Real-world gait changes preceding Parkinson's disease diagnosis in a population-scale UK biobank cohort.

Parkinson's disease (PD) has a long prodromal phase characterised by non-motor symptoms and subtle motor dysfunction, including changes in gait. While early gait abnormalities have been described in high‑risk populations, less is known about how real‑world gait behaviour changes prior to diagnosis in the general population. To investigate whether digital gait biomarkers derived from wrist‑worn sensors are associated with future Parkinson's disease diagnosis. The study comprised 73,413 UK Biobank participants who wore a wrist-worn device for seven days. Seventeen digital gait biomarkers were derived using the Watch Walk algorithm. Participants were followed for up to 10 years through linked electronic health records, and associations with time to PD diagnosis were assessed using Cox regression models adjusted for age and sex. Of the 73,294 participants without PD at the accelerometry assessment, 314 were diagnosed with PD during follow-up. Compared with those who did not develop PD, those who did had lower daily step counts (6370 vs. 4043-5585 steps/day), with greater differences observed closer to diagnosis, as well as slower walking speeds, altered step regularity, and reduced arm swing at baseline (all p < 0.001). Across groups stratified by time to PD onset, five gait measures (daily step count, maximum walking speed, step regularity, proportion of long walking bouts, and time spent walking with static arm positions) were consistently associated with subsequent PD diagnosis (all p < 0.001). Differences in real‑world gait behaviour were observable years before PD diagnosis in this large population cohort. These findings suggest that digital gait biomarkers may help characterise early motor changes detectable up to 6.8 years before a clinical diagnosis of PD.