I have a breast cancer database. I applied the Cox PH model to estimate the hazard ratio of covariates and breast cancer risk. When I used age at censor as the time scale and adjusted for age at baseline, I got different effect sizes of the association between the variable of interest and breast cancer risk, compared to the Cox ph model using age at baseline as the time scale without adjusting for any time-related variables. For example, the hazard ratio of height is 1.10(1.02,1.19) V.S. 1.15(1.06,1.23). Which time scale should I stick to? Thanks so much!
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Often it does not make sense to use age at censor as timeline. The timeline often start at inclusion time, then the time run until one observes an event or no event.
I found an introduction to survival analysis here.
The time-scale should always be chosen based on what is most relevant in your study - usually (calendar time, age, or time since a specific event (fx surgery)).
I hope this helps you.
