Abstract
Background/Objectives: Differentiating benign, premalignant, and early malignant vocal fold lesions is challenging. Perpendicular vascular changes (PVCs) per the European Laryngological Society (ELS) are key malignancy indicators. Enhanced contact endoscopy with narrow-band imaging (NBI-CE) visualizes intrapapillary capillary loops (IPCLs) at high magnification, independent of gross morphology. However, defining malignancy as any PVC increases sensitivity but lowers specificity-particularly in papillomas-whereas limiting malignancy to narrow-angle PVC improves specificity but risks false negatives and reduced reproducibility. Methods: We intraoperatively evaluated 146 histology-proven vocal fold lesions using NBI-CE. Six raters (three experienced otolaryngologists, three PhD students) classified vascular patterns. Two approaches were tested: (1) malignancy = narrow-angle PVC; (2) malignancy = any PVC. Outcomes were accuracy, sensitivity, specificity, and interrater agreement. Results: Approach (1) had higher specificity but lower sensitivity than (2) (~85% vs. ~70% specificity; ~50% vs. ~80% sensitivity). Accuracy did not differ significantly. Experienced raters showed higher interrater agreement and a more favorable sensitivity-specificity balance. Common errors were false positives in papillomas and false negatives in dysplasia/early carcinoma. Conclusions: PVC assessment with NBI-CE is feasible and informative. Choosing between "any PVC" and "narrow-angle only" entails a sensitivity-specificity trade-off and depends on lesion type and experience. Refined ELS descriptors and automated analysis may improve reproducibility and accuracy.