ODYSSEY-HCM: mavacamten cut NT-proBNP by 59 percent and still missed both primary endpoints

What was tried

Bristol Myers Squibb ran ODYSSEY-HCM, a Phase 3 randomized, double-masked, placebo-controlled study of mavacamten in symptomatic non-obstructive hypertrophic cardiomyopathy (NCT05582395). Dosing started at 5 mg once daily with adjustments guided by ejection fraction. The study opened on 14 December 2022, randomized 580 participants, 289 to mavacamten and 291 to placebo, and reached primary completion on 6 March 2025.

Entry required unexplained left ventricular hypertrophy of at least 15 mm, or 13 mm with a family history, on core laboratory reading. The defining criterion was hemodynamic: peak left ventricular outflow tract gradient below 30 mmHg at rest and below 50 mmHg with provocation. Participants were NYHA class II or III, and mimics such as Fabry disease and amyloidosis were excluded. Mean age was 56.4 years on drug and 55.7 on placebo, and 266 of 580 were female.

The registered reason for stopping reads: terminated because of lack of efficacy in the short term acute phase. The sponsor closed the long-term extension early, affecting 251 participants in one arm and 259 in the other.

The biological hypothesis

Mavacamten is a selective allosteric inhibitor of cardiac beta-myosin heavy chain, the product of MYH7. It reduces the fraction of myosin heads available for actin crossbridges, lowering hypercontractility and improving relaxation.

The target-disease link is about as strong as human genetics gets. Open Targets scores MYH7 against hypertrophic cardiomyopathy at 0.907, with genetic association at 0.969, clinical evidence at 0.948, and genetic literature at 0.860. In gnomAD the gene is loss-of-function constrained, observed-to-expected 0.56, interval 0.48 to 0.64. The molecule is equally credentialed. ChEMBL lists mavacamten at maximum phase 4 with first approval in 2022, earned in obstructive disease on EXPLORER-HCM.

The extrapolation is where the risk sat. In obstructive disease the drug relieves a mechanical obstruction that is itself the proximate cause of symptoms. In non-obstructive disease there is no gradient to relieve. The hypothesis was that hypercontractility and impaired relaxation drive symptoms anyway, so reducing contractility should still help. MAVERICK-HCM, the Phase 2 study in the same population, produced biomarker movement without a convincing functional signal. ODYSSEY-HCM tested whether that movement would convert.

What actually happened

It did not convert. Both primary endpoints missed.

Change from baseline in the KCCQ-23 Clinical Summary Score at Week 48 was 13.13 points on mavacamten and 10.38 on placebo, a least squares mean difference of 2.74, 95 percent confidence interval -0.08 to 5.56, p equal to 0.056. Change in peak oxygen consumption was 0.52 mL/kg/min against 0.05, a difference of 0.47, interval -0.03 to 0.98, p equal to 0.066. The trial prespecified 1.4 mL/kg/min as the clinically meaningful change. The observed difference is roughly a third of that, and its upper bound still falls short.

The pharmacology worked. NT-proBNP fell to 0.42 of baseline on mavacamten against 1.02 on placebo, an adjusted geometric mean ratio of 0.41, interval 0.36 to 0.47. That is a 59 percent reduction in a biomarker of myocardial wall stress, with an interval nowhere near the null.

Secondary measures moved slightly. The HCMSQ shortness of breath domain fell 0.70 points more on drug, interval -1.18 to -0.22. NYHA class improved in 36.33 percent against 31.62 percent, a risk difference of 5.01 percent, interval -2.25 to 12.23. Ventilatory efficiency did not separate, VE/VCO2 slope difference -0.81, interval -2.21 to 0.60.

Discontinuation for adverse events during the treatment period was 12 of 289 on mavacamten against 5 of 291 on placebo. The sponsor stated in April 2025 that no new safety signals were observed, and that the result does not change the approved product's profile in obstructive disease.

Failure mechanism, best guess

This is a population mismatch, not a target or molecule failure.

The evidence is the gap between the biomarker and the patient. A 59 percent reduction in NT-proBNP with a tight interval shows the drug reached cardiac myosin and changed ventricular loading. What did not move was how far patients could walk and how they felt.

The likeliest explanation is that in non-obstructive disease the symptom-limiting physiology is not hypercontractility. Candidate drivers include diastolic stiffness from established fibrosis, microvascular dysfunction, chronotropic incompetence, and atrial disease. None responds to lowering crossbridge availability, and some may worsen if stroke volume falls. The KCCQ result also carries a 10.38 point placebo response, compressing the detectable window.

