Amlitelimab worked in eczema and was dropped in alopecia areata

What was tried

Sanofi ran a Phase 2 proof of concept of amlitelimab, a fully human, non depleting monoclonal antibody against OX40 ligand (OX40L, gene TNFSF4), in adults with severe alopecia areata (NCT06444451). The study was randomized, quadruple masked, and three armed, testing two subcutaneous amlitelimab doses against placebo in patients with at least 50 percent scalp hair loss. The primary endpoint was change from baseline in the Severity of Alopecia Tool (SALT) score at Week 36 (ClinicalTrials.gov). It enrolled 166 patients and ran to its primary completion date of September 24, 2025. Sanofi then terminated it, recording the reason as a sponsor decision unrelated to any safety concern (ClinicalTrials.gov). Amlitelimab is the same molecule (ChEMBL ID CHEMBL4594569) that Sanofi has pushed to Phase 3 in atopic dermatitis (ChEMBL).

The biological hypothesis

OX40L on antigen presenting cells engages OX40 on T cells to deliver a late costimulatory signal that expands and sustains effector and memory populations (Croft, 2010, Webb et al., 2016). Block that signal with a non depleting antibody and you damp ongoing T cell driven inflammation without wiping out the cells. The thesis travels well in atopic dermatitis, where a Phase 2b study showed clear, durable benefit on eczema severity (Weidinger et al., 2025). Alopecia areata is also T cell mediated, so on paper the same lever should apply. The catch is the kind of T cell. Alopecia areata is driven mainly by CD8 cytotoxic lymphocytes acting through NKG2D and interferon gamma, with the collapse of hair follicle immune privilege reversible by JAK inhibition that interrupts that interferon circuit (Xing et al., 2014). OX40L costimulation is one input into T cell activation, not the dominant effector axis in that disease.

What actually happened

Sanofi did not post efficacy results, so the registry record carries only the sponsor decision language (ClinicalTrials.gov). The context is informative. The trial reached its Week 36 primary readout before being stopped, and across the same period Sanofi reported that amlitelimab met all primary and key secondary endpoints in its COAST 1 Phase 3 atopic dermatitis study and continued a five trial Phase 3 program in that indication (Sanofi, 2025). A sponsor that is accelerating a molecule in one indication and quietly closing a completed proof of concept in another is usually acting on the data it just saw. The evidence base supports that reading. OX40L association with atopic dermatitis scores 0.545 in Open Targets, with a genetic_association component of 0.732. The same target against alopecia areata scores 0.076, with no genetic component at all (Open Targets Platform).

Failure mechanism, best guess

The most parsimonious read is an indication mismatch, not a molecule problem. Amlitelimab engages OX40L and produces real clinical effect where OX40L costimulation feeds the dominant pathology, which in atopic dermatitis is a genetically supported, Th2 skewed process (Weidinger et al., 2025, Open Targets Platform). Alopecia areata runs through a CD8 cytotoxic, interferon dominated circuit where JAK inhibition, not costimulation blockade, has shown the clearest control (Xing et al., 2014). Upstream costimulation blockade can blunt T cell priming, but in a disease where established CD8 effectors and resident memory cells already sit at the follicle, removing one costimulatory input may not be enough to reverse the attack. The absence of any genetic_association signal for OX40L in alopecia areata, against a strong one in atopic dermatitis, is the quantitative version of that argument.

How to prevent this next time

Two quantitative tools would have framed this bet more honestly before 166 patients enrolled. First, a human genetics anchored indication score. Open Targets exposes a per indication genetic_association datatype, and a pre-specified rule (advance only indications above a defined genetic threshold, or require an explicit non genetic mechanistic case) would have flagged alopecia areata at 0.076 with no genetic term as a higher risk extension of a target validated at 0.545 in atopic dermatitis (Open Targets Platform). Second, a Bayesian predictive probability of success, with a biomarker enriched arm. Stratifying by a scalp CD8 and interferon signature, then computing:

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

where theta is the dose response on SALT and the prior reflects the historical base rate for costimulation blockers in CD8 driven disease, would have set a quantitative bar for continuation and identified, at interim, whether any responder subgroup existed. The single highest leverage change would have been a genetics anchored, biomarker enriched proof of concept that pre-committed to an interferon high subgroup and a Bayesian futility boundary, so a weak or absent signal stopped the study at interim rather than at full enrollment.

What this means for similar programs

OX40 and OX40L are crowded, with rocatinlimab and others working the same axis. The lesson is not that the pathway is invalid but that target validation does not transfer across indications for free. A costimulation blocker earns its strongest odds in diseases with a genetic anchor and a costimulation dominated mechanism. Extensions into T cell diseases that run on cytotoxic, interferon driven effectors should carry an explicit mechanistic argument and, ideally, a biomarker defined population, rather than borrowing confidence from the anchor indication. Sanofi's decision to keep investing in atopic dermatitis while closing the alopecia areata arm is a rational sort by mechanism.

Open questions

Did any amlitelimab dose move SALT at all in alopecia areata, and was there a biomarker defined subgroup that responded. Would combination with a JAK inhibitor, hitting both the priming and the interferon effector arm, change the calculus. Does OX40L blockade have a role earlier in disease, before resident memory CD8 cells entrench at the follicle. Until Sanofi releases the Week 36 data, whether amlitelimab had any biological effect in this indication stays unresolved.

Sources

1. ClinicalTrials.gov, A Study to Evaluate the Efficacy and Safety of Subcutaneous Amlitelimab Monotherapy Compared With Placebo in Adult Participants With Severe Alopecia Areata, NCT06444451, 2024 to 2026, https://clinicaltrials.gov/study/NCT06444451.

2. Weidinger S, Bieber T, Cork MJ, et al., Phase 2b randomized clinical trial of amlitelimab, an anti OX40 ligand antibody, in atopic dermatitis, Journal of Allergy and Clinical Immunology, 2025,.

3. Xing L, Dai Z, Jabbari A, et al., Alopecia areata is driven by cytotoxic T lymphocytes and is reversed by JAK inhibition, Nature Medicine, 2014,.

4. Webb GJ, Hirschfield GM, Lane PJL, OX40, OX40L and autoimmunity, a comprehensive review, Clinical Reviews in Allergy and Immunology, 2016,.

5. Croft M, Control of immunity by the TNFR related molecule OX40 (CD134), Advances in Immunology, 2010,.

6. Sanofi, Press release, amlitelimab met all primary and key secondary endpoints in the COAST 1 Phase 3 study in atopic dermatitis, 2025, https://www.sanofi.com/en/media-room/press-releases/2025/2025-09-04-05-00-00-3144170.

7. Open Targets Platform v4, TNFSF4 (ENSG00000117586) associations with atopic eczema (EFO_0000274) and alopecia areata (EFO_0004192), accessed 2026-06-15, https://platform.opentargets.org/target/ENSG00000117586.

8. ChEMBL, Amlitelimab (CHEMBL4594569), accessed 2026-06-15, https://www.ebi.ac.uk/chembl/web_components/explore/compound/CHEMBL4594569.