EEG recordings during visuo‐attentive task reduce sex bias in Alzheimer's disease diagnosis

INTRODUCTION Pathological cognitive decline affects roughly 15% of older adults, presenting relevant sex differences, being a stronger predictor of Alzheimer's disease in females. Yet, sex-specific neural markers that support a balanced early diagnosis remain limited. Event-related potentials (ERPs) derived from task-based electroencephalograph (EEG) offer a non-invasive window into cognitive processing, but their sex-specific diagnostic value in subjective cognitive decline (SCD) has not been explored. METHODS We recorded EEG during a visuo-attentive task in elders with SCD (n = 119) and age-matched healthy controls (n = 19) to extract ERP components reflecting stimulus encoding and decision processes. We investigated sex-specific associations between ERP features and task performance and trained separate machine learning models for SCD diagnosis in males and females. Diagnostic models were constructed using either clinical features alone or a combination of clinical and ERP features to quantify the added value of ERPs over standard assessments. RESULTS ERP analyses revealed distinct sex-specific associations between neural responses and behavioral performance, suggesting partially divergent neurocognitive mechanisms underlying attentional processing in aging males and females. Diagnostic models based solely on clinical data produced significantly unbalanced performance between sexes (area under the curve [AUC]: 0.75 males, AUC: 0.63 females; p < 0.00001). When ERP features were incorporated, classification accuracy improved in both groups and the sex imbalance was eliminated (AUC: 0.77 males, AUC: 0.75 females; p = 0.31). ERP features consistently demonstrated higher sensitivity to subtle cognitive alterations, particularly in females. DISCUSSION Task-related ERPs provide complementary, sex-specific neural information that mitigates diagnostic disparity arising from clinical assessments alone. Incorporating ERP features supports a more equitable early identification of cognitive decline and may improve screening strategies for populations usually under-recognized in preclinical Alzheimer's disease pathways.