RESUMEN
OBJECTIVES: The majority of patients with locally advanced larynx or hypopharynx squamous cell carcinoma are treated with organ-preserving chemoradiotherapy (CRT). Clinical outcome following CRT varies greatly. We hypothesized that tumor microRNA (miRNA) expression is predictive for outcome following CRT. METHODS: Next-generation sequencing (NGS) miRNA profiling was performed on 37 formalin-fixed paraffin-embedded (FFPE) tumor samples. Patients with a recurrence-free survival (RFS) of less than 2 years and patients with late/no recurrence within 2 years were compared by differential expression analysis. Tumor-specific miRNAs were selected based on normal mucosa miRNA expression data from The Cancer Genome Atlas database. A model was constructed to predict outcome using group-regularized penalized logistic ridge regression. Candidate miRNAs were validated by RT-qPCR in the initial sample set as well as in 46 additional samples. RESULTS: Thirteen miRNAs were differentially expressed (p < 0.05, FDR < 0.1) according to outcome group. Initial class prediction in the NGS cohort (n = 37) resulted in a model combining five miRNAs and disease stage, able to predict CRT outcome with an area under the curve (AUC) of 0.82. In the RT-qPCR cohort (n = 83), 25 patients (30%) experienced early recurrence (median RFS 8 months; median follow-up 42 months). Class prediction resulted in a model combining let-7i-5p, miR-192-5p and disease stage, able to discriminate patients with good versus poor clinical outcome (AUC:0.80). CONCLUSION: The combined miRNA expression and disease stage prediction model for CRT outcome is superior to using either factor alone. This study indicates NGS miRNA profiling using FFPE specimens is feasible, resulting in clinically relevant biomarkers.