Hi there,
I am using competing risk analysis with stcrreg to estimate the risk of neutropenia related to exposure to a drug (valganciclovir) in solid organ transplant recipients. The competing risk is death. I initially did not treat valganciclovir as time-varying but I received some feedback that I should.
My questions are:
1. Can I use valganciclovir as binary and still assess it as time-varying or should I ideally treat it that variable as continuous (i.e., days) if treating it as time-varying? It is just that I don't expect a significant increase in the risk by one unit of time which is one day. I mostly want to compare the rates of this complication in patients that receive this drug regardless of duration versus those who don't.
2.When I use the tvc command, should I should report the time variable SHRs in tvc only? I assume so as if I input it in both main and tvc, the only purpose would be to check the PH assumption. Is there any other way to check the PH assumption using stcrreg?
Example:
stcrreg age valgan ATG, compete(failure_var = 2) tvc(valgan) texp(ln(_t)) > Check PH
stcrreg age ATG, compete(failure_var = 2) tvc(valgan) texp(ln(_t)) > The actual model SHRs that I will be reporting
3. If I'm using stcrreg, do I also need to report cause-specific HRs using Cox? I am uncertain that this should always be done for competing risk regression.
4. Since I have the time in survival time, I'm using the ln of time texp(ln(_t)). How do I know if I should multiply this by a certain factor?
This was all based in https://www.stata.com/manuals/ststcrreg.pdf but wanted to make sure my interpretation is correct.
Thanks,
Sara
I am using competing risk analysis with stcrreg to estimate the risk of neutropenia related to exposure to a drug (valganciclovir) in solid organ transplant recipients. The competing risk is death. I initially did not treat valganciclovir as time-varying but I received some feedback that I should.
My questions are:
1. Can I use valganciclovir as binary and still assess it as time-varying or should I ideally treat it that variable as continuous (i.e., days) if treating it as time-varying? It is just that I don't expect a significant increase in the risk by one unit of time which is one day. I mostly want to compare the rates of this complication in patients that receive this drug regardless of duration versus those who don't.
2.When I use the tvc command, should I should report the time variable SHRs in tvc only? I assume so as if I input it in both main and tvc, the only purpose would be to check the PH assumption. Is there any other way to check the PH assumption using stcrreg?
Example:
stcrreg age valgan ATG, compete(failure_var = 2) tvc(valgan) texp(ln(_t)) > Check PH
stcrreg age ATG, compete(failure_var = 2) tvc(valgan) texp(ln(_t)) > The actual model SHRs that I will be reporting
3. If I'm using stcrreg, do I also need to report cause-specific HRs using Cox? I am uncertain that this should always be done for competing risk regression.
4. Since I have the time in survival time, I'm using the ln of time texp(ln(_t)). How do I know if I should multiply this by a certain factor?
This was all based in https://www.stata.com/manuals/ststcrreg.pdf but wanted to make sure my interpretation is correct.
Thanks,
Sara
