Avelumab plus lenvatinib in children with primary CNS tumors: an adult combination that did not translate

What was tried

EMD Serono Research and Development Institute ran NCT05081180, a single-arm, multicenter Phase 1/1b study of avelumab plus lenvatinib in children with primary CNS tumors, with Merck KGaA, Darmstadt, Germany as collaborator. The record was first posted on 2021-10-18 and the study started on 2021-12-03. Actual enrollment reached 17 participants aged 2 to 18 years. The design carried no allocation, used a sequential intervention model, and applied no masking.

Eligibility required a histologically confirmed high-grade primary CNS malignancy, measurable disease by Response Assessment in Neuro-Oncology criteria, and a Lansky or Karnofsky performance status of at least 50. Most participants had to have progressed after at least one prior systemic therapy. Children with diffuse midline glioma could enroll with no more than one prior systemic therapy or prior radiotherapy alone. Low-grade gliomas were excluded.

Part 1 carried two safety primary endpoints, Grade 3 or higher treatment-emergent adverse events under CTCAE version 5.0 and dose-limiting toxicities. Part 2 carried progression-free survival by the same criteria. Primary completion was 2026-02-11 and completion 2026-06-09, with the record updated on 2026-07-29 as terminated.

Both agents were approved drugs rather than novel candidates. ChEMBL records avelumab as an antibody, CHEMBL3833373, first approved in 2017, and lenvatinib as a small molecule, CHEMBL1289601, first approved in 2015.

The biological hypothesis

The combination paired PD-L1 blockade with multi-kinase VEGFR inhibition. Avelumab binds CD274, the gene encoding PD-L1, releasing an inhibitory brake on T cell recognition of tumor cells. Lenvatinib inhibits VEGFR2, encoded by KDR. The intended synergy is well established in adult oncology: antiangiogenic pressure is expected to normalize tumor vasculature, improve immune cell trafficking, and reduce myeloid-driven suppression, so that checkpoint blockade has a substrate to work on.

Prior pediatric exposure made this a defensible next step rather than a leap. Avelumab had been dose-escalated in children with refractory or relapsed solid tumors (Cancer Immunology, Immunotherapy 2022), and lenvatinib had been combined with everolimus in recurrent and refractory pediatric solid tumors including CNS tumors (Journal of Clinical Oncology 2020). Checkpoint blockade in glioma also had adult support, including neoadjuvant anti-PD-1 in recurrent glioblastoma (Nature Medicine 2019).

What the hypothesis lacked was target-to-disease evidence. Open Targets scores the CD274 association with central nervous system cancer at 0.1406, built from a literature datatype score of 0.9852 and a clinical datatype score of 0.1368. The same platform scores CD274 against diffuse midline glioma at 0.0389, and KDR against central nervous system cancer at 0.1354. Open Targets lists 13 distinct drug or clinical candidates against CD274 and 75 against KDR.

What actually happened

The sponsor terminated the study and posted a reason: the study was stopped due to a lack of clear evidence of anti-tumor activity with the combination treatment in the participants treated. No results have been posted to ClinicalTrials.gov for this record, so no response rate, progression-free survival estimate, or subgroup effect is available. This post reports none, because none exists in the source.

The safety picture must therefore be read from post-marketing data rather than from the trial. In openFDA FAERS, avelumab has 2211 total reports, 2132 serious, 412 with a death outcome. Lenvatinib has 15246 total, 13028 serious, 1995 with a death outcome, led by hypertension (1732). These are spontaneous reports across all approved uses and ages with no denominator, so they describe the class rather than this trial.

Failure mechanism, best guess

The most probable mechanism is a translational mismatch dressed as an efficacy question. Pediatric high-grade CNS tumors, and diffuse midline glioma in particular, are characteristically low in tumor mutational burden and sparse in T cell infiltrate. Adult glioma work has shown that response to checkpoint blockade tracks with specific immune and genomic features rather than with the presence of the target (Nature Medicine 2019), and hypermutation in gliomas has distinct mechanisms of its own (Nature 2020). A combination that depends on unleashing pre-existing T cell recognition has little to unleash in a tumor that was never immunologically engaged.

The VEGFR arm carries a second and partly opposing risk. VEGFR inhibition in glioblastoma upregulates CXCR4 through TGF-beta receptor signaling (Cancer Letters 2015), an axis associated with invasive escape rather than immune improvement, and single-agent VEGFR inhibition in recurrent glioblastoma produced limited activity (Journal of Neuro-Oncology 2016). Preclinical work in diffuse intrinsic pontine glioma models points toward TIM-3 rather than PD-1 or PD-L1 as the productive checkpoint (Cancer Cell 2023), which suggests the axis chosen here was wrong for this histology. A recent preprint reports that age-dependent tumor-immune interactions underlie immunotherapy response in pediatric cancer (doi:10.1101/2025.07.16.663652).

