BI 765049 in Asian solid tumours: the regional bridge closed at 21 patients while the B7-H6 programme moved on

What was tried

Boehringer Ingelheim ran NCT06091930, a Phase 1 open-label, non-randomised, multi-centre trial of BI 765049 given by repeated intravenous infusion, alone and with the anti-PD-1 antibody ezabenlimab, in Asian patients with advanced solid tumours expressing B7-H6. It opened on 16 February 2024, recorded a primary completion date of 15 August 2025 and a completion date of 2 February 2026, and enrolled 21 participants. The design had four parts: monotherapy escalation, monotherapy expansion, combination escalation and combination expansion. Primary endpoints were dose-limiting toxicities during the maximum tolerated dose evaluation period for parts one and three, and objective response by RECIST v1.1 for parts two and four.

Eligible tumours were advanced, unresectable or metastatic gastrointestinal, colorectal, pancreatic, liver, head and neck or lung cancer progressing after available therapy. Enrolment ran through two consent forms, the first covering B7-H6 tissue testing for every patient except those with colorectal cancer, the second covering the trial. Colorectal patients skipped the expression screen, which tells you the sponsor treated colorectal cancer as reliably B7-H6 positive and everything else as needing proof. The record lists the study as terminated with the posted reason "Sponsor decision", and no results are posted.

The biological hypothesis

B7-H6, encoded by NCR3LG1, is a ligand for the natural cytotoxicity receptor NKp30. It is largely absent from healthy adult tissue and induced on transformed cells, which is what makes it attractive as a T-cell engager target rather than a checkpoint (DOI 10.1016/j.lfs.2022.120709, DOI 10.3390/ijms251910326). Recent work placed it inside a wider surveillance loop, showing that B7-H6 also subjects activated T cell responses to NK cell control (DOI 10.1126/sciimmunol.adj7970). The sponsor's own preclinical package identified B7-H6 by membrane proteomics as a gastrointestinal tumour antigen with little to no normal tissue expression, and reported activity for a B7-H6 by CD3 IgG-like T-cell engager in humanised mouse models and in colorectal cancer precision-cut tissue slices (DOI 10.1158/1078-0432.CCR-22-2108). Combining B7-H6 engagement with IL-15 has been reported to clear chemotherapy-resistant solid tumour models (DOI 10.3389/fimmu.2025.1625813).

The support is expression biology, not human genetics. Open Targets holds 150 disease associations for NCR3LG1, and the five strongest are abnormality of the skeletal system at 0.504, hypertensive disorder at 0.444, type 2 diabetes mellitus at 0.363, diabetes mellitus at 0.351 and increased blood pressure at 0.279. Digestive system carcinoma (MONDO_0006181) does not appear at all, so the association score for this pair is zero. Open Targets lists no clinical candidate against NCR3LG1, and the tractability flags returning true are localisation and signal-peptide annotations rather than clinical precedent. BI 765049 has no ChEMBL molecule record.

What actually happened

The public record contains a status change and nothing else. Twenty-one participants enrolled against a four-part design whose expansion cohorts require far more, and the posted reason for termination is a two-word sponsor decision. No dose-limiting toxicity counts, no maximum tolerated dose, no response data and no pharmacokinetics have been released.

What the surrounding record does show is that the target survived the trial. The global first-in-human study of the same molecule, NCT04752215, enrolled 67 patients between May 2021 and December 2024 and is listed as completed. A separate Phase 1 dose escalation, NCT06882746, opened on 28 March 2025 in colorectal carcinoma, gastric carcinoma and pancreatic ductal adenocarcinoma with a planned enrolment of 135 and remains active and not recruiting, with a completion date on record of 9 September 2026. The Asian bridging study began before the global escalation had finished and closed after the newer study had opened.

Failure mechanism, best guess

This reads as strategic reprioritisation rather than a data-driven stop. Three facts point that way: the sponsor kept the molecule in a larger, newer, tumour-restricted escalation, the terminated study accrued 21 patients over 18 months, and no safety or efficacy result was released. A safety signal would ordinarily surface as an amended risk section or a specific stopping reason, and none is posted. The most economical reading is that a regional four-part study running in parallel with a global escalation became redundant once the sponsor reshaped Phase 1 around three gastrointestinal tumours, and was closed rather than amended. That is an inference from the sequence and status of three public records, not a sponsor statement.

How to prevent this next time

No endpoint-level data were released, so no quantitative reweighting is possible here. The available levers are design levers.

Do not open a regional bridging protocol before the global escalation reaches a recommended dose. NCT06091930 opened in February 2024 while NCT04752215 was still running, putting a four-part design in the field with no dose to bridge to.

Size the screening funnel from measured prevalence, not assumed positivity. The protocol required B7-H6 tissue testing for every tumour type except colorectal cancer, and an accrual of 21 over 18 months across six tumour types is the signature of a funnel that was never modelled.

Measure the soluble antigen before committing to the format. B7-H6 circulates in soluble form and rises in inflammatory states (DOI 10.1007/s10157-025-02705-9), and soluble antigen is a known sink for T-cell engagers. A pre-specified soluble assay would have been informative regardless of how the trial ended.

The single highest leverage change would have been to run one adaptive global escalation with regional cohorts written into it, rather than a separate regional four-part protocol that opened before the parent study had a dose and closed before it produced a single reportable endpoint.

What this means for similar programs

NCR3LG1 enters the Claidex graph with one recorded failure and a mechanistic risk score of 27, in the yellow band. Almost the whole score comes from the genetic deficit term at 15.0, because Open Targets carries no association between this target and digestive system carcinoma. Phase burden contributes 3.53 and saturation 2.72, which is what a single Phase 1 against an uncontested target looks like. For other B7-H6 programmes the reading is that the target is early rather than damaged. The mechanism has not been tested to failure in humans, and the one trial that stopped did so for reasons the sponsor described as its own decision.

Open questions

Were any dose-limiting toxicities observed in the 21 participants, and will they be reported through the parent programme? What fraction of screened non-colorectal patients met the B7-H6 expression threshold? Did the restriction of NCT06882746 to colorectal, gastric and pancreatic tumours follow from screening data generated in the terminated study?

Sources

1. • ClinicalTrials.gov, NCT06091930. https://clinicaltrials.gov/study/NCT06091930 - ClinicalTrials.gov, NCT04752215. https://clinicaltrials.gov/study/NCT04752215 - ClinicalTrials.gov, NCT06882746. https://clinicaltrials.gov/study/NCT06882746 - The immunoglobulin superfamily ligand B7H6 subjects T cell responses to NK cell surveillance. Sci Immunol, 2024. https://- The potential of B7-H6 as a therapeutic target in cancer immunotherapy. Life Sci, 2022. https://- Harnessing B7-H6 for Anticancer Immunotherapy. Int J Mol Sci, 2024. https://- Zhang W, et al. A Novel B7-H6-Targeted IgG-Like T Cell-Engaging Antibody for the Treatment of Gastrointestinal Tumors. Clin Cancer Res, 2022. https://- Dual T/NK cell engagement via B7-H6-targeted bispecific antibodies and IL-15. Front Immunol, 2025. https://- Association between elevated serum soluble B7-H6 and infection in hemodialysis patients. Clin Exp Nephrol, 2025. https://- Open Targets Platform, NCR3LG1 (ENSG00000188211). https://platform.opentargets.org/target/ENSG00000188211 - openFDA FAERS drug event endpoint, 28 August 2026. https://api.fda.gov/drug/event.json.