MON-600 Title: liver enzymes as predictive biomarkers for type 2 diabetes: a machine learning Approach

Abstract Disclosure: R. Ali: None. S. Sahu: None. S. Cherukuri: None. S. iftikhar: None. P. Chandra: None. S. Koppula: None. F. Safa: None. S. Vr: None. A. Palakurthi: None. Introduction: Type 2 Diabetes Mellitus (T2DM) represents a significant health challenge in India, adversely affecting liver health and contributing to Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD). Individuals with T2DM often present with elevated levels of liver enzymes, including Aspartate Aminotransferase (AST), Alanine Aminotransferase (ALT), and Gamma-Glutamyl Transpeptidase (GGT). These elevated enzyme levels are linked to hepatic steatosis and an increased risk of cardiovascular complications. However, their effectiveness as predictors of T2DM progression remains uncertain. This study utilizes machine learning techniques to evaluate these enzymes as potential early biomarkers for assessing the risk of T2DM. Methods: A retrospective analysis was performed on a dataset of 543 patients, including those with T2DM and healthy controls. Key metabolic indicators such as BMI, Waist-to-Hip ratio, lipid profiles, and fasting glucose levels were evaluated alongside liver enzymes (AST, ALT, GGT). Independent sample t-tests compared means, while logistic regression analyzed predictive power. A Random Forest model was constructed using Scikit-Learn (Python), applying an 80-20 train-test split. Model performance was assessed with Accuracy, ROC AUC score, Precision, and Recall metrics. Results: Patients with T2DM showed significantly higher levels of AST (p < 0.0001), ALT (p < 0.001), and GGT (p < 0.001) compared to controls. Fasting glucose was the strongest predictor of risk (p < 0.0001; odds ratio [OR]: 1.40), while liver enzymes also contributed to risk assessment. The Random Forest model achieved an impressive Accuracy of 98.2% and an ROC AUC score of 98.1%, highlighting the predictive power of combining metabolic and liver biomarkers. Discussion: These findings showcase the connection between elevated liver enzymes and T2DM complications, supporting their use as early biomarkers. Given the prevalence of MASLD in T2DM patients, incorporating liver enzyme levels in routine diabetes screenings could enhance early detection. The absence of a standardized, cost-effective risk assessment tool indicates a need for further validation through more extensive cohort studies. Conclusion: This study suggests that liver enzyme biomarkers can assist in early diabetes detection beyond standard metabolic indicators. Elevated liver enzymes may signal metabolic health issues to clinicians, and the high accuracy of machine-learning models that include hepatic markers provides a valuable opportunity for non-invasive diabetes risk screening. Future research should focus on refining these predictive models, expanding dataset sizes, and validating findings across diverse populations. Presentation: Monday, July 14, 2025