$RLFTF (related) thread on Yh-0 by Glunker
I did my own statistical analysis of the data provided in the Previous SSRN Peer Review report. Z-test of Proportions on the various values provided in Table 4 - Day 60 Mortality by Site of Care.
I notice that they never showed a statistic for the total number of Aviptadil vs Placebo patients and deaths across both Tertiary and Regional hospitals (Aviptadil 136 patients, 46 deaths [35.1%] / Placebo 65 patients, 37 deaths [56.9%]).
I ran the statistic for this and got a P-value of 0.00181. However, there is a difference in the actual totals shown in the Placebo Deaths column for Tertiary (23 vs 30 [true sum]) with asterisks noting it.
If you put the seeming 'incorrect' sum value of 23 in the results, the percent deaths is the reported value of 46.2%, and the P-value comes out to be 0.067...just off the mark of 0.05...so it wasn't reported as significant.
7 patients were pulled from the placebo death column for the tertiary hospitals as indicated by an asterisk note in the table.
The result with asterisks are not explicitly discussed in the document other than one sentence with the asterisk that said "However, a significant treatment * site interaction, opposite in sign with site of care was observed and, therefore, Wilks test was used to confirm changes in -2 log likelihood."
I am thinking that "a significant treatment site interaction" was probably death due to nosocomial infection, etc. If so, then all this does is establish even more credibility of the team by not allowing this to skew data in their favor.
Either way, if true counts are played, Aviptadil has a REALLY significant affect on probability of death (Avip=35.1% vs Placebo=56.9% at P<0.00181)...HUGE as we already know.
Just wanted everyone to feel comfortable with the 60 day results. Our team did not embellish...in reality they were extremely conservative in the reporting of their results.
