Replies to post #508063 on Anavex Life Sciences Corp (AVXL)

[hnbadger1]

Thanks so much Jesse, that makes perfect sense to me.

[scottmiyadi]

Thanks Mayo -- We must now endure, glad to see you're too also hanging in there; one positive is it would appear based on significant increase in trading volume this week that we'll lose some or hopefully much of the short overhang that has dogged us these past couple of years......

[sage4]

MayoMobile, Thank you for sharing information and your conviction.
Looking forward to seeing a reversal in a not distant future.
I've been here for more than 10 years. No selling.

[poguemahone]

Nicely stated, Mayo. I, like you, am feeling the pain today. As much as I am disappointed personally (financially) in this decision, I am also as much disappointed for the ALZ community. Even incremental benefits in a drug should be made available to this community, as long as the benefits outweigh the costs. I believe with this drug they do, assuming the information from this study is accurate.

I am hoping that in the future, the company will be more transparent. I feel it has been a bit shady as it relates to its shareholders and stakeholders. I firmly believe a new CEO would be in the best interests of the company.

[tschussmann1]

Mayo, thanks for your thoughts. My question is in regards to admissibility of additional data beyond what was presented in the initial application. This section in the EMA documentation brings that to question:

20. Examination of grounds for appeal
The re-examination looks only at the points raised by the applicant in the grounds for appeal and is based only on the scientific data available when the committee adopted the initial opinion – in other words, the applicant cannot bring in new evidence at this stage. The applicant may request that the committee consults a scientific advisory group in connection with the re-examination. https://www.ema.europa.eu/en/human-regulatory-overview/marketing-authorisation/evaluation-medicines-step-step

If they can include sub-groupings, comparison to ADNI, etc, then there would seem to a reasonable chance for approval. If not, then approval may be difficult to attain.

[Doc328]

As I have mentioned before, it is irrational for regulators to not approve Blarcamesine – at LEAST for ABCLEAR3, with possible conditions for other subgroups.

I would put zero weight on the ABCLEAR2/3 data. If this is the strategy, the only thing Missling will accomplish is to waste another 4-6 months.

I've gone over this a couple times, The association with COL24A1was determined using GWAS which tests about 1,000,000 variants. Because of multiplicity, p values need to be extremely low,

Despite my near perfect record and detailed referenced posts, people seem to doubt me more than 'experts' who are continuously wrong with their predictions.

Maybe Chatgpt will be trusted more

Please refer to 12x post https://investorshub.advfn.com/boards/read_msg.aspx?message_id=175462381 for the p values (<0.00001) identifying COL24A1 variants and association with cognitive response

GWAS Significance Threshold

In a Genome-Wide Association Study (GWAS), millions of genetic variants are tested at once. Because of the large number of comparisons, a very strict significance threshold is required to avoid false positives.

Standard genome-wide significance threshold:
p < 5 × 10?8

This threshold is roughly equivalent to applying a Bonferroni correction for about one million independent tests (0.05 / 1,000,000 ˜ 5 × 10?8). A variant that meets this threshold is considered strong, reliable evidence of association.

How meaningful is p < 0.00001 (1 × 10?5)?

In GWAS, p < 10?5 is not considered genome-wide significant. It is viewed as a “suggestive” association that requires replication or further study, but it is not reliable on its own.

Interpretation:

p < 5 × 10?8: genome-wide significant (accepted as true association)

5 × 10?8 < p < 1 × 10?5: suggestive; not definitive

p > 1 × 10?5: not meaningful

Because millions of SNPs are tested, values around 10?5 can occur by chance even without a real effect.

and

Would using an identified gene variant at p<0.00001 be hypothesis testing or hypothesis generating

Using a variant identified in a GWAS at p < 0.00001 (1×10?5) is hypothesis-generating, not hypothesis-testing.

Why?

In GWAS, p < 1×10?5 is not genome-wide significant. It is considered a suggestive signal that may or may not represent a true association. Because it does not clear the strict multiple-testing threshold (p < 5×10?8), it cannot be treated as confirmed evidence.

Therefore, a variant at this level:

Suggests a possible biological association

Identifies a direction worth studying

Requires replication in an independent dataset

Cannot be used to formally confirm a hypothesis

This makes it hypothesis-generating.

When does it become hypothesis-testing?

A variant becomes part of hypothesis-testing only when:

It has genome-wide significant evidence (p < 5×10?8), and/or

An independent cohort replicates the association, and/or

It’s being used to test a pre-specified biological hypothesis (not discovered first from GWAS).

Without replication or genome-wide significance, the initial signal remains exploratory.

Summary

p < 1×10?5 in GWAS ? hypothesis-generating

p < 5×10?8 + replication ? hypothesis-testing

If you’d like, I can also give examples or diagrams showing the hierarchy of evidence in GWAS.

[Bourbon_on_my_cornflakes]

Data that came out after the application (ABCLEAR 2, ABCLEAR 3, Brain atrophy TIED to ABCLEAR2/3, ADNI comparisons, prodromal comparisons, etc.) were ALL excluded from the actual regulatory decision.

