The 5-years overall survival (OS) for OSCC is 60%. TNM staging is group-based and it is limited in tailoring personalized therapy. GF on primary tumor is a mature method to identify prognostic markers, but is limited by availability of surgical specimen. RF, based on techniques being already part of the diagnostic phase, is feasible in all pts, but techniques are not yet defined. We recently defined a prognostic 48-genes signature (48-GS) in OSCC. Association and integration of GR and RF could pave the way to non-invasive assessment of prognostic models. We considered the series of OSCC pts with clinical stage III/IV from 5 European centers enrolled in the ongoing BD2Decide project.
Primary tumors were analyzed for gene expression by ClariomD (Affymetrix). Several clinical and GF were tested and their prognostic value was compared to TNM. Magnetic resonance imaging (MRI) radiomics was evaluated on diffusion weighted images acquired at different b-values (0, 50, 100, 500, 1000) and on ADC maps. A total of 641 features were computed on tumor volume, including First Order Statistics (FOS), Textural and Shape-and-size RFs. A cohort of 87 pts with available data for both GF and MRI RF was analyzed. According to 48-GS, pts were divided in 2 classes: high and low risk. Statistical difference of RF was evaluated in the 2 classes using Mann-Whitney test with FDR correction.
The 48-GS separated the OSCC OS significantly better than TNM (p < 0.001). When challenged against available demographic and clinical factors, this model retained a significant independent association in multivariate Cox regression analysis. Five RF of the FOS group computed on fast ADC map were statistically different between the two 48-GS classes (p < 0.05). External validation of association between RF and GF is ongoing.
We identified a prognostic model based on the 48-GS able to improve assessment of stage III/IV OSCC risk based on OS. Since significantly altered RFs mainly describe the variability of the intensity signals inside the tumor, their association with 48-GS could mirror biological differences.
This article is authored also by Synbrain data scientists and collaborators. READ THE FULL ARTICLE