median survival time using stci

[Laura Myles]

Dear Users,

I would like to estimate median survival time for a survival model in Stata15. My file has one line per record and 1 failure time. I was using the following approach:

stset surv, id(patid) failure(fail==1) scale(30.4)
sum _t, d
bysort sex: sum _t, d

A colleague has now mentioned that I should be able to use -stci- to achieve the same

stci, by(sex)

The numbers are not the same using -stci- so I am wondering if my original approach is wrong, although I am not fully sure how -stci- is estimating median time.

Thanks for your help,
Laura

[Marcos Almeida]

Providing results would entice a better approach.

That said, - stci - won't estimate the median survival time.

Rather, it estimates the confidence interval for the median survival time.

[Andrea Discacciati]

Laura, if you have censored survival times, using -summarise- to estimate the median survival time will give you an incorrect result (this is not 100% accurate, because it actually depends on how the censored survival times are distributed, but I'm digressing. Let's say that "in general", the estimated median survival time via -summarise- is incorrect).

-stci- estimates the median survival time "taking into account" censored survival times (similarly to what the KM does when estimates survival proportions. And, in fact, -stci- obtains the median from the KM survival curve (see the user's guide for more details). )

[Laura Myles]

Thank you both for your reply.

Andrea - could you say a bit more about what you mean by "-stci- estimates the median survival time "taking into account" censored survival times"? I assumed that summarise _t would also do that.

Is -stci- the best approach to estimate the median survival?

Thanks again!

[Andrea Discacciati]

No, -summarise- is unaware of the survival times being censored or not.

-stci- calculates the median survival time (and other percentiles) from the Kaplan-Meier estimates of the survival function or, as the user's guide puts it "The percentiles of survival times are obtained from S(t), the Kaplan–Meier product-limit estimate of the survivor function. The 25th percentile, for instance, is obtained as the minimum value of t such that S(t) 0.75."

The method implemented in -stci- is a nonparametric approach to estimate the median survival time (non-parametric because it is -again- based on the KM estimator which is nonparametric).
A second approach is o model the survival times using a certain parametric distribution (log-normal, Weibull, chi-squared etc etc) and to estimate the median survival time from there.
A third option (which is actually an extension of/related to the second one) is to follow a "flexible parametric" approach using the -stpm2- command (from SSC).

[suganthiny kumarasamy]

Median Survival time
stsum
stsum, by()