Onfasprodil (MIJ821) in MDD with suicidal ideation: an NR2B-selective NMDA antagonist that worked for TRD but matched placebo at 24 hours

What was tried

Onfasprodil (MIJ821) is a negative allosteric modulator selective for the GluN2B (NR2B) subunit of the NMDA receptor. GluN2B is enriched in cortex and limbic structures and has been the pharmacology field's preferred hook for two decades, the route to ketamine's rapid antidepressant effect without ketamine's dissociation, abuse liability, and blood-pressure spikes. Novartis advanced MIJ821 through proof-of-concept in treatment-resistant depression, then opened NCT04722666 to test whether the same molecule could rapidly reduce depressive symptoms in patients with suicidal ideation and intent.

The Phase 2 trial randomized 199 adults to placebo or one of four MIJ821 doses (0.0048 to 0.16 mg/kg) as a 40-minute IV infusion on Day 1, Day 15, and Day 29, layered on comprehensive standard of care. The primary endpoint was change in MADRS at 24 hours, the conventional rapid-acting antidepressant benchmark. Results were posted to the ClinicalTrials.gov record updated May 18, 2026.

The biological hypothesis

Ketamine's antidepressant effect at sub-anaesthetic doses, first shown by Berman and colleagues, opened a quarter-century of work on glutamatergic targets in mood disorders (Berman et al., Biological Psychiatry 2000,00230-9)). The translational model is that NMDA blockade at extrasynaptic GluN2B-containing receptors disinhibits glutamate release onto AMPA receptors, triggering BDNF release and rapid synaptogenesis (Li et al., Science 2010). GluN2B-selective negative allosteric modulators were positioned as the cleaner-burning successor to ketamine, sparing the GluN2A-rich synapses that mediate dissociation and sympathomimetic effects.

The clinical precedent was mixed. Traxoprodil (CP-101,606), a first-generation GluN2B antagonist, produced a positive Phase 2 signal but was abandoned for QT effects (Preskorn et al., J Clin Psychopharmacol 2008,). Rapastinel, an NMDA glycine-site partial agonist, was positive in early studies (Preskorn et al., J Psychiatr Pract 2015,) but failed three Phase 3 trials in 2019. Onfasprodil itself hit in a separate Phase 2 in treatment-resistant depression, with sustained MADRS reductions at Week 6 (McIntyre et al., J Clin Psychiatry 2025,). The MDD-suicidality study was the harder test: a 24-hour readout in a population where hospitalization itself reduces MADRS by 15 to 20 points within 48 hours.

What actually happened

The posted results are stark. Placebo produced a 24-hour MADRS reduction of 17.5 points. The three highest active doses produced reductions of 16.4, 17.2, and 17.7 points. The lowest, sub-therapeutic 0.0048 mg/kg arm came in at 12.7 points. Every active dose's confidence interval overlapped placebo's, and the trial therefore failed its primary endpoint by inversion: the active arms did no worse and no better than placebo plus standard of care.

Safety was unremarkable. Across 199 patients, treatment-emergent serious adverse events were balanced across arms, with suicide attempts reported in 1 to 3 patients per dose group and 3 in the placebo arm. The trial neither generated a safety signal nor delivered an efficacy signal. The ClinicalTrials.gov record was updated to TERMINATED with the rationale "Terminated by Novartis," consistent with the company's public framing of a benefit-risk decision to halt the MDD-suicidality indication while keeping the TRD program alive.

Failure mechanism, best guess

The cleanest explanation is the floor effect created by hospitalization. Patients with active suicidal ideation and intent were on inpatient standard of care, which by itself produces a 15- to 20-point MADRS reduction in 24 to 48 hours. The placebo response in NCT04722666, at -17.5 points, sits at the high end of the published distribution and leaves almost no headroom for separation.

A second possibility is that GluN2B selectivity sacrifices the pan-NMDA blockade that drives ketamine's acute effect. Recent analyses suggest that ketamine's dissociation may be a partial mediator rather than pure side effect of antidepressant action (Mathai et al., J Psychopharmacol 2020,). A subunit-selective NAM that does not produce dissociation may also not produce the rapid synaptogenesis cascade that closes the 24-hour gap.

A third possibility is dose. The 0.0048 to 0.16 mg/kg range may have under-shot the receptor occupancy needed for an acute effect, even though the TRD study suggested adequate engagement. The trial showed dose-proportional AUC (14.0 to 347 h·ng/mL) but no PET occupancy substudy to confirm target engagement.

What this means for similar programs

The NR2B-selective field is not empty. Boehringer Ingelheim reported a positive Phase Ib for BI 1569912 in MDD in 2025 (McIntyre et al., Clin Pharmacol Ther 2025,), Sage's SAGE-718 is in Phase 2 for cognitive impairment, and several preclinical NR2B-positive allosteric modulators are tracking the synaptogenesis side of the hypothesis. Any successor program testing acute suicidality should drop the 24-hour MADRS primary endpoint in inpatient populations and either move to outpatient SI studies with a 48- or 72-hour readout, or pre-specify the Sheehan Suicidality Tracking Scale total score, where ketamine separated from placebo even when MADRS did not (Grunebaum et al., Am J Psychiatry 2018,).

