CLINICAL SUBTYPE, LESION SIZE AND ANATOMICAL SITE AS DETERMINANTS OF ORAL EPITHELIAL DYSPLASIA IN LEUKOPLAKIA: A CROSS-SECTIONAL STUDY OF 330 PATIENTS FROM AN ARECA-NUT-ENDEMIC REGION OF SOUTH INDIA

Abstract

Background: Oral leukoplakia is the commonest oral potentially malignant disorder, and epithelial dysplasia remains its strongest histological predictor of progression. Whether chairside clinical features can identify lesions likely to harbour dysplasia has direct implications for biopsy prioritisation in resource-constrained settings. Objectives: To identify independent clinical determinants of dysplasia in oral leukoplakia, quantify grading reproducibility, and test whether lesion laterality corresponds to habitual betel quid placement side. Methods. Cross-sectional study of 330 consecutive patients with histopathologically confirmed oral leukoplakia at a tertiary dental hospital in Chennai, India, between January 2024 and July 2026. Lesions were graded using the WHO 2017 three-tier and WHO 2022 binary systems, with a random 20% subsample (n = 66) independently regraded by a second oral pathologist. Multivariable logistic regression identified independent predictors. Reporting follows STROBE. Results: Median age was 56 years (IQR 48–63); 251 patients (76.1%) were men. Dysplasia was present in 133 lesions (40.3%) and high-grade dysplasia in 40 (12.1%). Independent predictors of dysplasia were non-homogeneous subtype (adjusted odds ratio [aOR] 2.45, 95% CI 1.32–4.52), lesion size per 10 mm (aOR 2.43, 1.64–3.61), tobacco use (aOR 2.21, 1.16–4.23) and age per decade (aOR 1.58, 1.23–2.02); area under the curve (AUC) 0.805. For high-grade dysplasia, lesion size (aOR 3.73, 2.00–6.97), non-homogeneous subtype (aOR 3.28, 1.24–8.70) and age (aOR 1.96, 1.31–2.93) were independent (AUC 0.867). The crude association of tongue or floor-of-mouth location with dysplasia (odds ratio 3.49) attenuated to non-significance after adjustment for size and subtype. Agreement was substantial (weighted κ 0.80 three-tier; κ 0.77 binary). Among 94 patients with a unilateral lesion and lateralised habit, the lesion arose on the quid placement side in 90 (95.7%; κ = 0.91). Conclusions: Lesion size and clinical subtype, not anatomical site per se, independently predict dysplasia in this population. Site appears to act largely as a proxy for local carcinogen contact. Clinical assessment stratifies risk usefully but does not discriminate sharply enough to defer biopsy.