Abemaciclib in HPV-negative HNSCC AIM window trial halted after standard-of-care shift to neoadjuvant immunotherapy

What was tried

NCT04169074 (the AIM Trial) was a Phase 2 single-arm window-of-opportunity study at the University of Arizona evaluating short-course neoadjuvant abemaciclib (LY2835219, VERZENIO) in patients with HPV-negative, locally advanced stage III to IVA head and neck squamous cell carcinoma. Patients received 10 to 21 days of abemaciclib 150 mg twice daily before planned curative-intent surgery or definitive chemoradiation. The primary endpoint was quantitative change in tumor size on pre- and post-treatment imaging, with secondary endpoints around CDK4/6-pathway pharmacodynamics in tumor biopsies (Ki-67, retinoblastoma phosphorylation, CDKN2A loss correlates).

Abemaciclib is an FDA-approved CDK4/6 inhibitor for HR-positive HER2-negative breast cancer and high-risk early breast cancer in the adjuvant setting. The AIM design was a translational bet on extending CDK4/6 inhibition into a defined-genomics solid tumor where CDKN2A loss occurs in roughly 70 percent of HPV-negative cases (TCGA HNSCC 2015 analysis). It enrolled 7 of 20 planned patients before termination.

The biological hypothesis

HPV-negative HNSCC is dominated by CDKN2A inactivation and cyclin D1 amplification, both of which converge on hyperactivation of CDK4/6-mediated retinoblastoma phosphorylation and cell cycle entry. Preclinical work by Adkins and colleagues showed that CDK4/6 inhibition by palbociclib produced durable G1 arrest in HPV-negative head and neck cell lines and patient-derived xenografts but minimal effect in HPV-positive lines, where viral E7 sequesters Rb (Michel 2016, Adkins 2018). Subsequent work added that CDK4/6 inhibition modulates the tumor immune microenvironment by upregulating MHC class I and inducing interferon signaling (Goel 2017), generating a coherent rationale for window-of-opportunity testing in the locally advanced setting.

Abemaciclib was the natural choice given its broader kinase profile (CDK9 activity, stronger CDK4) and higher single-agent activity than palbociclib in non-breast tumors. The AIM design used a short-course pre-operative window to read out pharmacodynamics and biomarker change rather than overall response. With 20 patients and paired biopsies the trial was powered to detect a 50 percent reduction in Ki-67 with greater than 80 percent power at alpha 0.05.

What actually happened

Seven patients enrolled between February 2020 and mid-2024. Enrollment slowed once the SARS-CoV-2 pandemic disrupted head and neck oncology workflow, then stalled when the National Comprehensive Cancer Network updated the HNSCC guideline (NCCN v3.2024, June 2024) to incorporate neoadjuvant and adjuvant immune checkpoint blockade in locally advanced disease, following the KEYNOTE-689 readout. Once peri-operative pembrolizumab became standard of care for the AIM eligible population, asking patients to delay definitive treatment for an experimental CDK4/6 window was no longer ethically defensible. The trial was terminated in December 2024 with 7 of 20 patients enrolled (35 percent of target).

Failure mechanism, best guess

This is a strategic_reprioritization archetype driven by an external standard-of-care shift, not by a mechanistic or safety problem with abemaciclib. The CDK4/6 hypothesis remained intact through the trial period, and the seven enrolled patients should still yield interpretable Ki-67 and Rb-phosphorylation biomarker data even if the sample is underpowered for clinical endpoints.

Two structural decisions made the trial vulnerable. The protocol was activated as a single-site investigator-initiated study with a narrow patient pool and no contingency for landscape change. The 24-month accrual window overlapped with expected readouts of peri-operative immunotherapy trials: KEYNOTE-689 had been initiated in 2018 and its readout was widely expected by 2023 to 2024. A conservative feasibility analysis would have flagged that the standard-of-care window was closing.

How to prevent this next time

Two changes would have meaningfully altered the trajectory.

First, a competitive landscape red-team analysis with explicit standard-of-care horizon dating. Before activation, a competitive intelligence read on the trial population should have surfaced KEYNOTE-689, KEYNOTE-412, and KEYCHAIN as imminent readouts likely to redefine neoadjuvant management in locally advanced HNSCC inside the trial's accrual window. A red-team would have estimated the probability that peri-operative immunotherapy would become standard of care within 24 months of activation at greater than 60 percent and recommended either accelerating the timeline, combining abemaciclib with pembrolizumab in the design, or pivoting to a recurrent or metastatic population where standard-of-care shifts were less imminent.