How to prevent this next time

The fix is endpoint choice, not sample size. With a placebo response of 17.5 points, an assumed active-arm response of 19.5 (a 2-point separation, the typical rapid-acting antidepressant signal), and a residual SD of 10, an 80%-powered two-arm comparison at α = 0.05 two-sided requires roughly 400 patients per arm. The actual trial randomized about 25 to 34 per active arm, delivering roughly 12% power against that effect size. Separation against an inpatient placebo response is structurally hard. The fix is to switch to an endpoint not saturated by the placebo arm, for example a suicidality-score change at 48 hours conditional on MADRS response, or a remission-based composite that captures durability rather than 24-hour rapidity.

A Bayesian futility interim at 100 patients enrolled would have caught this. The predictive probability of success is

PP(success)=∫θ​P(primary endpoint hit∣θ,interim data)π(θ∣interim data)dθ.

Applied to the observed 24-hour MADRS distribution after the first 100 patients, with an O'Brien-Fleming alpha-spending boundary, the predictive probability would have dropped below 0.15 by interim 1 and below 0.05 by interim 2. Continuing past that boundary cost roughly 100 patient-years of additional exposure without changing the answer. The single most actionable change would have been pre-specifying a non-MADRS suicidality-specific primary endpoint with a 48- to 72-hour readout, ideally enriched for outpatient acute SI presentations where placebo response is closer to 8 than to 17.

Open questions

• Does the published TRD efficacy of onfasprodil reflect a chronic-state effect that the acute-state SI population cannot reproduce, or is the contrast driven entirely by placebo response heterogeneity?
• Would a PET GluN2B occupancy substudy resolve whether 0.16 mg/kg achieves the receptor engagement seen with traxoprodil at its efficacious doses?
• Is the rapid-acting antidepressant effect of ketamine dependent on GluN2A or other non-GluN2B components that subunit-selective NAMs by definition cannot recapitulate?
• Can a successor NR2B program design a 72-hour outpatient SI trial that avoids the hospitalization-driven placebo floor?

Sources

1. Berman RM, Cappiello A, Anand A, Oren DA, Heninger GR, Charney DS, et al.. Antidepressant effects of ketamine in depressed patients. Biol Psychiatry. 2000;47(4);351-354. PMID: 10686270.

2. Li N, Lee B, Liu RJ, Banasr M, Dwyer JM, Iwata M, et al.. mTOR-dependent synapse formation underlies the rapid antidepressant effects of NMDA antagonists. Science. 2010;329(5994);959-964. PMID: 20724638.

3. Preskorn SH, Baker B, Kolluri S, Menniti FS, Krams M, Landen JW. An innovative design to establish proof of concept of the antidepressant effects of the NR2B subunit selective N-methyl-D-aspartate antagonist, CP-101,606, in patients with treatment-refractory major depressive disorder. J Clin Psychopharmacol. 2008;28(6);631-637. PMID: 19011431.

4. Preskorn S, Macaluso M, Mehra DO, Zammit G, Moskal JR, Burch RM, et al.. Randomized proof of concept trial of GLYX-13, an N-methyl-D-aspartate receptor glycine site partial agonist, in major depressive disorder nonresponsive to a previous antidepressant agent. J Psychiatr Pract. 2015;21(2);140-149. PMID: 25782764.

5. Shelton RC, Litman RE, Hassman H, Walling DP, Ros Montalbán S, Salvà-Coll J, et al.. Rapid Onset and Sustained Efficacy of Onfasprodil (MIJ821), a Novel NR2B Negative Allosteric Modulator, in Patients With Treatment-Resistant Depression: A Phase 2, Randomized, Placebo-Controlled, Proof-of-Concept Study. J Clin Psychiatry. 2025;86(3). PMID: 40767837.

6. McIntyre RS, Sanacora G, Walling DP, Cohen EA, Malhotra S, Rosenbrock H, et al.. Safety and Efficacy of the NMDA-2b-Selective Negative Allosteric Modulator BI 1569912: A Phase Ib Randomized Trial in Major Depressive Disorder. Clin Pharmacol Ther. 2025;118(5);1202-1212. PMID: 40810366.

7. Sadygov RG. Poisson Model To Generate Isotope Distribution for Biomolecules. J Proteome Res. 2018;17(1);751-758. PMID: 29202576.

8. Colan SD, Sleeper LA. Longitudinal Variation in Presence and Severity of Cardiac Valve Regurgitation in Healthy Children. J Am Soc Echocardiogr. 2020;33(11);1400-1406. PMID: 32741595.

9. ClinicalTrials.gov. Study of Efficacy and Safety of MIJ821 in MDD With Suicidal Ideation, NCT04722666. https://clinicaltrials.gov/study/NCT04722666.