Unblinding
Here's my 2 cents on the unblinding...
I've created a spreadsheet with p-values based on various death rates in the placebo and LL arms. Based on the DSMC's recommendation (and assuming we only had 45 deaths, not the 50 NP eluded to in last call), I suspect we had 26-27 deaths in LL and 18-19 in placebo. Both of these put us very close to statistical significance (.0823-.1023). The difference between stat sig and not is 1-2 deaths (again, we don't know for sure the exact number of deaths and hope they give that info today).
My point is, we were very close to stat sig, and more importantly, if I'm correct, this means we showed at least a 25% reduction in deaths. This is huge and would mean we could save thousands of lives, and announce to the world that there is a drug more effective than current SOC.
For this reason, I believe we should take the statistical penalty and unblind the data. The FDA would have to react to an announcement that there is a better SOC than Remdesivir, especially if we are in the 25-30% improvement.
The other thing the spreadsheet shows is that a .005 penalty will (for the most part) not affect the outcome of a stat sig result, meaning the way the deaths affect stat sig, it shouldn't affect whether we hit the stat sig target or not. Adding .005 to most of the stat sig levels will not make LL not stat sig (assuming same death rates occur in both arms). If LL can save even more lives (esp after considering 42 day mortality - DSMC must have seen something to indicated people in placebo were on downward trend).
Lastly, it's unclear to me if UK and Philippines can get access to data without unblinding. If this is the case, it's another reason we should unblind.
Market Data 
Markets 
AI
