Acasunlimab in post-checkpoint NSCLC: a 17.5 month median, then a portfolio decision

What was tried

Genmab ran GCT1046-04 (NCT05117242), a multicenter, randomized, open-label Phase 2 trial of acasunlimab in relapsed or refractory metastatic non-small cell lung cancer, with Merck Sharp & Dohme supplying pembrolizumab. Acasunlimab, also known as GEN1046 and DuoBody-PD-L1x4-1BB, is a bispecific antibody with one arm against PD-L1 and one against the costimulatory receptor 4-1BB, encoded by TNFRSF9.

Eligibility required stage 4 disease with at least one prior line containing a PD-1 or PD-L1 antibody, PD-L1 expression in at least 1 percent of tumour cells, and ECOG performance status of 0 or 1. A dose-finding part was followed by randomization to three parallel, unmasked arms: acasunlimab monotherapy, acasunlimab plus pembrolizumab every 3 weeks, and acasunlimab plus pembrolizumab every 6 weeks. The primary outcome was objective response rate by RECIST v1.1 and actual enrollment was 125. The trial started on 27 October 2021, reached actual primary completion on 2 December 2024, and the record was updated to TERMINATED on 21 September 2026.

The biological hypothesis

Open Targets describes TNFRSF9 as the receptor for TNFSF9/4-1BBL, conveying a signal that enhances CD8-positive T cell survival, cytotoxicity and mitochondrial activity. Agonising it should restore tumour-specific T cells that checkpoint blockade alone no longer revives, which is exactly the population a post-checkpoint NSCLC trial enrolls.

The obstacle for this class was never potency. It was therapeutic index. Segal and colleagues reported dose-limiting hepatotoxicity in the integrated safety analysis of urelumab, and separately reported that single-agent utomilumab was well tolerated but only modestly active. Systemic 4-1BB agonism concentrates toxicity in the liver, where resident myeloid and T cell populations cluster the receptor without any tumour present.

The bispecific answer was conditional agonism, crosslinking 4-1BB only where PD-L1 is dense so that costimulation stays confined to tumour and tumour-draining tissue. Muik and colleagues characterized GEN1046 preclinically and reported Phase 1 results in 2022, and Capello and colleagues reported in 2025 that the Fc-inert format combined with PD-1 blockade potentiates antitumour immunity.

Open Targets scores TNFRSF9 against non-small cell lung carcinoma at 0.3141, built from literature evidence at 0.938 and somatic mutation evidence at 0.456, with a clinical component of 0.125. That is pharmacological and expression plausibility rather than human genetic validation.

What actually happened

Genmab and BioNTech reported Phase 2 results on 1 June 2024. In the randomized PD-L1-positive population, reported median overall survival was 5.5 months for monotherapy, 8.6 months for the every-3-week combination, and 17.5 months for the every-6-week combination, with a 12-month overall survival rate of 69 percent in the every-6-week arm. Confirmed objective response rates were 13, 18 and 17 percent and disease control rates were 50, 59 and 75 percent. The efficacy analysis included 62 centrally confirmed PD-L1-positive patients and the survival analysis 80. Treatment-related adverse events were described as primarily grade 1 and 2, with the every-6-week schedule showing lower grade 3 or higher rates, and transaminase elevations characterized as generally asymptomatic and manageable.

That signal launched ABBIL1TY NSCLC-06 (NCT06635824), a Phase 3 of acasunlimab plus pembrolizumab against docetaxel in checkpoint-experienced PD-L1-positive metastatic NSCLC.

On 29 December 2025 Genmab announced it would discontinue further clinical development of acasunlimab, citing a strategic focus on the most value-creating opportunities in its late-stage portfolio and a thorough assessment of the evolving competitive landscape. The same release stated that the clinical profile observed to date had been encouraging, and the registry record adds that the decision was not related to safety concerns.

openFDA FAERS held 20 reports naming acasunlimab when queried on 23 September 2026, all coded serious and 6 carrying a death outcome, led by febrile neutropenia at 7, hepatitis at 6 and COVID-19 at 4. These are spontaneous reports with no denominator, though the presence of hepatitis is consistent with the known class liability rather than against it.

Failure mechanism, best guess

This was not a biology failure. The archetype is strategic reprioritization, and three independent pieces of evidence hold that classification in place. The sponsor named portfolio and competitive-landscape reasons, the registry separately records that safety was not the driver, and the Phase 2 numbers taken at face value were competitive for a post-checkpoint NSCLC population where docetaxel remains a weak comparator.

The residual uncertainty sits in the durability of the survival estimate rather than in its direction. Clinical Trials Arena reported analyst commentary describing deterioration of the Phase 2 overall survival curve with longer follow-up, which is secondary attribution rather than a disclosed dataset. What the primary sources support is arithmetic: a 17.5 month median drawn from roughly 27 patients in one unmasked arm carries wide sampling variability, and no confidence interval was published alongside it. A Phase 3 committed against a point estimate inherits the risk sitting between that point and its interval.

