NBI-921352 in SCN8A-DEE: the only Nav1.6-selective clinical asset gets terminated in extension at n=8

What was tried

NBI-921352 is a small-molecule inhibitor of the voltage-gated sodium channel Nav1.6, encoded by SCN8A, originally developed at Xenon Pharmaceuticals as XEN901 and in-licensed by Neurocrine. NCT05226780 was the long-term open-label active extension of the Phase 2 SCN8A-DEE program (DEE2012). Eligible participants were children and adults who had completed 16 weeks of treatment in the parent trial. The extension allowed up to 164 weeks of continued NBI-921352 as adjunctive therapy in SCN8A-DEE, with a safety-and-tolerability primary endpoint focused on serious treatment-emergent adverse events, TEAEs leading to discontinuation, and fatal TEAEs. The actual enrollment was 8 participants. The primary completion date is recorded as 2025-11-12 with the study marked TERMINATED, "prematurely terminated due to sponsor decision."

The biological hypothesis

SCN8A encodes the alpha-1 subunit of Nav1.6, the dominant axonal sodium channel of mature cortical pyramidal neurons. Pathogenic missense variants in SCN8A produce a developmental and epileptic encephalopathy (DEE13) dominated by gain-of-function effects: increased persistent current, slowed inactivation, and elevated neuronal excitability. The Open Targets evidence channel profile for SCN8A x MONDO_0013801 is near-maximal: genetic_association 0.977, genetic_literature 0.836, animal_model 0.379, literature 0.077, overall 0.844. This is one of the better-validated single-gene targets in epilepsy [Open Targets v25].

The therapeutic rationale was sharper than usual. Existing antiepileptics that hit sodium channels (phenytoin, oxcarbazepine, lacosamide, cenobamate) have been shown clinically to reduce seizure burden in SCN8A-DEE more effectively than other classes, including a recent prospective cenobamate series in Epilepsia 2025 [doi:10.1111/epi.18257]. But all of those drugs are pan-sodium channel blockers and carry the predictable off-target burden of Nav1.1 inhibition (sedation, ataxia, and a contraindication in Dravet syndrome). Bagal and colleagues described NBI-921352 in 2022 as the first Nav1.6-selective molecule to enter the clinic, with selectivity over Nav1.1 sufficient to prevent seizures in Scn8a gain-of-function mice without the neurological liabilities of pan-blockers [Bagal et al. 2022, eLife, doi:10.7554/eLife.72468]. The Open Targets drug landscape against SCN8A lists 75 distinct molecules, but exactly one of them belongs to the Nav1.6-selective class. NBI-921352 was structurally and mechanistically alone.

What actually happened

The DEE2012 parent trial completed earlier and supported continued open-label dosing. The extension opened, enrolled eight participants, and ran for roughly two years before termination. No safety signal was disclosed in the why-stopped field. There is no FAERS file because the molecule is investigational [openFDA query 2026-05-30]. The closest readout in the public domain is the development timing: Xenon Pharmaceuticals' separate Nav1.6 inhibitor program (XEN901 originally, distinct chemistry from later assets) was returned and Neurocrine's external communications through 2024 and 2025 progressively de-emphasized SCN8A as a near-term commercial priority while elevating its CRF1 antagonist (crinecerfont) and broader neuropsychiatric portfolio. A terminated extension at n=8 is the typical end-state of a Phase 2 program that did not produce a placebo-adjusted seizure-reduction effect large enough to underwrite Phase 3 or the operational cost of multi-year long-term dosing in pediatric rare disease.

Failure mechanism, best guess

This is sponsor_decision with a thin layer of efficacy_failure underneath. Three pieces of evidence support that read. The parent Phase 2 has not been published in any peer-reviewed venue or presented at the major epilepsy congresses with a positive seizure-reduction headline. The selective Nav1.6 mechanism in mouse SCN8A gain-of-function models works cleanly, but human gain-of-function variants segregate into at least five electrophysiological subgroups, and a single channel-selective inhibitor cannot match the diversity of biophysical defects equally [Epilepsia 2024, doi:10.1111/epi.18118]. And the wider Nav1.6 chemistry generation has shifted to base-editing and antisense approaches that target the SCN8A transcript directly, with seizure-protective in vivo data published in Brain 2025 and Ann Neurol 2024 [doi:10.1093/brain/awaf120; doi:10.1002/ana.27053]. The competitive landscape has moved past selective small-molecule channel inhibition.

