Capricor Therapeutics has said, repeatedly and in writing, that its Phase 3 HOPE-3 trial of a cell therapy for Duchenne muscular dystrophy achieved statistical significance on its primary endpoint. The FDA, in briefing documents released Monday morning ahead of a Wednesday advisory committee meeting, says the same trial failed its prespecified primary and secondary endpoints, with no statistically significant difference between the drug and placebo at twelve months.
Those two statements are not describing different trials. They are describing the same trial, and the gap between them is the entire story. The stock fell roughly 40% within hours of the documents posting. Understanding how a company and its regulator can look at one dataset and reach opposite conclusions is the key to reading what happens Wednesday, and it matters far beyond one small biotech, because families are watching this one closely.
The disease is the context that makes this hard
Before the statistics, the stakes. Duchenne muscular dystrophy is a severe, X-linked genetic disorder that primarily affects boys, causing progressive muscle degeneration that damages skeletal, respiratory, and cardiac muscle. It affects roughly 15,000 people in the United States. There is no cure, treatment options are limited, and the deterioration of the heart muscle into cardiomyopathy and heart failure is the leading cause of death.
This context is not incidental. It is the pressure under which every judgment in this story is made. When a disease is fatal and options are few, the human case for approving something, anything, that might help is overwhelming, and the families living with the disease feel that urgency in a way no regulator's statistical standard can capture. That pressure is real and legitimate. It is also exactly the condition under which regulators have historically approved drugs that later turned out not to work, which is why the tension here is genuinely hard rather than a simple case of an agency being obstructive.
How the same data yields opposite verdicts
The contradiction resolves once you understand what changed between the trial's design and its analysis.
A rigorous clinical trial is supposed to specify, in advance and in writing, exactly what it will measure and how it will analyze the results. This document, the statistical analysis plan, exists to prevent a specific and well-documented form of self-deception. If researchers are allowed to decide how to analyze the data after they have seen it, they can, consciously or not, select the analysis that produces the most favorable answer. Locking the plan down beforehand is what makes a positive result trustworthy, because it means the goalposts were planted before anyone knew where the ball would land.
According to the FDA's account, that is not what happened here. The agency says changes were made to the statistical analysis plan after the randomized, double-blind portion of the trial had ended, during the open-label extension, and that those changes modified the endpoint definitions, the analytical methods, and the data-imputation strategy, meaning how missing data were handled. The FDA further states that the final analyses in the application included additional changes without an updated written plan, and that the relevant plan version was never submitted for the agency's review before filing, nor discussed or agreed upon.
That sequence, if accurate, explains the whole contradiction. Under the original prespecified analysis, the FDA says the trial failed. Under the revised analysis, the company reports success. Both can be true statements about the same raw numbers, because they are the product of different analytical choices, and the choices were changed after the data were seen. The company's "statistically significant" and the FDA's "failed" are not a factual dispute about what happened to the patients. They are a dispute about which analysis counts.
Why the FDA treats post-hoc changes as disqualifying rather than minor
To a non-specialist, adjusting an analysis after seeing the data can sound like reasonable refinement. To a statistician, it is close to the original sin of clinical research, and the reason is worth making concrete.
Imagine flipping a coin and only deciding what counts as winning after you see how the flips came out. You could almost always construct a rule under which you won. Changing endpoint definitions and analytical methods after unblinding is the clinical-trial version of that problem. It does not necessarily mean the drug does not work. It means the positive result can no longer be trusted to reflect the drug rather than the analytical choices, because the people making the choices already knew which choices would produce a win.
This is why the FDA's standard, stated in the briefing documents, is that substantial evidence of effectiveness generally requires adequate and well-controlled investigations that are convincing on their own terms. A result that depends on a post-hoc analysis plan is, by this standard, not convincing on its own terms, regardless of whether the underlying drug has a real effect. The problem is epistemic, not moral. Even if deramiocel genuinely helps, a trial analyzed this way cannot prove it, and the agency is not permitted to assume it.
This was already rejected once
The history sharpens the concern rather than softening it. This is not the FDA's first look at this drug.
