Command Palette

Search for a command to run...

ARGX-119 in ALS and the difference between holding a synapse and saving a neuron

NeurologyTranslational MismatchSeptember 16th, 2026·6 min read·10.5281/zenodo.20479005

argenx terminated its 60-participant Phase 2a of the MuSK agonist antibody ARGX-119 in amyotrophic lateral sclerosis after an interim review found continuation unlikely to show a clinically meaningful effect. The same antibody is entering Phase 3 in congenital myasthenic syndromes, where the causal lesion sits at the neuromuscular junction rather than upstream of it.

Mechanism Risk Score

ComponentPoints
Phase-weighted failure burden8.1 / 30
Archetype severity8.5 / 25
Temporal recency4.2 / 15
Genetic evidence deficit13.9 / 15
Programmatic saturation2.7 / 15

For MUSK in Amyotrophic lateral sclerosis, the Mechanism Risk Score is 37/100 (yellow band). The score is a failure-burden index derived from Claidex post-mortems on this target–disease pair, not a probability of approval.

Score is dominated by the genetic deficit component (13.92 of 15): the MUSK-ALS association in Open Targets carries no human genetic evidence and rests on animal model and literature evidence only.

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 (ARGX-119 (adimanebart) / MUSK / Amyotrophic lateral sclerosis): ARGX-119 in ALS and the difference between holding a synapse and saving a neuron

What was tried

argenx ran NCT06441682 (ARGX-119-2303), a Phase 2a double-blind, randomised, placebo-controlled study of ARGX-119 in adults with amyotrophic lateral sclerosis, followed by an active-treatment extension. It enrolled 60 participants across 9 sites, randomised in parallel to three intravenous dose levels or placebo, quadruple masked. It ran from 23 October 2024 to a primary completion date of 12 August 2026. ClinicalTrials.gov posted the termination on 15 September 2026, reason recorded verbatim as "The study was terminated following review of interim data indicating that continuation was unlikely to demonstrate a clinically meaningful treatment effect. The decision was made to minimize unnecessary participant burden."

Eligibility was tighter than most ALS trials. Participants met Gold Coast criteria, carried a TRICALS risk profile of at least -6.0 and below -2.0, and held a slow vital capacity of at least 60% predicted. The registered primary outcome was adverse events to week 96. The efficacy question sat in the secondaries, and the informative one was rate of change from baseline in motor unit number derived from electrophysiological muscle scan, read to week 24. ARGX-119, international nonproprietary name adimanebart, is recorded in ChEMBL as CHEMBL6068571, an antibody at maximum phase 2. No results have been posted.

The biological hypothesis

MuSK is the receptor tyrosine kinase that holds the neuromuscular junction together. Motor neuron derived agrin binds LRP4, LRP4 activates MuSK, and MuSK clusters acetylcholine receptors and maintains the apposition of nerve terminal to muscle endplate. Loss of that signalling dismantles the synapse.

In ALS, junction denervation is an early event. Endplates are vacated before motor neuron cell bodies die, and that distal disconnection produces weakness and eventually respiratory failure. The inference was direct: if the junction goes first, an agonist driving MuSK harder should slow its loss and buy function.

The preclinical case was unusually consistent. Cantor and colleagues treated SOD1-G93A mice with a MuSK agonist antibody after disease onset and reported slowed denervation, improved motor output, greater motor neuron survival and extended lifespan (doi:10.7554/eLife.34375). Sun and colleagues reproduced the benefit in a C9orf72 poly-PR model, where poly-PR disrupted the agrin to LRP4 interaction and attenuated MuSK activation (doi:10.1016/j.ymthe.2024.05.016). argenx then built ARGX-119 as a humanised agonist against human MuSK and showed it reversed relapse in a congenital myasthenia model (doi:10.1126/scitranslmed.ado7189).

What actually happened

The interim review found the effect unlikely to reach clinical meaningfulness and the sponsor stopped the study rather than complete it. No endpoint values, no motor unit number slopes and no ALSFRS-R data are in the public record, so nothing quantitative can be said about the size or direction of the observed effect. openFDA FAERS holds no adverse event records for ARGX-119 or adimanebart, as expected for an investigational antibody with no marketed exposure. The same query returned 7,770 records for olanzapine, so the null reflects the drug rather than a broken connector.

