Malignant Transformation Risk in Patients With Oral Leukoplakia and Type 2 Diabetes Mellitus

Oral leukoplakia (OLK) is the most common oral potentially malignant disorder, with a malignant transformation rate of up to 9.8%. Type 2 diabetes mellitus (T2DM) is associated with increasing cancer risk. However, risk factors for malignant transformation in OLK-T2DM patients remain undefined. This study aimed to identify these factors and develop prediction models integrating glycaemic and clinical indicators.

A retrospective nested case-control study was conducted in a real-world OLK-T2DM cohort (2013-2025). Cases (n = 55) with histologically confirmed malignant transformation to oral squamous cell carcinoma were matched 1:2 by age and sex with controls (n = 110) without transformation. Glycaemic control (fasting plasma glucose [FPG], haemoglobin A1c [HbA1c]) was the main exposure. Risk factors were analysed using univariable analyses, multivariable conditional logistic regression, multivariable stratified Cox regression models and machine learning models.

Suboptimal glycaemic control (HbA1c > 7% or FPG > 7.2 mmol/L) was strongly associated with malignant transformation, with each 1 mmol/L increase in FPG raising malignant transformation risk by 60% to 174% according to the results of multivariable stratified Cox regression models for the risk of malignant transformation. Moderate and severe oral epithelial dysplasia (adjusted hazard ratio = 399.43, P = .021) was also an independent predictor. By integrating glycaemic and clinical indicators, machine learning models achieved stable predictive performance (area under the receiver operating characteristic curve up to 0.78; accuracy up to 76.91%).

Suboptimal glycaemic control was independently associated with increased malignant transformation risk in OLK-T2DM patients after adjustment for key confounders. This association was consistent across both regression and machine learning models.

These findings underscore the importance of early monitoring and strict glycaemic management, and suggest that incorporating glycaemic indicators into risk prediction tools may help improve the identification and management of high-risk patients.