss the use of PFS and TTP in Oncology studies, paying special attention to both Biological and Biostatistical features of these metrics. For those who are new to this topic, you may notice that both Progression - Free Survival and Time to Progression are progression based metrics. In oncology that usually means tumor growth, tumor spread from local environment to other parts of the body or deterioration of disease status or even death (be careful with death as part of the variable as its only contained in the PFS). Some Oncology approaches have tried to set the PFS or TTP as a surrogate for OS which is something to discuss. PFS may be a better surrogate but never a real replacement for OS (its generally higher predictive power of OS in comparison with TTP, mostly due to the fact that PFS includes death as an event and in TTP its censored).
And from the last sentence, you may immediately notice a difference. PFS is a composite variable, in which 'event' is considered the progression of disease, but death is also considered a PFS specific event. So either progression or death will be statistically treated as 'event' in the Survival analysis.
On the other hand in TTP, only progression is treated as an 'event' and death is considered as censor specific point in the survival analysis of TTP.
While PFS and TTP are similar metrics, their are very different Statistically, one is composite (PFS) variable and other is not composite variable (TTP).
In reality in most Oncology scenarios, PFS will be a superior metric to TTP, but as i mentioned there are scenarios where TTP may be superior. Statistically speaking, PFS is much easier to implement as well. You may use it in the typical censor based Survival setting such as Cox Proportional Hazards models or you may use the Competing risks models, it will work well in most scenarios. For TTP to be effective in case of any death events occurring, you may need to use the Competing risks models and similar models as they are more flexible in accounting the death events. For PFS you don't need to account for mortality as its contained in the composite statistical variable.
Keep in mind that while PFS may be superior to TTP because its composite variable in most Oncology Clinical trials and RWE studies, it should usually be paired with other variables such as OS (overall survival), RFS, recurrence free survival and ORR (objective response rates) to make the assessment more complete. Further, since Oncology treatment applicability is highly dependent on lowering the adverse events, adverse event variables may also be important to complement the PFS and TTP in various scenarios. Thank you for reading!
by Darko Medin