Vic. firstly thank you so much for your time and being able to...

Vic. firstly thank you so much for your time and being able to assist a novice. The explanation regarding 'power' now make a lot more sense.

One last (hopefully) question that you may be able to assist with.
The dudes at the FDA required a trial size of about 1000 for an 80% power (I believe this is designed to produce a P value ~ 0.01).
Due to the trail design the initial 600 patient trial was completed 2 years ago, and the remaining 400 will have data in around 2016/7. Unusual it may be, but the FDA have allowed the pooling of the 3 trails to obtain the OS data.

Assuming for one minute that the toatl of 1000 will be successful and yield a P value of ~0.01.

If the initial data from the 600, showed an OS of > 4 months, what sort of P value would this yield?

I can't get it out of my head that if the data from the initial 600 is good enough then a claim of success in OS could be achievable. It probably is not that good, but I'd like to get a view on what it would have to be, to be considered an overall success.

As background they have released a progression free survival of 8 month over standard of care and as 90% die from liver mets, it is reasonable to assume a significant OS benefit will occur. The early small trails showed very large increases in OS - almost too good to be true.

Any comment you are able to make is appreciated - even if it is that I'm one crazy person