How to prevent this next time

Endpoint-level patient data are not public, so no Bayesian posterior or survival reconstruction is supportable. The levers below are qualitative.

Base-rate adjustment is the first. Randomized open-label Phase 2 arms of roughly 27 patients generate survival medians whose variability is large relative to the effect being chased. Asking what fraction of similarly sized survival signals in second-line NSCLC have replicated in Phase 3, and treating the observed median as a draw from a distribution rather than as the truth, changes the go decision without requiring new data.

Schedule effects need a mechanism, not just a number. The every-6-week arm beat the every-3-week arm on both survival and tolerability. Less frequent dosing improving both is plausible for a costimulatory agonist, where sustained receptor occupancy can drive activation-induced cell death, but an unmasked three-arm Phase 2 cannot separate a schedule effect from allocation noise.

Competitive landscape belongs in the design rather than only in the post hoc explanation. A commercial and futility gate firing before Phase 3 enrollment opens converts a late write-off into an early one.

The single highest leverage change would have been treating the 17.5 month every-6-week median as an interval rather than a point, and requiring a second randomized cohort to replicate it before committing Phase 3 resources.

What this means for similar programs

Claidex now holds three TNFRSF9 post-mortems. The other two are acasunlimab in cutaneous melanoma and PRS-344/S095012, a 4-1BB by PD-L1 bispecific in solid tumours, and both were also sponsor or portfolio decisions rather than efficacy or safety failures. Across this target the recurring cause of death is not a negative trial. It is a sponsor concluding that a plausible, tolerable, modestly active costimulatory agent does not clear the bar against a crowded checkpoint landscape.

The recomputed mechanism risk score for TNFRSF9 is 50, in the orange band, up from the 36 and 29 carried by the two earlier disease-specific rows. The increase comes from phase burden and recency rather than archetype severity, which stays low because every recorded failure is a portfolio decision.

Open questions

Was the every-6-week survival advantage a schedule effect or randomization noise? Only patient-level data would settle it.

How did the Phase 2 overall survival curve evolve with longer follow-up? The analyst commentary reported by Clinical Trials Arena implies updated data exist that have not been published.

Does the hepatitis signal in FAERS reflect the class liability the PD-L1-conditional design was built to avoid? Trial-level transaminase data have not been published in full.

Sources

1. • ClinicalTrials.gov, NCT05117242, record retrieved 23 September 2026. https://clinicaltrials.gov/study/NCT05117242 - Genmab, Portfolio Prioritization Update, 29 December 2025. https://ir.genmab.com/news-releases/news-release-details/genmab-portfolio-prioritization-update/ - BioNTech and Genmab, Investigational Acasunlimab in Combination with Pembrolizumab Demonstrates Meaningful Clinical Activity in Phase 2 Trial in Previously Treated Metastatic NSCLC, 1 June 2024. https://www.biontech.com/us/en/home/mediaroom/news/press-releases/2024/06/investigational-acasunlimab-duobodyr-pd-l1x4-1bb-combination.html - Clinical Trials Arena, Genmab shelves Phase III lung cancer candidate, 30 December 2025. https://www.clinicaltrialsarena.com/news/genmab-shelves-phase-iii-lung-cancer-candidate/ - Muik A, et al. Preclinical Characterization and Phase I Trial Results of a Bispecific Antibody Targeting PD-L1 and 4-1BB (GEN1046). Cancer Discov.

2. - Capello M, et al. Acasunlimab, an Fc-inert PD-L1x4-1BB bispecific antibody, combined with PD-1 blockade potentiates antitumor immunity. J Immunother Cancer.

3. - Yamamoto N, et al. A phase 1 dose-finding and pharmacokinetic study of acasunlimab alone or in combination with pembrolizumab. Int J Clin Oncol.

4. - Segal NH, et al. Results from an Integrated Safety Analysis of Urelumab, an Agonist Anti-CD137 Monoclonal Antibody. Clin Cancer Res.

5. - Segal NH, et al. Phase I Study of Single-Agent Utomilumab (PF-05082566), a 4-1BB/CD137 Agonist, in Patients with Advanced Cancer. Clin Cancer Res.

6. - Open Targets Platform, TNFRSF9 (ENSG00000049249) association with non-small cell lung carcinoma (MONDO_0005233), retrieved 23 September 2026. https://platform.opentargets.org/target/ENSG00000049249 - openFDA FAERS drug event endpoint, query on acasunlimab, retrieved 23 September 2026. https://api.fda.gov/drug/event.json - ChEMBL, CHEMBL5095337 (acasunlimab), retrieved 23 September 2026. https://www.ebi.ac.uk/chembl/compound_report_card/CHEMBL5095337/.