Overlying 3 sts graphs of different outcomes with different FU time

[Isabel Hostettler]

Hi there,

I am trying to put three sts graphs into one. There are three different events at follow-up I am interested in, which can be at different time points. I get the survival curves for each event:
Event 1:

stset TimetoICH_or_FU_years, failure 
sts graph

Event 2:

stset TimetoIS_or_FU_years, failure(IS_FU)
sts graph

Event3:

stset TimetoreccSAH_or_FU_years, failure(recurrent_cSAH) 
sts graph

This gives me three separate KM curves. I would like to get these exact curves all into one graph, not changing/combining anything and if possible labelling them accordingly (ICH on FU, IS on FU and recurrent cSAH on FU). As the event times differ the three time variable cannot be combined. I have checked previous posts and have tried different things including addplot, which didnt work.I have tried using something along the lines of the following posts:
https://www.statalist.org/forums/for...ts-of-the-data
and
https://www.statalist.org/forums/for...-than-combined

I am using stata 15.1.

Thank you very much for any suggestions.

Best,

Isabel

[Steve Samuels]

You tell us little about your data, so I'm assuming that all these statements were run on the same data set--i.e. that people were at risk for multiple events. If not, the modifications needed for the code below should be obvious.

tempfile t1 t2
stset TimetoICH_or_FU_years, failure(??)
gen event =1
save `t1', replace

stset, clear
stset TimetoIS_or_FU_years, failure(IS_FU)
replace event = 2
save `t2', replace

stset , clear
stset TimetoreccSAH_or_FU_years, failure(recurrent_cSAH)
replace event = 3

append using `t1'
append using `t2'
keep event _*  // _* are the variables created by stset
sts graph, by(event)

As FAQ 12 states, saying that something "didn't work" is not helpful. Please in future posts, show what didn't work and provide a sample of data if necessary for us to understand what you are dealing with.

[Isabel Hostettler]

Apologies, of course I can provide more informations. Everything was ran on the same data set and yes they are at risk for multiple/different events. We were interested in three outcomes=events: ICH=intracerebral haemorrhgae, IS=ischaemic stroke and recurrent cSAH= convexity subarachnoid haemorrhage
I have used:

tempfile t1 t2
stset TimetoICH_or_FU_years, failure(ICH_FU)
gen event =1
save d1, replace

stset, clear
stset TimetoIS_or_FU_years, failure(IS_FU)
replace event = 2
save d2, replace

stset , clear
stset TimetoreccSAH_or_FU_years, failure(recurrent_cSAH)
replace event = 3

append using d1
append using d2
keep event _*                                           
sts graph, by(event)

And it worked very well, as attached. Again, apologies, will include more details on what exactly didnt work next time. Just in case its still of matter: used the following:

use *yourfile
stset TimetoICH_or_FU_years, failure(ICH_FU)
sts gen surv = s  
gen group = 1 
save d1, replace

use *yourfile
stset TimetoIS_or_FU_years, failure(IS_FU)    
sts gen surv = s  
gen group = 2 
save d2, replace

use *yourfile
stset TimetoreccSAH_or_FU_years, failure(recurrent_cSAH)    
sts gen surv = s  
gen group = 3 
save d3, replace

use d1, clear
append using  d2
save d4, replace
append using d3
save d5, replace

keep in 1/2
keep group _t  surv
replace _t = 0
replace surv = 1
replace group = _n 
append using d5

separate surv, by(group) gen(s)
sort group _t surv
scatter d1 d2  d3 d4, c(J J) ms(i i)

If I replace the last line by

sts graph, by (group)

, it again works like you suggested yo I dont need this bit I suppose at all:

 keep in 1/2
keep group _t  surv
replace _t = 0
replace surv = 1
replace group = _n 
append using d5

separate surv, by(group) gen(s)
sort group _t surv