Quantifying impact loading for osteoporosis prevention: a study on the relationship between ground reaction forces...

High-impact exercise can strengthen bones, but clinical use is limited because mechanical loading is difficult to measure outside laboratories. We validated a wrist-worn accelerometer against force-plate ground reaction forces and found that wearable data reliably reflect impact loading. This approach may enable practical monitoring of bone-stimulating exercise in real-world settings.

Mechanical loading must surpass a minimum intensity threshold to stimulate osteogenic adaptation, yet clinicians lack practical tools to quantify impact dose outside laboratory settings. Wearable accelerometers could enable safe, scalable monitoring of bone-strengthening exercise, but their validity for estimating osteogenic loading remains uncertain. This study examined whether wrist-worn accelerometry can support approximate estimation and coarse classification of platform-derived impact loading during maximal countermovement jumps in postmenopausal women.

Thirty-eight women were enrolled; 37 contributed 125 valid CMJs with concurrent force-platform and wrist-accelerometer recordings. The primary model predicted peak platform-derived vertical acceleration from peak wrist vector-magnitude acceleration and body mass. Coefficient inference used linear regression with participant-clustered robust standard errors. Predictive performance and four-class osteogenic classification (< 2 g, 2- < 3 g, 3- < 3.9 g, ≥ 3.9 g) were evaluated using leave-one-subject-out cross-validation. Fuller anthropometric models and residual-based outlier filtering were examined as sensitivity analyses.

In the primary reduced model, wrist vector magnitude and body mass were both independently associated with peak platform acceleration. In leave-one-subject-out validation, the model explained 44.3% of the variance (R² = 0.443), with MAE = 5.83 and RMSE = 7.99 m·s⁻². Mean bias was 0.05 m·s⁻², with 95% limits of agreement from - 15.66 to 15.77 m·s⁻². Four-class classification accuracy was 0.544, with quadratic weighted κ = 0.639; misclassifications occurred predominantly between adjacent bins. The fuller anthropometric model did not improve out-of-sample performance.

Wrist-worn accelerometry provided limited precision for per-jump estimation of platform-derived loading, even under constrained laboratory conditions. However, it retained moderate utility for coarse, bin-based classification of impact intensity during maximal CMJs performed with hands on hips. Its most defensible application is session-level monitoring and conservative progression in home- and community-based osteogenic exercise programs, rather than replacement of laboratory force-platform measurement.