Second, a Bayesian stopping framework with explicit posterior probability of success on pharmacodynamic endpoint, not just clinical endpoints. With a Beta(2, 6) prior for the probability of a clinically meaningful Ki-67 reduction (informed by Adkins 2018 PDX data), interim analysis at n equals 5 with observed responder-fraction 4/5 would have updated the posterior to a probability of success greater than 80 percent on the pharmacodynamic endpoint. Using:

PP(success)=∫θ​P(final success∣θ,interim data)⋅π(θ∣interim data)dθ

the team could have closed early with a positive pharmacodynamic readout before standard-of-care change made the trial unfinishable. The single highest leverage change would have been pre-specifying a Bayesian interim that allows early success-based stopping on a pharmacodynamic endpoint when the standard-of-care landscape is unstable.

What this means for similar programs

CDK4/6 inhibition in HNSCC remains biologically plausible. Negative readouts in metastatic settings (palbociclib plus cetuximab, Adkins 2019) show monotherapy and simple doublets are insufficient, but combinations with PI3K/mTOR, STAT3, or immunotherapy retain mechanistic logic (PMID 38735381, 39791215). The CDK6 Mechanism Risk Score is 47 in the yellow band, driven by accumulated Phase 2 failures and 13 distinct CDK4/6 programs in Open Targets, not by a definitive mechanistic veto.

Investigator-initiated window designs in settings with evolving standard of care should bake in a Bayesian early-stopping rule and a 12-month landscape sensitivity check. For next-generation CDK4-selective inhibitors (atirmociclib, PF-07220060), HNSCC should be entered as a combination with pembrolizumab rather than monotherapy.

The lerociclib EQRx shutdown (NCT05085002) and the AIM termination together suggest single-agent CDK4/6 inhibition has limited room left outside hormone receptor-positive breast cancer. The next round of CDK4/6 trials should default to combination designs.

Open questions

Will the seven enrolled AIM patients yield a biomarker readout on Ki-67 reduction and Rb phosphorylation that informs the broader CDK4/6 in HNSCC question, or is the sample too small for any pharmacodynamic conclusion? Does CDKN2A copy-number loss versus mutational inactivation predict abemaciclib response differently in HPV-negative HNSCC? Will the planned combination of abemaciclib with pembrolizumab in recurrent or metastatic HNSCC (NCT04088032) replicate the AIM mechanistic rationale in a population insulated from neoadjuvant immunotherapy displacement?

Sources

1. • ClinicalTrials.gov NCT04169074 record (status TERMINATED, reason "Standard of care therapy for Stage III-IVA HNSCC updated in June to include neoadjuvant + adjuvant immunotherapy", actual enrollment 7). - NCCN Head and Neck Cancers Guideline v3.2024, June 2024 update incorporating peri-operative immunotherapy. - KEYNOTE-689 (NCT03765918) peri-operative pembrolizumab in locally advanced HNSCC, primary readout 2024. - Michel L et al. PD0332991 (palbociclib) preclinical activity in HPV-negative HNSCC. Oncotarget, 2016.- Adkins D et al. Palbociclib plus cetuximab in platinum-resistant HNSCC. Clinical Cancer Research, 2019.- Goel S et al. CDK4/6 inhibition triggers anti-tumour immunity. Nature, 2017.- TCGA Network. Comprehensive genomic characterization of head and neck squamous cell carcinomas. Nature, 2015.- Adkins D et al. Palbociclib, a Selective CDK4/6 Inhibitor, in Recurrent/Metastatic HNSCC. JCO Precision Oncology, 2025.- ChEMBL CHEMBL3301610 (abemaciclib), small molecule, max phase 4. - openFDA FAERS, abemaciclib reactions through April 2026: 5,004 serious reports, 857 deaths, top events include diarrhoea (n=5,708), fatigue (n=2,322), nausea (n=1,975). - Open Targets v23.12: CDK6 target page, CDK6-HNSCC association score 0.429, CDK4-HNSCC 0.453. Available from: https://clinicaltrials.gov/study/NCT04169074.