Esmethadone (REL-1017) in major depressive disorder: a low-affinity NMDA blocker whose Phase 3 program stopped at interim re-estimation

What was tried

REL-1017, also called esmethadone or dextromethadone, is the S-enantiomer of methadone and an orally dosed small molecule (ChEMBL CHEMBL350719, small molecule, maximum reported phase 3). Its sponsor tested it as an adjunctive treatment for major depressive disorder in a confirmatory Phase 3 program. NCT06011577 (study REL-1017-304) randomized participants who were already taking a standard antidepressant to REL-1017 or placebo, with change from baseline to Day 28 in the Montgomery-Asberg Depression Rating Scale (MADRS) as the primary endpoint (ClinicalTrials.gov, NCT06011577). The companion study REL-1017-302 (NCT04855747) used the same 28-day MADRS design and planned to enroll 280 participants (ClinicalTrials.gov, NCT04855747). Both studies were terminated by the sponsor. Study 304 stopped after enrolling 27 participants. Study 302 stopped at roughly 75 percent of its planned enrollment, at 236 participants, following a prespecified interim analysis for sample size re-estimation (ClinicalTrials.gov, NCT06011577 and NCT04855747).

The biological hypothesis

Esmethadone is a low-potency uncompetitive antagonist of the N-methyl-D-aspartate (NMDA) glutamate receptor, the same receptor class that ketamine and esketamine act on (Bettini et al., 2022). The obligatory GluN1 subunit of that receptor is encoded by GRIN1 (Ensembl ENSG00000176884). The rationale followed the glutamatergic model of depression, in which brief NMDA receptor blockade is thought to trigger downstream AMPA receptor activation and synaptogenesis and to produce antidepressant effects faster than monoaminergic drugs. Open Targets scores the GRIN1 to major depressive disorder association at 0.59 overall, and that number is carried almost entirely by clinical evidence at 0.96 rather than by human genetics, where the genetic association score is 0.00 (Open Targets Platform). The design bet was that esmethadone's weaker channel block would preserve the antidepressant signal while avoiding the dissociation and abuse liability that constrain ketamine.

What actually happened

Study 304 was terminated after 27 participants and study 302 after 236, both at the prespecified interim sample size re-estimation, a checkpoint used when a sponsor is uncertain whether the planned sample can detect the target effect (ClinicalTrials.gov, NCT06011577 and NCT04855747). The primary endpoint in both was the 28-day MADRS change from baseline (ClinicalTrials.gov). A long-term open-label analysis reported that esmethadone was generally well tolerated across extended dosing (Fava et al., 2025). Pharmacokinetic and drug-drug interaction studies described a predictable oral profile with CYP-mediated metabolism (Ferri et al., 2024). Post-marketing adverse event reports for esmethadone remain sparse because the drug was investigational, with isolated reports of dizziness, panic attack, and somnolence in the FDA FAERS database and too few cases for disproportionality analysis (openFDA FAERS). The pattern is a compound with a clean tolerability record whose adjunctive antidepressant effect was not large enough, or not separable enough from placebo, to justify continuing the confirmatory studies.

Failure mechanism, best guess

Two mechanisms fit the record. The first is effect size against placebo. Adjunctive antidepressant trials carry a high placebo response, and a low-affinity NMDA channel blocker may produce a true drug effect that is real but small relative to that placebo movement. The interim sample size re-estimation is the tell, because that checkpoint is triggered when the observed treatment effect runs smaller than the design assumed (ClinicalTrials.gov, NCT06011577). The second is the depth of target engagement. Ketamine's antidepressant action is tied to a degree of NMDA blockade that also produces dissociation, and esmethadone was engineered to avoid that, so the same property that improves its safety may cap its efficacy. The Open Targets profile reinforces the concern, because the GRIN1 to MDD association rests on clinical evidence at 0.96 with zero human genetic support (Open Targets Platform), which means the target is validated by drugs that worked rather than by genetics that anchor the mechanism to the disease.

How to prevent this next time

Two quantitative tools would have sharpened the continue-or-stop decision. The first is a Bayesian predictive probability of success computed at each interim, defined as:

PP(success)=∫θ​P(final success∣θ,interim data)⋅π(θ∣interim data)dθ

Under this framework the program continues only while the predictive probability of a positive confirmatory readout stays above a preset threshold, which converts the sample size re-estimation trigger into an explicit stopping rule rather than a sponsor judgment call. The second is an honest power calculation anchored to a realistic adjunctive effect. For a true adjunctive MADRS separation near 2 points against a placebo standard deviation near 8, the standardized effect is about 0.25, and detecting it at 90 percent power with two-sided alpha of 0.05 requires roughly 336 participants per arm, far more than the roughly 140 per arm implied by the 280-participant plan (ClinicalTrials.gov, NCT04855747). A placebo-run-in enrichment design would further cut placebo-response variance before any confirmatory commitment. The single highest leverage change would have been to size a placebo-run-in-enriched Phase 2b that estimated the adjunctive effect with tight confidence intervals before opening parallel Phase 3 studies.

What this means for similar programs

NMDA receptor modulation in depression now shows a split record. Ketamine and esketamine work with meaningful dissociation, while lower-affinity and subunit-selective approaches have repeatedly failed to separate from placebo, including the GluN2B-selective onfasprodil in MDD with suicidal ideation already documented in the Claidex graph. The Mechanism Risk Score for GRIN1 stands at 38 of 100 (yellow band), driven by the phase-weighted burden of a terminated Phase 3 and the efficacy-failure archetype, with the NMDA receptor already carrying 31 clinical-stage molecules in Open Targets as broader landscape context. New glutamatergic antidepressant programs should treat the safety-versus-efficacy tradeoff as the central design question rather than a secondary one.

Open questions

• Did the terminated studies show any signal in prespecified subgroups such as higher baseline severity, and will the sponsor release the interim MADRS estimates?
• Is there a tolerated dose above the tested range where esmethadone reaches ketamine-like efficacy without dissociation, or is that window empty?
• Would a monotherapy design rather than an adjunctive one have produced a larger separation by removing the background antidepressant response?

Sources

1. • ClinicalTrials.gov, NCT06011577 (REL-1017-304): https://clinicaltrials.gov/study/NCT06011577 - ClinicalTrials.gov, NCT04855747 (REL-1017-302): https://clinicaltrials.gov/study/NCT04855747 - Open Targets Platform, GRIN1 (ENSG00000176884): https://platform.opentargets.org/target/ENSG00000176884 - ChEMBL, esmethadone (CHEMBL350719): https://www.ebi.ac.uk/chembl/compound_report_card/CHEMBL350719/ - Bettini E et al., 2022, Pharmaceuticals (Basel, Switzerland).- Ferri N et al., 2024, Drugs in R&D.- Ferri N et al., 2024, Drugs in R&D.- Fava M et al., 2025, The Journal of clinical psychiatry.- openFDA FAERS drug event API: https://open.fda.gov/apis/drug/event/.