Rivoceranib in adenoid cystic carcinoma: RM-202 closed three years after its data, on an endpoint the disease rarely moves

What was tried

Elevar Therapeutics ran RM-202 (NCT04119453), an open label, single arm Phase 2 study of oral rivoceranib 700 mg once daily in recurrent or metastatic adenoid cystic carcinoma of all anatomic sites of origin. The study opened on 22 January 2020 and enrolled 80 participants in the United States and South Korea. Primary completion and study completion were both recorded on 28 June 2023. The record was updated on 30 July 2026 to overall status TERMINATED, with the sponsor citing redirection of the rivoceranib development plan.

The design carried no randomisation, no masking and no control arm. Two co-primary endpoints were specified, both objective response rate by RECIST 1.1, one by investigator and one by blinded independent review committee. Eligibility required documented progression within the preceding 6 months, defined as at least a 20 percent increase in measurable lesions or the appearance of new lesions. Rivoceranib is the free base of apatinib (ChEMBL CHEMBL3186534, small molecule, molecular weight 397.48, synonyms apatinib and YN968D1), recorded in ChEMBL at maximum phase 3.

The biological hypothesis

Rivoceranib inhibits kinase insert domain receptor, KDR, also called VEGFR2 (Ensembl ENSG00000128052). Open Targets classifies KDR within the tyrosine protein kinase VEGFR family and flags approved drug tractability for both small molecule and antibody modalities. Adenoid cystic carcinoma is a slow growing salivary and secretory gland malignancy with a strong tendency to perineural spread and late haematogenous metastasis, and it has no systemic therapy approved by the United States Food and Drug Administration. Angiogenesis blockade was the only class with repeated single agent activity signals in this setting, so a more selective VEGFR2 agent was a reasonable next step.

The genetic support for that step is thin. Open Targets scores the KDR to adenoid cystic carcinoma association at 0.121, built from a literature component of 0.117 and a clinical component of 0.192, with no human genetic datatype contributing. The hypothesis was pharmacological rather than genetic, and the design carried no enrichment biomarker.

What actually happened

Posted results on ClinicalTrials.gov give an intent to treat objective response rate of 13.8 percent (95 percent CI 7.1 to 23.3) by investigator and 8.8 percent (95 percent CI 3.6 to 17.2) by blinded central review, both across all 80 enrolled participants. Median duration of response was 14.9 months (95 percent CI 4.9 to not reached) among 11 investigator assessed responders and 7.2 months (95 percent CI 3.5 to 8.4) among 7 centrally assessed responders. Median time to progression was 9.2 months by investigator and 9.3 months by central review. Progression free survival rates were 74.1 percent at 6 months, 37.3 percent at 12 months and 8.9 percent at 24 months by investigator. Overall survival was 74.7 percent at 12 months and 51.5 percent at 24 months.

Every one of the 80 treated participants had a treatment emergent adverse event and 32 had a serious adverse event. The peer reviewed report of the same study (Hanna et al., Clinical Cancer Research 2023) analysed 72 efficacy evaluable patients and reported an investigator response rate of 15.3 percent, a central rate of 9.7 percent, median progression free survival of 9.0 months by both readers, grade 3 or higher treatment related adverse events in 56 patients (70.0 percent) and hypertension in 34 (42.5 percent). Registry and publication figures differ because the registry reports the intent to treat set of 80 and the publication the efficacy evaluable set of 72.

Failure mechanism, best guess

This trial did not fail on its data. It completed, reported and then sat for three years before the registry status changed, and the stated reason is portfolio redirection rather than safety or futility. The archetype is strategic reprioritisation.

The design question matters more than the termination. A pooled analysis of 17 prospective trials of 10 VEGFR inhibitors in 560 patients with recurrent or metastatic adenoid cystic carcinoma reported an objective response rate of 6 percent (95 percent CI 3 to 12), a stable disease rate of 82 percent (95 percent CI 74 to 87) and a 6 month disease control rate of 54 percent (95 percent CI 45 to 62). Response is close to the noise floor in this disease, and RM-202 chose response as its primary readout without a control arm. The observed 13.8 percent sits above the pooled 6 percent but its confidence interval overlaps the lenvatinib subgroup estimate of 14 percent (95 percent CI 7 to 25) from the same meta analysis, so the result does not separate rivoceranib from the class it belongs to.

