Deramiocel Year Two: What a Delayed Start Can and Can't Prove

For young men with Duchenne muscular dystrophy who no longer walk, the arms are the last place independence lives. Feeding, steering a wheelchair and using a phone all depend on how far a hand can reach. That is why Capricor Therapeutics rebuilt its approval case for deramiocel, an infused cell therapy, around upper limb function. On October 5, at the World Muscle Society congress in Hiroshima, it presented a second year of data from the HOPE-3 trial. The FDA's decision date is November 22. The extension data are the newest evidence in the file, and they come from a design that deserves careful reading.

The randomized year

HOPE-3, published in The Lancet, randomized 106 participants aged 10 or older to deramiocel (54) or placebo (52), given by intravenous infusion every three months. Participants were non ambulatory or late ambulatory, took stable corticosteroids and had measurable but impaired arm function. The primary endpoint was the percentage change in total Performance of the Upper Limb 2.0 (PUL 2.0) score at 12 months. The deramiocel group did better by 4.55 percentage points (95% CI 0.47 to 8.63, P = 0.029), and the paper reports a safety profile similar to placebo. Capricor funded the trial.

The paper also documents a rocky protocol history. The FDA had previously issued a complete response letter on an application that relied on earlier phase 2 data. After an FDA meeting in August 2025, protocol version 9.0 in September 2025 returned the primary endpoint to the PUL 2.0 score, away from a left ventricular ejection fraction endpoint proposed in version 8.0 that was never implemented. Version 9.0 also changed the analysis from absolute to percentage change. Enrollment ended in May 2024, the statistical analysis plan was finalized on November 23, 2025, and unblinding followed on November 25. The authors state these changes preceded unblinding. They remain late changes, and a reviewer reading a result with a lower bound of 0.47 is entitled to notice them.

What the open label year added

After the 12 month blinded period, 98 participants entered an open label extension, 49 from each original arm, and everyone received deramiocel. Capricor reports these results in score points. In the group that had been on placebo, the average change in PUL 2.0 was minus 2.05 during the placebo year and minus 0.49 in the first year on deramiocel. The difference was 1.56 points (95% CI 0.47 to 2.64), which the company calls a 76% reduction in the rate of decline. In the group that had been on deramiocel from the start, decline was minus 0.95 in year one and minus 0.89 in year two, essentially unchanged.

The company adds a comparison with a natural history model that predicts each patient's decline from their starting score, age and ambulatory status. At month 24, the observed change was minus 3.76 for the group with 12 months on the drug and minus 3.42 for the group with 24 months, against predicted declines of minus 5.35 and minus 5.88. The release notes that the placebo group tracked the model during its untreated year (minus 2.7 observed, minus 2.9 predicted), and that the comparison is descriptive, with no matching and no statistical testing. Month 24 assessments covered 42 participants in one group and 40 in the other.

Why a delayed start is suggestive, not decisive

The delayed start design is clever, and it answers a real objection: if the benefit appeared only because of time passing, patients who started late should not catch up. Seeing the placebo group's decline slow after switching, while the early group held its pace, is the pattern a true effect would produce. It is also a pattern that other forces can produce. Patients who stay in an extension are survivors of the first year, with 49 of 52 placebo patients continuing. Everyone knows they are on active drug, and strength tests that rely on effort can respond to expectation. And disease progression is not perfectly steady: a year of fast loss can be followed by slower loss for reasons unrelated to treatment, which is why the natural history comparison matters and why a model, rather than a matched control group, is a weaker anchor than it sounds.

What the FDA is weighing

The advisory committee voted 9 to 3 that the evidence did not support effectiveness for cardiomyopathy, and in a separate discussion on upper limb function, the committee's feedback was directionally supportive of the HOPE-3 evidence, according to Capricor. On August 24, the FDA extended its review from August 22 to November 22, after Capricor filed what the agency treated as a major amendment: 24 month extension data, robustness analyses and a refined proposed indication focused on upper limb function. The agency accepted it, citing the unmet need in Duchenne. Our fourth quarter FDA calendar lists this decision with the other late 2026 files.

I read the situation this way: one randomized trial with a significant but fragile primary result, a protocol that changed late, an advisory vote against the original indication, and an open label year that is consistent with benefit and cannot establish it. None of that makes the drug ineffective, and the people affected have almost nothing that slows arm decline. It does mean the November decision will rest on judgment about how much uncertainty is acceptable in a disease with few options, not on a clean result. For why the interval's lower bound matters, see our guide to reading a study.