Across 36,941 participants — 35,822 healthy adults, 1,046 with Mild Cognitive Impairment (MCI), and 73 with dementia — the self-administered PROTECT Cognitive Test System achieved an AUC of 0.966 for dementia detection, with 90.90% sensitivity and 87.80% specificity. An eight-task battery covering memory, attention, and executive function also robustly identified MCI, predicted 24-month cognitive decline in both amnestic and non-amnestic MCI phenotypes, and correlated strongly with plasma biomarkers p-tau217 and Neurofilament Light chain (NfL) in a 46-person subsample.
The scale of this validation — nearly 37,000 participants — is unusually large for a cognitive screening tool and represents a meaningful advance over smaller clinical validation studies. Most comparable digital tools have been tested in hundreds, not tens of thousands, making PROTECT's breadth genuinely notable. The strong alignment with p-tau217, now considered among the most diagnostically informative Alzheimer's biomarkers, adds biological credibility beyond psychometric performance alone. Practically, a self-test, scalable system could transform primary care triage and clinical trial recruitment pipelines — two persistent bottlenecks in dementia research and diagnosis.
Caveats are significant, however. The dementia cohort is small (n=73), limiting confidence in subgroup analyses. The biomarker correlation subsample (n=46) is far too small to draw firm conclusions. The community-recruited sample may skew toward health-engaged, digitally literate adults, potentially inflating real-world performance. Critically, this is a preprint not yet peer-reviewed — findings should be considered preliminary until independent scrutiny confirms them. Incremental-to-confirmatory overall, with genuine scalability promise.