Estimating running economy using wearable trunk accelerometry: a comparative analysis of three footwear conditions in highly trained runners

Wearable accelerometry offers a practical methodology for analysing running biomechanics, a key determinant of running economy (RE). This study investigated whether accelerometer-derived features could estimate RE in highly trained runners. Eighteen male participants (age: 26.5 ± 3.4 years; body mass index: 21.1 ± 1.2 kg/m2; 5 km personal best: 15 min 26 s ± 1 min) completed one laboratory visit, with RE assessed during 5-min treadmill runs at 12 km/h and 16 km/h across three shoe conditions. Metabolic parameters were quantified via indirect calorimetry, while a tri-axial accelerometer measured biomechanical features. A data-driven two-step variable-selection procedure was combined with a linear mixed model to estimate RE from accelerometer data. Estimation accuracy was assessed with mean absolute error (MAE) and Bland-Altman plots. A three-way repeated-measures ANOVA was used to compare estimated and actual RE across speed and shoe conditions. At 12 km/h, six biomechanical variables explained 35.4% of RE variance, whereas at 16 km/h, two variables accounted for 35.1%. While estimated and actual RE did not significantly differ at either speed (p = 1.00), this does not demonstrate equivalence. Relative MAE was 3–4% and Bland-Altman plots showed wide relative limits of agreement (~9–10%), indicating limited individual-level accuracy. Accelerometry could estimate RE within this sample on a group-level, but its individual-level accuracy and sensitivity to footwear effects are limited.