BACKGROUND: Digital health technologies show promise for improving the measurement of Parkinson's disease in clinical research and trials. However, it is not clear whether digital measures demonstrate enhanced sensitivity to disease progression compared to traditional measurement approaches. METHODS: To this end, we develop a wearable sensor-based digital algorithm for deriving features of upper and lower-body bradykinesia and evaluate the sensitivity of digital measures to 1-year longitudinal progression using data from the WATCH-PD study, a multicenter, observational digital assessment study in participants with early, untreated Parkinson's disease. In total, 82 early, untreated Parkinson's disease participants and 50 age-matched controls were recruited and took part in a variety of motor tasks over the course of a 12-month period while wearing body-worn inertial sensors. We establish clinical validity of sensor-based digital measures by investigating convergent validity with appropriate clinical constructs, known groups validity by distinguishing patients from healthy volunteers, and test-retest reliability by comparing measurements between visits. RESULTS: We demonstrate clinical validity of the digital measures, and importantly, superior sensitivity of digital measures for distinguishing 1-year longitudinal change in early-stage PD relative to corresponding clinical constructs. CONCLUSIONS: Our results demonstrate the potential of digital health technologies to enhance sensitivity to disease progression relative to existing measurement standards and may constitute the basis for use as drug development tools in clinical research.
- Publikační typ
- časopisecké články MeSH
BACKGROUND: Increasingly large and complex biomedical data sets challenge conventional hypothesis-driven analytical approaches, however, data-driven unsupervised learning can detect inherent patterns in such data sets. METHODS: While unsupervised analysis in the medical literature commonly only utilizes a single clustering algorithm for a given data set, we developed a large-scale model with 605 different combinations of target dimensionalities as well as transformation and clustering algorithms and subsequent meta-clustering of individual results. With this model, we investigated a large cohort of 1383 patients from 59 centers in Germany with newly diagnosed acute myeloid leukemia for whom 212 clinical, laboratory, cytogenetic and molecular genetic parameters were available. RESULTS: Unsupervised learning identifies four distinct patient clusters, and statistical analysis shows significant differences in rate of complete remissions, event-free, relapse-free and overall survival between the four clusters. In comparison to the standard-of-care hypothesis-driven European Leukemia Net (ELN2017) risk stratification model, we find all three ELN2017 risk categories being represented in all four clusters in varying proportions indicating unappreciated complexity of AML biology in current established risk stratification models. Further, by using assigned clusters as labels we subsequently train a supervised model to validate cluster assignments on a large external multicenter cohort of 664 intensively treated AML patients. CONCLUSIONS: Dynamic data-driven models are likely more suitable for risk stratification in the context of increasingly complex medical data than rigid hypothesis-driven models to allow for a more personalized treatment allocation and gain novel insights into disease biology.
- Publikační typ
- časopisecké články MeSH
