If you're asking how exactly to calculate it, that's a question for the stats folks. But for simplification purposes, the recent enrollees are considered to have survived for however many months they have been randomized. For an extreme example, let's say that 99 patients were randomized on September 30, 2007, and the interim look occurs on Sept 30, 2008. Assume all of the Sept 07 pts are still alive on 9/30/08. So 66 pts in the treatment arm have 12 months survival and 33 pts in the control arm have 12 months. That's going to mess up the p value for the treatment arm, because there is no separation vs the control arm. It's obviously not as bad as if it would be the 33 control pts all surviving on 9/30/08 and all 66 treatment pts dying one month earlier. But it's still going to hurt the p value.
