Command Palette

Search for a command to run...

Navacaprant and the end of the kappa antagonist thesis in major depression

OtherEfficacySeptember 15th, 2026·6 min read·10.5281/zenodo.20479005

Neumora terminated the 904-participant open-label extension of navacaprant after the drug failed to separate from placebo on MADRS in all three KOASTAL Phase 3 studies. It is the second selective kappa opioid receptor antagonist to fail in Phase 3 for major depressive disorder in eighteen months, on a target whose Open Targets association carries no human genetic evidence.

Mechanism Risk Score

ComponentPoints
Phase-weighted failure burden19.0 / 30
Archetype severity15.8 / 25
Temporal recency7.2 / 15
Genetic evidence deficit6.3 / 15
Programmatic saturation15.0 / 15

For OPRK1 in Major depressive disorder, the Mechanism Risk Score is 63/100 (orange band). The score is a failure-burden index derived from Claidex post-mortems on this target–disease pair, not a probability of approval.

MRS 63/100 (ORANGE). Two Phase 3 programmes against OPRK1 in major depressive disorder are on file, both efficacy failures: aticaprant (VENTURA, March 2025) and navacaprant (KOASTAL, June 2026). Open Targets v26.06 scores the OPRK1 to MDD association at 0.584, composed of clinical evidence 0.936 and literature 0.482 with no genetic_association contribution, so the genetic deficit term carries 6.25 of a possible 15. Saturation is near maximum at 14.99 of 15 because Open Targets records 38 drug and clinical candidates against OPRK1. The MRS is a structured summary of the documented failure record recomputed from the Claidex claims table, not a prediction of future trial outcomes.

This score does not predict whether the next trial will succeed. It flags how heavy the documented mechanistic failure record is before a new program is justified.

Primary figure supporting this claim (Navacaprant / OPRK1 / Major depressive disorder): Navacaprant and the end of the kappa antagonist thesis in major depression

What was tried

Neumora Therapeutics ran NCT06029439, a 52-week open-label extension of navacaprant (NMRA-335140, formerly BTRX-335140 and CYM-53093) at 80 mg once daily in adults with major depressive disorder. It enrolled 904 participants, all rollovers from the KOASTAL Phase 3 studies NMRA-335140-301, -302 and -303. The design was single group and unmasked, and the primary outcome was safety and tolerability over 54 weeks. It ran from 10 November 2023 to 17 August 2026. ClinicalTrials.gov posted the termination on 14 September 2026, reason recorded verbatim as "NMRA-335140 did not achieve the primary endpoint across three randomized controlled trials. Program discontinued due to lack of efficacy in MDD."

Navacaprant is a selective small molecule kappa opioid receptor antagonist, recorded in ChEMBL as CHEMBL4592045 at maximum phase 3. The extension was never the experiment that mattered. It was the safety container around three pivotal trials, and it closed because those trials produced no drug worth dosing.

The biological hypothesis

The dynorphin and kappa opioid receptor system is a well characterised aversion circuit. Kappa agonism produces dysphoria, reduced reward seeking and social withdrawal. The inference that ran through fifteen years of translational psychiatry was the mirror image: blocking the receptor should restore reward processing and so treat anhedonia, the symptom domain monoaminergic antidepressants address least well.

That inference had experimental support. A 2020 fast-fail proof-of-mechanism trial reported that kappa antagonism moved a reward-related neuroimaging readout in the expected direction (Nature Medicine 2020). Navacaprant was characterised as selective and free of partial agonist activity (Neuropharmacology 2024), and its Phase 2 MDD study appeared in 2025 (doi:10.1097/JCP.0000000000001967).

What the hypothesis never had was human genetics. Open Targets (v26.06) scores the OPRK1 to major depressive disorder association at 0.584, decomposing into clinical evidence 0.936 and literature 0.482, with no contribution from the genetic_association datatype. The association exists because people ran trials and wrote papers about the idea, not because human variation at OPRK1 tracks depression risk. The score was measuring the field's enthusiasm back to itself.

What actually happened

The three randomised studies produced no separation from placebo on the Montgomery-Asberg Depression Rating Scale at week 6.

