Okay, I am not a trained statistician but in many ways this is...

Okay, I am not a trained statistician but in many ways this is an exercise in logical thinking.
The first thing to do is identify what we are trying to predict.
As I understand it, we are trying to predict what is the ratio of patients in the active group and in the control group who are enrolled in the trial but who have not yet reached day 30 at the date of data cut-off surviving to day 30.
Presumably the statisticians will run a computer simulation that looks for correlations between survival at day 30 and measurements taken at baseline and/or at days 7, 14 and 21. For example, such a simulation may show a very strong correlation between a steady reduction in one or more of the inflammatory markers (IL-6 and TNF-alpha) on days 7,14 and 21 (endpoints 24-26) and survival at day 30. One could also look for other correlations between reductions in the severity of ARDS (endpoints 11-13) on those days and survival at day 30.
With the very sophisticated computer programs used to run these kinds of simulations there may be many different correlations involving a large number of different measurements across the patient population from which inferences may be drawn as to what proportion of the two groups will survive at day 30.
Thoughts?
Cheers.