Role of gene signatures combined with pathology in classification of oropharynx head and neck cancer.

Treatment personalisation remains an unmet need in oropharynx cancer (OPC). We aimed to determine whether gene expression signatures improved upon clinico-pathological predictors of outcome in OPC. The clinico-pathological predictors, AJCC version 7 (AJCC 7), AJCC 8, and a clinical algorithm, were assessed in 4 public series of OPC (n = 235). Literature review identified 16 mRNA gene expression signatures of radiosensitivity, HPV status, tumour hypoxia, and microsatellite instability. We quality tested signatures using a novel sigQC methodology, and added signatures to clinico-pathological variables as predictors of survival, in univariate and multivariate analyses. AJCC 7 Stage was not predictive of recurrence-free survival (RFS) or overall survival (OS). AJCC 8 significantly predicted RFS and OS. Gene signature quality was highly variable. Among HPV-positive cases, signatures for radiosensitivity, hypoxia, and microsatellite instability revealed significant underlying inter-tumour biological heterogeneity, but did not show prognostic significance when adjusted for clinical covariates. Surprisingly, among HPV-negative cases, a gene signature for HPV status was predictive of survival, even after adjustment for clinical covariates. Across the whole series, several gene signatures representing HPV and microsatellite instability remained significant in multivariate analysis. However, quality control and independent validation remain to be performed to add prognostic information above recently improved clinico-pathological variables.