Dear all, I have a question regarding a survival analysis.
I thank you all in advance.

I have to compare 5 drugs, adjusting for covariates.
It is a multicenter study where most of the failures occures at 6, 12, 24 and 36 months. After the 36-month visit the patient is censored.

The proportionality of the hazards holds for all the covariates but NOT for the "drug" variable (it is quite evident also from the attached graph).

Array

The problem is:

to take into account the multi-centre structure of the studio, I have used a shared frailty in the model.

However, as a result of this, STATA does NOT allow me to stratify for the "drug" variable.

I wonder how to fix this.

Or can the problem of no-proportionality of hazards be solved in another way?

What I get from STATA is:

xi: stcox age i.sex i.adverseevents i.efficacy i.pathology, frailty(gamma) shared(hospital) nolog forceshared strata(drug)

option strata(drug) not allowed with shared frailty models

Thanks again.
Gianfranco