Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 5 down vote favorite
2

Hello to all the biostatisticians and epidemiologists out there,

I have looked up and down for a standard definition of recurrence-free survival, and the issue I'm having is determining if the standard includes DEATH as an event or not. I have seen some studies censor at death and others include death along with recurrences as events. If death is traditionally included as an event, what is the rationale for that? It certainly is not a recurrence. Also, is there a difference between relapse-free survival and recurrence-free survival?

Thanks for your help!

share|improve this question
I don't know the answers but would like to comment. Relapse-free and recurrence-free sure sound the same. I of course agree that death can not be considered a recurrent event unless you count someone whose heart stopped and was revived. I think death makes sense to use as a censoring time because it prevent the recurrence of many future events for the subject. – Michael Chernick Jul 26 '12 at 22:41

2 Answers

up vote 2 down vote accepted

As you have seen, both definitions are acceptable, mostly because both options have problems. Even the FDA does not seem to express a preference. Quoting from Guidance for Industry Clinical Trial Endpoints for the Approval of Cancer Drugs and Biologics (DFS is Disease Free Survival):

The definition of DFS can be complicated, particularly when deaths are noted without prior tumor progression documentation. These events can be scored either as disease recurrences or as censored events. Although all methods for statistical analysis of deaths have some limitations, considering all deaths (deaths from all causes) as recurrences can minimize bias. DFS can be overestimated using this definition, especially in patients who die after a long period without observation. Bias can be introduced if the frequency of long-term follow-up visits is dissimilar between the study arms or if dropouts are not random because of toxicity. Some analyses count cancer-related deaths as DFS events and censor noncancer deaths. This method can introduce bias in the attribution of the cause of death. Furthermore, any method that censors patients, whether at death or at the last visit, assumes that the censored patients have the same risk of recurrence as noncensored patients.

share|improve this answer
I suppose the take-away is that there is no convention and, most importantly, no one will strongly object if I use one definition or the other. – JJM Jul 27 '12 at 15:06
up vote 2 down vote

The issue of death in any non-fatal disease is a problem in many survival analysis studies. I have generally seen this handled two ways, and have not seen anyone arguing that there is a particular dominant or unified standard.

The first way is to consider death to be one of your outcomes. So your event becomes "Event OR All-cause Mortality". This may be more intuitive and, conveniently, if you think your disease may be driving some of those deaths, may capture additional information. It's also the most amenable to a straightforward survival analysis.

The second method is to treat death as a competing risk. Then, the usual approach is to treat death-events as censored for the outcome of interest at time of death, and analyze from there. This essentially asks "What would your hazard of death have been, had you not died for some other reason?". Especially if you think death is something of an "uninteresting" outcome, this may be a workable approach.

As both are pretty approachable using standard software, I'd consider doing both, and seeing if your estimates end up being substantially different.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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