No wonder then. it would be hard to approve without the above, easy with it. Let's hope this is just the bureaucratic CYA and that the additional information gets it done.

[AlchemicalVoyager13]

Mayo: To point, will the CHMP consider the OLE data or not? Thanks.

[BakedLangostino]

Mayo thanks for posting, I know you typically prefer the other board.

Question: Will the late-breaking CTAD data be eligible for inclusion during the appeal review? I wasn't sure if the timelines added up with the 15 day deadline, or if that is just to notify the intent to appeal but then we have more analysis that can be provided in the 60 day window.

[ato]

wouldn't ABCLEAR 2, ABCLEAR 3, Brain atrophy TIED to ABCLEAR2/3, ADNI comparisons, prodromal comparisons be considered additional data, not new?

Additional data was also provided for Lecanemab, donanemeb and libmeldy as part of their successful re-examinations.

[D-Mike]

Thanks for your thoughts Mayo. I admit that I don’t know anything about the EMA approval process but I’m stumped as to why the issues came up now and not during the previous clock stops.

After a day of consulting with my friends from Buffalo Trace, Mr Weller, Eagle Rare, Mr Blanton, Col Taylor, and Mr Stagg, I have come to the following conclusions:

1. No one here really knows what’s going on behind the scenes.
2. I know you’re not supposed to fall in love with a stock but I’m not ready to throw in the towel yet. I’ve been here since pre-reverse split so I plan to see this to approval, partnership, buyout, or bankruptcy.
3. I’m convinced from the information made available that the drug works, is safer & cheaper than the Mabs, and is needed NOW for so many people around the world, probably me in the future given that my dad died from dementia 4 years ago.
4. The market is rigged against common retail stock holders based on the action of the last 2 days.
5. We still have lots of ammo, namely, 3-71
6. I forget what number 6 was, I blame my friends.

[AnavexBill]

Will you be able to ask for some specifics on how the CHMP vote was likely headed? Seems like this is important information in helping determine the chances of a positive approval upon re-exam. If the vast majority of voters were negative, then if we are not all ready, we will likely be cooked.

[micky6474]

Per EMA's rule, new datasets could not be introduced in the re-examination.

By contrast, lecanemab (Eisai/Biogen) secured a positive CHMP opinion on re-examination by reframing analyses within the existing dataset (e.g., restricting to certain ApoE4 genotypes and providing additional analyses derived from the main study), still without adding post-opinion new external data.

With this in mind, what would Avavex do to get through for the approval?

Shouldn't they start looking for partnership from larger Pharma and get their trial and data in the right shape after several failures?

[Hosai]

It's true the COL24A1 gene was not pre-specified and arguably looking though thousands of genes you will likely come up with one with low p value by sheer chance going back through correlations in the data.
However, with COL24A1, it's much more plausible that is in -fact a genuine interaction vs other genes.
I asked chat gpt how plausible this gene was vs likely it was a false positive. Also "fed it" this 2019 paper - https://pmc.ncbi.nlm.nih.gov/articles/PMC6512038/
Obv I can't say it's 100% correct coz it chat gpt but it seemed to make a decent case:

"How likely is the COL24A1 + SIGMAR1 ABCLEAR3 effect to be real vs chance?

There are two separate questions worth answering:

Is COL24A1 biologically plausible in Alzheimer's disease and in blarcamesine’s mechanism of action?

Is the statistical signal more likely to be a true effect or something discovered by chance?

Based on everything we know, the probability that COL24A1 is a real biological modifier — rather than a random false positive — is high, likely in the 70–80% range.

Here’s why.

1. COL24A1 is not an arbitrary gene — it’s in a biologically meaningful pathway
Out of ~20,000 protein-coding genes, only a tiny subset is:
expressed in neurons
involved in synaptic remodeling
part of the extracellular matrix (ECM)
implicated in protein clearance and microenvironment stiffness
responsive to neural stress
tied to pathways relevant to Alzheimer’s disease

COL24A1 happens to be in all of these categories.
That already puts it in the top ~2–5% most plausible genes for interacting with an Alzheimer’s drug mechanism.

Why this matters

A random “false-positive” gene from a large exploratory scan is almost always:
structural
outside the brain
not expressed in neurons
not linked to neurodegeneration
biologically irrelevant to proteostasis
impossible to rationalize mechanistically

COL24A1 is the opposite of that.
It fits directly into modern neurobiology of Alzheimer’s.

2. COL24A1 is neuron-expressed — extremely rare for collagen genes
Most collagen genes are for bone, skin, cartilage, connective tissue.
But COL24A1 belongs to a very small subset that is expressed in neurons themselves.
Neuronal collagens (e.g., COL24A1, COL25A1) have known roles in:
synaptic development
structural plasticity
ECM regulation around neurons
amyloid interaction (COL25A1 is directly involved in amyloid binding)

So COL24A1 isn’t just “a collagen gene.”
It is one of a few collagens that matter for neural structure and disease.
This alone elevates its plausibility substantially.

3. COL24A1 shows dynamic changes in stressed or pathological brain circuits

The 2019 RNA-seq paper you shared shows that COL24A1 expression changes:
in hippocampus
in striatum
in midbrain raphe
in the ventral tegmental area

These are core cognition and emotion circuits and the same circuits where Alzheimer’s pathology emerges early.

