CTX-471 in post-checkpoint solid tumours: the combination arm that could not have answered its own question

What was tried

CTX-471 is a fully human IgG4 agonist antibody against CD137, the costimulatory receptor encoded by TNFRSF9. Compass Therapeutics ran it through a first-in-human study, NCT03881488, open from 17 May 2019 to 21 May 2025, with an actual enrolment of 100 patients across 10 United States sites. The design was open label, non-randomised and sequential, with no masking. Merck, Sharp and Dohme supplied pembrolizumab as a collaborator.

Two parallel arms each ran a dose escalation and a dose expansion. The monotherapy arm admitted any metastatic or locally advanced malignancy and tested CTX-471 at 0.1, 0.3, 0.6 and 1.2 mg/kg every two weeks. The combination arm added pembrolizumab 400 mg every six weeks and restricted expansion to non-small cell lung cancer, small cell lung cancer and melanoma at the 0.3 and 0.6 mg/kg levels. Every patient had progressed after at least two doses of an approved PD-1 or PD-L1 inhibitor given over at least 12 weeks. The primary endpoint in both arms was toxicity, not response.

Compass closed the combination arm and let the monotherapy arm finish. The registry gives the reason directly: the combination arm "did not yield an efficacy signal". Results appeared 9 October 2026.

The biological hypothesis

CD137 is a TNF receptor superfamily member that, on ligation by TNFSF9, sustains CD8 T cell survival, cytotoxicity and mitochondrial activity. The clinical case for agonising it rested on two earlier antibodies that both stalled. Urelumab produced dose-limiting hepatotoxicity and was pushed down to flat doses of 3 and 8 mg, where its final combination results showed limited activity. Utomilumab reached Phase 3 but behaved as a weak agonist. A 2024 review names hepatotoxicity as the barrier that has held the class back, and points to epitope, affinity and IgG subclass as the levers that might clear it.

CTX-471 was built against exactly that reading. Its preclinical characterisation reported a unique epitope conserved across human, cynomolgus monkey and mouse, FcγR-dependent IFN-γ induction in vitro at a level between the two clinical-stage antibodies, and curative monotherapy activity in syngeneic models including tumours of roughly 500 mm3, with no hepatic toxicity at very high doses. The distinguishing claim was monotherapy potency against large established tumours, not combination synergy.

Human genetic support is absent. Open Targets scores the TNFRSF9 association with neoplasm at 0.209, from literature mining at 0.993, animal models at 0.371 and clinical evidence at 0.286, with a genetic association component of zero. For non-small cell lung carcinoma it falls to 0.093.

What actually happened

The monotherapy arm produced five confirmed responses in 73 response-evaluable patients, a rate of 6.8 percent, and they were not evenly spread. The 0.3 mg/kg level gave four responses in 31 patients, the 0.6 mg/kg level one in 32, and neither 0.1 mg/kg in three patients nor 1.2 mg/kg in seven gave any. Median duration of response at 0.3 mg/kg was 24.43 months across four responders, observed full range 1.88 to 46.97 months.

The combination arm produced zero responses in 20 patients, across all three dose levels tested. Median progression-free survival ran 3.05 to 3.55 months in monotherapy and 1.84 to 2.89 months in combination, and median overall survival 9.41 to 14.70 months against 4.05 to 8.65 months. Those arm-level differences cannot be read as treatment effects, because the arms were non-randomised and admitted different tumour types.

Safety was unremarkable for the class. Cytokine release syndrome was serious in two of 36 patients at 0.6 mg/kg monotherapy and one of nine at 0.3 mg/kg combination. Serious transaminase elevations appeared once each at 0.6 mg/kg monotherapy, and non-serious alanine aminotransferase increases in three of 33, three of 36 and two of seven monotherapy patients at 0.3, 0.6 and 1.2 mg/kg. No hepatotoxicity signal emerged of the kind that constrained urelumab. CTX-471 has no openFDA adverse event records, as expected for an unmarketed agent.

Failure mechanism, best guess

The stated reason is accurate as far as it goes. The question is whether the combination arm could have produced a signal at all.

Zero responses in 20 patients is consistent with any true response rate up to 13.9 percent, taking the exact one-sided 95 percent upper bound as 1 minus 0.05 raised to the power 1/20. The monotherapy arm in the same study ran at 6.8 percent. The combination arm therefore could not have excluded a response rate equal to, or roughly double, what the drug already achieved on its own. A null result from that arm carries almost no information about the combination hypothesis.

