I couldn't post this earlier for some reason so I'll start a new thread.
Since the 1st interim look will happen at 220 deaths, the way for the trial to be unblinded would be if at least 129 of those will be from the control group and less than 91 of those from the vaccine group (41% improvement). The trial began in May 2010 and reached the midway point in enrollment in 2Q of 2012 (340 patients). It is scheduled to reach full enrollment (680 patients) by the end of 2013 since trial enrollment begins slow and picks up speed over time.
If we look at when the 129th control group patient would have probably been enrolled, there would have been twice as many enrolled in total (258). This would likely have been around Oct. or Nov. of 2011 based on current enrollment data and projections. Also, the 91st vaccine group patient may have been enrolled around Apr or May 2011 (182nd overall). If you believe that the control group will have a median OS of around 16-18 months then there would likely be 129 control group deaths around Jan-Mar 2013 roughly. Also, if the vaccine group is likely to live around 24-27 months then the 91st vaccine group death would not happen until May-July 2013. Therefore there would be greater than 129 of the 220 deaths from the control group at the first look if it happens in 1Q 2013.
First I think hitting the 40% less deaths at 220 Int1 is possible. I also believe that hitting 30% improvement at Int2 with 340 deaths is probable. The total enrollment in the trial is scheduled to be 720 . The phase 2 results had increasing improvement starting at 37% at one year and reaching greater than 100% at three years , this was expected survival since unlike the phase 3 the phase2 had no control . The phase3 will have a cross section of time frames so we are all guessing!
One of the great reasons that clinical trials are set up the way they are is that you can get data in a 2-3 years that normally it would take 4-5 years to get. When you look at the final OS data for a trial you compare vaccine group OS to control group OS and come up with a % difference. A quicker way to get the same data (roughly) would be to pick a number of deaths and compare how many of them were from the control group and how many were from the vaccine group and come up with a % difference. At first blush it seems like these numbers could be really different but when you think about it there's really only a difference of number of patients being evaluated in each case (statistical significance).
What I'm saying is that whatever the improvement in OS is at the 1st interim look is what's likely to be the same % difference as what the final data will show. Don't get this confused with 1, 2, and 3 year improvements which will obviously show a greater improvement in time but that's because you're comparing apples and oranges. The latter can be exaggerated with smaller numbers, the formal can't be. Example would be if only 5 people survived in a control group after 5 years and 20 people survived in the vaccine group after 5 years, this is a 300% improvement. But it hardly explains how good the drug works if there are hundreds of people in the trial.
They'll enroll up to 722 in order to evaluate 680 patients. The 37% improvement was after 1 year when the vaccine group had both 100 and 300 million doses and comparing that to what that group would likely have done based on historic data, which I'll admit isn't ideal. With everyone getting the high dose in the phase 3 trial plus more maintence doses I believe the vaccine group will live longer than the phase 2 trial patients did. We're not using 1, 2, and 3 year improvements anymore so 37%, 59%, and 121% isn't comparable to this trial. We're looking at improvements after a certain number of deaths which is likely to be between 16-27 months after enrolled (probably 18) so I bet that improvement is likely to be somewhere between 37% and 59%, but possibly more due to the higher dosing.