Thanks all for your sharingand great analysis on this thread....

Thanks all for your sharingand great analysis on this thread.

There has been a lotof discussion on ‘background’ mortality rates. Various articles seem to putmortality rates for those hospitalised with COVID-19 in the range of 30-60%,with age, comorbidities and being on a ventilator being predictive factorspushing to the upper end of this range. With so much information out there, I’velooked for a way to try and narrow down the mortality rate for those eligiblefor the trial.

I’m not sure if thishas already been posted, but the following link is to a Science Direct articlepublished 23 October which provides a detailed analysis of factors contributingto mortality with respect to a number of characteristics. It is based on data from March-Aprilin Spain. Whilst the standard of care may have moved on, I found the analysis onmortality rates in relation to SpO2/FiO2 ratios quite helpful.

https://www.sciencedirect.com/science/article/pii/S0091674920310277

The NIH / MSB trialstipulates enrolment criteria relating to moderate or severe ARDS indicated by:

· PaO2/FiO2 >100 mmHg and ≤200 mmHg, onventilator settings that include PEEP ≥5 cm H2O OR

· PaO2/FiO2 ≤100 mmHg on ventilator settings thatinclude PEEP ≥5 cm H2O

The following plotshows this cut off and the box plots for this study and I’ve drawn a lineacross the enrolment criteria of <200mmH

Scaling roughly withrespect to the interquartile ranges, it seems:

ICU survivors below200mmHG are about 40% x 28 = ~11

ICU non-survivors below 200mmHG are about75% x 36 = 27.

So, in this example, outof the pool of those in ICU who would have met the trial’s , SpO2/FiO2, we mightestimate a mortality rate just over 70%.

These box plots show very few cases for severe <100mmHG. Perhaps this helps explain why the trial takes so much time to recruit as it is very selective.

I have previously posted a Chi Square model assuming a 30% mortality rate. I’ve run this again with a 60% background mortality rate for the placebo group and selecting p-values of 0.2 and 0.0005 as boundaries for potential early termination of the trial. This is different to the Bayesian analysis, which is how the trial is being reviewed and what @kervio has presented and which is more accurate, but I'll post it anyway.



Please don't rely on this for any investment decisions, but for my part, I'm estimating

· a 20% chance of the trial being stopped at the45% stage

· a 50% chance of the trial being stopped at the60% stage

· quite a strong chance of the trial achieving statisticalsignificance at completion.

Good luck to all andfingers crossed for some good news in the next couple of weeks!