Tolerability points the same way. Grade 3 or higher treatment related events in 70.0 percent of patients exceed the pooled class figure of 53 percent (95 percent CI 42 to 64), and the pooled dose reduction rate for the class is 59 percent (95 percent CI 40 to 76). A regimen that requires dose reduction in most patients to deliver a response in roughly one in eight is a hard commercial proposition with no confirmatory Phase 3 in progress.

How to prevent this next time

Endpoint level comparator data do not exist for this single arm study, so the useful levers here are design levers rather than a Bayesian re-analysis.

First, size the study against the published base rate rather than against a generic Simon threshold. Computed here from sourced inputs, a one sided exact binomial test of a null response rate of 6 percent (the Hoff 2024 pooled estimate) against the observed 13.8 percent has 68 percent power at n equal to 80 and reaches 83 percent power at n equal to 90, rejecting at 10 or more responses. The trial as run was underpowered against its own class base rate.

Second, pick an endpoint the disease can move. Stable disease was the modal best response in 82 percent of pooled patients, so 6 month disease control benchmarked to the pooled 54 percent, or progression free survival against a randomised observation arm, carries far more information per patient than response rate.

Third, build an enrichment hypothesis before opening the study. RM-202 used progression within 6 months as its only biological filter, so it could not tell a responder subgroup from chance.

The single highest leverage change would have been to randomise RM-202 against a control arm with progression free survival as the primary endpoint, because in a disease where 82 percent of patients have stable disease as their best response, a single arm response rate cannot distinguish drug effect from natural history.

What this means for similar programs

KDR is a saturated target. Open Targets lists 75 distinct clinical stage or approved agents against it, and the Claidex graph now holds two KDR failures, this one and the lenvatinib plus pembrolizumab Phase 2 in melanoma brain metastases. The recomputed KDR mechanism risk score is 60 of 100, band orange, from phase burden 13.94, archetype severity 11.05, recency 7.25, genetic deficit 13.19 and saturation 15.00. The saturation term is at its ceiling. New KDR programs in rare tumours should assume that the class effect, not the molecule, will dominate the readout.

Open questions

Where the rivoceranib development plan was redirected is not stated in the registry record. Whether the RM-202 responders shared any molecular feature is not reported. Whether 51.5 percent overall survival at 24 months reflects drug effect or the indolent natural history of the enrolled population cannot be answered from a single arm study.

Sources

1. • ClinicalTrials.gov, NCT04119453, RM-202 protocol record and posted results, last update posted 30 July 2026. https://clinicaltrials.gov/study/NCT04119453 - Hanna GJ, Ahn MJ, Muzaffar J, et al. A Phase II Trial of Rivoceranib, an Oral Vascular Endothelial Growth Factor Receptor 2 Inhibitor, for Recurrent or Metastatic Adenoid Cystic Carcinoma. Clin Cancer Res 2023;29(22):4555-4563.- Hoff CO, Manzi J, Lazar Neto F, Ferrarotto R. Vascular Endothelial Growth Factor Receptor Inhibitors for Recurrent or Metastatic Adenoid Cystic Carcinoma: A Systematic Review and Meta-Analysis. JAMA Otolaryngol Head Neck Surg 2024;150(7):587-597.- Wagner VP, Ferrarotto R, Vargas PA, et al. Drug-based therapy for advanced adenoid cystic carcinoma: Current landscape and challenges based on an overview of registered clinical trials. Crit Rev Oncol Hematol 2023.- Galot R, Machiels JH. State of the art and future directions in the treatment of metastatic adenoid cystic carcinoma. Curr Opin Oncol 2026.- Open Targets Platform, target KDR (ENSG00000128052), association with adenoid cystic carcinoma (MONDO_0004971), clinical candidate list, retrieved 1 August 2026. https://platform.opentargets.org/target/ENSG00000128052 - ChEMBL, rivoceranib, CHEMBL3186534, retrieved 1 August 2026. https://www.ebi.ac.uk/chembl/compound_report_card/CHEMBL3186534/ - openFDA drug adverse event endpoint, query on apatinib and rivoceranib as medicinal product, retrieved 1 August 2026. https://api.fda.gov/drug/event.json.