FWIW, the only way A273 gets approved is if they can successfully convince CHMP that the ABCLEAR1/S1RWT data should be used for conditional approval. (any thought that the OLE with its 70% dropout or the posthoc ABC2 and 3 matters is laughable).
I just took another look at the JPAD paper to try to get a better feel for the power calculations for the original study to try to determine if the study had any power if one considered 70% N (i.e instead of mITT of 462 we have mITT_S1R of N=323 and instead of pooled treatment/placebo mITT of N=298 and 164 we have N=208 and 114) to account for S1R WT
From the paper we see this:
Sample size and power calculations were based on a simulation approach with several planned scenarios and assuming co-primary endpoints (ADAS-Cog13 and ADCS-ADL). The sample size calculation assumes the mean difference between either blarcamesine arm and placebo of 1.5 points (SD=4.5) in the ADAS-Cog and ADCS-ADL with at least 90% power using a two-sample t-test with alpha = 0.05 (2-sided). For the calculation of power concerning co-primary endpoints, conservatively assuming that power can be independently calculated [33], this will achieve at least 80% power for two endpoints. A 33% dropout rate was considered in estimating the sample size based on earlier studies. Therefore, 509 participants would need to be enrolled to allow for an anticipated 342 completers, i.e., 228 patients per combined treatment and 114 per placebo arm, respectively.
So, when the study was set up they had data from the gantenerumab (missed primary CDRSB but squeaked by for secondary ADAS-Cog in one P3 and barely missed) and other AD trials that reported ADAS-Cog CDR-SB and other measures to help guide. For independent power calculations of 90%, for ADAS-Cog and ADCS-ADL, Anavex used a mean difference of 1.5 and a std dev of 4.5 - the mean of 1.5 was reasonable as gantenerumab 2 1000 patient P3 studies showed 1.25 and 1.28 however std dev was 9-10. It turned out (calculating from the provided CI 95% range that the std dev was around 8 - close to historical, close to lecanemab but far far away from the power assumptions. Someone at Anavex royally screwed up when they determined power or else (more likely) they decided to pick a std dev that would lead to the size study they wanted to undertake. 90% power using 1.5 and sd 4.5 could be obtained with a sample size around 432 hence 450 growing to 509 sounds ok, however, when a realistic sd=8 is used the trial size should have been around 1300!!! Not that it matters much as ADCS completely missed, but the realistic Std dev would have been 7-8 based on historical studies and as was observed in the Anavex trial.
Of course since the study was poorly designed by grossly underestimating variance, it would be even further underpowered for the 70%S1R WT subset.
Market Data 
Markets 
AI
