luza studios/E+ via Getty Images
There’s some regulatory drama going on in the ALS area. There aren’t a lot of things that have worked out in that field, which is, of course, very bad news for everyone involved. For example, Biogen (BIIB) and Ionis (IONS) just dropped an antisense drug candidate because of a complete miss on its clinical trial endpoints. There are no really good animal models for neurodegenerative diseases, so every clinical trial is even more of a shot in the dark than usual and all too often these are the results.
That’s where Amylyx (AMLX) comes in with their AMX0035. That’s a fixed-dose combination of two drugs, sodium phenylbutyrate and tauroursodeoxycholic acid (taurursodiol). Unpacking that, sodium phenylbutyrate is a compound with several different effects. Its most well-known activity is to increase nitrogen excretion in the kidneys, and the drug is thus given to babies with urea-cycle disorders, where there is an immediate need to shed such compounds because normal urea formation and excretion are disrupted. (The mechanism is the metabolism of the phenylbutyrate to phenylacetate, which then conjugates to glutamine, with the resulting phenylacetylglutamine being excreted. It’s already a natural nitrogenous metabolite waste product, and this just turns up the volume on that pathway.)
Second, phenylbutyrate is known to be a sort of chemical chaperone for protein folding, with effects on (for example) mutant CFTR protein in cystic fibrosis. That same effect has shown up with a number of other misfolded proteins in vivo, and it also seems to reduce protein aggregation. For this reason, phenylbutyrate has been studied in the clinic for ALS, Huntington’s, and other such conditions, although it doesn’t look like any of these have progressed.
The other component of AMX0035, taurursodiol, is a bile acid. It’s been used for several liver and biliary diseases, although it’s only approved in the US for biliary cirrhosis. But it has effects well past that system because the compound seems to have effects on apoptosis pathways. That’s led to its study in ALS, Huntington’s, Parkinson’s, Alzheimer’s, and so on, in an attempt to ward off the neuronal death phenotype seen in these diseases. My impression is that the apoptosis effects are certainly real, but that the mechanisms behind them have not been worked out in detail (and for good reasons – it’s a complex area indeed).
So, Amylyx is trying both of these at the same time; that’s their shot at an ALS therapy, and it certainly seems worthwhile, although the main thing you would hope to see is a slowing of disease progression. And that’s where the arguing starts. The company has completed its trial and believes that it has shown evidence of efficacy, but as you will see from this FDA briefing document for Wednesday’s advisory committee meeting, there are a lot of issues to be discussed around this (see page 47 of that document for a summary).
Among these is (1) the rating scale (ALSFRS-R) used in the trial to assess the progression of the disease, and whether that progression can be assumed to be linear or a more complex curve. If you make the linear assumption, then the drug shows statistically significant effects (barely) on the rate of decline of function, but if you don’t, that significance can disappear immediately. It’s worth noting that Amylyx themselves analyzed the data both ways and came to that conclusion as well. The FDA’s position is that “The Applicant’s primary analysis (slope analysis) assumes linearity of ALSFRS-R over time. However, linearity over time is not established for the ALSFRS-R in patients with ALS and exploratory analyses raise questions about the validity of the linearity assumption.”
There’s also (2) a question of patient deaths and how they are accounted for in the data: two of the trial participants died in the drug-treatment arm of the study, but their ALSFRS-R data are not included in the analysis. This was not a huge trial (89 in the treatment arm, 48 in the placebo/standard-of-care group), so that really could affect the numbers, particularly since the results were so close to being non-significant in general. The primary analysis by Amylyx does not account for patient deaths at all, actually. Added to that, (3), there’s a problem with that “standard of care” part. There were fourteen patients in the AMX0035 treatment group who were also getting edaravone or riluzole, the only two drugs that are approved for ALS, but only two of the 48 in the placebo group were on either drug. Neither of those is game-changing therapy, but they showed enough effect to be approved for the disease, and if significantly more patients were taking them in the treatment arm, you would have to worry about the disease progression numbers being affected. Having this sort of imbalance in the two arms of the study seems like a serious mistake.
There’s more (4): both patient groups are missing around 17% of the data from week 24 of the study, and the FDA is concerned that “unverifiable missing data assumptions” could have skewed the analysis as well. The trial also (5) had a randomization problem – the first 18 patients were all assigned to the AMX0035 treatment group, apparently because of a shipping problem that made placebo doses unavailable at the start of the trial, and they then assigned the next 9 to the placebo group. The FDA notes that “The unblinded (Data and Safety Monitoring Board (DSMB) statistician became aware of this at the first DMSB meeting and then attempted to adjust the pre-planned randomization schedule to fix this problem. The Applicant’s reported analyses are for the “as treated” groups for the first 27 patients rather than for the “as-randomized” groups“.
It looks like every time the FDA’s own statisticians tried to correct for these issues, it pushed the results of the trial well back into non-significance as originally defined. Another way to look at this, and it’s not a very complimentary one, is that Amylyx has presented their data in what seems to be the only way to make the trial show a successful result. The agency says that it is “challenging to interpret” the statistical significance after all this, and particularly in such a small sample. As noted here at Endpoints, the company is also making a lot out of a survival benefit number in an open-label extension of the trial, but that analysis has plenty of problems of its own.
But now we get to the regulatory decision itself. As we have seen in the past, the FDA is completely capable of approving this drug anyway. This is in many ways quite similar to the Sarepta (SRPT) and Aduhelm situations: terrible neurological disorders with very few treatment options, and a drug comes along that maybe, sorta, kinda might have shown benefit if you line up the numbers in the most favorable way possible and turn the lights down. And in each case, the companies involved did indeed line up the numbers in that way, while the FDA’s own statisticians have said (and are saying here) that the positive results just don’t hold up if you apply normal standards to the data. As mentioned, the AdComm meeting is tomorrow, and the FDA decision itself will come in June. There’s a lot of very aggressive advocacy coming from patient groups on this (naturally enough), and I do not envy the FDA their job of making this decision. My own take is that the clinical trial was not well run and that it has enough statistical issues to deny approval of the drug pending better data, which can only come from a new trial. It seems likely that a larger, better-run effort would in fact show that AMX0035 does no good for ALS patients. But the FDA might approve it anyway (with a post-marketing trial requirement, you’d think) so that everyone can feel better about ALS.
And, in the long run, I think this new standard of “You have to show safety and efficacy, unless of course you really, really want this drug to approved anyway” is a bad idea. I don’t think you can bend the rules just a little and look away from the data just a little, and keep getting away with it.
Disclosure: None
Editor’s Note: The summary bullets for this article were chosen by Seeking Alpha editors.
Be the first to comment