Category: Technology
Objective: To develop composite digital measures that could be used to track disease progression in early PD patients, and to determine whether they can outperform MDS-UPDRS.
Background: Recent work demonstrates that many Digital Health Technology (DHT)-derived movement features can differentiate PD patients from healthy controls (HC) and show change over time. However, single digital features are unlikely to capture the multi-faceted nature of PD progression, nor provide higher sensitivity in progression tracking compared to MDS-UPDRS which is a composite clinical scale. Composite digital measures could potentially improve progression tracking, though methods of constructing such measures have not yet been fully explored.
Method: Longitudinal clinical scale data and DHT data (collected via two DHT platforms: APDM and Clinical Ink (CI)) from 82 early-PD and 50 healthy subjects enrolled in the WATCH-PD study were used for analysis. A comprehensive machine learning-based framework was developed to identify DHT features that showed disease progression and construct composite digital measures for longitudinal tracking. This framework consisted of linear mixed effects model-based feature screening, univariate association analysis, cross-validation based feature selection, and subsequent Penalized Generalized Estimating Equations (PGEE) based composite measure development. Progression tracking performance was evaluated using the effect size of progression slopes between PD and HC cohorts.
Results: 53 APDM features, and 23 CI features showed progression over 12 months in early PD patients, all with smaller effect sizes compared to MDS-UPDRSIII. In contrast, composite digital measures, constructed from (1) 14 APDM features, (2) 12 CI features, and (3) 20 APDM+CI features, all showed progression trends with larger effect sizes than MDS-UPDRS. Among these, the APDM-based composite digital measure achieved the largest effect size, >2-fold larger than MDS-UPDRSIII.
Conclusion: Composite digital measures can significantly enhance disease progression tracking compared to existing clinical assessment in early PD patients. A novel, machine learning-based methodology framework was developed, enabling identification of digital features to include in the construction of composite digital measures for improved disease progression tracking.
References: 1. Pagano, G. et al. Trial of Prasinezumab in Early-Stage Parkinson’s Disease. New Engl J Med 387, 421–432 (2022).
2. Adams, J. L. et al. Using a smartwatch and smartphone to assess early Parkinson’s disease in the WATCH-PD study. npj Park.’s Dis. 9, 64 (2023).
3. Sotirakis, C. et al. Identification of motor progression in Parkinson’s disease using wearable sensors and machine learning. npj Park.’s Dis. 9, 142 (2023).
4. Adams, J. et al. Using a Smartwatch and Smartphone to Assess Early Parkinson’s Disease in the WATCH-PD Study – 12-month results. (2024) doi:10.21203/rs.3.rs-3793129/v1.
To cite this abstract in AMA style:
S. Zhai, J. Ren, J. Shen, S. Chatterjee, Y. Xu, A. Liaw, V. Svetnik, O. Patil, ER. Dorsey, J. Adams, M. Dockendorf, R. Baumgartner. Developing Composite Digital Measures for Tracking Parkinson’s Disease (PD) Progression using a Comprehensive Machine Learning-based Framework [abstract]. Mov Disord. 2024; 39 (suppl 1). https://www.mdsabstracts.org/abstract/developing-composite-digital-measures-for-tracking-parkinsons-disease-pd-progression-using-a-comprehensive-machine-learning-based-framework/. Accessed October 5, 2024.« Back to 2024 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/developing-composite-digital-measures-for-tracking-parkinsons-disease-pd-progression-using-a-comprehensive-machine-learning-based-framework/