The Contentious Path of AMX0035 to FDA Approval

AMX0035, a combination of sodium phenylbutyrate and taurursodiol marketed as Relyvrio, became one of the most polarizing drug approvals in recent FDA history when it received the green light for amyotrophic lateral sclerosis in September 2022. The approval rested on a single Phase 2 trial of 137 patients, and the story that followed, a failed Phase 3 confirmatory trial and voluntary market withdrawal, has reshaped conversations about how regulatory agencies should handle fatal diseases with few treatment options.

What AMX0035 Is and How It Was Supposed to Work

AMX0035 pairs two existing compounds: sodium phenylbutyrate, a drug long used to treat certain metabolic disorders, and taurursodiol (also known as tauroursodeoxycholic acid, or TUDCA), a bile acid. Each targets a different cellular stress pathway thought to contribute to nerve cell death in ALS. Sodium phenylbutyrate addresses stress in the endoplasmic reticulum, the cellular compartment responsible for folding proteins correctly. Taurursodiol works on mitochondria, the structures that supply energy to cells. The rationale was that hitting both pathways simultaneously might slow the disease more effectively than targeting either one alone.

Animal research has supported this two-pronged concept. In rats with stroke-induced brain damage, AMX0035 reduced markers of both endoplasmic reticulum stress and mitochondrial dysfunction, and it lowered the death rate of oligodendrocytes, the cells that insulate nerve fibers.¹ Whether this preclinical activity translates cleanly to people with ALS was always the question.

The CENTAUR Trial That Started Everything

The pivotal human evidence for AMX0035 came from CENTAUR, a randomized, placebo-controlled Phase 2 trial that enrolled 137 people with ALS at 25 sites across the northeastern United States. Over six months, participants receiving the drug lost function more slowly than those on placebo, as measured by the ALSFRS-R, a standard 48-point scale that tracks activities like speaking, swallowing, walking, and dressing. The drug group declined by about 1.24 points per month compared with 1.66 points per month for placebo, a difference of 0.42 points per month that reached statistical significance.²

That difference, while statistically real, was modest. The trial was small, with just 137 patients.³ Critics pointed out that a trial this size produces fragile results: a handful of patients doing unexpectedly well or poorly can push a borderline finding over or under the threshold for statistical significance. Supporters countered that the signal was consistent and pointed in the right direction for a disease with essentially no effective treatments.

Survival Data That Raised the Stakes

What made the CENTAUR story more compelling than many Phase 2 results was the follow-up survival analysis. After the randomized phase ended, all surviving participants were offered open-label access to the drug. When researchers tracked long-term outcomes, the group originally assigned to AMX0035 lived a median of 25 months, compared with 18.5 months for those who started on placebo. That 6.5-month difference was statistically significant.⁴

For a disease that typically kills within three to five years of diagnosis, a potential gain of more than half a year mattered enormously to patients. But the survival analysis came with its own limitations. It was not the trial’s original primary endpoint. The open-label phase, where everyone received the drug, made it impossible to maintain a clean comparison between treated and untreated groups over the long term. These kinds of post-hoc analyses are suggestive but not definitive, and the field knew it.

A Fractured Advisory Committee and the FDA’s Decision

The path to FDA approval was anything but smooth. When the agency’s advisory committee first considered the data, the vote was not clearly supportive. In an unusual move, the FDA reconvened an advisory panel, and public testimony from ALS patients and their families played a significant role in the proceedings.⁵ The emotional weight of those testimonies was immense. ALS patients, many speaking through assistive devices, described watching their bodies fail and pleaded for any treatment that might slow the decline.

The approval that followed drew criticism from multiple directions. Some observers argued that the FDA had bent its usual evidentiary standards under political and emotional pressure. A peer-reviewed analysis later described the approval process as “questionable,” noting that the agency reconvened advisory commissions to obtain favorable decisions and designated the drug as a new molecular entity, a classification that granted it market exclusivity.⁶ The new molecular entity designation was particularly contentious because AMX0035 combines two substances that were already individually available, raising questions about whether that label was scientifically appropriate or primarily served commercial interests.

Others, including many in the ALS community, defended the decision. They pointed to the survival benefit, the favorable safety profile, and the brutal reality that people with ALS do not have time to wait for perfect data. Bayesian statistical frameworks were invoked to argue that even with the small sample size, the totality of evidence was sufficient to justify access.³ The underlying tension was philosophical as much as scientific: how much uncertainty is acceptable when patients are dying and asking for help?

What Happened After Patients Started Taking It

Once AMX0035 reached the market under the brand name Relyvrio, real-world experience quickly revealed practical challenges beyond the scientific debate. At one U.S. ALS center, among 73 patients with available data, 54 started the drug and 19 never did. Among those who never initiated treatment, the most common reasons were personal choice or out-of-pocket costs.⁷ Of those who did start, roughly 44% eventually stopped, with side effects being the primary reason for discontinuation.

