Prediction of the Risk of Bone Mineral Density Decrease in Type 2 Diabetes Mellitus Patients Based on Traditional...

Area of Science:

• Endocrinology
• Gerontology
• Data Science

Background:

• Type 2 diabetes mellitus (T2DM) is associated with an increased risk of bone mineral density (BMD) loss.
• Current risk assessment models for BMD decreases in T2DM patients often lack integration of body composition data.
• There is a need for advanced predictive tools to identify T2DM individuals at higher risk of osteoporosis.

Purpose of the Study:

• To develop and compare machine learning (ML) models with traditional logistic regression for predicting BMD decreases in T2DM patients over 50 years of age.
• To incorporate body composition parameters into ML models for enhanced risk prediction.
• To evaluate the performance and clinical applicability of ML-based nomograms versus conventional methods.

Main Methods:

• A cross-sectional study of 450 T2DM patients (aged >50) divided into normal and decreased BMD groups.
• Development of prediction models using traditional multivariate logistic regression and six ML algorithms (including XGBoost).
• Construction of nomograms for male and female patients, with performance evaluated using Area Under the Curve (AUC) and decision curve analysis.

Main Results:

• The Extreme Gradient Boost (XGBoost) model demonstrated superior performance.
• XGBoost achieved higher AUCs (0.800 for males, 0.880 for females) compared to logistic regression (0.722 for males, 0.876 for females).
• Decision curve analysis indicated greater net benefits using XGBoost models for risk prediction in both sexes.

Conclusions:

• XGBoost models show higher accuracy and clinical utility in predicting BMD decreases in T2DM patients over 50 compared to traditional logistic regression.
• The developed ML models offer a promising tool for personalized risk assessment of bone loss in this population.
• Further validation in larger prospective studies is recommended to confirm these findings.