estimating 5 year survival

[Arnaud Lesegretain]

Hi,
I have a dataset with survival months and event (dead/alive). I have been able to stset it and calculate various survival statistics from STATA menu. However, I could not find out how to estimate 5 year survival rate in my data set? Seems simple on the face of it but just cannot figure this out. Many thanks Arnaud

[Maarten Buis]

You can read that of the survivor function. If you type sts list, the survivor function gets displayed on screen. You then look up the survivor function for 60 months (5 years). If no event happend at exactly 60 months, then that will not appear in your table, but this function is a step function, so the time point immediatly preceding 60 is the estimate you are looking for. For example:

. sysuse cancer
(Patient survival in drug trial)

. sts list

        Failure _d: died
  Analysis time _t: studytime

Kaplan–Meier survivor function

             At                  Survivor      Std.
  Time     risk   Fail   Lost    function     error     [95% conf. int.]
------------------------------------------------------------------------
     1       48      2      0      0.9583    0.0288     0.8435    0.9894
     2       46      1      0      0.9375    0.0349     0.8186    0.9794
     3       45      1      0      0.9167    0.0399     0.7930    0.9679
     4       44      2      0      0.8750    0.0477     0.7427    0.9418
     5       42      2      0      0.8333    0.0538     0.6943    0.9129
     6       40      2      1      0.7917    0.0586     0.6474    0.8820
     7       37      1      0      0.7703    0.0608     0.6236    0.8656
     8       36      3      1      0.7061    0.0661     0.5546    0.8143
     9       32      0      1      0.7061    0.0661     0.5546    0.8143
    10       31      1      1      0.6833    0.0678     0.5302    0.7957
    11       29      2      1      0.6362    0.0708     0.4807    0.7564
    12       26      2      0      0.5872    0.0733     0.4304    0.7145
    13       24      1      0      0.5628    0.0742     0.4060    0.6931
    15       23      1      1      0.5383    0.0749     0.3821    0.6712
    16       21      1      0      0.5127    0.0756     0.3570    0.6483
    17       20      1      1      0.4870    0.0761     0.3326    0.6249
    19       18      0      2      0.4870    0.0761     0.3326    0.6249
    20       16      0      1      0.4870    0.0761     0.3326    0.6249
    22       15      2      0      0.4221    0.0786     0.2680    0.5684
    23       13      2      0      0.3572    0.0788     0.2087    0.5083
    24       11      1      0      0.3247    0.0780     0.1809    0.4771
    25       10      1      1      0.2922    0.0767     0.1543    0.4449
    28        8      1      1      0.2557    0.0753     0.1247    0.4093
    32        6      0      2      0.2557    0.0753     0.1247    0.4093
    33        4      1      0      0.1918    0.0791     0.0676    0.3634
    34        3      0      1      0.1918    0.0791     0.0676    0.3634
    35        2      0      1      0.1918    0.0791     0.0676    0.3634
    39        1      0      1      0.1918    0.0791     0.0676    0.3634
------------------------------------------------------------------------

If we want to know the one year survival rate then that is 0.5872 (we can read that in the table at 12 months), and if we want to know the 1.5 year survival rate then that is 0.4870 (18 months is not in the table, but it is a step function so the closest month below 18 months is the row we want to look at, in this case 17 months).

[Arnaud Lesegretain]

thank you Maarten

[Arnaud Lesegretain]

Hi, a quick follow-up question: is there a way to automate this across many variables and create a nice table in Stata vs having to stlist each covariate, manually look and having to assemble table manually?

[Maarten Buis]

I don't understand: you don't stlist a covariate; the time is till event is not a covariate/explanatory/independent/x-variable but the explained/dependent/y-variable.

Maybe you want to know the 5-year survival rates for different groups?

[Maarten Buis]

Here is an example for the one-year survival rate

tempfile temp
sysuse cancer, clear
sts list, by(drug) saving(`temp')
frame create tab
frame change tab
use `temp'
drop if time > 12
bys drug (time) : keep if _n == _N
decode drug, gen(row)
mkmat survivor lb ub, matrix(res) rownames(row)
matlist res, format(%9.2g)
frame change default
frame drop tab

[Arnaud Lesegretain]

Hello Maarteen,

I tried using the code above for 2 groups of variables : Prognostic Factor first (low, intermediate, high) and Race (3 values);
tempfile temp
sts list, by(prog race) saving(`temp')
frame create tab
frame change tab
use `temp'
drop if time > 24
bys prog race (time) : keep if _n == _N
decode prog race, gen(row)
mkmat survivor lb ub, matrix(res) rownames(row)
matlist res, format(%9.4g)
frame change default
frame drop tab

Stata is telling me:
decode prog race, gen(row)
too many variables specified

Am I doing anything wrong or is this truly a limitation?
Is there a way around this?