1

I am using Stata and completing a competing risks regression with secondary cancer diagnosis as the failure and death as a competing risk.

I am not sure if I am using the stset command correctly. The code I am using is this:-

  stset diagtime, time0(diagnosisdate1) origin(time diagnosisdate1) exit(diagnosisdate2) failure(fail==1)

Where "diagtime" is the time between primary and secondary diagnosis and fail == 1 is the occurrence of a secondary diagnosis.

I need to specify death as a competing failure for when I run the regression but not sure if this should be specified as death alone, or death as well as no second diagnosis.

sophierw
  • 11
  • 1

1 Answers1

1

A delayed response, but in case others find it helpful.

I can't speak to the t0 and origin options being correct without seeing the dataset. For the fail option, though: regardless of what type of competing risks model you're estimating, the stset format is what you have. To strip down to the key parts:

stset diagtime, failure(fail==1)

Because fail==1 represents your event of interest--secondary diagnosis.

If you're using stcrreg, you must specify the competing event as an option. Say death (your competing event) is represented by iAmDeath==1. The stcrreg syntax would be:

stcrreg [varlist] [if] [in], compete(iAmDeath==1)

For competing risks with any other type of canned survival model in Stata, you're implicitly taking a latent approach to competing risks. That means you're treating all events other than the 'primary' one of interest as right censored. Ergo, there is nothing additional you must do, beyond setting stset 's fail option correctly (i.e., to your primary event of interest, as you do in your stset statement).

SMzgr356
  • 83
  • 6