The program did not stop. argenx stated in January 2026 that adimanebart will enter Phase 3 for congenital myasthenic syndromes, with proof-of-concept work continuing in ALS and spinal muscular atrophy. What ended was the ALS hypothesis in the form this trial tested.

Failure mechanism, best guess

This reads as translational mismatch rather than a target that fails to engage. The evidence separates cleanly by where the causal lesion sits. Open Targets scores the MUSK association with amyotrophic lateral sclerosis at 0.072, assembled entirely from animal model evidence at 0.528 and literature co-mention at 0.260, with no human genetic association component. The ALS genes that do carry human genetic evidence are motor neuron intrinsic: SOD1 at 0.944, TARDBP at 0.939, C9orf72 at 0.673. DOK7, behind the congenital myasthenic syndrome indication ARGX-119 is advancing into, carries genetic association evidence at 0.900.

In DOK7 congenital myasthenic syndrome the junction is the disease, and restoring MuSK signalling addresses the causal lesion. In ALS the junction is a readout of a motor neuron dying for reasons upstream of MuSK, so a receptor is pushed harder at the far end of a process that begins elsewhere. Mouse models of ALS are dominated by rapid, synchronised distal denervation, which flatters an endplate-directed intervention in a way a heterogeneous human cohort does not.

A second contributor is dosing confidence. The first-in-human study states that no pharmacodynamic marker existed for MuSK dimerisation by adimanebart, so human dose predictions came from the minimum anticipated biological effect level in nonclinical studies (doi:10.1002/jcph.70208). A null at interim therefore cannot be cleanly separated into wrong target and insufficient engagement.

How to prevent this next time

Endpoint-level data are absent, so the levers here are qualitative. No power calculation, posterior or effect estimate is presented, because the public record does not carry the inputs.

The first lever is base-rate adjustment by evidence composition. An association carried by animal models and literature co-mention, with zero human genetic support, should be priced below one anchored in human variation, even when the mouse data are as clean as they were here. MuSK and ALS sat at 0.072 with that composition while the same antibody's other indication sits behind DOK7 genetics at 0.900.

The second lever is causal position. The question is not whether a target is involved in the disease but whether it sits upstream or downstream of the lesion that kills the patient. Early is not the same as causal.

The third lever is a target engagement biomarker preceding the efficacy trial. Without a pharmacodynamic readout for MuSK activation, this study could not separate a failed hypothesis from an underdosed one, and the graph records an ambiguity rather than an answer.

The single highest leverage change would have been to establish a human pharmacodynamic marker of MuSK activation before running an efficacy study in ALS, so that the interim futility result could be read as evidence about the hypothesis rather than about the dose.

What this means for similar programs

The Claidex failure graph now carries MUSK at an MRS of 37, band yellow, with the genetic deficit component at 13.92 of a possible 15. That single component is most of the score. The target is not saturated and only one Phase 2 program has failed. What drives the reading is that human genetics contributes nothing to the ALS association.

The read across is to junction and muscle directed approaches in ALS more broadly. Preserving the synapse is a coherent goal where the synapse is the primary lesion, and a bet on disease modification where it is not. Programs aimed at distal compartments in motor neuron disease should be asked which situation they are in, and asked to answer with human evidence. For congenital myasthenic syndromes the implication runs the other way: a futility signal in ALS says close to nothing about an indication where the genetic case is anchored in human variation and the pharmacology matches the lesion.

Open questions

What did the motor unit number data show? An agonist that shifted the electrophysiological readout without shifting function would mean something different from one that shifted neither.

Was target engagement achieved at the doses tested, and by what measure, given that no pharmacodynamic marker existed at first-in-human?

Does the ALS proof-of-concept work argenx says is continuing take a different form, such as a biomarker-selected population?

Sources

Related failure claims

Linked claims sharing target, indication, or failure mechanism.