Published: June 06, 2026, 15:02 UTC
The diagnosis of Alzheimer’s disease has moved toward biological definitions based on cerebrospinal fluid (CSF) biomarkers and amyloid PET imaging. But lumbar punctures are invasive, and PET scans are expensive. A new study in GeroScience asks whether a simpler, cheaper, and non-invasive measure , the sleep EEG , can detect the same biological subgroups that CSF testing identifies.
The research team, led by investigators at IRBLleida and Universitat de Lleida in Spain, enrolled 42 patients with mild-to-moderate Alzheimer’s disease and 58 cognitively unimpaired older adults as controls. All participants underwent overnight polysomnography, and the Alzheimer’s patients also had CSF drawn for measurement of core biomarkers: amyloid-beta 42 (Aβ42), phosphorylated tau 181 (p-tau181), total tau (t-tau), and neurofilament light chain (NfL).
Sleep EEG signals from four scalp channels were processed and characterized using a wide range of quantitative features: linear measures, spectral power across frequency bands, and non-linear complexity metrics. The resulting dataset was high-dimensional, so the team used principal component analysis to reduce it to a compact 30-component representation that still explained 92.4 percent of the variance.
When this qEEG representation was fed into a clustering algorithm (Gaussian mixture models), it distinguished cognitively normal controls from Alzheimer’s patients with high fidelity. More importantly, it identified three distinct subgroups within the Alzheimer’s group. These subgroups showed graded differences in CSF biomarker profiles, including the key p-tau181/Aβ42 ratio and NfL levels, indicating that the sleep EEG was capturing real biological heterogeneity rather than random noise.
The idea that sleep brain activity might reflect Alzheimer’s pathology is not new. Amyloid and tau accumulation are known to disrupt sleep architecture, slow-wave activity, and spindle density. But this study goes a step further: it suggests that the specific profile of EEG changes can stratify patients into biologically meaningful subtypes, potentially identifying who might respond to different treatments or progress at different rates.
The study is relatively small (42 patients), and the findings need replication in larger, more diverse cohorts. The EEG features were derived from only four channels, and full-head high-density EEG might reveal even richer stratification. The cross-sectional design cannot track whether these subgroups predict clinical progression.
For sleep medicine, the finding adds to a growing argument that the sleep EEG is not just a tool for diagnosing sleep disorders. It may also serve as a window into neurodegenerative disease, accessible in any clinic with a PSG setup.
Source: Gaeta AM, Gallego Viñarás A, Barbé F, et al. Quantitative sleep EEG identifies CSF core biomarker-related subgroups in Alzheimer’s disease. GeroScience, 2026. DOI: 10.1007/s11357-026-02266-z