I am trying to make sense of what disease-free survival is (still new to medical research).

I was reading this article on the survival of patients after liver transplantation. In it, the author includes KM curves where one is overall survival (OS) and the other, disease-free survival (DFS). My initial understanding of DFS is survival time of patient for non-recurrence cohort. However, if that's the case then, the survival curve should be higher in DFS as compared to OS. The plots show OS being having better survival probability than DFS.

Does DFS mean:

  • Still looking at death as event of interest, in addition patients that had recurred would have time-to-event as time to recurrence instead of time to death? or..
  • Still looking at death as event of interest, in addition patients that had recurred would be censored?
  • $\begingroup$ If I understand it correctly, it's your first option. In slightly different terms: Definition of an event in OS: death. Definition of an event in DFS: diagnosis with disease or death. The definition of censoring applies to both OS and DFS equally: a patients survival time is censored, if the event did not occur during the observation time. $\endgroup$
    – LuckyPal
    Aug 20, 2020 at 8:31

1 Answer 1


Disease-free survival (DFS) does not use death as the event of interest. The event for DFS is the first clinical evidence of the return of disease. That means that the individual has lived at least that long, so overall survival (OS) for such patients will be no shorter than DFS. Thus the DFS curve tends to show "worse" survival than OS, but that's because they are measuring different things.

There are a couple of potential "gotchas" with DFS. One is the specific timing, as the return of disease is often detected at a scheduled clinical follow-up but the actual return of disease could have been at any time since the previous clinical visit. So DFS values aren't always very precise, and might better be handled as interval-censored data.

The second is how any particular study handles a death that happens before clinically detected return of disease. If death is found at autopsy to have been from that disease then the time of death could be treated as the time of disease return. If death is from another cause the time of death is an event for overall-survival analysis but is usually a censoring time for DFS analysis: the time to return of disease would have been longer than than the time to death.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.