A very important number form the conference yesterday, to a very good question 
was that of 87 deaths so far.
We don’t know how many were “inside” the 28 days’ time-frame and how many outside. We don’t know either if it includes some more after the 28 days of the first half. Only NP and the AMAREX statistician know this if it is assumed that they are reported promptly (and accurately).
However this has implications as, accounting for reported and current SOCs the number should had slowed down somewhat. We had 45 in the first half and, apparently around 42 in the second half. I do suspect that some of these are SEAs from the first 195 patients (and hopefully exclusively from the 65 on SOC).
In ay case, I had already made some calculations for 83 total which now seems a good number to work with (assumes a total of 4 deaths outside 28 days). However, I run for personal (reference) reasons some calculations with 87 deaths total as well. Not wanting to swamp the Hangout again with more graphics and numbers (some others have already made some relevant calculations) they can be found here for those interested.
https://drive.google.com/file/d/19hcv4lYe3-sg...sp=sharing
But the significance of the number raising is that it will make the power of the trial better. The designer(s) of the trial (AMAREX-FDA) underpowered it slightly. This is understandable for cost and expediency reasons, however if the number of deaths is low the power of the trial diminishes respectively. Not sure what is the “acceptance” threshold for FDA in this case (in non-pandemic times is 80%-90%) but running post-hoc analysis with 41.3 % reduction of deaths we are in the 80% range. As a reference, for 41.1% reduction and 61 total deaths (33 Vx/28 SOC), power is app 61%.
Of course, the higher number of deaths is statistically good provided they don’t reflect the Vx arm, we still have to show a good improvement. A 33% mortality reduction should suffice for EUA and approval imo provided the FDA does not demand a “high” power. 41.3% reduction will meet this as well.
was that of 87 deaths so far. We don’t know how many were “inside” the 28 days’ time-frame and how many outside. We don’t know either if it includes some more after the 28 days of the first half. Only NP and the AMAREX statistician know this if it is assumed that they are reported promptly (and accurately).
However this has implications as, accounting for reported and current SOCs the number should had slowed down somewhat. We had 45 in the first half and, apparently around 42 in the second half. I do suspect that some of these are SEAs from the first 195 patients (and hopefully exclusively from the 65 on SOC).
In ay case, I had already made some calculations for 83 total which now seems a good number to work with (assumes a total of 4 deaths outside 28 days). However, I run for personal (reference) reasons some calculations with 87 deaths total as well. Not wanting to swamp the Hangout again with more graphics and numbers (some others have already made some relevant calculations) they can be found here for those interested.
https://drive.google.com/file/d/19hcv4lYe3-sg...sp=sharing
But the significance of the number raising is that it will make the power of the trial better. The designer(s) of the trial (AMAREX-FDA) underpowered it slightly. This is understandable for cost and expediency reasons, however if the number of deaths is low the power of the trial diminishes respectively. Not sure what is the “acceptance” threshold for FDA in this case (in non-pandemic times is 80%-90%) but running post-hoc analysis with 41.3 % reduction of deaths we are in the 80% range. As a reference, for 41.1% reduction and 61 total deaths (33 Vx/28 SOC), power is app 61%.
Of course, the higher number of deaths is statistically good provided they don’t reflect the Vx arm, we still have to show a good improvement. A 33% mortality reduction should suffice for EUA and approval imo provided the FDA does not demand a “high” power. 41.3% reduction will meet this as well.
13
