controlling for variables

i really dont know how to ask but i hope some of you will understand and answer my question.

suppose that i was solving a survival analysis problem where i have to assess treatment effect, controlling for any important variables.

and say i had a already reduced model T*censor(0)= trt + x1+x2+trt*x1+trt*x2+x1*x2, assuming that i am using SAS.

if i want to assess the treatment effect holding x1 and x2 fixed at their means, my teacher told me to center x1 and x2 at the means and run proc phm and use test statement for trt. that is, set x1=x1-(mean of x1) and x2=x2-(mean of x2) and run proc phm.

but i do not see at all why this is all i have to do to hold them fixed at the means. i see that the model has been shifted and centered at the means but i think x1 and x2 still vary.

or is it because when SAS tests the effect of trt, it automatically sets all the other covariates = 0? or does it have something to do with the statistic used for the test?

please help me understand it.

thank you.