Early warning for diabetes: AI model identifies prediabetes risk with high accuracy

By Priyanjana Pramanik, MSc.Reviewed by Susha Cheriyedath, M.Sc.Dec 10 2025

By combining oxidative stress biology with advanced machine learning, researchers show how a simple blood-based antioxidant measure can significantly sharpen prediabetes risk prediction and support earlier, more targeted prevention strategies.

Study: Artificial intelligence model as a tool to predict prediabetes. Image Credit: CI Photos / Shutterstock

In a recent study published in the journal Scientific Reports, researchers developed a pattern neural network (PNN) model that combined a novel measure of total antioxidant status with traditional indicators to improve predictions of prediabetes for Indian adults. The PNN outperformed support vector machines, k-nearest neighbors, and logistic regression models trained on the same dataset, achieving an accuracy of 98.3%. Waist circumference and antioxidant status showed the strongest predictive power, according to model-derived feature importance, with BMI also contributing meaningfully to classification performance.

Growing Need for Accurate Prediabetes Detection

Prediabetes is a critical early stage characterized by elevated blood sugar levels that have not yet progressed to diabetes. Each year, about 5–10% of individuals with prediabetes develop diabetes, while a comparable proportion return to normal glucose levels. Because progression is not inevitable, early detection is essential for preventing type 2 diabetes and its associated long-term complications.

Traditional diagnostic approaches rely on blood-based tests and clinical evaluation, but these methods can be time-consuming, costly, and sometimes limited in their ability to predict individual risk. As data-driven tools have advanced, AI allows researchers to combine data from multiple sources and has emerged as a promising alternative for early disease detection.

AI-based prediction models offer multiple benefits, including higher diagnostic accuracy, individualized risk profiles, and earlier intervention. These advancements could significantly lower healthcare costs by preventing disease progression.

Integrating Oxidative Stress Markers Into AI Models

Researchers developed an AI model specifically optimized for prediabetes prediction using real-world clinical data from Indian adults. Unlike earlier studies, the researchers aimed to identify not only the most accurate model but also one that aligns closely with clinically relevant biomarkers, including oxidative stress indicators that may reflect underlying pathophysiology.

This pilot study included 199 adults aged 18 to 60, classified as either prediabetic (n = 100) or healthy controls (n = 99) based on glycated hemoglobin (HbA1c) levels. After an overnight fast, 6 mL of peripheral blood was drawn. Biochemical tests included HbA1c, fasting glucose, and lipid profile measurements using standardized enzymatic assays. High-density lipoprotein (HDL), low-density lipoprotein (LDL), and very low-density lipoprotein (VLDL) values were calculated.

A key addition to this dataset was the measurement of total antioxidant status, with antioxidant activity expressed as percentage scavenging potential. Healthy individuals typically show 20–60% of the total.

A total of 14 features, including demographic, clinical, biochemical, and oxidative stress markers, were used to train a Pattern Neural Network with 14 input nodes, 10 hidden nodes, and one output node. Data were randomly partitioned into training, validation, and test sets, followed by preprocessing steps such as normalization, outlier removal, and handling missing values. Model performance was compared with other AI models and logistic regression. Pearson correlation and descriptive statistics were used to examine relationships among variables and assess feature relevance prior to model training.

Key Biomarkers Distinguishing Prediabetes Profiles

Of the 14 measured variables, six showed significant differences between individuals with and without prediabetes: age, body mass index (BMI), waist circumference, antioxidant activity, oral glucose tolerance test (OGTT), and HbA1c levels. Individuals with prediabetes had notably lower antioxidant capacity, indicating higher oxidative stress, and displayed higher values for key metabolic indicators such as HbA1c and glucose responses.

Boxplot analyses reinforced these group differences by revealing distinct distributions for HbA1c, OGTT, and lipid markers, with fasting glucose showing some distributional differences despite the group comparison not reaching statistical significance. Some parameters showed positively skewed distributions, suggesting clustering of abnormal values in the prediabetes group. Correlation testing highlighted moderate associations between BMI and waist circumference, and modest associations between anthropometric measures and fasting glucose, which together capture overlapping but non-redundant aspects of metabolic risk.

PNN Model Demonstrates Superior Predictive Accuracy

The PNN model trained on these variables demonstrated highly accurate classification. It achieved 97.9% accuracy on the training set and 95.2% on both the testing and validation sets. Overall accuracy across all datasets was 98.3%, with perfect precision, strong recall, and F1 Scores. Compared with other models, the PNN consistently outperformed the alternatives, achieving the highest area under the curve (AUC) and the strongest error minimization.

Implications for Early Risk Stratification

This study successfully integrated total antioxidant status into an AI-based prediabetes prediction model for an Indian population, highlighting oxidative stress as an important and often overlooked risk marker with potential mechanistic relevance to disease development rather than merely a correlational feature.

The findings confirm that waist circumference, BMI, glucose markers, and antioxidant capacity are among the most informative predictors, aligning with evidence from other populations. The PNN delivered superior accuracy compared with traditional machine-learning models and demonstrated strong potential as a rapid, low-cost screening tool pending external validation in independent cohorts.

Strengths include the comprehensive set of biochemical and clinical features and the introduction of oxidative stress measures, which add biological depth to risk assessment. However, the single-center design, modest sample size, and cross-sectional nature limit generalizability and the ability to track changes over time.

Overall, the PNN offers a robust framework for early detection and risk stratification in prediabetes. Future research should validate the model in larger, multi-site cohorts and explore integration with longitudinal clinical data for prospective clinical and public-health applications while formally assessing real-world feasibility and performance stability.