Replies to post #509354 on Anavex Life Sciences Corp (AVXL)

[Hoskuld]

The approval is not about ADL so your post is moot.

[Hosai]

"“No further confirmatory conclusions are possible… where the primary null hypothesis cannot be rejected.”
(Section 5.4)
ADL failed ? ABCLEAR3 cannot be confirmatory"

Well co primary ADAS COG p value under 0.01, Secondary CDR SB p value 0.01. Dr Jin claimed this makes the trial a success and he has 27 years FDA experience judging which neurology NDA'S to approve. I imagine in terms of this guidance when they say "primary endpoint failing" they are imagining just one primary endpoint, and it fails but the applicant is relying purely on secondaries etc.

"“One or more additional trials should usually be conducted.”
(Section 5.4) Post-hoc = new RCT required"

It does also say things like "unless effect size exceptionally strong", which ABCLEAR 2 and 3 clearly was. Also Anavex should actually have more data in terms of ABCLEAR 2/3 from the placebo group who moved onto to take the drug in OLE. They should be able to look at the 50 odd who were ABCLEAR 3 vs the 50 odd who were non-ABCLEAR 3 and how they fared when they both started taking the drug at week 48 up to weeks 144/192. Can adjust for any baseline differences via co-variate analysis. If ABLCEAR 3 outperforms nonABCLEAR 3 again then this is replicated evidence of additional benefit in this subgroup vs wider population.

"“Credibility depends on biological plausibility and replication.”
(Section 4.6)
ABCLEAR3 has none of these. "

But it does have biologial plausibility? I don't have a strong sense of the science in this regard, but from chat gpt:

"ECM remodeling — COL24A1 encodes a collagen involved in extracellular matrix structure, especially in tissues that remodel under stress.

High neuronal expression — It is expressed in cortex and hippocampus, the same regions most affected in Alzheimer’s disease.

Autophagy–ECM link — ECM stability influences neuronal stress handling and autophagy efficiency, which is directly relevant to blarcamesine’s SIGMAR1–autophagy mechanism.

Disease-responsive gene — Independent 2025 Nature Communications work shows COL24A1 is a robust, disease-regulated gene across large patient datasets (not random noise).

Functional impact of the variant — The missense mutation (R641H) likely impairs ECM stability, so the WT version being the better responder is mechanistically coherent, not arbitrary."

[Doc328]

Correct. Statistical hierarchy is a b*tch

[boi568]

You are attempting to prove that the COL24A1 WT results cannot be considered by the EMA for its blarcamesine MAA. To get there, you cite to a scenario of EMA guidance (5.4) where "The clinical data presented fail to establish statistically persuasive evidence in the primary analysis population but there is interest in identifying a subgroup where a relevant treatment effect and compelling evidence of a favourable risk-benefit profile can be assessed."

The problem with this argument is that its underlying assumption is untrue -- the clinical data show the primary endpoint ADAS-Cog13 met stat sig with a p value of under .025, as did the major secondary endpoint, CDR-SB. The ADL miss is not disqualifying, since this was an early AD trial, per NIA-AA criteria accepted by the EMA (and used by Anavex in the 2b/3). Further, if it were true that the ADL miss were disqualifying, CHMP would have already issued a negative opinion and not indicated to Anavex interest in receiving more data analysis, such as biomarkers, in a re-examination.

You also misattribute (to Section 4.6) and misunderstand how the term "credibility" is used by the EMA. It is a function of several elements, each contributing to the level of credibility achieved. The EMA guidance recognizes elsewhere that, for example, post hoc data may be considered in the context of a single trial, even though "replication" is also an element of credibility.