Power calculation for time to event study

[Megan Moreton]

Hi,

I am powering an RCT where the primary endpoint is time-to-response from initiation of treatment (treatment vs. control).

The mean time-to-response in the control arm is 5 days & we estimate a clinically important elongation in response would be 7 days. The SD is 10 days & the expected event rate is ~100%. For the power calculation, I am currently using power twomeans, but I am unsure whether this fully captures the time-to-event nature of the study:

power twomeans 5 12, sd(10)

Would power cox be more appropriate? If so, how would I apply this command?

Thank you for any help,
Megan

[Megan Moreton]

Would the following be appropriate - to detect a significant improvement in the median outcome of 5 days to 12 days (ie, HR or 0.42) I could use the following:

power logrank, hratio(0.43)

Is there any way to incorporate uncertainty?