Dr. Hahn's comment on the gold standard coincides with a meaningful p-value and I am sure he got this 35% of reduction of deaths from his statisticians.
With this improvement in the interim
the p-value would be higher than 0.05, however, with the complete enrollment (390 patients) a 36.4% reduction of deaths produces a p-value of 0.03794 (a 35% reduction will be almost exactly on 0.05
Please see below in lower section of table for full enrollment assuming 30 more deaths (I use binomial distributions)..
However, due to the pandemic and lack of options with many drugs failing one would expect that a lower death reduction should be accepted. Heck, even 20% should be OK by Pete's sake !!!
The interim <0.005 would be achieved with around 50% reduction. This is an extremely difficult bar to achieve and I think (imo) that this is roughly what the board will be looking at to stop the trial.
And it is a very good question to ask !!!