Botensilimab reported activity in checkpoint-refractory melanoma and the study closed anyway

What was tried

NCT05529316, run by Agenus as study C-800-23, was a multicohort, open-label, randomized Phase 2 trial of botensilimab in advanced cutaneous melanoma refractory to prior checkpoint inhibitor therapy. It opened on 12 December 2022, reached an actual enrollment of 150 participants, and recorded actual primary completion on 27 April 2026. The primary endpoint was objective response rate, assessed up to 40 months.

The trial ran four cohorts. Part 1 gave botensilimab alone, Cohort A to patients refractory to PD-(L)1 therapy and Cohort B to patients refractory to both PD-(L)1 and anti-CTLA-4. Part 2 repeated that split with botensilimab plus balstilimab, an anti-PD-1 antibody, dosed at 75 mg and 450 mg every three weeks.

Botensilimab is an Fc-enhanced anti-CTLA-4 antibody, ChEMBL CHEMBL4650435, recorded in Open Targets as an antibody at maximum clinical stage Phase 3. The ClinicalTrials.gov record was updated on 20 October 2026 with the status TERMINATED and the reason "Sponsor decision".

The biological hypothesis

Conventional anti-CTLA-4 antibodies block the receptor. Botensilimab was engineered to add Fc-mediated effector function on top of that blockade. Chand and colleagues reported in 2024 that this Fc enhancement produced activity in tumours poorly responsive to conventional immunotherapy. Kim and colleagues provided the mechanistic detail the same year, showing that the Fc-enhanced molecule binds FcγRIV in mice, the ortholog of human FcγRIIIA, and selectively depletes intratumoral regulatory T cells through phagocytosis by tumour-associated macrophages while sparing peripheral regulatory T cells.

That mechanism makes the drug dependent on the myeloid compartment of the tumour, not only on T cell biology. The same paper reported that FcγRIIIA expression on tumour-associated macrophages in human glioblastoma was highest at diagnosis.

Open Targets scores the CTLA4 association with cutaneous melanoma at 0.484, with a genetic association component of 0.743 and a clinical component of 0.186. The prediction under test was that adding effector function would recover responses in patients whose disease had already progressed through checkpoint blockade, including through conventional anti-CTLA-4.

What actually happened

The trial generated data and reported them. At ASCO 2026, abstract 9543, presented by Michael B. Atkins on 31 May 2026, Agenus released results for the botensilimab plus balstilimab population, N=36. Confirmed objective response rate was 22 percent. Median overall survival was 16.6 months, two-year overall survival was 42 percent, and the clinical benefit rate at 24 weeks was 33 percent. Median duration of response was not reached, and 86 percent of responders remained in response at 12 months. Grade 3 or higher treatment-related adverse events occurred in 36 percent, with no treatment-related deaths reported.

The subgroup refractory to both PD-(L)1 and anti-CTLA-4, n=14, reported a 29 percent response rate, two-year overall survival of 64 percent, median overall survival not reached, and all responders still in response at 12 months. No confidence intervals were given for any melanoma figure, and all of it comes from a company release rather than a peer-reviewed report.

For comparison, the same combination in relapsed or refractory microsatellite stable metastatic colorectal cancer, reported by Bullock and colleagues in 2024, produced a confirmed response rate of 17 percent (17 of 101 evaluable, 95% CI 10 to 26 percent), disease control in 61 percent (62 of 101, 95% CI 51 to 71 percent), median progression-free survival of 3.5 months, and treatment-related adverse events in 89 percent of 148 treated patients.

Five months after the melanoma data were presented, the registry recorded the study as terminated for sponsor decision.

Failure mechanism, best guess

The archetype is strategic reprioritization, and the evidence points to capital allocation rather than a negative result. In its second quarter 2026 report Agenus stated that it "concentrates its development resources on ROBBIN", a global randomized Phase 3 of neoadjuvant botensilimab plus balstilimab in high-risk stage II and III microsatellite stable colon cancer, planned at roughly 850 patients with event-free survival as the primary endpoint and first dosing expected in the first quarter of 2027. The same report disclosed cash of 18.7 million dollars at 30 June 2026, down from 35.0 million at the end of the first quarter, before roughly 85 million dollars gross from a July private placement. It also recorded the discontinuation of planned funding for BATTMAN, a separate Phase 3 in refractory colorectal cancer, stating that the decision "reflected financing and development priorities" and was not driven by enrollment, efficacy or safety. That release does not state that melanoma was deprioritized, and no company statement reviewed here links the melanoma termination to the ROBBIN decision.