A separate single-center report found that about 13% of patients who never started treatment cited high cost or insurance denial as the barrier.⁸ Relyvrio’s list price at launch was roughly $158,000 per year, a figure that drew sharp criticism given the uncertain evidence base. For a drug whose benefit was debated among experts, the pricing felt especially aggressive to critics who viewed the approval itself as premature.

The side effect profile in practice was not dangerous but was burdensome. The drug came as a powder dissolved in liquid, and its taste was widely described as unpleasant. Gastrointestinal problems, particularly nausea and diarrhea, were the most frequently reported issues. For people already struggling with swallowing difficulties from ALS, the large volume of liquid required to take the medication added another layer of hardship. These tolerability issues help explain why discontinuation rates were high even among motivated patients desperate for effective treatment.

PHOENIX and the Collapse of the Evidence

Everything hinged on the Phase 3 confirmatory trial, PHOENIX, a global study designed to validate the CENTAUR findings in a larger, more diverse population. When the results came in, they were devastating. PHOENIX failed to meet its primary endpoint: there was no statistically significant difference between AMX0035 and placebo on the ALSFRS-R at 48 weeks. The p-value was 0.667, nowhere close to significance.⁹ No secondary endpoints showed a benefit either.

Perhaps the most telling detail: even when researchers looked specifically at a subset of PHOENIX participants who matched the CENTAUR trial’s enrollment criteria, they found no significant difference. This undermined the argument that CENTAUR’s signal was real but limited to a particular patient subgroup. The larger, better-powered trial simply did not replicate the earlier result, and the failure was not marginal. A p-value of 0.667 means the drug performed essentially identically to placebo.

For the scientific community, PHOENIX was a sobering reminder of why regulators usually require Phase 3 confirmation before approval. Small trials produce statistically significant results more often by chance than large ones do. The CENTAUR result, while genuine at the time of analysis, may have been a statistical artifact amplified by a small sample, normal variability in ALS progression, and the particular patients who happened to enroll.

Voluntary Withdrawal From the Market

To its credit, Amylyx Pharmaceuticals did not try to keep the drug on the market after PHOENIX failed. The company initiated the process with both the FDA and Health Canada to voluntarily withdraw Relyvrio’s marketing authorization.⁹ The drug was pulled from pharmacy shelves, and patients were transitioned off treatment.

The withdrawal created its own kind of anguish. Patients who believed they were benefiting from the drug, whether through a real biological effect or through the psychological power of taking an active treatment for a disease that otherwise offers so little, had to stop. Some ALS advocacy groups pushed back against the withdrawal, arguing that even a small or uncertain benefit was better than nothing. Others acknowledged that continuing to sell a drug at $158,000 a year without evidence of efficacy was untenable both ethically and practically.

It is worth noting that European regulators had reached a different conclusion much earlier. The European Medicines Agency declined to approve AMX0035, judging the CENTAUR data insufficient on its own to support marketing authorization. That decision, criticized at the time by some patient advocates as overly conservative, looked prescient after PHOENIX.

The Broader Debate About Approving Drugs for Fatal Diseases

The AMX0035 saga sits at the center of an ongoing tension in drug regulation. For diseases like ALS, where the average patient lives only a few years after diagnosis and available treatments offer limited benefit, the usual drug development timeline can feel inhumane. Patients understandably want access to anything that might help, and they resent being told to wait for larger trials that may take years to complete.

Regulators, for their part, have mechanisms designed to speed access: accelerated approval pathways, breakthrough therapy designations, and expanded access programs. But the AMX0035 case illustrates the risk of those mechanisms. A drug approved on thin evidence may not just fail to help; it can drain financial resources from patients and insurers, expose patients to side effects without compensating benefit, and consume clinical bandwidth that might be directed elsewhere. When the evidence collapses, public trust in the approval process itself takes a hit.

The reconvened advisory committee was a particular sore point. Regulatory agencies must balance patient advocacy with scientific rigor, and the perception that emotional testimony swayed a scientific decision troubled many researchers. At the same time, dismissing patient testimony as mere emotion ignores the legitimate insight that people living with a disease bring to risk-benefit calculations. A healthy volunteer in a statistics class and a person losing the ability to breathe may reasonably weigh the same uncertain data differently.

