Cumulative incidence curves competing risk analysis differs when using stcompet and stcurve

[Kim Vaarts]

Dear all,
I am making a competing risk analysis in STATA. I read that I should use the Fine and Gray method using stcurve(after stcrreg) in STATA. However, in the STATA literature I read about another type of competing risk analysis by the Aalen-Johansen method using stcompet in STATA. However, I get two different curves when using both methods. See below. Which method should I use for making a cumulative incidence curve?

I am comparing thee different types of drugs. The cumulative incidence is the response (time in months) of the patients when using the drugs.

For the Aalen-Johansen method stcompet I use the following STATA code and get the following Curve (see picture1):
gen drug = 0
replace drug= 1 if med==”A”
replace drug = 2 if med==”B”
replace drug= 3 if med==”C”

quietly stset dftime, failure(failtype == 1)

stcompet cif = ci, compet1(2) by(drug)
gen cif_drug = cif if failtype== 1 & drug==1
gen cif_drug = cif if failtype==1 & drug==2
gen cif_drug = cif if failtype==1 & drug==3
twoway line cif_drug_* _t, connect(step step) sort xlabel(0(12)125) xtitle(months)

For the Fine and Gray method stcurve(after stcrreg) I usde the following STATA code and get the following curve (see picture2):
quietly stset dftime, failure(failtype == 1)
stcrreg i.drug, compete(failtype== 2)
stcurve, cif at1(drug=1) at2(drug=2) at3(drug=3)