A prognostic gene expression signature for oropharyngeal squamous cell carcinoma

Background: Robust prognostic stratification of patients with oropharyngeal squamous cell carcinoma (OPSCC) is important for developing individualized treatment plans. This study was conducted to develop and validate a clinically feasible prognostic classifier based on transcriptome-wide gene expression profiles. Methods: Tumor tissues were collected from 208 OPSCC patients treated at Washington University in St. Louis and 130 OPSCC patients treated at Vanderbilt University, used for model training and validation, respectively. OPSCC patients (n = 70) from the TCGA cohort were also included for independent validation. Based on RNA-seq profiling data, Cox proportional hazards regression analysis was performed to identify genes associated with disease outcomes. Then, Lasso-penalized multivariate survival models were constructed to identify biomarker genes for developing a prognostic gene signature. Findings: A 60-gene signature was identified by RNA-seq profiling analysis. Computed risk score of the gene signature was significantly predictive of 5-year overall survival of the training cohort (Hazard ratio (HR) 28·32, P = 4·3E-41). Subgroup analysis stratified by HPV status revealed that the signature was prognostic in HPV-positive OPSCC patients (HR 30·55, P = 7·0E-37) and was independent of clinical features. Importantly, the gene signature was validated in two independent patient cohorts, including the TCGA cohort (HR 3·94, P = 0·0018) and the Vanderbilt cohort (HR 8·50, P = 5·7E-09) for overall survival. Interpretation: The prognostic gene signature is a robust tool for risk stratification of OPSCC patients. The signature remains prognostic among HPV-positive OPSCC patients. Funding: National Institutes of Health.