There is a mechanistic reading that makes the portfolio move coherent rather than arbitrary. If botensilimab depends on FcγRIIIA-expressing tumour macrophages, as Kim and colleagues described, then a treatment-naive neoadjuvant setting should favour it over a heavily pretreated refractory one. Kasi and colleagues had already reported an unusual rapid regression pattern with neoadjuvant botensilimab plus balstilimab in mismatch repair proficient colorectal cancer in 2023. Moving to the setting where the required myeloid substrate is most abundant is the scientifically defensible version of this decision.

The design left the melanoma question unanswerable either way. A 150-patient study split across four single-arm cohorts produced a 36-patient combination readout with no randomized comparator and no confidence intervals. A 22 percent response rate in that population is interesting and uninterpretable at the same time.

How to prevent this next time

No randomized comparison exists for either melanoma cohort, and no dispersion is reported, so a quantitative recalculation is not supportable here. The qualitative levers are the ones that apply.

Allocation discipline comes first. Four single-arm cohorts across two monotherapy and two combination questions guaranteed that no individual readout would be large enough to force a decision. Deciding the combination question first and spending the full enrollment on it would have produced one interpretable number.

Biomarker enrichment comes second and follows directly from the mechanism. If activity requires FcγRIIIA-positive tumour macrophages, that density is the obvious stratification variable, and no cohort in this trial is reported as having measured it. A program built on an Fc-dependent mechanism that never measures the Fc receptor is leaving its own hypothesis untested.

A pre-specified go threshold comes third. Setting a response rate that would trigger continuation, against a defined benchmark in post-checkpoint melanoma, converts a sponsor decision into a protocol decision and removes the ambiguity that now surrounds this termination.

The single highest leverage change would have been to run one randomized combination cohort powered against a defined control instead of four single-arm cohorts that split 150 patients into readouts too small to act on.

What this means for similar programs

Engineered-Fc antibodies are a growing class, and this program illustrates their central design requirement. The effector mechanism depends on host cells inside the tumour, so setting and line of therapy are not interchangeable the way they often are for a blocking antibody. Programs in this class should treat myeloid composition as a selection variable from first-in-human onward.

ChEMBL curates six molecules with CTLA-4 mechanisms, ipilimumab, tremelimumab, cadonilimab, erfonrilimab, zalifrelimab and quavonlimab, and botensilimab is not among them, which puts at least seven distinct programs against this target. Target saturation is high and the differentiating claim for any new entrant now has to be mechanistic rather than positional.

Open questions

What did the botensilimab monotherapy cohorts show? Only the combination population has been reported, and the monotherapy arms address whether the Fc enhancement works without PD-1 blockade.

Will the melanoma data be published with confidence intervals and cohort-level detail? Everything available is a company release describing a poster.

Does tumour FcγRIIIA-positive macrophage density predict response? The mechanism predicts it, the ROBBIN trial is positioned as though it is true, and no reported cohort has tested it.

Sources

1. Journal articles below were retrieved via PubMed. All sources accessed 21 October 2026. - ClinicalTrials.gov, NCT05529316 record. https://clinicaltrials.gov/study/NCT05529316 - Agenus, Phase 2 BOT+BAL melanoma data at ASCO 2026, abstract 9543. https://www.biospace.com/press-releases/agenus-presents-phase-2-bot-bal-melanoma-data-showing-durable-responses-and-meaningful-survival-in-advanced-checkpoint-refractory-melanoma-at-asco-2026 - Agenus, second quarter 2026 results and ROBBIN Phase 3. https://finance.yahoo.com/healthcare/articles/agenus-reports-second-quarter-2026-113000238.html - Bullock AJ, et al. Nature Medicine 30(9), 2024, 2558-2567. https://- Chand D, et al. Cancer Discovery 14(12), 2024, 2407-2429. https://- Kim KS, et al. Neuro-Oncology 26(11), 2024, 2044-2060. https://- Kasi PM, et al. Oncogene 42(44), 2023, 3252-3259. https://- Open Targets, CTLA4 (ENSG00000163599) and cutaneous melanoma (MONDO_0005012) association. https://platform.opentargets.org/evidence/ENSG00000163599/MONDO_0005012 - ChEMBL, botensilimab CHEMBL4650435 and CTLA-4 mechanism records against target CHEMBL2364164. https://www.ebi.ac.uk/chembl/ - openFDA drug adverse event endpoint, counts for botensilimab. https://open.fda.gov/apis/drug/event/.