Risk Assessment of Oral Leukoplakia by Individual Dysplasia Features

• Published in: Journal of oral pathology & medicine Vol. 54; no. 7; pp. 537 - 546
• Main Authors: Soares, Andresa Borges, Warnakulasuriya, Saman, Silveira Terra Junqueira, Letícia, Freitas, André Luis Santana, Schneider, Amanda, Silva, Alan Roger Santos, Sperandio, Marcelo
• Format: Journal Article
• Language: English
• Published: Denmark Wiley Subscription Services, Inc 01.08.2025
• Subjects: Adult; Aged; Aged, 80 and over; Cell Transformation, Neoplastic - pathology; Dysplasia; epithelial dysplasia; Female; Humans; Inflammation; Leukokeratosis; Leukoplakia, Oral - pathology; machine learning; Male; Middle Aged; Mouth Mucosa - pathology; Mouth Neoplasms - pathology; Nucleoli; OPMD; oral cancer; oral leukoplakia; Precancerous Conditions - pathology; random forest; Risk Assessment; oral leukoplakia; random forest; machine learning; OPMD; epithelial dysplasia; oral cancer
• ISSN: 0904-2512; 1600-0714; 1600-0714
• DOI: 10.1111/jop.13633

ABSTRACT Objectives Malignant transformation (MT) risk assessment in oral leukoplakia (OL) relies on tissue sample and oral epithelial dysplasia (OED) grading. The aim of this study was to evaluate the role of each OED feature in predicting MT in OL. Subjects and Methods Ninety‐nine OL patients were selected (81 non‐transforming and 18 with MT). All OED features were individually scored for each case. Follow‐up data were obtained from both local and regional cancer registries. Spearman correlation, logistic regression, and Kaplan–Meier were tested with MT as outcome. Thresholds for number of features indicating risk and predictive values (PV) were calculated. A random forest (RF) model was built to assess the relevance of each feature. Results Loss of epithelial cohesion, increased nucleoli, and inflammation were the associated with MT (p ≤ 0.016). Combining these three features yielded high specificity (95%) and PV (50% positive and 85% negative). Using a six‐feature threshold to establish risk reached 54%, 26% and 90%, and 72%, respectively. Specifying and counting OED features proved crucial to establishing risk. Conclusion Loss of epithelial cohesion, increased size and number of nucleoli, and inflammation are key risk features and sum of OED features is the most useful predictor of MT.