Clinical validation of a mesenchymal gene expression signature for prognosis and treatment decision making in early-stage oral cavity squamous cell carcinoma.

6091 Background: Treatment for human papillomavirus (HPV)-negative oral cavity squamous cell carcinoma (OCSCC) includes surgical resection of the primary tumor, usually with elective or therapeutic neck dissection with adjuvant radiation or chemotherapy and radiation based on risk stratification. While therapeutic advances have been modest at best for over a century, the addition of chemotherapy to radiation has demonstrated benefit for the highest risk patient groups, although recurrence rates remain high for advanced-stage patients. Even more concerning, some early-stage patients demonstrate life-threatening relapses in the absence of biomarker guidance. Using retrospective analysis, our group and others have identified gene expression signatures that reveal molecular subtypes (classical, atypical, basal and mesenchymal) that could identify early-stage patients most likely to recur and are likely candidates for therapy escalation, such as chemotherapy plus radiation. We hypothesize that the mesenchymal (MS) gene expression subtype is associated with worse overall survival (OS) in early-stage, node-negative OCSCC compared to non-mesenchymal subtypes. Herein we present further clinical validation of our previously described 88-gene OCSCC classifier (Mayhew et al, 2022; PMID: 36147910). Methods: As part of a prospectively designed retrospective study, residual tumor samples and paired clinical data, including n-stage and survival, were collected for patients with node-negative OCSCC (IRB# 201706088 at Washington University in St. Louis). Using the NanoString nCounter platform, RNA was isolated and counts were generated according to the manufacturer’s specifications. Gene expression data were used to assign molecular subtypes, and associations with overall survival (OS) were investigated using Kaplan-Meier plots and cox models in patients who were pathologically node negative. Results: A total of n=137 patients had node-negative OCSCC and sufficient tumor quantity to generate molecular subtype results. Consistent with our previous findings, the MS subtype was strongly associated with inferior survival in both univariate (HR = 3.24, p = 0.0019) and multivariable (HR = 3.26, p = 0.0026) analyses, confirming a potential role for gene expression-based subtyping as a biomarker for prognostication and informing clinical treatment decisions. Conclusions: These findings confirm that the MS subtype is associated with significantly worse survival in surgically resected, early-stage OCSCC which typically demonstrate favorable outcomes. Also, these findings strongly suggest value in gene expression subtyping for prognostication and treatment decision-making in OCSCC patients and warrant the further development of the 88-gene classifier as a diagnostic test for this clinically actionable patient population.