The FDA Rejected Capricor's Duchenne Therapy and Called the Trial Data Unreliable
The FDA went beyond a routine rejection, publishing a statement that cites serious data integrity violations in Capricor's pivotal Duchenne trial. The company says it stands behind its data.
The FDA issued a complete response letter to Capricor Therapeutics for deramiocel, its allogeneic cardiosphere-derived cell therapy for Duchenne muscular dystrophy, on Nov. 20. A second rejection would have been notable. What made the night extraordinary was the rest: the agency published its own statement, and it went further than a finding that the evidence fell short. It said the correspondence "identifies serious data integrity violations in the conduct of Study HOPE-3 that cannot be remediated post-hoc."
The FDA said its Bioresearch Monitoring program inspected Capricor and found unauthorized post hoc changes to the statistical analysis plan, violations of study blinding procedures, and additional protocol violations. It also issued a Regulatory Responsibility Letter dated Nov. 20. The message was blunt: "flexibility in regulatory approach cannot come at the cost of the foundational requirement that the data underlying an approval decision be reliable."
Capricor disputed that reading. Chief executive Linda Marban said deramiocel "has been studied in two randomized, double-blind, placebo-controlled trials, HOPE-2 and HOPE-3, both of which met their primary endpoints," and that the company stands "behind the rigor of our clinical program and the integrity of our data." The shares fell nearly 80% in after-hours trading to $1.63 from a Friday close of $7.59.
What the trial actually showed
HOPE-3 randomized 106 patients, 54 to deramiocel and 52 to placebo, aged 10 or older, with quarterly infusions to 12 months. The trial's published primary result was a least-squares mean difference in total Performance of the Upper Limb 2.0 percentage change from baseline of 4.55% at 12 months, 95% CI 0.47 to 8.63, p=0.029. Safety was similar to placebo, and the Capricor-funded study ran in The Lancet in August.
It is not what the agency's benchmark produced. Under the pre-specified plan in the FDA's briefing document, SAP 1.1 dated July 19, 2022, mean change in PUL 2.0 at month 12 was -1.73 for deramiocel and -2.39 for placebo, a difference of 0.66 points, 95% CI -0.45 to 1.77, p=0.24. Re-analyzed with SAP 3.0's model but pre-specified missing-data handling, the difference was 2.53%, p=0.21.
The disagreement concentrates on two patients. Only four of 106 missed month 12, but the FDA said handling of two placebo subjects had an outsized influence: one withdrew after an adverse event and had the worst observed value imputed, and one took a prohibited medication and had months 6 to 12 dropped. Both adjustments favored deramiocel, while missing data from three treated patients were assumed to match completers.
Blinding, and the HOPE-2 precedent
Hypersensitivity reactions occurred in 22 of 53 deramiocel patients, 41.5%, versus 8 of 52 on placebo, 15.4%, and 11 placebo patients without such reactions in the blinded period developed them once treatment became open label. The FDA flagged possible functional unblinding. The approved blinding plan called for the blinded vendor, Pentara, to write the statistical analysis plan, but SAP 2.0 and 3.0 were written by the applicant; SAP 3.0, dated Nov. 24, 2025, was never submitted and was finalized a day before database lock.
HOPE-2 shadows the claim that both trials met their endpoints. The FDA says it was designed for 84 patients but enrollment stopped after 20, 16 of whom completed, and the data were unblinded to the applicant at that point. Under the model pre-specified before unblinding, it showed no significant effect on the primary endpoint or any secondary endpoint. Cardiac data follow the pattern: left ventricular ejection fraction changed by -0.041 percentage points, p=0.97, pre-specified, while Capricor reported a ranked ANCOVA difference of 11.65, p=0.041, an analysis the FDA said excluded nearly 22% of randomized patients.
Capricor's case, and what remains open
Marban contested that at the July 29 advisory committee. She argued SAP 3.0 was finalized before unblinding while the FDA used SAP 1.1, which she called an unfinished draft, and that Capricor submitted SAP 2.0 without FDA feedback. FDA biostatistician Tingting Zhou said the agency used SAP 1.1 because it was the plan in place when the study began, and cited late changes without adequate scientific justification. More than two dozen trial investigators and physicians signed a letter urging approval. The committee still voted 9 to 3 against. The Muscular Dystrophy Association called the decision a blow and said it would seek to learn more about the integrity findings.
Capricor reported about $198.5 million in cash and securities at Sept. 30, which it says covers at least 12 months, and is reviewing costs after cutting about 10% of staff in September. This is its second rejection in two years; the first, in July 2025, also cited manufacturing deficiencies. Deramiocel holds Orphan Drug, RMAT and Rare Pediatric Disease designations, and the company has begun regulatory engagement in Europe and Japan. The FDA statement came weeks after it launched a Rare Disease Framework Initiative promising fit-for-purpose review, and the agency's briefing document notes that no therapy is approved for DMD cardiomyopathy. The FDA has now said in public that flexibility has a floor. Whether a result this sensitive to two patients in a 106-patient trial can clear it is the question Capricor has to answer next.