Field-Based Assessment of Running Biomechanics Using Global Positioning System/Inertial Measurement Unit:...

Weber, JA, Peeling, P, Hart, NH, and Newton, RU. Field-based assessment of running biomechanics using global positioning system/inertial measurement unit: Establishing construct validity in elite athletes. J Strength Cond Res XX(X): 000-000, 2026-To examine the between-session reliability and construct validity of global positioning system (GPS)/inertial measurement unit (IMU)-derived biomechanical variables during field-based running in elite athletes. Fifty-four professional male Australian Football League players completed repeated 40-m runs across 1 competitive season (3,943 trials). Running speed for the 20-40 m segment was quantified via 10-Hz GPS and center-of-mass displacement metrics derived from a 100-Hz IMU. Between-session reliability (n = 25, mean (k) = 2.95) was assessed using a linear mixed-effects model with velocity as a fixed covariate, expressed as intraclass correlation coefficient (ICC) (absolute agreement), SEM, and minimal detectable change95, with 95% confidence intervals from athlete-cluster bootstrap. Five candidate models were compared; Multivariate Adaptive Regression Splines was selected based on lowest cross-validation root mean squared error (RMSE). Validity was assessed over 100 iterations of athlete-stratified cross-validation (80/20 split), with agreement quantified by ICC, RMSE, and athlete-level Bland-Altman analysis. Velocity-adjusted reliability ranged from moderate to good-to-excellent across 12 variables (ICC = 0.65-0.92), with 8 meeting good-to-excellent thresholds (ICC ≥0.75). Multivariate Adaptive Regression Splines accounted for 93% of variance in running speed (R2 = 0.93; RMSE = 0.20 m·s-1; ICC = 0.96 ± 0.004). Athlete-level Bland-Altman analysis indicated negligible bias (-0.01 m·s-1; LoA: -0.243 to +0.223 m·s-1) with no proportional bias (p = 0.369); limits narrowed in athletes with adequate data volume (n = 47; -0.207 to +0.189 m·s-1). Mechanics-speed relationships persisted after accounting for within-athlete clustering (residual between-athlete ICC = 0.43), and the most influential predictors were consistent with established biomechanical determinants of speed. These findings support GPS/IMU-derived biomechanical variables as field-based indicators of running mechanics. Pending replication and criterion validation, this approach may offer practitioners an accessible framework for monitoring performance and rehabilitation.