AI is increasingly shaping how clinicians identify people at risk of chronic disease, and new research suggests it may significantly improve how prediabetes is detected. In a recent study involving Indian adults, researchers developed an advanced AI-based model that predicts prediabetes with high accuracy by incorporating markers of oxidative stress alongside conventional clinical risk factors.
Prediabetes is a metabolic state in which blood glucose levels are elevated but not yet high enough for a diagnosis of diabetes. Because many individuals with prediabetes progress to type 2 diabetes without intervention, early identification is considered critical for prevention. However, existing prediction models often rely only on routine clinical and biochemical measures, which may miss important underlying biological processes.
Understanding Prediabetes Prediction Through AI
The study enrolled 199 adults aged 18 to 60 years and classified them into control and prediabetes groups using HbA1c levels. Researchers trained a novel Pattern Neural Network, a type of artificial intelligence model, using 14 input variables. These included age, sex, BMI, waist circumference, lipid levels, glucose measures, haemoglobin, and total antioxidant scavenging potential, a marker reflecting the body’s ability to counter oxidative stress.
Oxidative stress has long been implicated in the development of insulin resistance and metabolic disease, but it has rarely been integrated into AI-based risk models. By combining antioxidant status with established metabolic indicators, the researchers aimed to capture a more complete picture of prediabetes risk.
High Accuracy and Clinical Relevance
The AI model demonstrated a validation accuracy of 98.3%, outperforming commonly used machine learning approaches such as support vector machines, k-nearest neighbours, and logistic regression. Among all variables analysed, antioxidant scavenging potential and waist circumference emerged as the most influential predictors of prediabetes risk.
The model generated a risk score that reliably distinguished individuals at increased likelihood of progressing toward diabetes. According to the researchers, this level of accuracy suggests the approach could be clinically useful for early screening, particularly in populations at high risk of diabetes.
Why Antioxidant Status Matters
The findings support the growing view that oxidative stress plays a meaningful role in early metabolic dysfunction. Lower antioxidant capacity may reflect an imbalance that contributes to impaired glucose regulation, even before overt diabetes develops. Including this biological signal appears to enhance the model’s predictive performance beyond standard laboratory measures alone.
This research represents the first AI-based prediabetes prediction model in an Indian cohort to explicitly link antioxidant status with diabetes risk. If validated in larger and more diverse populations, the approach could support earlier identification of at-risk individuals and enable more targeted lifestyle or clinical interventions.
Reference
Yesupatham A et al. Artificial intelligence model as a tool to predict prediabetes. Sci Rep. 2025;15(1):43421.
