Sex-Specific Dietary Predictors of Blood Glucose Identified Through Decision Tree Modeling in Adults

Diabetes mellitus is a global public health crisis, with cases projected to rise to 1.3 billion by 2050. Lifestyle interventions are crucial in preventing and managing Type 2 diabetes. This study used a machine learning approach to explore the relationship between dietary components and fasting blood glucose in a young adult population, with a focus on potential sex-specific differences.

This cross-sectional study analyzed data from 288 young adults (195 females, 93 males; mean age 23 years). Participants provided two-day diet records, and their fasting capillary blood glucose was measured. Machine learning was used to predict blood glucose based on a variety of dietary variables, including fiber, macronutrient proportions, and fat types. Energy expenditure was used as a proxy for energy intake. Models were created for the overall population, males, and females.

In the overall population, the most important predictor of fasting blood glucose was fiber intake. For females, the most important predictor was energy expenditure, followed by fat quality (linoleic to alpha-linolenic acid ratio and saturated fat intake). For males, the most predictive factor was the percentage of calories from fat, followed by alpha-linolenic acid intake.

The findings suggest that predictors of blood glucose differ between males and females, highlighting the need for sex-specific strategies in blood glucose management. The models emphasize the importance of increasing fiber intake, maintaining a healthy energy intake, and improving fat quality by prioritizing essential fatty acids. This approach can be used to inform personalized dietary recommendations for the prevention and management of diabetes.