I have question regarding kaplan meier curve to compare two curve

I have Event variable ( which the patient have relapse or not ) the total number is 51 we have 16 events ( relapse ) and 35 censored . and when I run the kaplan meier : progression free survival analysis which is = 49% by the end of period of our observation.

so I want to group this survival by the "type of relapse" The variable I used for grouping is: Type of relapse the total number is 16 ether (regional recurrence (n=15) or Distant recurrence ( n=1))

the problem here when I run the kaplan meier curve by using the type of relapse as grouping variable ( which include the information of only 16 patients who relapsed ( regional recurrence (n=15) or Distant recurrence ( n=1) ) so the censored patient where excluded and the result from two curves ( reginal or distains = 0% by the end of study instead of 49% )

so the question is what the best way to compare the two curves (regional recurrence Distant recurrence ) ? it's okay to be 0 and exclude the censored


1 Answer 1


In this particular situation, with only 1 distant recurrence, you aren't going to be able to say much useful in the comparison of recurrence times between loco-regional and distant recurrence. I'd recommend showing the recurrence-free survival curve for all individuals and highlighting the event (drop in the survival curve) for the 1 distant recurrence.

More generally, if loco-regional and distant recurrences could be considered mutually exclusive events, the best approach is to model them as competing risks, as explained for example in the R competing survival events vignette. The vignette notes:

A common mistake with competing risks is to use the Kaplan-Meier separately on each event type while treating other event types as censored.

That argues against your suggestion of simply censoring the observation time of the distant recurrence.

Further complicating this situation is that loco-regional and distant recurrences aren't necessarily mutually exclusive in practice. You might find one first on physical examination, but then find the other with technologies like PET scans. Then you need to apply your understanding of the subject matter and think carefully about the details of the hypothesis you want to test.


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.