Independent Research on TUDCA and What It Suggests

One component of AMX0035, taurursodiol (TUDCA), has been studied independently in ALS outside the Amylyx program. A Phase 2b proof-of-concept trial tested TUDCA alongside riluzole, the longstanding standard treatment. The results were positive: TUDCA-treated patients declined about seven points less per year on the ALSFRS-R compared with riluzole alone, a difference that corresponded to an estimated four to five months of additional median survival. The proportion of patients classified as responders was 87% in the TUDCA group versus 43% in the placebo group.¹⁰

This is interesting because it suggests that at least one of AMX0035’s two components may have genuine biological activity in ALS, even if the combination product failed to prove its worth in a large trial. Independent TUDCA research is ongoing, with a dedicated trial protocol designed to generate the kind of rigorous Phase 3 data that CENTAUR could not provide. If TUDCA alone demonstrates efficacy in a well-powered trial, the AMX0035 story may look less like a complete dead end and more like a case where the wrong formulation or the wrong development path obscured a real signal.

That said, the independent TUDCA data also came from a relatively small study, so the same caution applies. The field has learned, painfully, that promising Phase 2 results in ALS have a poor track record of surviving Phase 3 confirmation. Dozens of compounds that looked encouraging in small trials have failed in larger ones. ALS is a heterogeneous disease, with wide variation in how fast patients decline, which makes it particularly prone to producing false-positive results in underpowered studies.

How ALS Heterogeneity Complicates Clinical Trials

One of the underappreciated lessons from the AMX0035 experience is how much the natural variability of ALS itself confounds drug development. Some patients decline rapidly and die within a year; others live a decade or longer. The rate of progression can shift unpredictably, with periods of relative stability followed by sharp drops. This variability means that small trials are almost guaranteed to have imbalanced groups, where one arm happens to contain more fast progressors or slow progressors than the other, no matter how carefully randomized.

In a 137-person trial like CENTAUR, even modest imbalance can generate a statistically significant difference that has nothing to do with the drug. A global trial like PHOENIX, with hundreds of participants across many countries, dilutes that noise. The fact that PHOENIX found absolutely no signal is strongly suggestive that CENTAUR’s positive result was an artifact of its size rather than evidence of a real treatment effect.

This problem is not unique to AMX0035. The ALS field has been struggling with it for decades. Enrichment strategies, where trials select for patients with a particular genetic subtype or rate of progression, are one approach to reducing variability. Biomarker-based endpoints, like levels of neurofilament light chain in blood, are being explored as alternatives to clinical rating scales, which rely on subjective patient reports and can be noisy. These methodological advances may eventually make ALS trials more efficient, but they were not in place for CENTAUR or PHOENIX.

The Cost of Hope

The financial dimension of AMX0035 deserves attention separate from the scientific questions. At roughly $158,000 per year, Relyvrio was priced at a level that assumed meaningful clinical benefit. Once the drug was approved, insurers were generally expected to cover it, though the real-world data showed that cost and insurance denial kept some patients from starting treatment.⁸ The total revenue generated during the roughly two years the drug was on the market is difficult to calculate precisely, but it represented significant spending on a treatment whose benefit was always uncertain and ultimately unsupported.

The “new molecular entity” designation that AMX0035 received compounded the cost issue. Both sodium phenylbutyrate and TUDCA are available individually, and TUDCA can be purchased as a dietary supplement at a fraction of the cost. Granting market exclusivity to a combination of two known substances effectively prevented patients from accessing cheaper alternatives through compounding pharmacies, even though the combination itself had not proven superior to either component alone. For critics, this was the clearest evidence that commercial interests had outrun the science.⁶

None of this means Amylyx acted in bad faith. The company agreed to conduct the Phase 3 trial as a condition of approval and withdrew the drug when that trial failed. That sequence, approval with a confirmatory commitment followed by honest withdrawal, is exactly how the accelerated approval framework is supposed to work. But it leaves behind real financial and emotional costs borne by patients and the healthcare system during the interval between premature approval and subsequent failure. Whether those costs are acceptable is a question that different stakeholders answer very differently depending on where they sit.

What Patients Took Away From the Experience

For the ALS community, the AMX0035 arc was whiplash. The approval was celebrated as a victory for patient advocacy and a sign that regulators were finally listening to people living with terminal illness. The withdrawal felt like a betrayal, even though the scientific case for it was clear. Trust, once given, is hard to retract cleanly.

The high discontinuation rates seen in real-world practice add another dimension. Even among patients who wanted desperately to believe in the drug, nearly half of those who started it at one center stopped, mostly because of side effects.⁷ The gap between wanting an effective treatment and tolerating the one that was available proved substantial. This is a recurring theme in ALS, where the physical burdens of taking a medication can collide with the physical burdens of the disease itself, and the math only works out if the drug is genuinely helping.

Some patients and caregivers have channeled their frustration into advocacy for better trial designs, more transparent regulatory processes, and increased federal funding for ALS research. Others have grown more skeptical of the pharmaceutical industry and the FDA alike. Both reactions are understandable. The AMX0035 story did not create the desperation that drives ALS patients to accept uncertain treatments, but it did put that desperation under a spotlight and forced everyone involved to confront its consequences.