Test-Retest reliability and measurement error of the uncontrolled manifold analysis: A step towards the clinical translation

Abstract

The uncontrolled manifold (UCM) analysis has gained broad application in biomechanics and neuroscience for investigating the structure of motor variability in functional tasks. The UCM utilizes inter-trial analysis to partition the variance of elemental variables (e.g., finger forces, joint angles) that affect (VORT) and do not affect (VUCM) a performance variable (e.g., total force, end-effector position). However, to facilitate the translation of UCM into clinical settings, it is crucial to demonstrate the reliability of UCM estimates: VORT, VUCM, and their normalized difference, ΔV. This study aimed to determine the test–retest reliability using the intraclass correlation coefficient (ICC3,K), Bland-Altman plots, the standard error of measurement (SEM), and the minimal detectable change (MDC) of UCM estimate. Fifteen healthy individuals (24.8 ± 1.2 yrs old) performed a finger coordination task, with sessions separated by one hour, one day, and one week. Excellent reliability was found for VORT (ICC3,K = 0.97) and VUCM (ICC3,K = 0.92), whereas good reliability was observed for ΔV (ICC3,K = 0.84). Bland-Altman plots reveled no systematic differences. SEM% values were 24.57 %, 26.80 % and 12.49 % for VORT, VUCM and ΔV respectively, while the normalized MDC% values were 68.12 %, 74.30 % and 34.61 % for VORT, VUCM and ΔV respectively. Our results support the use of UCM as a reliable method for investigating the structure of movement variability. The excellent measurement properties make the UCM a promising tool for tracking changes in motor behavior over time (i.e., effects of interventions in prospective studies).