Behind that sits a dose selection problem. Responses clustered at 0.3 mg/kg and thinned at 0.6 and 1.2 mg/kg, the non-monotonic pattern the 4-1BB literature predicts when FcγR-mediated crosslinking drives agonism and excess systemic receptor occupancy degrades it. The combination arm inherited the 0.3 and 0.6 mg/kg levels from the monotherapy escalation, where they had been chosen on tolerability, not on any pharmacodynamic measure of CD137 crosslinking.

How to prevent this next time

The data supports one quantitative lever and several qualitative ones. Sizing a combination expansion to 20 patients when the comparator is a 6.8 percent monotherapy rate guarantees an uninformative answer. A futility boundary set against the monotherapy rate, rather than against zero, would have shown before the first combination patient enrolled that the arm could not reach a decision.

The qualitative levers follow from the same record. Biomarker enrichment: nothing selected patients by CD137 expression, FcγR genotype, or any measure of T cell costimulatory capacity, which leaves the 6.8 percent monotherapy rate undecomposed. Base-rate adjustment: with two prior clinical-stage CD137 agonists stalled and no human genetic anchor, the prior probability of combination benefit was low enough to justify spending the available patients on resolving the monotherapy dose response instead. Red-teaming the arm structure: the preclinical package argued for monotherapy activity against large tumours, and the combination arm tested the hypothesis the molecule was least differentiated on.

The single highest leverage change would have been to size the combination expansion against the 6.8 percent monotherapy response rate it was meant to beat, since zero of twenty cannot exclude a true rate of 13.9 percent.

What this means for similar programs

Claidex now holds four TNFRSF9 failures. Acasunlimab was shelved in post-checkpoint non-small cell lung cancer and again in melanoma, PRS-344/S095012 was stopped in Phase 1 in solid tumours, and CTX-471 joins them. Counting urelumab and utomilumab from the Open Targets record, five distinct programmes against this target now carry a documented stop. The mechanism risk score for TNFRSF9 rises to 62 of 100, in the orange band, with the genetic deficit component at 10.29 of 15 reflecting an association that carries no genetic evidence at all.

The molecule is not dead. Compass currently lists CTX-471 in a Phase 1 basket study in NCAM-positive neuroendocrine neoplasms, with a Phase 2 start planned. That is the right read of this result: what failed was an indication and a combination strategy, not necessarily the antibody.

Open questions

Whether the five monotherapy responses concentrated in one histology is unknown, because the response-evaluable subset has not been published by tumour type. No pharmacodynamic data has been posted, which leaves the difference between 0.3 and 0.6 mg/kg unexplained. Whether the combination arm closed on a prespecified futility rule is not stated. The long responses at 0.3 mg/kg, one past 46 months, have no published correlate.

Sources

1. • ClinicalTrials.gov. NCT03881488, A Phase 1, Open-Label, Multiple-Ascending Dose Study of CTX-471. Results first posted 9 October 2026. https://clinicaltrials.gov/study/NCT03881488 - Eskiocak U and colleagues. Differentiated agonistic antibody targeting CD137 eradicates large tumors without hepatotoxicity. JCI Insight, 2020.https://- Singh R and colleagues. 4-1BB immunotherapy: advances and hurdles. Experimental and Molecular Medicine, 2024.https://- Khushalani NI and colleagues. Final results of urelumab, an anti-CD137 agonist monoclonal antibody, in combination with cetuximab or nivolumab in patients with advanced solid tumors. Journal for ImmunoTherapy of Cancer, 2024.https://- Claus C and colleagues. The emerging landscape of novel 4-1BB (CD137) agonistic drugs for cancer immunotherapy. mAbs, 2023.https://- Open Targets Platform. TNFRSF9 (ENSG00000049249) associations and clinical candidates, accessed 13 October 2026. https://platform.opentargets.org/target/ENSG00000049249 - Compass Therapeutics. Pipeline, accessed 13 October 2026. https://www.compasstherapeutics.com/pipeline/ - Claidex failure graph: acasunlimab-gen1046-tnfrsf9-pdl1-4-1bb-nsclc-phase2-strategic-termination, acasunlimab-tnfrsf9-pdl1-4-1bb-bispecific-melanoma-phase2-strategic-termination, prs-344-s095012-tnfrsf9-4-1bb-pdl1-solid-tumors-phase1-2-sponsor-termination.