Background: Conventional diagnostic tests for dementia are invasive, expensive, and/or not

scalable, hampering early detection. Whilst blood-based biomarkers are beginning to enter

clinical practice, their utility for determining which individuals will develop clinically

expressed disease is limited. The Early Detection of Neurodegenerative diseases (EDoN)

Initiative is investigating the potential utility of digital measures to support the early

detection of dementia at scale.

Methods: EDoN formed partnerships with four international cohorts: Predictors of COgnitive

DECline using digital devices (CODEC-II; UK); Boston University Alzheimer’s Disease Research

Center Health Outreach Program for the Elderly (BU ADRC-HOPE; USA); Western Australia

Memory Study (WAMS; Australia); and Healthy Brain Aging Program cohort (HBA; Australia).

Cohorts collected standard clinical, biofluid, and neuroimaging data alongside EDoN digital

measures from healthy controls, and participants with subjective cognitive decline (SCD) or

mild cognitive impairment (MCI). Participants used two or more of the EDoN wearables/apps

for 12 months: electroencephalographic sleep headband, wrist-worn activity tracker,

cognitive testing smartphone app, and passive smartphone monitoring app.

Results: The four cohorts enrolled a total of 432 individuals (n=261 female; 60%), of whom

202 (47%) were healthy controls, 109 (25%) had SCD, and 114 (26%) had MCI. Participants’

mean (SD) age was 70.0 (9.3) years, with 15.0 (3.3) years of education. Sixteen participants

withdrew; the remaining 416 (96%) completed data collection. The numbers of participants

using the wearables/apps were: sleep headband (n=313; 72%); wrist-worn activity tracker

(n=366; 85%); cognitive testing app (n=342; 79%); and passive app (n=374; 87%). Qualitative

interviews explored the acceptability of wearables/apps and identified future priorities.

Details of investigations (e.g., neuroimaging, blood-based biomarkers) included in the

dataset will be presented at the conference.

Conclusion: Ongoing data harmonisation and feature extraction will maximise the

interoperability of data across cohorts. Data will be made available to the research

community via a cloud-based trusted research environment. Raw data will be (minimally)

processed to extract useful signals and ensure ease of use for researchers to conduct further

analyses. Digital measures have the potential to detect dementia symptoms early and show

promise as a frontline screening tool to support remote identification of candidates for

further pathology-specific biomarker testing.