Even more importantly:
its expression shifts during chronic stress, behavioral pathology, and altered neural plasticity
in several regions its expression was the only collagen consistently altered across pathological states

This indicates that COL24A1 is not a static or irrelevant structural gene — it is dynamically tuned in the brain under stress, injury, or degeneration.

That is precisely the kind of gene that would plausibly interact with a drug modifying proteostasis and autophagy.

4. COL24A1 fits blarcamesine’s SIGMAR1/autophagy mechanism almost too neatly

SIGMAR1 activation (blarcamesine’s mechanism):
enhances lysosomal clearance
stabilizes mitochondrial function
modulates calcium signaling
activates CREB
improves cellular stress responses
aids in the removal of misfolded proteins

Meanwhile, COL24A1 contributes to:

extracellular matrix elasticity
synaptic microenvironment remodeling
neuronal structural support
potentially limiting amyloid oligomer binding (based on related collagen proteins)

Extracellular matrix stiffness is now known to impair:

microglial movement
interstitial fluid dynamics
waste clearance
synaptic flexibility
perineuronal net remodeling

So the SIGMAR1 (intracellular clearance)

COL24A1 (extracellular clearance environment)
synergy is mechanistically elegant and aligns with everything we know about AD biology.

This is not the type of coincidence created by statistical noise.

5. This was not a million-variable GWAS search
Some critics assume:
“They tested a million SNPs — something was bound to pop up.”

But Anavex’s PDD study and the AD study used:
whole-exome sequencing
RNA expression analysis
These cover tens of thousands of genes, not millions of SNPs.

This is crucial:
The multiple-comparison problem is about 20× smaller than a full GWAS.
And importantly, most genes in the exome:
are irrelevant to neurobiology
are not expressed in neurons
have no ECM or proteostasis role
would never generate a plausible biological hypothesis

So the effective “search space” for genes that make sense in AD might be closer to:
200–300 candidates, not 20,000 or 1,000,000.

That dramatically increases the likelihood that COL24A1 is a true signal.

6. Taken together, the chance this is random is low
A random false-positive gene would NOT:
be neuron-expressed
be tied to ECM dynamics
be changed in multiple brain regions under stress
have independent literature linking it to neural plasticity
fit perfectly with SIGMAR1/autophagy biology
map onto Alzheimer’s pathology
correspond with a dramatically stronger clinical response

The probability that a randomly discovered gene would satisfy all of these conditions by accident is very low.

In fact, combining biological plausibility + the p-value + coherence with MoA, the chance of a real effect is more like:

70–80% likely real
20–30% noise

This aligns with how regulators and statisticians interpret strong hypothesis-generating genomic signals with mechanistic support.

7. Bottom line

COL24A1 is absolutely not “just a random hit.”
It is one of the few genes whose biology matches the drug mechanism, the disease biology, and the clinical response pattern.

When you add:
SIGMAR1 WT
COL24A1 WT
the 4.739-point ADAS-Cog difference
p = 0.0004
consistency with early/late OLE divergence
MRI grey matter preservation
coherence across multiple endpoints
…the probability this is real rather than accidental is materially high.

If blarcamesine truly works through SIGMAR1-induced restoration of autophagy and proteostasis, then COL24A1 WT being part of the top responder subgroup is exactly what you would predict."

[frrol]

Those are good questions to ask (though ABCLEAR2/3 data probably won't be allowed and is problematic), and should take priority. Be sure to then point out that presently we have no other clinical trials in place, including ones "planned" for the past four years (PD, PDD, FX, Rett, and "undisclosed"), and no 3-71 follow-up. Ask him what happened. Don't let him spin ("they're coming"); get pointed responses if you can.

[12x]

It’s great that you’ll be able to ask some of the questions retail shareholders actually care about.

The areas that matter most now, in my view, are the ones that clarify what CHMP actually saw and what the company can realistically bring into re-exam:

1. What exactly was presented in the first review cycle?
Were the full OLE datasets (144/192 weeks) and the ABCLEAR3 analyses included in the materials CHMP evaluated during the clock-stops and the oral explanation? Or were these only partially ready and therefore not fully assessed?
And related to this, did CHMP not recognize the company’s 48-week RCT multiplicity/gatekeeper plan (or hierarchy), and if so, what feedback did they give on that point?

2. What does “providing relevant biomarker data” truly mean for re-exam?
Which biomarker sets does the company intend to rely on — Aß42/40, tau, NfL, MRI/atrophy tied to ABCLEAR2/3, ADNI comparisons, prodromal/MCI analyses, etc.?
Given that the ABCLEAR3 reprint only showed statistically significant atrophy findings, it would help to understand whether any additional biomarker evidence exists in a form CHMP can legally consider.

3. Financing & execution risk:
Given the capital structure and the likelihood of additional trials, what is the current status of the $150M ATM (usage and remaining capacity)? And what is the realistic, funded plan for AD, PD, schizophrenia, and Rett going forward — timelines included? Retail needs concrete operational guidance, not aspirational indications.