proc power compute sample size for survival curves

guanlinchang — Fri, 21 Aug 2020 02:57:12 GMT

I am trying to perform a sample size calculation in SAS for a two sample time to event case.

Here is the situation:

1. Assume both sample follows exponential distribution
2. Assume a given constant hazard ratio under alternative hypothesis, we call hr (group 2 vs group 1)
3. We will use logrank test.
4. Given accrual time a, and follow up time f
5. Also given the exponential hazard for group 1, called it exph1
6. Assume the sample size ratio between the two group is 1:1
7. required nominal power is p

Now my code looks like this:

proc power;
twosamplesurvival test=logrank
accrualtime = a
followuptime = f
refsurvexphazard= exph1
hazardratio = hr
power = p
/* eventstotal = . /*events total */
/* ntotal= . /*total sample size */
; 
run;

you can uncomment either eventstotal = . or ntotal=. depending on whether you want to compute the requested number of events, or the actual total sample size.

They should not be the same consider by the end of follow up, if the event does not happen, then the subject will be right censored.

However I am always getting the same number for events total and total sample size. What did I do wrong here?

I actually know how to compute this by hand, and my hand calculation for requested event number is very close to SAS output (SAS gives a slightly larger value but very close), however my total sample size is much larger than the event number.

I could not disclose any particular initial value for the parameters above due to confidential reasons. Could someone help? Would really appreciate that.

topic proc power compute sample size for survival curves in Statistical Procedures

proc power compute sample size for survival curves