ICT01 in advanced solid tumors: EVICTION-2 met its endpoints and was closed by an acquisition

What was tried

EVICTION-2 (NCT05307874) tested ICT01, a humanized anti-BTN3A monoclonal antibody, in combination with low dose subcutaneous interleukin-2 and, in later arms, pembrolizumab. The sponsor was ImCheck Therapeutics, an Ipsen company. The study opened on 4 May 2022 as a two part, open label, non-randomized Phase 1/2 design across five sites in France, Germany and the United Kingdom, and enrolled 56 participants.

Eligibility required failure of at least two lines of systemic therapy, ECOG performance status of 1 or better, at least one measurable lesion by RECIST 1.1, and a baseline tumor biopsy plus willingness to undergo on-study biopsies. Permitted tumor types were metastatic colorectal, ovarian, castration-resistant prostate and pancreatic ductal adenocarcinoma, and refractory melanoma.

Eleven arm groups were registered. Five escalating ICT01 dose levels were paired with low dose IL-2, one early arm used high dose IL-2, two arms added a day 8 booster dose on top of a primer dose, and three arms added pembrolizumab. ICT01 was given intravenously on day 1 of every 21 day cycle, and subcutaneous IL-2 on days 1 to 5 of cycles 1, 2 and 3 only.

The registered primary outcome was treatment-emergent adverse events over one year. Secondary outcomes were the change from baseline in circulating Vγ9Vδ2 T cell numbers by flow cytometry, and disease control rate defined as stable disease or better by RECIST 1.1.

The biological hypothesis

Butyrophilin 3A1 is a cell surface protein that reports intracellular phosphoantigen accumulation to Vγ9Vδ2 T cells. Structural work in Nature in 2023 showed that phosphoantigens bind at the interface between the intracellular domains of BTN3A1 and BTN2A1, gluing the two molecules together to license T cell receptor engagement (doi:10.1038/s41586-023-06525-3). Later work refined this into a heteromer and multimer model (doi:10.1038/s41590-024-01892-z, doi:10.1016/j.immuni.2025.05.011). An agonist antibody that locks BTN3A into the activating conformation should recruit an innate-like cytotoxic population that does not depend on peptide-MHC presentation or neoantigen load.

That property is the attraction. Pancreatic adenocarcinoma, microsatellite stable colorectal cancer and castration-resistant prostate cancer are poorly served by checkpoint blockade, which requires an existing αβ T cell response to release. Work in Science in 2020 also showed that BTN3A1 coordinates αβ and γδ responses, and that its engagement can suppress αβ T cells while activating γδ cells (doi:10.1126/science.aay2767). Low dose IL-2 was added to sustain the expanded γδ compartment, and pembrolizumab to release the αβ compartment in parallel.

What actually happened

The registry records actual study completion on 9 October 2025 and gives this reason for termination: the study was terminated on a sponsor decision unrelated to safety or efficacy, it met the key primary and secondary endpoints, and two patients were still on treatment at the decision point, preventing protocol-defined completion. The termination record was posted on 30 July 2026. No results have been posted to ClinicalTrials.gov.

Thirteen days after the recorded completion date, on 22 October 2025, Ipsen announced an agreement to acquire ImCheck Therapeutics for 350 million euros upfront and up to 1 billion euros in total consideration. That announcement named a single lead indication for ICT01: first line acute myeloid leukemia in patients ineligible for intensive chemotherapy, supported by interim data in 45 patients presented at ASCO 2025 in combination with venetoclax and azacitidine. The release described BTN3A overexpression across melanoma, urothelial, colorectal, ovarian, pancreatic and lung cancer, but cited no solid tumor clinical data.

Read against the protocol, the phrase "met the key primary and secondary endpoints" carries less weight than it appears to. The primary endpoint was adverse event incidence. One secondary endpoint measured circulating γδ T cell counts, which earlier ImCheck disclosures had already shown to move in a dose dependent way. The only efficacy secondary endpoint was disease control rate, which counts stable disease. None of these could distinguish a durable antitumor effect from target engagement without benefit.

Failure mechanism, best guess

This was a portfolio decision, not a biology readout. The archetype is strategic_reprioritization. An acquirer paying up to 1 billion euros for a single asset concentrated the program on the indication with the strongest interim dataset, and a 56 patient open label study with no randomized comparator and no posted efficacy signal did not compete for that capital.

