Digital gait biomarkers in Parkinson's disease: susceptibility/risk, progression, response to exercise, and prognosis
Comprehensive Overview
Published in npj Parkinson's Disease in 2025, this review article asks whether measurements of how a person walks, captured automatically by wearable sensors and cameras rather than assessed subjectively by a clinician, could help detect Parkinson's disease earlier, track its progression, gauge whether exercise is helping, and predict who is at risk of falling. By the time Parkinson's is diagnosed from visible tremor, stiffness, and slowness, 50 to 70 percent of the brain cells that produce dopamine are typically already lost, so earlier warning signs are badly needed. The authors reviewed published studies using wearable inertial sensors, pressure-sensitive mats, and optical camera systems to quantify five domains of walking, including overall speed and stride length, rhythm, step-to-step variability, upper-body stability such as arm swing and trunk motion, and left-right asymmetry, organizing the evidence around four practical uses: identifying risk before diagnosis, tracking disease change over time, measuring exercise response, and predicting future falls. The topline finding was that different gait features proved useful for different purposes, with upper-body measures like arm swing showing the most promise for early risk detection and walking speed and stride length proving most useful for tracking disease progression and predicting falls.
Outcomes and Implications
The single most important finding was that gait variability, meaning how inconsistent a person's stride length or timing is from step to step, emerged as a sensitive marker across all four contexts examined, though it is not specific enough on its own to distinguish Parkinson's disease from other causes of unsteady walking. Upper-body characteristics, particularly reduced and asymmetric arm swing and abnormal trunk motion, showed the most promise for flagging risk before diagnosis, including in people who carry Parkinson's-linked genetic mutations or have a sleep disorder called REM sleep behavior disorder that often precedes the disease by years. Pace-related measures, such as walking speed and stride length, proved most useful for tracking how the disease progresses and for predicting falls, while measures of dynamic stability, including trunk steadiness and the time spent with both feet on the ground, tended to worsen as the disease advanced but could improve with structured aerobic exercise. Aerobic gait training consistently improved walking speed and distance across the studies reviewed. However, the authors identified a major limitation: the field lacks standardized testing protocols and an agreed-upon minimum set of gait measurements, making it difficult to compare results across research groups or translate any single biomarker into routine clinical use. They call for future studies using a common, agreed-upon protocol to validate specific gait biomarkers so that objective, wearable-sensor-based walking assessments can eventually complement or replace subjective clinical rating scales in both research and everyday patient care.