R help - Sep 2010 - extending survival curves past the last event using plot.survfit

Hello,

 

I'm using plot.survfit to plot cumulative incidence of an event.
Essentially, my code boils down to:

 

cox <-coxph(Surv(EVINF,STATUS) ~ strata(TREAT) + covariates, data=dat)

surv <- survfit(cox)

plot(surv,mark.time=F,fun="event")

 

Follow-up time extends to 54 weeks, but the last event occurs at week
30, and no more people are censored in between.  Is there a direct way
to extend the curves with a horizontal line to the end of follow-up (54
weeks), rather than stopping at the time of the last event (30 weeks)?

 

I've pretty much exhausted my search options and didn't see this in a
previous thread, so would really appreciate any help or thoughts!

 

Thanks in advance!

Amy

 


	[[alternative HTML version deleted]]

On Sep 22, 2010, at 8:15 PM, Krambrink, Amy M wrote:
> Hello,
>
>
>
> I'm using plot.survfit to plot cumulative incidence of an event.
> Essentially, my code boils down to:
>
> cox <-coxph(Surv(EVINF,STATUS) ~ strata(TREAT) + covariates, data=dat)
>
> surv <- survfit(cox)
>
> plot(surv,mark.time=F,fun="event")
>
> Follow-up time extends to 54 weeks, but the last event occurs at week
> 30, and no more people are censored in between.  Is there a direct way
> to extend the curves with a horizontal line to the end of follow-up  
> (54
> weeks), rather than stopping at the time of the last event (30 weeks)?
Not sure if it's the approved method, but this works:
 > ?survfit.object
 > ?survfit   # to get a working example since you did not provide one
 > lsurv2 <- survfit(Surv(time, status) ~ x, aml, type='fleming')
 > plot(lsurv2, lty=2:3, xmax=300)  # drats, no effect of xmax
 > str(lsurv2)      # so see the structure of the survfit object
 > lsurv2$time[21] <- 300         #add a time value
 > lsurv2$n.censor[21] <- 1       # mark as censoring time
 > lsurv2$strata[2] <- 11         # add to count of group 2
 >
 > plot(lsurv2, lty=2:3, xmax=300)  # horizontal line to 300 for group
2
>
>
>
> I've pretty much exhausted my search options and didn't see this in
a
> previous thread, so would really appreciate any help or thoughts!
>
>
>
> Thanks in advance!
>
> Amy
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
-- 
David Winsemius, MD
West Hartford, CT

On Sep 22, 2010, at 8:15 PM, Krambrink, Amy M wrote:
> Hello,
>
>
>
> I'm using plot.survfit to plot cumulative incidence of an event.
> Essentially, my code boils down to:
>
> cox <-coxph(Surv(EVINF,STATUS) ~ strata(TREAT) + covariates, data=dat)
>
> surv <- survfit(cox)
>
> plot(surv,mark.time=F,fun="event")
>
> Follow-up time extends to 54 weeks, but the last event occurs at week
> 30, and no more people are censored in between.  Is there a direct way
> to extend the curves with a horizontal line to the end of follow-up  
> (54
> weeks), rather than stopping at the time of the last event (30 weeks)?
The survfit.coxph function only records the survival curve at the death
times, so there is no data in the "surv" object about time 54.  (This
will change in my next bundle of updates, which are currently under
test.)  You will need to update the survival curves' output by hand.
For now this would be the easiest code for multiple strata, at least
that I can think of offhand.

plot(surv, mark.time=F, fun='event', xlim=c(0, 54))
for (i in 1:length(surv$strata)) { #number of curves
	temp <- surv[i]
	lines(c(max(temp$time), 54), 1- rep(min(temp$surv),2))
	}

Terry Therneau