Kaplan Meier survival description output - no median result

[Amaranta Losey]

hi all,
I am using Stat 16 on Mac.
I am calculating Kaplan Meier survival curves for 2 groups of patients who do or do not have a genetic mutation.
But after I set the survival data (stset), I am getting what seems like an error in my descriptive statistics for the survival data.
Ex: nothing listed for 25%, 50%, 75% survival times.
Is this because my failure rate is low compared to the time at risk?
Is there any way around this to determine the median survival time for each group?
Thank you!

stsum, by(hyper)

failure _d: recur2 == 1
analysis time _t: betweendx2censor

| Incidence Number of |------ Survival time -----|
hyperm~h | Time at risk rate subjects 25% 50% 75%
---------+---------------------------------------------------------------------
No | 352,899 .0000878 260 . . .
Yes | 87,166 .0001262 63 1943 . .
---------+---------------------------------------------------------------------
Total | 440,065 .0000954 323 . . .

[Amaranta Losey]

Same ?error? when I use stci:

stci

failure _d: recur2 == 1
analysis time _t: betweendx2censor

| Number of
| subjects 50% Std. Err. [95% Conf. Interval]
-------------+-------------------------------------------------------------
Total | 323 . . . .

[Clyde Schechter]

Yes, of course. What these analyses are telling you is that fewer than 25% of your patients have died by the latest time of observation in your study. Fortunate for them; for you, perhaps not so much. No, there is nothing you can do to get these values as it is simply anybody's guess how long it will be before that many of the patients die.

If you are just playing around with the data for fun, or perhaps trying to get an estimate to use in planning some future study, I suppose you could fit some plausible parametric survival model to the data and then use the parametric formula of the model to calculate an extrapolated 25th, 50th, and 75th percentile based on that model. But you would be basing those calculations on an unreliable extrapolation from fitting an arbitrarily assumed parametric model to data restricted to a small part of its left tail. I don't think anybody would take those results seriously.

[Amaranta Losey]

thank you!