Wearable tracking of walking and non-walking as progression markers in early Parkinson’s disease
Abstract
IMPORTANCE
Wearable-based measures of walking (as proxy for physical activity) may quantify disease progression and modification thereof in early-stage Parkinson’s disease (PD).
OBJECTIVES
Establishing the validity of digital measures of walking and non-walking in PD.
DESIGN
Retrospective longitudinal analyses of data from cohorts within 3 larger studies, consisting of wearable sensor, demographic, and clinical data collected during 2017-2023, with 1-2 year follow up.
SETTING
Three independent multicenter cohort studies.
PARTICIPANTS
People with PD, and age/sex matched non-PD cohort.
EXPOSURES
None.
MAIN OUTCOMES AND MEASURES
Digital measures’ test-retest reliability, analyzed using intraclass correlation coefficients across consecutive monthly-aggregated data. Digital measures’ sensitivity: ability to detect within-participant changes, analyzed over 24 months using linear mixed-effect models, and analyzed as effect-size changes-from-baseline comparing 1- and 2-year longitudinal Cohen’s-d (mean and 95% CIs) vs conventional clinical endpoints. Analyses replicated in two independent PD cohorts (internal validation and external evaluation). Compared within-participant changes between PD and non-PD cohorts using linear mixed-effect model slopes.
RESULTS
We analyzed 57 digital measures (51 individual, 6 composite) in a development cohort (N=171), selecting 32 (26 individual, 6 composite) for further study based on their sensitivity and test-retest reliability. During internal validation (N=101), 20 measures could detect statistically significant within-participant changes and 7 showed larger 2-year effect-size changes than conventional clinical measures; non-walking bout (NWB) duration (12.4% yearly change; 2-year Cohen’s-d 0.623 [95% CI: 0.461,0.811]) and 95th percentile of NWB duration (17.1% yearly change; 2-year Cohen’s-d, 0.623 [95% CI: 0.461,0.811]) performed best. Measures could detect significant and persisting changes from baseline at 10 months. During external evaluation (N=67), 15 measures could detect statistically significant within-participant changes and 12 showed larger 1-year effect-size changes than conventional clinical measures; 12 measures showed significantly greater change in people with PD than in matched non-PD individuals (N=171).
CONCLUSION AND RELEVANCE
Internal validation and external evaluation of 32 digital measures that quantified walking and non-walking behaviors in patients with early-stage PD showed that they could have greater sensitivity to detect longitudinal changes than conventional measures, and that these changes were disease-specific (e.g., separate from aging), making them candidates for disease-specific progression markers.
Key Points
Question
Can wearable sensor-based digital measures of physical activity and mobility serve as markers of disease progression in early-stage Parkinson’s disease (PD)?
Findings
In two independent longitudinal cohorts of people with PD, digital measures detected statistically-significant changes in walking and non-walking behaviors after 1 and 2 years of follow-up; additionally, a comparison between people with and without PD (from a third cohort) showed that these changes were disease-specific. Compared with MDS-UPDRS-based conventional metrics, measures of non-walking behavior showed greater effect size (such as mean non-walking bout duration, with an annual increase of 12.4% and a 2-year Cohen’s-d of 0.623).
Meaning
Wearable sensor-based digital measures can detect and quantify disease-specific changes in walking and non-walking behaviors over time in people with early-stage PD.
Related articles
Related articles are currently not available for this article.