We sought to determine the clinical meaningfulness of actigraphy metrics as they relate to typical motor symptoms of multiple system atrophy (MSA).

We recruited 18 patients with clinically probable MSA and continuously monitored their physical activity using PAMSys^TM sensors (BioSensics LLC, Newton, MA, USA) during a 12-month, 5-visit study. We applied Spearman’s rank correlations to determine associations between baseline clinical measures and sensor-derived parameters of posture, locomotion, and postural transitions over a one-week period. Additionally, we developed machine learning models to investigate the ability of sensor-derived measures to predict clinical scores.

Total walking time was negatively correlated with tandem walk (rho=-0.705) and Timed Up and Go (TUG) (rho=-0.811), while sedentary time was positively correlated with tandem walk (rho=0.626) and TUG (rho=0.597). We found a negative association between daily step count and walking episodes with tandem walk and TUG. Additionally, we identified positive relationships between average sit-to-stand and stand-to-sit durations with UMSARS-II (rho=0.722, 0.628), the motor section of NNIPPS-PPS (rho=0.690, 0.689), and TUG (rho=0.644, 0.596). Finally, regression models established successful prediction of clinical scores, with TUG demonstrating the highest explained variance.