Predicting unreported clinical trial efficacy using an ex-vivo tumor platform: Pembrolizumab+pt+5-FU vs. extreme in recurrent head and neck squamous cell carcinoma.

Background: PD-1 inhibition has shown promise in second-line, platinum (Pt) resistant HNSCC (e.g. KEYNOTE-055). To improve the outcome of PD-1 blockade, clinical trials are testing immunotherapy combinations such as, KEYNOTE-048, an unreported trial combining a PD-1 inhibitor (Pembrolizumab) with Pt/5-fluorouricil (5-FU) vs. cetuximab/Pt/5-FU (EXTREME) as frontline therapy for recurrent or metastatic HNSCC. A lack of preclinical models to profile immunotherapy has made efficacy prediction difficult. Methods: Using fresh tumor biopsies from HNSCC patients with metastatic or locally advanced disease (N = 39), we used CANscript, a clinically-validated ex-vivo tumor platform that preserves the native patient tumor and immune contexture. We tested three treatment schedules: Pembro, Pembro/Pt/5-FU, cetuximab/Pt/5-FU; reflecting treatment arms from the yet unreported KEYNOTE-048. Using phenotypic reflex to drug, ex-vivo, paired with a proprietary algorithm, we predicted clinical response (M-Score). We used PD-L1 status (IHC) and gene expression profiling to understand mechanistic changes following therapy. Results: We predicted, using M-Score, an ORR of 12.8% with Pembro, correlating to results from KEYNOTE-040 (14.6%). The EXTREME regimen showed 30.7% ORR, consistent with NCT00122460 data (36%). We predicted no benefit with Pembro/Pt/5-FU vs. EXTREME (30.7% vs. 30.7%). We also profiled ORR based on Pt type (Carboplatin vs. Cisplatin) with greater ORR from Pembro/Carbo/5-FU (43.7%) vs. Pembro/Cis/5-FU (21.7%). In contrast, we predicted EXTREME alone results in similar ORR regardless of Carbo or Cis therapy (31.3% vs. 30.4%). Conclusions: These results demonstrate the ability of the CANscript system to profile clinical response to immunotherapy. Our efforts may better inform clinical trial design thereby increasing the potential for trial success. If prediction for response to the novel combination in KEYNOTE-048 correlates, this would provide the first evidence that a human tumor model can be used to profile immunotherapy response, while also providing translational insights into mechanisms of response and resistance.