Long, I agree that starting the clock early makes the difference in median stand out less. I even posed that long ago. But the P value is unchanged by adding a constant time offset.
Why not add the 3.8 months to the Optune MOS instead? Now one has Optune at 23.4 months vs DCVax (blended) at 23.1 months.
And why add/subtract the median time from surgery when you should be using the mean unless you have raw data? If you look at the Stupp trial as an example for time between surgery and adjuvant, you will see that a decent number of patients take far more than the nominal 3 months to begin adjvant treatment. I would gladly bet you that the blended median in DCVax-L from randomization is less than 20.0 months due to this effect.
Only 24 months after last patient enrolled in Optune trial. (Their last update)
Only 10.3 % alive in the blended population. (In the DCVax-L trial, approximately over two and-a-half times that number of patients were alive at a similar time in its trial.)
Here was the last Optune update of patients still alive at 24 months after last patient enrolled.