Abstract
Purpose: Extranodal extension on radiology (iENE) is reported in single-center studies to be negatively prognostic in human papillomavirus-mediated oropharyngeal cancer (HPV + OPC) and is a major eligibility criterion for surgical treatment. However, studies report widely varying sensitivities, specificities, and interobserver correlation. In this research the prognostic power, sensitivity, and specificity of iENE in HPV + OPC in real-world practice are determined. Methods and Materials: A retrospective cohort of 821 consecutive subjects with p16 + OPC, treated with surgery and/or chemoradiation therapy (CRT), from 13 multinational secondary hospitals in 9 countries between January 1, 1999 and December 31, 2020 was analyzed. The main outcomes were sensitivity, specificity, and overall survival (OS). Assessors were blinded to outcomes. Results: Six hundred thirty-eight patients were included in the final analysis. A total of 109 of 394 (27.7%) with no iENE had ENE on histopathology (pENE), and 109 of 192 (56.8%) of patients with pENE were misclassified as having no iENE. iENE sensitivity and specificity were 44.5% (95% CI, 37.8%-51.4%) and 87.6% (95% CI, 84.1%-90.6%), respectively, and varied significantly between centers. Negative predictive value was 75.3% (95% CI, 72.3%-77.5%). Subgroup analyses showed significantly increased sensitivity and specificity if patients had both computed tomography (CT) and magnetic resonance imaging (MRI): 84.6% (95% CI, 65.1%-95.6%, P < .001) and 94.5% (95% CI, 82.3%-99.4%, P = .022), respectively, compared with only CT or MRI alone. Specialist radiologists showed better specificities (89.14%; 95% CI, 85.69%-91.99% vs 46.67%; 95% CI, 21.27%-73.41%, P < .001) and similar sensitivities to nonspecialists. On multivariable analysis, iENE positivity was not a statistically significant independent predictor of OS (adjusted hazards ratio [aHR], 1.50 [95% CI, 0.97-2.32; P = .071]) or disease-free survival (aHR, 1.41; 95% CI, 0.95-2.09; P = .089). Two proposals for amended TNM staging did not yield large improvements. Conclusions: In current real-world practice, iENE showed widely varying and modest accuracy, and was not independently prognostic of outcomes in HPV + OPC. iENE accuracy and prognostic power increased significantly by using combined CT and MRI scanning, experienced head and neck radiologists and more inclusive diagnostic criteria. Validated consensus diagnostic criteria and protocols are urgently needed to enhance the clinical utility of iENE. Until then, clinicians should be cautious about making treatment decisions based on iENE.