Command Palette

Search for a command to run...

Eftilagimod alfa in metastatic breast cancer: AIPAC-003 stopped at the lead-in, and the Phase 3 it was meant to feed never opened

OncologySponsorAugust 1st, 2026·6 min read·10.5281/zenodo.20479005

Immutep terminated AIPAC-003, a Phase 2/3 study of the soluble LAG-3 protein eftilagimod alfa with weekly paclitaxel in HER2-negative metastatic breast cancer, recording sponsor decision and an actual enrolment of 72 against a design that placed a placebo-controlled overall survival comparison after the dose optimisation lead-in. The predecessor Phase 2b, AIPAC, had already missed its primary progression free survival endpoint with a median of 7.3 months in both arms. Open Targets scores LAG3 against breast cancer at 0.234, of which the genetic evidence contribution is zero.

Mechanism Risk Score

ComponentPoints
Phase-weighted failure burden19.6 / 30
Archetype severity12.2 / 25
Temporal recency9.0 / 15
Genetic evidence deficit13.4 / 15
Programmatic saturation10.5 / 15

For LAG3 in HER2-negative or HER2-low metastatic breast cancer, the Mechanism Risk Score is 65/100 (orange band). The score is a failure-burden index derived from Claidex post-mortems on this target–disease pair, not a probability of approval.

MRS 65/100 (ORANGE), recomputed target-level across all three LAG3 failures on file: favezelimab in Hodgkin and B-cell lymphoma, eftilagimod alfa in HER2-negative metastatic breast cancer, and the relatlimab-nivolumab Phase 3 RELATIVITY-123 in microsatellite stable metastatic colorectal cancer. Components: phase burden 19.63/30, archetype severity 12.16/25, recency 9.02/15, genetic deficit 13.45/15, saturation 10.48/15. The score moved from 46 to 65 because RELATIVITY-123 is the first LAG3 failure driven by a negative readout rather than a portfolio decision, which raises archetype severity from 3.84 to 12.16, and because a Phase 3 adds 4.0 to phase weight. The genetic term uses the LAG3 to colorectal carcinoma association of 0.1032, which carries no human genetic evidence datatype and is contributed entirely by literature. Saturation reflects six Open Targets clinical programmes against LAG3.

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 (Eftilagimod alfa / LAG3 / HER2-negative or HER2-low metastatic breast cancer): Eftilagimod alfa in metastatic breast cancer: AIPAC-003 stopped at the lead-in, and the Phase 3 it was meant to feed never opened

What was tried

Immutep S.A.S. ran AIPAC-003 (NCT05747794), a two part study of eftilagimod alfa added to weekly paclitaxel in metastatic breast cancer. Part one was an open label dose optimisation lead-in at 30 mg and 90 mg subcutaneous eftilagimod alfa with paclitaxel 80 mg per square metre, given on the same day. Part two was a double blind, randomised, placebo controlled Phase 3 comparison with overall survival as the primary endpoint. The registry lists both the optimal biological dose and overall survival among the primary outcome measures, which places the survival question after the dose question rather than alongside it.

Two populations were eligible. The first was hormone receptor positive, HER2 negative metastatic breast cancer that had progressed on at least one line of endocrine based therapy and was indicated for chemotherapy. The second was triple negative disease in patients who were not candidates for PD-1 or PD-L1 therapy in the first line metastatic setting. The study started on 22 May 2023, recorded actual primary completion on 30 April 2026 and actual completion on 30 June 2026, and the record was updated on 30 July 2026 to overall status TERMINATED with the reason given as sponsor decision. Actual enrolment was 72 participants.

Eftilagimod alfa is a recombinant soluble LAG-3 fusion protein (ChEMBL CHEMBL4594557, synonyms IMP321 and Immufact).

The biological hypothesis

LAG3 (Ensembl ENSG00000089692) is best known as an inhibitory checkpoint on exhausted T cells, and five of the six clinical stage LAG3 agents listed by Open Targets are antibodies that block it. Eftilagimod alfa does the opposite. It is a soluble LAG-3 protein that binds MHC class II on antigen presenting cells and acts as an agonist of that receptor, driving dendritic cell and monocyte activation upstream of the T cell response rather than releasing a brake on the T cell itself.

The rationale for pairing it with weekly paclitaxel was that chemotherapy induced tumour cell death supplies antigen, and an MHC class II agonist given on the same day converts that antigen into a primed T cell response. Open Targets scores the LAG3 to breast cancer association at 0.234, composed of a literature datatype score of 0.883 and a clinical datatype score of 0.341. No genetic datatype contributes to the association, so the target rationale in this indication rests on expression, pharmacology and prior trial activity.