How to prevent this next time

A program targeting a near-maximally genetically validated gain-of-function channelopathy with a first-in-class selective inhibitor should commit to two design elements before opening Phase 2.

First, biomarker-enriched stratification by SCN8A variant electrophysiology. Patients carry mutations that fall into separable functional bins (persistent-current dominant, slowed-inactivation dominant, hyperpolarized-shift). The Phase 2 should pre-specify a primary analysis in the bin where mouse model effect size is largest, and treat broader-population effect as a secondary endpoint. This is biomarker enrichment, not subgroup data dredging.

Second, an explicit Bayesian posterior probability calculation that ties Phase 2 interim seizure-reduction effect to the conditional probability of a successful Phase 3:

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

Pre-specifying a continuation threshold of PP >= 0.25, against a prior calibrated from cenobamate's published responder rates in SCN8A-DEE (n=14, approximate 50 percent responder rate at six months), would have given Neurocrine a structured decision point at the end of the parent study rather than an attritional extension that cost two years of clinical operations and recruited only eight participants.

The single highest leverage change would have been pre-specifying a variant-electrophysiology-stratified primary endpoint in the parent Phase 2 protocol.

What this means for similar programs

The Nav1.6-selective small-molecule path now has visible execution risk. Sponsors developing antisense (e.g. STK-001 class for SCN1A, by analogy) or base-editing approaches for SCN8A-DEE should accelerate program timelines: the small-molecule competitive shadow has thinned. Sponsors holding selective sodium-channel assets in other monogenic epilepsies (SCN1A, SCN2A, KCNQ2) should pressure-test whether their planned Phase 2 endpoints are powered against variant-stratified responder rates rather than population mean seizure reduction. The broader lesson is that strong genetic validation does not buy permission to skip electrophysiologically stratified clinical design.

Open questions

Why was the parent DEE2012 Phase 2 result never published or presented? Will Neurocrine return the asset to Xenon or out-license to a rare-disease specialist? Are there bedside-to-bench data linking variant electrophysiology bins to NBI-921352 response in the eight extension participants that should still be analyzed and published? Will the FDA accept a single Phase 3 in a Nav1.6-selective successor molecule if the sponsor commits to a pre-specified variant-enrichment design?

Sources

1. • ClinicalTrials.gov record NCT05226780, primary completion 2025-11-12, status TERMINATED. https://clinicaltrials.gov/study/NCT05226780 - Open Targets Platform v25, queried 2026-05-30. SCN8A (ENSG00000196876) x MONDO_0013801: overall 0.8438, genetic_association 0.9769, genetic_literature 0.8365, animal_model 0.3790, literature 0.0774. https://platform.opentargets.org/disease/MONDO_0013801 - Open Targets Platform v25, queried 2026-05-30. SCN8A drugAndClinicalCandidates count = 75 (only NBI-921352 listed as Nav1.6-selective at clinical stage). - ChEMBL v34, molecule CHEMBL4650313 (zandatrigine / NBI-921352 / XEN901), max_phase 2.0. Mechanism: sodium channel protein type VIII alpha subunit inhibitor. https://www.ebi.ac.uk/chembl/explore/compound/CHEMBL4650313 - Bagal SK et al. NBI-921352, a first-in-class, NaV1.6 selective, sodium channel inhibitor that prevents seizures in Scn8a gain-of-function mice. eLife 2022.- Johannesen KM et al. Genotype-phenotype correlations in SCN8A-related disorders reveal prognostic and therapeutic implications. Brain 2022.- Riva A et al. Patients carrying pathogenic SCN8A variants can be classified into five subgroups. Epilepsia 2024.- Petriti U et al. Cenobamate as add-on treatment for SCN8A developmental and epileptic encephalopathy. Epilepsia 2025.- Bunton-Stasyshyn RK et al. Allele-specific editing of a dominant SCN8A epilepsy variant. Ann Neurol 2024.- Tidball AM et al. Persistent Na+ current couples spreading depolarization to seizures in Scn8a gain-of-function mice. Brain 2025.- OpenFDA FAERS query for "NBI-921352" / "zandatrigine" on 2026-05-30 returned no records (investigational, not marketed).