A secondary consideration is that the effect may be real and small. Both primary p values sit just above 0.05 and both estimates favor drug. A trial sized for 1.4 mL/kg/min was never powered to resolve 0.47.

How to prevent this next time

No patient-level data were released, so a formal re-analysis is not possible. Three levers apply.

Biomarker enrichment first. NT-proBNP moved decisively while function did not, so the biomarker measured the drug rather than the process limiting these patients. The trial should have stratified on what limits exercise, for example late gadolinium enhancement burden or myocardial perfusion reserve, and named in advance which stratum could benefit.

Base-rate adjustment second. Genetic support for MYH7 in hypertrophic cardiomyopathy is near the ceiling at 0.907, and it did nothing to protect this trial. Association scores validate the target-disease pair, not the mechanism-population pair. The relevant prior here was the Phase 2 result in the same population, which showed biomarker movement without function.

Red-team third. MAVERICK-HCM was the warning, and the confirmatory trial went ahead on a symptom and exercise co-primary. A pre-mortem asking what would have to be true for non-obstructive patients to feel better without a gradient to relieve would have surfaced diastolic and microvascular physiology as the binding constraint, and argued for a mechanistic Phase 2b first.

The single highest leverage change would have been treating the MAVERICK-HCM biomarker-only result as a negative signal about the population rather than a positive one about the drug, and requiring an enriched, mechanism-matched Phase 2b before Phase 3.

What this means for similar programs

Cardiac myosin inhibition remains validated in obstructive disease. What this result constrains is extension to phenotypes without obstruction. Open Targets lists three clinical-stage or approved molecules against MYH7, mavacamten, danicamtiv, and omecamtiv mecarbil, and the class now has a boundary: benefit tracks the mechanical problem, not the genotype.

Post-marketing data support the same caution. openFDA FAERS returns 5083 reports for mavacamten or Camzyos, 2146 coded serious and 127 reporting a death outcome. The most frequent reaction terms are fatigue at 352, dyspnoea at 344, atrial fibrillation at 325, and ejection fraction decreased at 181. Lowering contractility in hearts that do not need it lowered has a cost.

The Claidex Mechanism Risk Score for MYH7 now stands at 43, yellow band. Genetic deficit contributes only 1.40 of a possible 15, the lowest component, while phase burden contributes 16.69 and saturation 8.26. That profile describes a target nobody should abandon and a population question the field has answered twice.

Open questions

Does a low fibrosis burden subgroup on cardiac magnetic resonance show a peak oxygen consumption benefit, and will the sponsor report it?

Was Week 48 long enough for reverse remodeling to translate into functional capacity, given the extension closed early?

How much of the 10.38 point placebo KCCQ improvement is regression to the mean in a symptom-selected cohort?

Sources

1. • ClinicalTrials.gov, NCT05582395, ODYSSEY-HCM, protocol and posted results. https://clinicaltrials.gov/study/NCT05582395 - Bristol Myers Squibb, Provides Update on Phase 3 ODYSSEY-HCM Trial, April 2025. https://news.bms.com/news/details/2025/Bristol-Myers-Squibb-Provides-Update-on-Phase-3-ODYSSEY-HCM-Trial/default.aspx - Mavacamten in Symptomatic Nonobstructive Hypertrophic Cardiomyopathy. N Engl J Med, 2025. https://- Ho CY et al. Evaluation of Mavacamten in Symptomatic Patients With Nonobstructive Hypertrophic Cardiomyopathy (MAVERICK-HCM). J Am Coll Cardiol, 2020. https://- Olivotto I et al. Mavacamten for treatment of symptomatic obstructive hypertrophic cardiomyopathy (EXPLORER-HCM). Lancet, 2020. https://- Echocardiographic Changes With Mavacamten in Nonobstructive Hypertrophic Cardiomyopathy: Exploratory Insights From the ODYSSEY-HCM Trial. J Am Coll Cardiol, 2025. https://- Mavacamten in Symptomatic Nonobstructive Hypertrophic Cardiomyopathy: Design, Rationale, and Baseline Characteristics of ODYSSEY-HCM. JACC Heart Fail, 2025. https://- Open Targets Platform, MYH7 (ENSG00000092054) and hypertrophic cardiomyopathy (MONDO_0005045): association score, evidence types, tractability, genetic constraint, clinical candidates. https://platform.opentargets.org/target/ENSG00000092054 - ChEMBL, mavacamten, CHEMBL4297517, maximum phase 4, first approval 2022. https://www.ebi.ac.uk/chembl/compound_report_card/CHEMBL4297517/ - openFDA FAERS drug event endpoint, queries for Camzyos and mavacamten, accessed 19 September 2026. https://api.fda.gov/drug/event.json.