Before the Conclusion | After Approval: Drugs Reconsidered · Case two
For a slowly progressing disease, waiting another year for more reliable research results may mean one more checkup. For someone with amyotrophic lateral sclerosis, or ALS, a year may instead bring irreversible loss of function: movements once possible become difficult, and speech, swallowing, and breathing may gradually be affected.[1]
For that reason, both requiring patients to wait and allowing a drug with uncertain efficacy onto the market carry costs.
Relyvrio's story presents this difficulty especially clearly. In 2022, the drug was approved in the United States to treat ALS. Less than two years later, a larger trial failed to confirm the expected efficacy, and the company announced the start of market withdrawal. The question it leaves is how much confidence the positive findings available at the time should have inspired.
What the early trial actually found
Relyvrio's development code was AMX0035. It combines sodium phenylbutyrate and taurursodiol. The development hypothesis was that intervening in cellular stress and mitochondrial injury could help protect nerve cells. That offered a reason to conduct research, but the original US label also explicitly stated that its exact mechanism of action in ALS was unknown.[2,3]
The initial pivotal study was CENTAUR. It randomized 137 patients to the drug or placebo and followed them for 24 weeks, primarily comparing the rate of decline on the Revised Amyotrophic Lateral Sclerosis Functional Rating Scale, or ALSFRS-R.[2]
The scale covers functions such as speech, swallowing, walking, daily activities, and breathing. Its maximum score is 48; lower scores generally indicate greater impairment. It measures patient function, rather than a biomarker changing only on a laboratory report. This story therefore differs from a dispute over “an improved marker without knowing whether patients benefit.”
The published study reported an average monthly decline of approximately 1.24 points in the drug group and 1.66 points in the placebo group. At 24 weeks, the estimated between-group difference was 2.32 points. The early signal was therefore slower functional decline, rather than restoration of all the abilities patients had already lost.[2]
For patients continually losing function, slowing decline may itself be meaningful. But this difference was estimated from a limited number of patients over limited follow-up, leaving considerable uncertainty. An average scale difference also cannot be directly translated into the same additional period of independent living for everyone.
Questions already existed alongside the positive results
The FDA review documents show that the evidence supporting approval was not uncontested. The primary analysis did not jointly account for function and death in the agency's generally preferred manner. Statistical results also changed under different assumptions about missing data. Other functional or respiratory secondary endpoints did not provide consistent statistical confirmation.[4]
These issues have practical meaning. In an ALS trial, a missing follow-up score may reflect withdrawal or discomfort, but also severe illness or death. How those unmeasured scores are handled affects the comparison between groups. Statistical methods need to reflect patients' actual experiences as closely as possible, rather than calculating results only for those who remain to complete assessments.
Longer follow-up also produced a positive survival signal. This made the case for approval more attractive, but the extension allowed open-label treatment and could not indefinitely preserve the original double-blind, placebo-controlled conditions. The FDA therefore cautioned in its public materials that these post hoc, exploratory survival analyses required careful interpretation.[1,4]
In other words, the choice in 2022 was not simply between “no evidence at all” and “full certainty.” There was a set of findings worth taking seriously, alongside clear limitations. The dispute concerned whether they were already sufficient to support widespread use.
Why the FDA approved it anyway
The FDA ultimately concluded that, given the severity and fatal nature of ALS and the substantial unmet treatment need, it could accept greater residual uncertainty than usual and exercise regulatory flexibility in granting approval. Its review documents set out this balance quite explicitly.[4]
An easily confused distinction matters here: Relyvrio received traditional US approval. It should not be placed directly in the category of accelerated approvals based on surrogate endpoints. The FDA subsequently also listed it explicitly under the traditional approval pathway in background materials for another ALS review.[5]
The later failure of PHOENIX therefore cannot simply be described as “failure of the confirmatory trial required under accelerated approval.” PHOENIX was a phase 3 study that had begun before market entry to provide stronger evidence of efficacy. Its results could change the judgment about the drug even though they did not fall under that same statutory approval arrangement.
Traditional approval does not mean every subsequent question has disappeared. The central issue in this case is how regulatory flexibility should be used in a serious disease, and how rigorous verification should continue after uncertainty has been accepted.
The larger trial did not reproduce the early hope
PHOENIX enrolled 664 patients and ran for 48 weeks. In March 2024, the company announced that it had not met the prespecified primary endpoint. Change from baseline in ALSFRS-R score at week 48 did not differ statistically significantly between drug and placebo. The P value for the primary comparison was 0.667, and the prespecified secondary endpoints also did not reach statistical significance.[6]
A P value cannot be interpreted as the probability that a drug works. Nor can one number alone establish that every tiny effect is impossible. The more defensible statement is that this larger study with longer observation did not provide the evidence of functional benefit previously hoped for.
Patients and study conditions cannot be identical in an early and a later trial. That permits investigation of why results differed, but “the populations were different” alone cannot exclude unfavorable findings from the assessment. If a drug is thought to help only a narrower population, that population needs to be explicitly defined and supported by evidence capable of testing the claim.
