Stage-specific digital health technology biomarkers enhance diagnostic and early progression detection in Parkinson's disease.

BACKGROUND

Digital health technology measurements show promise as objective biomarkers in Parkinson's disease. However, their sensitivity and consistency across early disease stages remain poorly defined, which currently limits their utility in clinical trial design.

OBJECTIVES

The study aims to evaluate the effectiveness of various DHT measures within tremor, bradykinesia, and axial symptom domains, in differentiating early stage PD from healthy controls and in monitoring short-term disease progression.

METHODS

In this study, we examined a range of tremor, bradykinesia, and axial symptom measures across age-matched healthy volunteers (n = 45), newly diagnosed Parkinson's disease patients (n = 54), and participants after 12 months of follow-up (n = 40) from WATCH-PD, a multicenter, observational study.

RESULTS

Our findings reveal distinct categories of functional measures: some effectively differentiate healthy controls from patients with recent Parkinson's diagnosis but show limited sensitivity to early progression, while others are insensitive to initial diagnosis yet capture longitudinal change. Models trained on disease-stage specific feature sets were most effective for their intended task, with the performance gain over combined features being more pronounced for progression detection (∆AUC = 0.15, ∆Cohen's d = 0.72).

CONCLUSION

These findings underscore the need to align composite digital biomarker design and feature selection to both disease stage and clinical objective, and suggest that adaptive, symptom- and side specific DHT measures may enhance sensitivity in trial population selection and short-term progression monitoring.