Median Follow-up time

[frederik baumgermany]

Hello,
could somebody give me an advice on how to calculate median follow-up time?
Thank you very much for your help.
Diagdat = Date of diagnosis
lBeo = last observation

* Example generated by -dataex-. To install: ssc install dataex
clear
input int(Diagdat lBeo) byte Status long id
20313 22022 1   18230
18396 19362 0   14134
20481 22022 1   18551
19423 22022 1   16346
18976 22022 1   15389
18365 21655 0 8350302
18044 20103 0   13457
19424 22116 1   16342
20178 20943 0 8384912
19309 20181 1 8366678
end
format %td Diagdat
format %tdnn/dd/CCYY lBeo
label values Status Status
label def Status 0 "alive", modify
label def Status 1 "died", modify

stset lBeo, id(id) origin (Diagdat) failure(Status=1) scale (30.25)

[Joseph Coveney]

isid id
quietly summarize _t, detail
display in smcl as text "Median follow-up (months): " as result %3.0f r(p50)

[frederik baumgermany]

Thank you very much for your quick help. This helped a lot, but I'am not sure wether this is the optimal approach to calculate it.
Is there also a Method to calculate reverse Kaplan-Meier estimate as suggested by Schemper and Smith (Control Clin Trials, 1996,17:343–346) and explained for R on this website:
https://publicifsv.sund.ku.dk/~tag/T...lAnalysis.html ( Chapter: Estimation of the median follow-up time)

[Joseph Coveney]

. . . I'am not sure wether this is the optimal approach to calculate it.
Is there also a Method to calculate reverse Kaplan-Meier estimate . . .

Are you looking to estimate median survival time? I'm not sure anymore what it is that you're seeking.

You asked for median follow-up interval, and that's what I showed you.

Follow-up and survival are supposed to be independent; one of the assumptions.

[Joseph Coveney]

OK. They're looking for potential follow-up time, and estimate it by reversing the sense of the censoring (failure) variable.

I'm not able to download in a readable format the GBSG2 R dataset that they reference in order to verify my crack at the Stata equivalent code, and I'm not going to circumvent the paywall for the Schemper and Smith paper, but you can get an idea of what to do from this PharmaSUG proceedings paper from last year by industrial scientists Nikita Sathish and Chia-Ling Ally Wu.

[Joseph Coveney]

Is there also a Method to calculate reverse Kaplan-Meier estimate as . . . explained for R on this website

Yes.

As mentioned, flip the sense of the censoring (failure) variable.

So, instead of

stset lBeo, id(id) origin (Diagdat) failure(Status=1) scale (30.25)

you'd

isid id
quietly stset lBeo, /* id(id) */ origin(Diagdat) failure(Status==0) scale(30.25)
stci

[frederik baumgermany]

Thank you so much for your help!!