Let's assume that the 180th event occurs on July 1, 2008 and the DSMB unblinds for the first interim look. Let's also assume that 120 patients had enrolled by July 1, 2005, approximately two years after the first patient enrolled.
I think this highly unlikely - Given the Provost enrollment figures of Feb 06 which implied less than 100 patients. I can accept that those number might be as much as 6 months behind. But seems unlikely that there were 120 by July 2005. FWIW.
I wonder how much the extra time beyond three years the long-term survivors are followed would help the p value.
I concede it is possible this effect *may* be an important factor in an analysis out in 2010. Where perhaps 60% of the patients might be followed for more than 3 years. But in 2008 I'd expect less than 25% to have been followed for that long - so I'd expect substantially less effect.
Looking at some of the simulations, I'm starting to wonder if it would be better for the company if the interim unblinding occurs in the 210-240 event range instead, and reserved alpha is 0.02...
Were I them I would assign the minimum reasonable alpha (i.e. the p value expected for that number of patients if the HR was the same as in 9901 straight log rank) just to have a look at the HR and be able to say "The HR is x and even at 1/2 that value in the final analysis the p value would be stat sig."
Clark
News 
Market Data 
Discover 
