On July 30, 2026, a boy with Duchenne muscular dystrophy was not represented in the room. His cardiomyopathy was. Capricor Therapeutics had brought its allogeneic cardiosphere-derived cell therapy, deramiocel (CAP-1002), before the FDA’s Cellular, Tissue and Gene Therapies Advisory Committee (CTGTAC) with a pivotal Phase 3 dataset, a prior Complete Response Letter already on the books, and a disease population with no cardiac-specific approved option. The committee voted 9 against, 3 for. No abstentions. That margin does not happen when a panel is merely uncertain. That margin happens when the evidence architecture has a structural problem the sponsor failed to fix before walking into the room.

What does a 9-3 vote against a therapy that met its primary endpoint tell us about how sponsors are managing statistical rigor on accelerated-approval pathways? More than Capricor would like.

A CRL That Should Have Changed Everything

Start the timeline at July 9, 2025. The FDA issued a Complete Response Letter for Capricor’s Biologics License Application for deramiocel. The agency’s stated objection: the BLA did not meet the statutory requirement for substantial evidence of effectiveness, and additional clinical data were needed. Chemistry, Manufacturing, and Controls deficiencies were also cited, according to Parent Project Muscular Dystrophy. That CRL was a documented, formal signal that FDA reviewers were not satisfied with what the sponsor had submitted. Sponsors who receive a CRL and then proceed to an AdCom within twelve months face a narrow window to rebuild credibility with reviewers. The analytical package must be airtight. Every statistical decision must be pre-specified, documented, and defensible in real time. Capricor appears to have walked into the July 30 CTGTAC meeting with at least one significant unresolved liability in that package.

The HOPE-3 Phase 3 study, which Capricor ran for deramiocel in non-ambulatory Duchenne patients, did achieve statistical significance on its primary endpoint: the Performance of Upper Limb measure (PUL v2.0), with a p-value of 0.03, according to Nippon Shinyaku’s investor disclosure. The key secondary cardiac endpoint, Left Ventricular Ejection Fraction, also met pre-specified thresholds. On paper, that is a study that met its endpoints. In the committee room, it was not enough, because the dispute that consumed the meeting was not about whether PUL moved. It was about how the analysis was constructed and whether the endpoints the sponsor emphasized reflected pre-specified hypotheses or post-hoc repositioning.

That distinction is fatal under accelerated approval standards.

When Post-Hoc Becomes the Story

According to Pharmaceutical Technology, the CTGTAC vote against deramiocel centered explicitly on disputes over statistical analysis and post-hoc endpoints. The committee’s concern was not that the therapy showed no signal. The concern was that the analytical frame Capricor used to present effectiveness was not the analytical frame pre-specified before data unblinding. That is not a semantic problem. Under 21 CFR Part 601 and the FDA’s own accelerated approval framework as revised by the Consolidated Appropriations Act of 2023, which increased FDA authority to require confirmatory evidence and tightened the standard for “reasonably likely to predict” clinical benefit, the agency’s tolerance for analytical flexibility has narrowed considerably.

When a sponsor shifts emphasis toward endpoints or subgroup results after seeing data, even partially, panels interpret that as a signal that the pre-specified primary analysis did not tell the story the sponsor needed. The committee’s 9-3 vote is consistent with a panel that believed it was being asked to endorse a conclusion built on a foundation that shifted after unblinding. Whether Capricor actually conducted post-hoc analysis in a way that violated its Statistical Analysis Plan, or whether the presentation of the data merely created that impression, is a distinction the company now has to litigate in its response to the FDA. But the AdCom record is the operational record, and that record shows nine panelists were not persuaded.

The BioSpace account of the meeting describes it as “chaotic,” which is a word that should appear nowhere in a sponsor’s post-meeting debrief as a surprise. An AdCom meeting devolves into statistical confusion when the sponsor’s briefing document and the FDA’s review memorandum are not aligned on the analytical hierarchy. That misalignment begins months earlier, in Type B and Type C pre-submission meetings, where sponsors are supposed to lock down exactly which endpoints count, in which order, under which inferential framework. If panelists were confused about what the primary analysis was, the confusion was planted in the IND record long before July 30.

This is not Capricor’s first time in this specific bind. The July 2025 CRL already flagged effectiveness as the core deficiency. A sponsor that receives that signal and does not rebuild its analytical package from the Statistical Analysis Plan outward is not preparing for an AdCom. It is hoping the committee will give it credit for effort.

