Tumor Inflammation Confounds Radiographic Determination of Extranodal Extension in HPV-Associated Head and Neck...

Background/Objectives: Pathologically identified extranodal extension (pENE) predicts outcome and guides the intensity of adjuvant therapy in HPV-associated HNSCC. ENE determined by imaging (iENE) is a criterion for upstaging HPV+ HNSCC in the AJCC Cancer Staging System, 9th edition; however, its prognostic value is less than that of pENE, varies between studies, and the correlation between pENE and iENE is less than optimal. This study explores biological characteristics of tumors that can decrease the predictive value of iENE. Methods: Pre-treatment CT scans from patients with HPV-associated HNSCC were evaluated for iENE status. RNA-Seq data from primary tumor samples were analyzed using edgeR v4.6.2 and transcriptional differences were categorized by ssGSEA. Machine learning was applied to refine gene expression changes associated with iENE and survival was analyzed with log-rank tests. Results: Recurrence-free survival did not differ between patients with and without iENE. Both ML-ENE status and enrichment for inflammatory signaling were associated with improved RFS (p < 0.01 and p < 0.0001). When stratified by the ML-ENE signature or the inflammatory signature, iENE status did not further separate outcomes, indicating that its prognostic information was largely captured by these gene sets. These stratified comparisons were exploratory and limited by subgroup size. Conclusions: These analyses suggest that tumor biology associated with an inflamed tumor microenvironment may confound determination of ENE by imaging and that accounting for tumors with inflammatory signaling could improve the correlation of iENE with outcomes.