Published September 30, 2026, in JMIR Aging, this systematic review examined 20 reports on EEG biomarkers, classification and prognosis in subjective cognitive decline and mild cognitive impairment. The evidence base included up to 3,853 deduplicated participants, with 2,927 classified as having subjective cognitive decline, mild cognitive impairment, Alzheimer's disease or another dementia.
Where the signal is promising Ten reports evaluated prediction or artificial-intelligence models, nine were nondiagnostic observational studies and one was prognostic. Across the literature, spectral, connectivity and complexity measures showed research potential for distinguishing groups and characterizing cognitive decline.
The clinical boundary The reviewers found no conventional diagnostic-accuracy study with a prespecified clinical threshold. They rated certainty very low across all four evidence bodies. Diverse EEG methods, participant definitions, small or selected samples and limited external validation mean these signals are not ready to diagnose an individual or replace a clinical assessment.
What better evidence would look like Progress will require prospective representative cohorts, standardized acquisition and preprocessing, clear separation of training and test data, prespecified thresholds and external validation. That path could turn promising group-level observations into evidence that is genuinely useful at the point of care.
Source and disclosure Full text checked. The review reports no conflicts of interest. It was supported by the National Natural Science Foundation of China and Hainan Province, with no funder role reported. Prepared with AI assistance and source-checked October 1, 2026.