Associations Between Physical Activity, Bone Mineral Density, and Low Bone Mass in U.S. Adolescents Aged 12–19 Years: A SHAP‑Based Machine‑Learning Study of NHANES 2011–2018

Background: Adolescence is a critical period for skeletal development and peak bone mass (PBM) acquisition. Low bone mineral density (BMD) during this stage increases the risk of osteoporosis and fractures later in life. While physical activity (PA) is known to benefit bone health, its dose-response relationship and predictive potential remain insufficiently explored. Methods: This study utilized data from the 2011–2018 National Health and Nutrition Examination Survey (NHANES), including 3,411 U.S. adolescents aged 12–19 years. Associations between PA (measured in MET-min/wk), BMD, and low bone mass were assessed using multivariate regression, restricted cubic spline (RCS) analysis, and machine learning approaches (LASSO, Boruta, AdaBoost). Results: Higher PA levels were nonlinearly associated with increased lumbar spine, pelvic, and total BMD (plateauing at ~ 1,440 MET-min/wk) and linearly associated with reduced risk of low bone mass (10.6% risk reduction per 1,000 MET-min/wk). Subgroup analyses revealed stronger protective effects among adolescents with BMI 25–30 and in certain racial/ethnic groups. AdaBoost outperformed other models in prediction performance, while SHAP (based on the Random Forest model) and LIME identified age, BMI, creatinine, and PA as key predictors. Conclusions: Moderate-to-vigorous PA is associated with improved BMD and reduced risk of low bone mass among adolescents. These findings support individualized, stratified PA interventions and provide a data-driven foundation for adolescent bone health promotion strategies.