The design made the study easy to close. Across eleven arms and 56 participants, the average arm held roughly five patients, computed from the registry counts rather than a reported allocation. No arm was powered to estimate a response rate with useful precision, and the endpoint hierarchy placed safety first and disease control last. A study built that way generates a defensible safety and pharmacodynamic package and almost no evidence that survives a portfolio review.

The biology carries a quieter second risk. Open Targets scores the BTN3A1 to neoplasm association at 0.113, and the whole score comes from literature co-mention at 0.933, with a human genetic association score of 0. Open Targets lists no known drugs or clinical candidates against BTN3A1. A target with no genetic anchor and no precedent asset is not disqualified, but every claim of clinical benefit must be earned inside the trial.

How to prevent this next time

No endpoint level efficacy data was ever posted, so the available data cannot support a Bayesian or power calculation. The applicable levers are qualitative and structural.

First, endpoint honesty. A Phase 1/2 whose only efficacy measure is disease control rate cannot answer the question a sponsor asks at portfolio review. One arm carrying a prespecified objective response rate estimate and a defined futility boundary converts a safety study into a decision-grade study.

Second, arm discipline. Eleven arms across 56 patients spreads the sample below the threshold at which any single regimen can be evaluated. Fewer arms with more patients each would have produced an interpretable dataset on the same budget.

Third, biomarker enrichment. The protocol required baseline and on-study biopsies from every participant, so the tissue existed to define a BTN3A-high population prospectively. Selecting on target expression rather than line of therapy alone would have concentrated any real effect.

Fourth, base rate adjustment. Sponsors should assume an asset acquired for a lead hematology indication will lose its solid tumor arms, and either finish those arms before a transaction or treat them as exploratory.

The single highest leverage change would have been to run two arms instead of eleven, with a prespecified response rate estimate in a BTN3A-selected population, so the solid tumor question had an answer before the acquisition made it moot.

What this means for similar programs

Three lessons carry across. Agonist antibodies against innate-like effector axes are attractive because they bypass the requirements of checkpoint blockade, and that same novelty means no genetic prior and no class precedent to borrow from. Registry language stating that a termination was unrelated to efficacy should be read against the endpoint list, because meeting a safety primary endpoint is compatible with a wide range of efficacy outcomes. Programs built around a single lead asset carry acquisition risk as a design constraint, and secondary indications should be sequenced accordingly.

Open questions

What was the disease control rate by tumor type, and did any of the five permitted types separate from the others? Did the pembrolizumab arms show anything the IL-2 arms did not? Was BTN3A expression measured on the required biopsies, and did it track with γδ infiltration or disease control? Will Ipsen post results for a completed study whose sponsor changed hands?

Sources

1. • ClinicalTrials.gov, NCT05307874, EVICTION-2 full study record, retrieved 31 July 2026. https://clinicaltrials.gov/study/NCT05307874 - Open Targets Platform API v26.6.3, target BTN3A1 (ENSG00000026950), association scores for MONDO_0005070, retrieved 31 July 2026. https://platform.opentargets.org/target/ENSG00000026950 - Ipsen press release, 22 October 2025, Ipsen to acquire ImCheck Therapeutics. https://www.ipsen.com/press-release/ipsen-to-acquire-imcheck-therapeutics-expanding-its-leadership-in-oncology-strengthening-its-pipeline-3170788/ - Yuan L, et al. Phosphoantigens glue butyrophilin 3A1 and 2A1 to activate Vγ9Vδ2 T cells. Nature, 2023 (). https://- Payne KK, et al. BTN3A1 governs antitumor responses by coordinating αβ and γδ T cells. Science, 2020 (). https://- Fulford TS, et al. Vγ9Vδ2 T cells recognize butyrophilin 2A1 and 3A1 heteromers. Nature Immunology, 2024 (). https://- Zhang M, et al. Structures of butyrophilin multimers reveal a plier-like mechanism for Vγ9Vδ2 T cell receptor activation. Immunity, 2025 (). https://- Mamedov MR, et al. CRISPR screens decode cancer cell pathways that trigger γδ T cell detection. Nature, 2023 (). https://- ImCheck Therapeutics, SITC presentation release on first clinical demonstration of γδ T cell activation with ICT01. https://www.imchecktherapeutics.com/all-press-release/news/imcheck-presents-at-sitc-first-clinical-demonstration-of-gamma-delta-t-cell-activation-and-tumor-infiltration-with-positive-safety-for-ict01/.