KOASTAL-1 randomised 383 participants (191 navacaprant, 192 placebo) and reported MADRS change from baseline of -12.5 against -12.5, a least squares mean difference of 0.0 with p=0.993. Its anhedonia secondary endpoint, the Snaith-Hamilton Pleasure Scale, gave -5.8 against -5.5, LSMD -0.3, p=0.648. KOASTAL-2 randomised 430 participants (217 and 213): -12.2 against -12.0, LSMD -0.3, p=0.813. KOASTAL-3 randomised 422 (212 and 210): -10.1 against -10.8, LSMD 0.7, p=0.480, a point estimate favouring placebo. A pre-specified post-optimization analysis in 426 participants gave -12.1 against -12.1, LSMD 0.0, p=0.976.

Safety was not the problem. Neumora described navacaprant as safe and generally well tolerated, and KOASTAL-1 reported no serious adverse events, with headache at 6.8 percent against 7.3 percent on placebo and diarrhoea at 5.2 percent against 2.1 percent. The openFDA adverse event database returned no records for navacaprant, as expected for a compound that never reached market. Neumora announced the KOASTAL-2 and -3 results on 15 June 2026 alongside discontinuation of the programme and a cut of roughly 35 percent of its workforce.

Failure mechanism, best guess

This was an efficacy failure and the archetype is clean. Target engagement was not in dispute, tolerability was not in dispute, and the compound was selective. The mechanism did not move the endpoint.

Two features matter. Placebo arms improved by a mean of 11.8 MADRS points across the three studies, computed from the reported values of -12.5, -12.0 and -10.8. A drug chasing an increment on top of that has little room, and nothing in KOASTAL suggests navacaprant found any. Second, the only positive signal in the programme was a sex subgroup in KOASTAL-1, where female participants showed a SHAPS LSMD of -2.3 at p=0.015 against an overall LSMD of -0.3. It sat inside a study whose primary endpoint had already failed, was one comparison among many, and did not replicate.

How to prevent this next time

Endpoint-level data exist for the primary outcome only, so the levers below are qualitative. No posterior, power curve or confidence interval is computed here, because the public record does not carry the variance estimates that would make one honest.

The first lever is base-rate adjustment against evidence type. An Open Targets association carried entirely by clinical and literature evidence, with zero genetic support, should not be priced like one anchored in human variation. OPRK1 and MDD sat at 0.584 with that composition, and three simultaneous Phase 3 studies was full price for a partially supported hypothesis.

The second lever is biomarker enrichment specified before the pivotal trials rather than discovered inside them. The reward-processing readout from the 2020 fast-fail work existed and was never made an enrolment criterion, so KOASTAL recruited major depressive disorder broadly and looked for the anhedonia signal afterwards.

The third lever is sequencing. Running the three studies in parallel meant the KOASTAL-1 null could not inform the other two, so three studies produced one piece of information. The fourth is red-teaming the subgroup, since a lone sex-stratified secondary endpoint at p=0.015 inside a failed trial is exactly what adversarial review exists to price correctly.

The single highest leverage change would have been to make the reward-processing biomarker from the proof-of-mechanism work a pre-specified enrolment gate in a single adequately powered study, and to read that study out before committing the other two.

What this means for similar programs

Navacaprant is the second selective kappa opioid receptor antagonist to fail in Phase 3 for major depressive disorder within eighteen months. Johnson and Johnson stopped the VENTURA programme for aticaprant in March 2025 citing insufficient efficacy in adjunctive MDD, filed in the Claidex graph as aticaprant-oprk1-mdd-anhedonia-ventura-phase3-efficacy-failure. Two independent, selective, well tolerated compounds, two negative Phase 3 readouts, no human genetic anchor. That is a class signal, not a molecule signal.

The Mechanism Risk Score for OPRK1 recomputed to 63 after this insert, moving the target from yellow into orange. Saturation sits near maximum because Open Targets records 38 drug and clinical candidates against OPRK1, ten touching depression, including both antagonists that reached Phase 3 and failed. Any new kappa antagonist in mood disorders inherits the burden of explaining what its predecessors got wrong, and selectivity is not an answer, because both had it.

Open questions

Did navacaprant achieve the receptor occupancy its preclinical work implied at 80 mg once daily, and will occupancy data from KOASTAL be published? Was the KOASTAL-1 female subgroup examined in pooled KOASTAL-2 and -3 data? Does the kappa hypothesis survive in a narrower population, such as anhedonia selected by an objective reward task rather than a self-report scale, or has it now been tested adequately in humans to close the question? The extension enrolled 904 participants over nearly three years with no posted results record, and those safety data should be released.

Sources

Related failure claims

Linked claims sharing target, indication, or failure mechanism.