Digital mobility outcomes showed responsiveness and clinical convergence without sample size advantage in Parkinson disease

A digital mobility outcome can change over time without representing a meaningful treatment benefit or reducing trial sample-size requirements. We examined these distinct properties in ambulatory Parkinson disease. We analysed 524 participants from the observational Mobilise-D Clinical Validation Study, with up to five seven-day lower-back sensor assessments over 24 months. The primary outcome was the weekly 90th percentile (P90) of walking speed in bouts longer than 30 s, a measure of the faster end of real-world walking. Mixed models assessed longitudinal change; within-participant models assessed clinical co-change. Sample-size planning was post hoc. The primary outcome declined by − 0.0164 m/s/year (95% CI − 0.0226 to − 0.0103; 2,081 observations). Walking-bout steps and stride length also declined. A one-baseline-SD within-participant speed reduction accompanied 1.10-point and 1.79-point worsening in MDS-UPDRS Parts II and III and a 0.92-point reduction in Late-Life Function and Disability Instrument function. At 24 months, 359 participants had the primary outcome. Missing-not-at-random sensitivity scenarios preserved the decline direction when unobserved values were assumed worse, but its magnitude remained uncertain. Under a 24-month, 80%-power scenario targeting a 30% slope reduction, the primary outcome required approximately 3,450 participants per arm, compared with 2,150 for MDS-UPDRS II and 2,550 for Part III. The examined lower-back sensor-derived outcomes captured small group-level changes in mild-to-moderate ambulatory Parkinson disease. Longitudinal responsiveness and clinical convergence did not establish treatment responsiveness, patient-perceived benefit, or a sample-size advantage over both clinical comparators. The parent Mobilise-D Clinical Validation Study was prospectively registered with the ISRCTN registry (ISRCTN12051706) on 10 December 2020.