AVXL - Anavex: The Pediatric Rett Trial Failure Is Disastrous

2024-01-05 05:17:06 ET

Summary

• Anavex's Alzheimer's trial failed one of its two co-primary endpoints due to a faulty application of statistical analysis.
• The company's stock nearly doubled after publishing preclinical data on blarcamesine's effects on rats despite its limited relevance to human trials.
• Anavex's Rett Syndrome trial showed no statistical significance in improving patient outcomes, leading to a significant drop in the company's stock.

In my October article , I discussed the statistical basis for saying that the Anavex Life Sciences ( AVXL ) Alzheimer's trial was a failure. This trial of blarcamesine, the company said, demonstrated slowing of cognitive decline in a key phase 2b/3 study conducted in ~500 participants with early symptoms of Alzheimer's, including mild cognitive impairment and mild dementia. But I discussed how a faulty application of the so-called Bonferroni correction obfuscated the fact that the trial failed one of its two co-primary endpoints. While it may sound unfair that failing one endpoint makes the whole trial a failure, that's how statistics works; and that's why these are called "co" primary endpoints.

Later that month, the company published preclinical data from the journal Neurobiology of Aging that showed blarcamesine was able to prevent cognitive decline… in rats. Specifically, there was a decline in the disease biomarker amyloid beta in rats. Doubtless, this made the senile murine population a happy lot, but at this late stage of the game, when the company is already doing - and failing - human trials in Alzheimer's, I don't see what benefit rat studies may have on an FDA approval.

Interestingly, the stock nearly doubled on this rat data in the next two months. AVXL has a 66% retail ownership, and this doubling on irrelevant data is one reason I have a strict 75-25 rule of ownership, which basically says that if a stock is owned more than 25% by retail owners, that's bad news. Because us retail buyers are a volatile lot. Institutional buyers are, one hopes, less ignorant. They, one hopes, do their homework better.

That Anavex clearly understands this phenomenon shows up from their next press release - the announcement that they have "started" the process of an MAA filing. I read company press releases all day. I have never seen a sample like this. I read press releases which say that an MAA application has been filed. I read press releases which say the CHMP has approved the application. I have never, ever, seen one that talks about "starting" the application process. Yet the stock went up on this "news."

In another demonstration of the beautiful use of the English language, company management in their November earnings call said this about the timeline of their Rett Syndrome trial:

…we are on track to release this data once we have it…

Earlier, in August, they had told us that they would release this data in 2023. Now, claiming delay due to "an additional safety follow up," they said the data release would be delayed. To be fair, though, they did release this data on Jan 2, which isn't a big delay. However, given the data that was released, one can understand why the company might have been reluctant to release it.

This data was from the EXCELLENCE trial, designed to evaluate ANAVEX 2-73 in 92 children aged 5-17 with Rett syndrome. There were two co-primary endpoints, Rett Syndrome Behaviour Questionnaire (RSBQ), and Clinical Global Impression-Improvement Scale (CGI-I). RSBQ is a behavior assessment, while CGI-I is a clinician's professional, if subjective, scoring. RSBQ is a Rett focused measure that includes questions related to various behaviors, skills, and symptoms commonly observed in individuals with Rett Syndrome, such as hand behaviors, mood, social interactions, and communication skills. CGI-I is used in various psychiatric conditions, and it is a subjective rating by a clinician, often based on their clinical judgment and observations, regarding how much the patient's condition has improved or worsened since the beginning of treatment or intervention.

The molecule failed on both endpoints. While the RSBQ score showed an improvement, there was no statistical significance. The CGI-I score was an abject failure.

Here's some of the data :

In an ad-hoc analysis, using the predefined mixed-effect model for repeated measure (MMRM) method, after 12 weeks of treatment, ANAVEX®2-73-treated patients improved LS Mean (SE') -12.93 (2.150) points on their RSBQ total score compared to LS Mean (SE') -8.32 (2.537) points in placebo-treated patients. The LS Mean difference (SE') of -4.61 (2.439) points between treated and placebo groups did not reach statistical significance (n=77; p=0.063).

Interestingly, this "ad hoc" analysis only shows 77 patients, whereas the trial supposedly had 92. Even with a smaller number of patients, there was no stat sig.

In my May 2022 article , I defended Anavex against certain criticisms of another Rett syndrome trial, the AVATAR trial in adult patients. I consider myself fair-minded, and I thought the criticisms lacked basis. The criticism centered around changing the endpoint from RSBQ to RSBQ-AUC two weeks before the data announcement. I tried to lend some support to the company's explanation that while the changes were done well before the data announcement, they were not properly updated on the clinical trials' registry in time.

For the pediatric trial, however, there is really nothing left to discuss. The trial was a failure. How much of an effect the pediatric Rett data will have on the rest of the pipeline is anyone's guess, but for now, the stock has taken a terrible beating, going down losing 40% overnight.

The company did try to explain the poor show by claiming a very strong placebo effect, but their reasons sounded hollow. Here's what they said:

Although data analysis is ongoing, the early conclusion is that the placebo rate could have been higher in the study due to a slight imbalance in disease severity at baseline, across the treatment arms, and the 2 to 1 drug to placebo randomization ratio.

How slight? Slight imbalances at baseline naturally occur in every clinical trial, and can cause what is called chance bias, but a large enough population is designed to average out those imbalances and reduce their impact on the placebo effect. It is also important to understand which baseline characteristics differed strongly, because that may tell us whether those variables had any strong associations with outcome. This trial had 92 patients, which isn't very large, but a decent enough number nevertheless. Also, I do not get how a 2 to 1 drug to placebo randomization hurts the balance. I mean, this is pretty standard practice, done to actually produce a stronger treatment effect, and not meant to produce what happened here, its exact opposite.

Bottomline, AVXL becomes an even more difficult company to defend with this trial failure. While they have some cash - $149.2mn to be exact - they just have that one drug, and multiple dubious trial results. I am staying very far away from this stock.

For further details see: