On 29 July 29, 2026, the same clinical trial data produced opposite verdicts. The FDA's Cellular, Tissue, and Gene Therapies Advisory Committee voted 9-3 that available evidence did not support the effectiveness of Capricor's deramiocel for Duchenne muscular dystrophy (DMD) cardiomyopathy. On the same day, data from the HOPE-3 trial was published in The Lancet reporting a statistically significant primary endpoint with deramiocel slowing upper-limb decline by 54% versus placebo.
The divergence is not about the data itself. It is about which version of the statistical analysis plan (SAP) governs interpretation. This situation raises questions that extend beyond a single therapy. For rare-disease programs operating with small, genetically heterogeneous patient populations, the structural fragility exposed by HOPE-3 carries lessons about trial design, endpoint architecture, and the operational decisions that shape evidence quality long before database lock.
The disagreement centers on which version of the trial's SAP should control interpretation. The FDA's briefing document relied on SAP version 1.1, which was in place when HOPE-3 began. Under SAP v1.1, the primary endpoint (PUL 2.0, measuring upper-limb function) showed a mean difference of 0.66 points, failing to reach statistical significance.
Capricor and The Lancet publication used SAP version 3.0, dated November 24, 2025. Under this version, the same primary endpoint reached significance. The company asserts that SAP v3.0 was finalized before database lock and unblinding, and should control interpretation.
FDA biostatistician Tingting Zhou cited "late changes to the SAP that lacked adequate scientific justification" and "a lack of robustness" in the analysis. Capricor's chief executive officer, Linda Marbán, characterized the FDA's reliance on SAP v1.1 as "Your professor grading your term paper on an early draft you have never even submitted."
This is not the first regulatory challenge for deramiocel. The therapy received a Complete Response Letter in July 2025 based on data from the earlier HOPE-2 trial.
Beyond the SAP dispute, the advisory committee raised concerns about the fragility of the efficacy findings. Panelists characterized the results as "very fragile," reflecting the inherent challenges of trials in rare, genetically heterogeneous diseases.
DMD presents significant cohort heterogeneity. Patients carry thousands of distinct dystrophin mutations, and disease severity is further modulated by genetic factors such as SPP1 and LTBP4. With approximately 106 patients in HOPE-3 spanning variable presentations, the margin between statistical significance and non-significance narrows considerably.
Functional unblinding added another layer of concern. Hypersensitivity reactions occurred in 42% of deramiocel recipients versus 15% of placebo recipients, raising the possibility that patients and investigators could infer treatment assignment.
The endpoint architecture also contributed to complexity. The primary endpoint measured skeletal function, while the sought indication was cardiomyopathy. The key cardiac secondary endpoint (left ventricular ejection fraction) did not reach significance and broke the hierarchical testing chain. This misalignment between endpoint and indication complicates the interpretability of results for regulatory purposes.
Importantly, peer review and regulatory review answer different questions. The Lancet validated the HOPE-3 design and analysis as scientifically sound and publishable. The FDA asks whether the evidence is robust and prespecified enough to support approval.
Both evaluations can be internally consistent and still diverge. Upper-limb data were described by University of Iowa biostatistics professor and advisory committee member Christopher Coffey as "a little more favorable," though he added, "I'm not convinced that [deramiocel] works."
The advisory committee vote is non-binding, and the PDUFA target action date remains August 22, 2026.
The issues surfaced by HOPE-3 are not unique to this program. SAP governance, endpoint-indication alignment, and cohort heterogeneity are recurring challenges across rare disease and precision medicine trials. When small sample sizes meet genetic variability, unexplained variance increases and statistical margins shrink.
The question for sponsors is where leverage exists to address this fragility. Much of the variance in rare-disease trials originates upstream, from imprecise patient characterization, inconsistent eligibility confirmation, and limited long-term follow-up that obscures durable treatment effects.
Sano's platform is built around this operational premise. By integrating genetic testing, biomarker-confirmed eligibility, and long-term participant engagement within a single system, the approach aims to reduce the unexplained variance that makes small cohort signals fragile.
The PDUFA date will determine the regulatory outcome for deramiocel, but the broader questions raised by HOPE-3 will persist.
For sponsors developing therapies in genetically heterogeneous rare diseases, the case illustrates how SAP governance, endpoint selection, and cohort composition interact to shape evidence interpretability. These are design decisions with regulatory consequences. They are also operational decisions that can be influenced by how patients are identified, characterized, and engaged.
Reducing unexplained variance at the cohort level, before database lock, may offer the most durable path to interpretable results.
Get in touch to discuss how upstream patient characterization and long-term engagement can strengthen evidence quality in precision medicine trials.