survival analysis up to specific timepoint

[Gernhardt Rheinholzen]

Dear all,

I have survival data with a follow-up of 30 months, however I would like to compare survival until 24 months. I would like to do so using Kaplan-Meier curves and log-rank testing.

I was trying around already with the stsplit command, but so far did not get the hang of it, does anyone have suggestions for me?

Best

[Clyde Schechter]

The way to do this is to censor at 24 months any observation that has not failed up to that point. So if your original survival time variable is called time and your original failure variable (0 = censored, 1 = failed) is called fail, you can do something like this:

gen time2 = min(time, 24)
gen fail2 = fail
replace fail2 = 0 if time > 24 & !missing(time)
stset time2, failure(fail2) // PERHAPS OTHER OPTIONS

and then run your survival analysis command (-stcox-, -stcrreg-, -sts test-, whatever).

[Paul Lambert]

Clyde's solution should work well, but perhaps an easier option is to use the -exit()- option of -stset-

stset time, failure(fail=1) exit(time 24)