Stats duded wanted, page-3

Thanks pal.
I'll try and lay it out as best I can.

The company in Q is SRX. They have just announced the top line results of a clinical trail and are due to present at ASCO in 10 days time.

The progession free survivcal of cancer in the liver was a 64% improvemnt with a P value of .002. This suggests the trial was overpowered by a considerable amout. There were just under 600 on the trail, about 50/50 to the control arm. This was a secondary end point, so there is no ethical aspect of an overpowered trial!

SRX have stated that to yield a statistically significant value for overall survival (the big game in town) they need to combine the data with two other trials where the data will not be avilable until later in 2016, for a total trial population of about 1000, 50/50 to control arm. This is based on an 80% power.

My undersatnding is that a P value of 0.05 is the go for clinical trails (less than 5% chance of the result being a random event), and that an 80% power should bring the P value down to about 0.01.

The value we have been given is that they are looking for about an additional 3 months OS

So the big question in my head is this. If for instance the OS from the current trial was say in excess of 6 months, would this yield a P value of about 0.01 for a 3 month advantage, or am I mixing apples and pears?

If the OS was in excess of 6months what P value could be concluded with only 60% of the expected population data available.

If there is any other info that would assist please shout.

And if I really am barking up te wrong tree, then apologies for wasting your time :-(

Cheers
CAFA