What actually happened

No results are posted for AIPAC-003 and no efficacy or safety data from it appear in Immutep's quarterly reporting. What is on the record is the shape of the study. A trial whose Phase 3 component was designed to test overall survival against placebo closed with 72 participants, a number consistent with the dose optimisation lead-in alone. Immutep's Q4 FY26 report, dated 30 July 2026, states that the last patient follow-up visit occurred during the quarter and that the trial was closed effective 30 June 2026. The registry codes the same event as a termination on sponsor decision, so the sponsor and the registry describe the same closure in different terms.

The context is a different trial. On 13 March 2026 Immutep announced that TACTI-004, a Phase 3 of eftilagimod alfa with pembrolizumab and chemotherapy in first line non small cell lung cancer designed for roughly 756 patients, would be discontinued after an independent data monitoring committee recommended stopping for futility. The same Q4 FY26 report records that a proposed investigator initiated Phase 2 of neoadjuvant eftilagimod alfa in early breast cancer is on hold pending the root cause analysis for TACTI-004.

The predecessor trial supplies the quantitative record. AIPAC (NCT02614833) randomised 114 patients to eftilagimod alfa plus paclitaxel and 112 to placebo plus paclitaxel. Median progression free survival, the primary endpoint, was 7.3 months in both arms and the endpoint was not met. Median overall survival was 20.4 months with eftilagimod alfa and 17.5 months with placebo (hazard ratio 0.88, P equal to 0.197). Pharmacodynamic effects were confirmed, with increases in absolute lymphocyte, monocyte, CD4 and CD8 counts and in plasma interferon gamma and CXCL10.

Failure mechanism, best guess

The archetype is sponsor decision. Nothing in the registry or the company disclosures points to a safety signal or an interim futility call inside AIPAC-003 itself.

The mechanistic reading is that AIPAC-003 was built on a pharmacodynamic result rather than a clinical one. AIPAC demonstrated that eftilagimod alfa does what an MHC class II agonist should do to circulating immune cells, and it did not demonstrate that this translates into disease control, since progression free survival was identical at 7.3 months in both arms. The 2.9 month overall survival difference was not significant globally and became significant only in exploratory subgroups defined after the fact. Building a survival powered Phase 3 on that foundation put the burden of proof on a hypothesis that the randomised data had already declined to support once, and TACTI-004 then failed a futility analysis on the same molecule in a different tumour with a different partner.

How to prevent this next time

Endpoint level data from AIPAC-003 do not exist, so the levers are qualitative and design based rather than quantitative.

First, separate pharmacodynamic proof from clinical proof when deciding to advance. AIPAC returned strong biomarker movement and a null primary endpoint, and only the first of those two was carried into the design of the successor trial.

Second, treat post hoc subgroups as hypothesis generating and require prospective replication before they carry a Phase 3. The AIPAC subgroups that showed overall survival benefit were exploratory, and none of them appears as a randomisation stratum or a prespecified enrichment criterion in the AIPAC-003 registry entry.

Third, adjust the base rate for what the class has already shown. Six clinical stage LAG3 agents exist and only one, relatlimab, is approved, and the mechanism at issue here is an agonist rather than a checkpoint blocker, so borrowing prior probability from LAG-3 antagonist success is not valid.

The single highest leverage change would have been to require a prospectively defined biomarker enriched population with progression free survival replication before committing to a survival powered Phase 3, rather than carrying a null Phase 2b primary endpoint forward on the strength of exploratory subgroups.

What this means for similar programs

LAG3 now carries two Claidex recorded failures, this one and the favezelimab Phase 1/2 in Hodgkin lymphoma recorded on 29 July 2026. The recomputed LAG3 mechanism risk score is 46 of 100, band yellow, from phase burden 12.91, archetype severity 3.84, recency 7.30, genetic deficit 11.49 and saturation 10.48. Both recorded failures were sponsor decisions rather than efficacy readouts, which keeps the archetype term low and means the score understates how much clinical evidence has actually accumulated against this target in solid tumours.

Open questions

Whether AIPAC-003 completed its dose optimisation objective and identified an optimal biological dose is not stated in the public record. Whether any efficacy data from the 72 participants will be posted or published is unknown. Whether the TACTI-004 root cause analysis implicates the molecule, the combination partner or the trial population has not been disclosed.

Sources

Related failure claims

Linked claims sharing target, indication, or failure mechanism.