How to prevent this next time

No endpoint-level data were posted for this trial, so no Bayesian update, futility boundary, or power calculation can be honestly computed from it. The available levers are qualitative.

Biomarker enrichment is the first. The trial enrolled on histology and performance status, with no requirement for PD-L1 expression, T cell infiltration, or tumor mutational burden. Where Open Targets scores the target-to-disease link at 0.1406, and 0.0389 for the diffuse midline glioma subset, an unenriched single-arm design cannot distinguish a wrong target from a wrong population.

Base-rate adjustment is the second. Claidex now records two CD274 program failures, and Open Targets lists 13 CD274 candidates and 75 KDR candidates. A crowded target with weak disease-specific evidence should start from a low prior on activity in a new histology, not from adult precedent.

A red-team review is the third. The CXCR4 upregulation literature and the TIM-3 preclinical work were both available before this trial completed, and an adversarial reading would have surfaced that the antiangiogenic partner might work against the immune objective here.

The single highest leverage change would have been to gate expansion on a prespecified immune-contexture biomarker, measured in the first six children, so that absent T cell infiltration ended the study in months rather than after four years and 17 participants.

What this means for similar programs

The Claidex Mechanism Risk Score for CD274 in pediatric primary high-grade CNS tumors stands at 34, in the yellow band, from contributions of 4.70 for phase burden, 9.84 for archetype severity, 4.25 for recency, 12.89 for genetic deficit, and 2.50 for saturation. Genetic deficit is the largest single term, which is the signature of a well-drugged target carried into a disease where its evidence is thin. Claidex now holds two CD274 entries, this one and an earlier advanced-solid-tumor failure scored at 23.

Programs pairing an approved checkpoint antibody with an approved kinase inhibitor in pediatric CNS tumors should treat adult combination precedent as insufficient. The relevant question is not whether PD-L1 is present but whether the tumor has an adaptive immune response to release. Sponsors running B7-H3 or GD2 cell-therapy approaches in the same histologies follow a different logic that does not depend on pre-existing T cell recognition, and this termination does not bear against them.

Open questions

Was PD-L1 expression or immune infiltrate measured in any of the 17 participants, and if so, was it uniformly low? Did diffuse midline glioma participants differ from the broader high-grade group? What interim looks were specified before the February 2026 primary completion? Will results be posted, given that the record currently shows none?

Sources

1. • ClinicalTrials.gov API v2, record NCT05081180, accessed 2026-07-30. https://clinicaltrials.gov/study/NCT05081180 - Open Targets Platform GraphQL API v4, target ENSG00000120217 (CD274) and ENSG00000128052 (KDR), associations against MONDO_0002714 and EFO_0020983, accessed 2026-07-30. https://platform.opentargets.org/target/ENSG00000120217 - ChEMBL API, CHEMBL3833373 (avelumab) and CHEMBL1289601 (lenvatinib), accessed 2026-07-30. - openFDA drug event API (FAERS), avelumab and lenvatinib queries, accessed 2026-07-30. - Avelumab in paediatric patients with refractory or relapsed solid tumours: dose-escalation results from an open-label, single-arm, phase 1/2 trial. Cancer Immunology, Immunotherapy, 2022.- A phase I/II study of lenvatinib plus everolimus in recurrent and refractory pediatric solid tumors, including CNS tumors. Journal of Clinical Oncology, 2020.- Neoadjuvant anti-PD-1 immunotherapy promotes a survival benefit with intratumoral and systemic immune responses in recurrent glioblastoma. Nature Medicine, 2019.- Immune and genomic correlates of response to anti-PD-1 immunotherapy in glioblastoma. Nature Medicine, 2019.- Mechanisms and therapeutic implications of hypermutation in gliomas. Nature, 2020.- TIM-3 blockade in diffuse intrinsic pontine glioma models promotes tumor regression and antitumor immune memory. Cancer Cell, 2023.- Phase II study of tivozanib, an oral VEGFR inhibitor, in patients with recurrent glioblastoma. Journal of Neuro-Oncology, 2016.- VEGFR inhibitors upregulate CXCR4 in VEGF receptor-expressing glioblastoma in a TGF-beta receptor signaling-dependent manner. Cancer Letters, 2015.- Age-dependent tumor-immune interactions underlie immunotherapy response in pediatric cancer. bioRxiv preprint, 2025.