Should you adjust the model for interaction for another independent variable?

Dear all,
I'm analyzing a big dataset of hospitalized patients and I use Cox regression analysis to estimate associations of different parameters with survival. I have detected that there is a significant interaction between age and sex and survival, i.e. females do worse in young patients and males do worse in older patients. My primary focus are comorbidities and their association with survival and I would like to present age and sex adjusted estimates of hazard ratio for a number of comorbidities to describe these relationships. I tend to do that since a lot of comorbidities are present in young patients and therefore do not significantly afect survival (they are compared to older patients without comorbidities and difference is "canceled out" in unadjusted analyses).
My question is since age and sex interaction with survival exists, do I need to include age*sex interaction term in models where I investigate age and sex adjusted HR estimates of different comorbidities? What information would I get with it?
Thank you in advance, Marko


Less is more. Stay pure. Stay poor.
If this is a true interaction, the two terms are conditional on each other and you would not look at their base effects. You would keep the interaction term in the model, in that it allows you to look are all four combinations of the variables (e.g., 0,0; 0,1; 1,0; 1,1).