The Structural Failure Behind the Vote

The Capricor outcome sits inside a broader enforcement pattern the FDA has been tightening for two years. The DIA Global Forum’s May 2025 analysis of accelerated approval in gene therapy identifies a recurring failure mode: sponsors treat the accelerated pathway as a shortcut to approval rather than as a conditional commitment requiring confirmatory infrastructure. The Consolidated Appropriations Act of 2023 gave FDA expanded authority to withdraw accelerated approvals more efficiently when confirmatory trials fail to materialize or when the original effectiveness evidence weakens under scrutiny. Capricor’s situation is the front end of that problem: the original evidence did not survive the scrutiny of a single AdCom meeting.

The structural read here goes beyond one company. Cell and gene therapy sponsors entering rare disease AdComs are routinely presenting data from trials with small sample sizes, surrogate endpoints, and patient populations where randomized controlled trials are operationally difficult to run. HOPE-3 enrolled a limited number of non-ambulatory Duchenne patients in a disease where every patient lost to follow-up materially shifts the statistical picture. Under those conditions, the Statistical Analysis Plan is not a regulatory formality. It is the only instrument that protects a sponsor from the exact accusation Capricor faced on July 30: that the analysis was shaped by the outcome rather than the outcome measured by the analysis. A pre-specified SAP, locked before unblinding, with an inferential hierarchy documented in a Type C meeting with FDA, is the evidentiary armor that stops a nine-vote rejection.

Capricor did not appear to have that armor fully in place. And the FDA’s pre-submission process, which allows sponsors to request clarity on analytical expectations at every phase, did not prevent this outcome, either. The agency’s Type C meeting framework explicitly permits sponsors to seek agreement on statistical methodology before pivotal readouts. If that alignment happened here, the committee’s confusion about post-hoc endpoints is a failure of documentation. If it did not happen, the failure is operational negligence at the protocol-design stage.

For sponsors running rare disease cell therapy programs right now, the Capricor AdCom record demands a specific operational response. Before the next PDUFA date or AdCom scheduling notice, pull your Statistical Analysis Plan and map every endpoint you intend to present to a numbered hypothesis in that document. If you cannot draw a straight line from your planned presentation to a pre-specified, pre-unblinding commitment, you are building the same vulnerability Capricor walked into on July 30. Beyond that, any sponsor whose analytical hierarchy might invite a post-hoc accusation needs to request a Type C meeting with FDA to lock the inferential framework in writing, before a single interim result reaches the sponsor’s statisticians. The FDA’s Type C meeting mechanism exists precisely for this alignment, and using it is not a sign of weakness. Failing to use it, and then losing a 9-3 AdCom vote, is.

For FDA-watchers, the question the CTGTAC vote leaves open is whether the agency will now use this outcome to publish clearer statistical methodology standards for rare disease cell therapy BLAs, or whether sponsors will continue navigating this through informal pre-submission interactions with no binding record. The AJMC’s account of the July 30 meeting makes clear that the statistical dispute was substantive enough to drive nine of twelve panelists to a negative vote. That level of committee consensus on a methodological objection is a signal the agency’s review division should respond to with guidance, not silence.

Capricor now faces the task of rebuilding a BLA package for a disease community that has been waiting years for a cardiac option, with a committee record that documents statistical confusion as the reason for rejection. Watch for the company’s formal response to the AdCom outcome and whether FDA issues a follow-up guidance document on SAP standards for rare disease cell therapies in Q4 2026. That guidance, if it comes, will tell you whether the agency learned the same lesson the committee just taught Capricor, or whether the next sponsor will have to learn it the same way.

References

  1. FierceBiotech — “FDA adcomm shoots down Capricor’s Duchenne cell therapy after meeting marked by statistical dispute”
  2. Parent Project Muscular Dystrophy — “Capricor Therapeutics Receives Complete Response Letter Regarding Deramiocel (CAP-1002)”
  3. Nippon Shinyaku — HOPE-3 Phase 3 study primary and key secondary endpoint results for deramiocel
  4. Pharmaceutical Technology — “FDA AdCom Shuns Capricor’s DMD Cell Therapy Amid Data Dispute”
  5. BioSpace — “FDA Advisers Vote Against Approval of Capricor’s DMD Therapy in Chaotic AdComm Meeting”
  6. DIA Global Forum — “Accelerated Approval as the New Norm in Gene Therapy for Rare Diseases” (May 2025)
  7. AJMC — “FDA Advisory Panel Votes Against Approval of Deramiocel for DMD”
  8. Facet Life Sciences — “Type C FDA Meetings” (March 2025)
Website |  + posts

Moe Alsumidaie is Chief Editor of The Clinical Trial Vanguard. Moe holds decades of experience in the clinical trials industry. Moe also serves as Head of Research at CliniBiz and Chief Data Scientist at Annex Clinical Corporation.