For medical judgment, the importance of the new result also extends beyond adding another paper. It directly tested the central promise that had supported clinical use: could patients' functional decline be reliably slowed? When stronger research did not support that promise, confidence in it had to fall accordingly.
Why withdraw without a major new safety alarm?
When the company announced the PHOENIX findings, it did not identify a major new safety problem as the reason for withdrawal. A drug therefore need not first cause catastrophic toxicity to lose its justification for continued use.[6]
Even if common adverse effects are mainly gastrointestinal, patients still bear the burden of taking medication, paying for treatment, and spending time. Belief in a therapy may also alter their other choices. Without credible clinical benefit, the absence of a particularly severe new risk is not, on its own, a reason to continue use.
On April 4, 2024, Amylyx announced the start of voluntary market-withdrawal procedures in the United States and Canada and stopped providing the drug to new patients. Existing patients could transfer to a free-drug program after discussion with their doctors. The company's 2024 annual report later stated that the last free supplies covered the period through early 2025. This was therefore a time-limited transition arrangement, not an indefinitely open free-treatment program.[7,9]
The US legal process followed these commercial and patient arrangements. Formal FDA documents record that the company later requested voluntary withdrawal of approval and waived the opportunity for a hearing. US approval was ultimately withdrawn on August 29, 2025. An account of this history needs to distinguish the date market withdrawal was announced from the date approval was formally withdrawn.[8]
Reconsidering “one more option for patients”
Supporters of earlier approval were concerned that patients could not wait until every question was answered. Those calling for stronger evidence were concerned that patients likewise had no spare time to invest in a treatment that might ultimately fail to demonstrate help. Both concerns arose from real patient circumstances.
Relyvrio's later failure cannot, by itself, establish that everyone making judgments in 2022 was careless. Assessing that decision requires returning to the information then available: were evidence limitations honestly presented, was the uncertainty accepted proportionate to the possible benefit, and was subsequent research advanced promptly?
Once new results appear, however, maintaining the same confidence requires new reasons. Preserving hope for patients does not mean requiring the evidence to uphold the original promise forever. One way research helps patients is by identifying as early as possible which directions deserve investment and which should be stopped or redesigned.
In my judgment, the value of a system offering earlier treatment opportunities depends on equally serious subsequent verification. Research should preferably have begun before market entry. Important results should be disclosed promptly, and both favorable and unfavorable evidence should inform judgment. If the original rationale fails, a workable patient transition must also be arranged.
Relyvrio shows that the cost of waiting must be considered, but it cannot replace proof of a drug's benefit. Respecting patients' limited time also means promptly acknowledging when a treatment has not withstood further testing.
Costs and responsibility after withdrawal require separate discussion. Free supply addresses continued access to treatment for a period. It does not automatically reimburse patients' prior spending or constitute compensation for harm. In the series' closing article, “After a Drug Is Withdrawn, Who Bears Patients' Losses?”, we will compare these different arrangements separately.
The next article turns to a drug that relieved pain but left the market because of serious safety risks: Vioxx. It raises the question of when we should stop using a treatment that really does have an effect.
Main sources
[1] FDA: Snapshot of the trial supporting Relyvrio's original approval. Includes disease background, the functional scale, the sample, and limitations of the survival analysis. The page reflects information at the time of original approval.
[2] Paganoni et al., New England Journal of Medicine, 2020: The CENTAUR randomized trial.
[3] FDA, 2022: Original Relyvrio label. Includes the unestablished mechanism of action and safety information.
[4] FDA, September 29, 2022: Summary review supporting approval. Includes analytical limitations, regulatory flexibility, and the benefit–risk assessment.
[5] FDA, September 27, 2023: ALS review background materials. Explicitly identifies Relyvrio as approved through the traditional pathway.
[6] Amylyx, March 8, 2024: PHOENIX phase 3 topline results. This article reports results from the original announcement and does not present it as a full peer-reviewed paper.
[7] Amylyx, April 4, 2024: Withdrawal decision and free-drug transition arrangements.
[8] FDA / US Federal Register, August 29, 2025: Formal withdrawal of Relyvrio approval.
[9] Amylyx, 2024 annual report: Records of withdrawal, final free-drug supplies, and revenue adjustments. Free supply, channel returns, and patient compensation must be understood separately.
Sources checked through October 28, 2026. This article is for discussion of medical and scientific information and does not provide individual medical advice. Discuss specific treatment changes with your doctor.
Beyond the conclusion, there are still questions worth asking.
Medicine · Science · Ideas and reading
Follow Before the Conclusion to stay curious and think independently together.
Text and image usage
The text of this article was originally written by Before the Conclusion. You are welcome to share the article's webpage link. Without separate permission, do not republish or excerpt the text by copying, reposting, redistributing screenshots, or similar means. Unless otherwise marked, article illustrations may not be separately republished, copied, modified, or used commercially. Rights in cited materials belong to their respective rights holders.