Machine learning incorporating trabecular bone score improves lumbar fracture risk stratification in type 2 diabetes

Conventional fracture risk assessment tools often fail to accurately predict the risk of lumbar fragility fractures in individuals with type 2 diabetes mellitus (T2DM). A study conducted at Jinjiang Hospital analyzed 261 T2DM patients over the age of 50, finding that a machine learning (ML) model incorporating trabecular bone score (TBS) significantly improves the prediction of lumbar fragility fracture risk.

The model, which included six features such as lumbar spine bone mineral density (LS-BMD), age, body mass index (BMI), gender, and fasting blood glucose, was trained using ten different machine learning algorithms and evaluated using 5-fold cross-validation and an independent test set. The multivariate integrated model performed better than models using either LS-BMD or TBS alone.

Among the ten algorithms, logistic regression emerged as the top performer, with an independent test set AUC of 0.818. The study highlights the crucial role of TBS as a microstructural predictor in assessing lumbar fragility fracture risk in T2DM patients, and demonstrates the potential of ML models to enhance fracture risk stratification in this population. The research was supported by grants from Jinjiang Municipal Hospital and Huaqiao University.