Abstract
BACKGROUND: An osteoporotic fragility fracture occurs every three seconds, with a particularly high incidence after the age of 65, reflecting a substantial decline in bone mass. Given the limitations of dual-energy x-ray absorptiometry (DXA) in early-stage bone mineral density (BMD) assessment, we aim to employ ultrasound-based BMD evaluation within community populations to gain a deeper understanding of the age at which bone mass reduction begins and the associated risk factors. METHODS: We conducted a cross-sectional study of 15,052 individuals from routine health check-ups at Beijing Tsinghua Changgung Hospital (2017-2024). BMD was assessed through ultrasound, with body composition measured in 4,999 participants using multi-frequency bioelectrical impedance analysis. Key risk factors were identified via least absolute shrinkage and selection operator (LASSO) regression. Logistic Regression, Support Vector Machine (SVM), Random Forest, and Extreme Gradient Boosting (XGBoost) were used to predict bone mass reduction risk. Models were evaluated with 5-fold cross-validation. Model performance was assessed using the area under the Receiver Operating Characteristic (ROC) curve (AUC). SHAP plots were employed for interpretability. The best model was deployed in a Shiny web application for real-time prediction. RESULTS: Among 15,052 individuals, 55.7% had normal bone mass, 43.0% had osteopenia, and 1.3% had osteoporosis. Bone mass was significantly associated with gender, age, body mass index (BMI), and metabolic markers (P < 0.001). Age increased with decreasing bone mass: normal (43 years), osteopenia (53 years), and osteoporosis (65 years). In 4,999 participants, osteopenia and osteoporosis were linked to higher fat mass index (FMI) and metabolic markers. Group medians in the osteopenia/osteoporosis fell within reference ranges, yet some individuals had values close to either limit. No associations were found between smoking or drinking status and BMD. Bone mass reduction rose sharply from 27.2% to 53.4% between ages 30-59. ROC analysis showed age as a predictor of bone mass reduction with optimal cutoffs at 47 years for males and 49 years for females. LASSO regression identified age, gender, high-density lipoprotein cholesterol (HDL-C), triglycerides (TG), glycated hemoglobin (HbA1c) and FMI as key factors. XGBoost achieved the highest AUC (0.734). Gender-stratified analysis showed that in males, age, HDL-C, FMI, and FBG were significant factors (XGBoost AUC = 0.687), while in females, age, TG, and FMI were key factors (XGBoost AUC = 0.770). CONCLUSION: This study found a high prevalence of bone mass reduction among Chinese adults aged 30-59 years. FMI and age showed significant associations with reduced bone mass. Furthermore, even when HDL-C, LDL-C, TG, and HbA1c were near the reference limits but within normal ranges, their variations were associated with bone mass reduction, which may serve as an early warning indicator. These results underscore the potential utility of community-based bone health monitoring and offer epidemiological insights for comparable aging populations.