Ok for non statisticians out there I will explain the point:
let’s assume you have a trial with 1 patient per arm. One Leronlimab patient. One placebo patient. Lets assume, out of pure chance, the Leronlimab arm’s patients survived and the placebo’s arm patient died. Bam you obtain a -100% mortality reduction in favor of Leronlimab. Champaign! Excepted...it is just pure noise because the probability to obtain such a value assuming this is just chance (null hypothesis) is huge. Redoing this trial you could end up with a 0% reduction or even a +100% increase. Because we dont have enough patients to cancel out the noise.
This is what statistical significance is all about: discarding noise.
This is exactly what I studied today based this time on the 63 disclosed patients. The noise is still huge. Getting a -25% reduction, or a +25% increase on the other side is VERY likely, and in many scenarios MOST likely. I could not find a single scenario where it is unlikely. And this doesn't even deal with the datamining performed by Cytodyn (they have given here the BEST outcome they found, so the pvalue is severely biased as it is the best pvalue out of dozens of them...pretty bad “champion” case).
The reason is simple: 63 patients is way too small to distinguish between pure chance (Leronlimab is saline) and a real effect. They need WAY MORE than that.
This is why the DSMC requested the new additional 75% interim to perform, I quote Nader Pourhassan, a “sample size reassessment”. They knew and communicated the issue way back then in October! But Nader refused that, and lost precious time now. Now everything has to be redone. Thats what I have written on this board many times, warning about this DSMC bad omen . “Sample size reassessment” is a dreaded word for any statistician...
News 
Market Data 
Discover 
AI
