Can we share data? - Kinematic consistency during walking in three different treadmill-based laboratories

Abstract

Background: Three-dimensional gait analysis is crucial for diagnosis and treatment planning. Treadmill-based laboratories efficiently collect 3D gait data over many consecutive steps. Pooling/sharing data across treadmill-based laboratories could enhance clinical utility. However, the inter-laboratory consistency of gait kinematics from treadmill-based systems is unknown. Research question: How consistent are lower-limb kinematics of healthy subjects measured in three different treadmill-based gait laboratories? Methods: Eighteen volunteers (14 women; 27 +/- 9 years; BMI 24 +/- 3 kg/m(2)) walked in three treadmill-based laboratories (Motek Medical, The Netherlands) within one week. Per laboratory, participants completed 3-minute walking trials (0.9, 1.1, 1.3 m/s) wearing a non-weight-bearing harness and identical clothes and shoes. The same marker-set (Human-Body Model 2) and virtual reality configurations were used. Statistical Parametric Mapping was used to compare time-normalized kinematic curves of the lower-limb, averaged over 40 steps, between laboratories. Root mean square differences (RMSD) calculated over periods of the gait cycle with statistically significant differences were considered clinically meaningful when > 5 degrees. Results and significance: Kinematics curves from all laboratories followed similar patterns. Only 17 % of all curves displayed clinically relevant differences. These differences included more knee flexion in laboratory 2 compared to the others (RMSD 6.0-8.6 degrees) and less hip flexion in laboratory 3 compared to laboratory 2 (all speeds) and to laboratory 1 (1.3 m/s; RMSD 5.4-6.4 degrees). Reported differences are likely due to varying operator protocols rather than to the measurement system. The findings indicate that inter-laboratory data sharing using such infrastructure is possible but training to align protocols is essential.