Good evening to all, Thanks to everyone contributinggreat...

Good evening to all,

Thanks to everyone contributinggreat analysis here. This post is to share my own attempts at hypothetical modelling to try and assess / guess the chances of the Covid-19 ARDS trials being successful, and at what interim stage.

Assumptions

We know that the trialwas not stopped for futility nor for overwhelming efficacy

We might guess that:

-Backgroundmortality rate for placebo arm is >30%

	https://www.inquirer.com/health/coronavirus/coronavirus-covid-19-ventilator-patients-survival-rates-increase-20200703.html
1	https://www.practiceupdate.com/content/covid-19-mortality-rate-for-intubated-adults-lower-than-previously-reported/102029

-Limit tostop for futility is when p-value > 0.2

	https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6442021/
1

-Limit tostop for overwhelming efficacy is when p-value <0.0005

	https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5549398/

Use of a model

In the absence of any data from which me might infer characteristics, Ihave used a simple Chai Square test to derive p-values for various scenariosfor mortality rates for Rem-L based on possibly observed outcomes from theunblinded 30% interim analysis ranging from 0-10 patients having passed away. Thisis probably not the right tool but, hey, it’s hypothetical anyway and is simple to understand.

I have rounded figures conservatively in my modelling.

Modelling

Here are some calculations for a scenario where 8 Rem-L patients passed away is plotted below.This scenario may well have led to the trial being stopped for futility:

Here are some calculations for a scenario where 7 Rem-L patients passed away is plotted below.This is perhaps a conservative/pessimistic case for the purpose of the trial, but where it was not stopped for futility:

Here is a table with permutations for several scenarios / mortalityrates:

Based on the assumptions made, it would have been very difficult for thetrial to have had found to have overwhelming efficacy at the 30% state. i.e. Nomore than 1 person would have had to have passed away.

My conclusions
I think we seem to be in the band where the mortality rate is between 4% (2 Rem-L patients passed away by last week) and 15% (7 Rem-L patients passed away). If the assumptions are correct and the background / Rem-L mortalityrates are consistent:

I think there is still only a low (say 15%) chance that the trial willbe stopped at the 45% interim stage.

I think there is an even (50%/50%) chance that the trial will be stoppedat the 60% interim stage.

I think it is very promisingthat the full trial will ultimately succeed.

Disclosures

I am not a medical or statistical professional, just an amateur, interestedengineer.

I am substantially (well at least for me) invested in MSB and have been following this stockall year.

GLTAH!