How do I get pointwise confidence Intervals for the covariate-adjusted cumulative incidence function (CIF) after STCRREG?

[Yachi Tsuneoka]

Hello,

I would like to get 95%CI for pointestimates for adjusted cumulative incidence function curves that were produced by "stcrreg" and "stcurve".
I could get pointestimates for CIF, but cannnot produce 95%CIs for the pointestimates by the following.
Would anyone help with producing 95%CI following to the codes below?

*********
*(1) Fine and Gray
stset days, failure(evt_cif==1)
stcrreg i.main_var i.drug, compete(evt_cif==2)
stcurve, cif at1 (main_var=1) at2 (main_var=2) at3 (main_var=3) range(0 150) ylabel(0 (0.1) 1.0) outfile(CIF_FIG1, replace)
*********


For Cox model, I could get pointwise 95%CI by using "survci" as explained by Cefalu(2011).
So I wonder if there are ways to do similar as following, using "stcrreg" for competing-risk regression models of Fine and Gray.

*********
*(2) COX
stset year, failure(event==1)
sts graph, by(main_var) adjustfor (agegroup drug)
sts list, by(main_var) adjustfor (agegroup drug) /*only point estimates and no 95%CI*/

stcox agegroup druggroup if main_var==1
survci, survival at1 (agegroup=0, drug=0) outfile (main_1, replace failonly)

stcox agegroup druggroup if main_var==2
survci, survival at1 (agegroup=0, drug=0) outfile (main_2, replace failonly)

stcox agegroup druggroup if main_var==3
survci, survival at1 (agegroup=0, drug=0) outfile (main_3, replace failonly)
********

Thank you,
Yachi

[Giorgio Roccaro]

I am having the same trouble. I can generate point estimates for CIF following stcurve, cif command and creating an outfile. Subsequently the "ipolate1" command will allow me to obtain point estimates at specified times. However, I am having trouble generating the 95% CI around the point estimates. I do not believe stcompet is the answer to this, but I may be wrong.

Thank you