The original application, built on the earlier Phase 2 HOPE-2 trial and its extension, received a Complete Response Letter in July 2025. The agency concluded the application lacked substantial evidence of effectiveness, and the briefing documents note that the earlier data showed no evidence of effectiveness on skeletal or cardiac function by the relevant measures. HOPE-3 was supposed to be the answer to that rejection, the larger, randomized, placebo-controlled trial that would supply the proof the first submission lacked.
So the stakes of the analysis-plan question are heightened. The trial that was meant to rescue a previously rejected drug is now itself the subject of concerns about whether its positive result was constructed after the fact. That does not mean the drug fails to work. It does mean the second attempt has run into a version of the same problem as the first, which is a difficult position from which to win an advisory committee's confidence.
The genuinely hard part, stated fairly
It would be a mistake to treat this as a simple story of a company massaging data and an agency catching it. The honest version is more uncomfortable, and both sides deserve their strongest case.
The company's position has real substance. Duchenne is fatal, the drug appears safe, the mechanism is biologically plausible, and Capricor argues that its accumulated evidence across multiple studies, including long-term extension data, points consistently in the same direction. In a disease this devastating, there is a serious argument that regulators should weigh the totality of evidence and the desperation of patients rather than applying the strictest statistical standard, and that a safe therapy with suggestive benefit is worth approving when the alternative is nothing. Patient advocates make exactly this argument, and they are not wrong to make it.
The FDA's position also has real substance, and it is not bureaucratic obstruction. The agency's job is to ensure that approved drugs actually work, because approving ineffective ones carries costs that are easy to overlook: families spend money and hope on a therapy that does nothing, patients endure infusions and side effects for no benefit, the real disease goes untreated by something that might have helped, and the incentive to develop genuinely effective drugs erodes when an ineffective one can reach the market on a sympathetic analysis. Compassion that approves things that do not work is not actually compassionate, and the statistical standards exist precisely because good intentions have repeatedly produced bad approvals.
Both of those are true at once. That is what makes Wednesday's meeting genuinely difficult rather than a formality.
What Wednesday actually decides
The advisory committee is not being asked whether Duchenne is terrible or whether these families deserve help. Everyone agrees on both. It is being asked a narrower and harder question: whether the HOPE-3 results, given the changes to the analysis plan, constitute substantial evidence of effectiveness under the legal standard the FDA must apply.
A biotech reporter framed the stakes of the briefing documents precisely before they posted: they would reveal whether the committee was being convened to approve the drug or to explain the rejection a second time. The documents, as posted, lean unmistakably toward the harder path for the company, which is why the stock reacted as sharply as it did.
The committee's vote is advisory, not binding, and the FDA's formal decision date is August 22. But the pattern is not encouraging for approval. A drug rejected once for insufficient evidence has returned with a trial whose positive result the agency attributes to post-hoc analytical changes rather than to a prespecified win. The company will argue the totality of the evidence and the gravity of the disease. The agency will argue that a result you can only reach by changing the rules after seeing the data is not a result you can rely on.
For the families watching, the cruelty of this situation is real and worth naming plainly: the standards that feel like obstacles are the same standards that protect them from spending their limited time and hope on something that does not work. A drug that truly helped would clear them. The tragedy is that when a trial is analyzed this way, no one, not the company, not the FDA, not the families, can actually tell from the data whether this drug is the one that helps. That uncertainty, not a villain, is the hardest thing in the room on Wednesday.
Primary sources
- Investing.com for the roughly 40% single-day share decline, the FDA briefing documents' account of the HOPE-2 basis for the original application showing no evidence of effectiveness, the prior Complete Response Letter for lack of substantial evidence and unfavorable benefit-risk, and the substantial-evidence standard.
- Wall St Engine's summary of the briefing documents stating that HOPE-3 failed its prespecified primary and secondary endpoints with no statistically significant difference from placebo at 12 months, and detailing the post-unblinding changes to endpoint definitions, analytical methods, and data-imputation strategy.
- NeurologyLive for the disease background, the July 2025 CRL rationale, the HOPE-3 design and the company's reported topline results, and CEO Linda Marbán's statements on the totality of evidence and the endpoint dispute.
- Capricor's SEC filings and press releases for the company's characterization that HOPE-3 achieved statistical significance on its primary and key secondary endpoints, the August 22, 2026 PDUFA date, and disease epidemiology.
- STAT biotech reporting for the framing of what the briefing documents would reveal about the purpose of the advisory committee meeting.