06-12-2012 05:32 PM
I am trying to get hazard ratio for a risk factor. Individuals with exposure had no events and individuals without exposure have 9 events. I have follow up time. when i use proc phreg, I don't get hazard ratio for risk factor. it comes up as "."
is there any other way to get HR in this case?
any help will be appreciated.
06-12-2012 10:09 PM
No. The hazard ratio is not defined when you don't have any events in one group; your sample size is too small for PHREG. You might be able to force PHREG to provide an estimate by arbitrarily creating an event in an observation in the exposed group, but the estimate will be quite unstable.
Computing the required sample size for a survival analysis is pretty tricky, you need both an estimate of the event rate per group and how much censoring you will have (the latter is really hard to get a handle on). PASS has a tool for it, but few other packages do.