Cognitive event related potentials as functional biomarkers in Alzheimer's disease

Clinically useful biomarkers are necessary to improve diagnosis of Alzheimer's disease, and to aid in drug development. We report a new approach to identifying functional biomarkers that combines dense array EEG techniques with a cognitive probe task that targets brain functions and structures vulnerable in AD. EEG reflects the summation of many cortical processes operating simultaneously, and may be sensitive to disease-specific variations. However, efforts to identify AD markers based on individual ERP (event-related potential) components have not produced sensitivity and specificity profiles adequate for clinical application. Rather than rely on individual components, our approach takes advantage of multiple components that, in combination, may provide a more reliable marker of the presence of AD pathology in the brain. We examined a range of ERP components and behavioral data based on a crossmodal associative memory task designed to elicit activity in the hippocampus, which is the primary predilection site for AD pathophysiology. The 20-minute memory task was completed by a group of patients newly diagnosed with mild AD (N=65, pre-treatment), and healthy age-matched controls (N=65). We examined ERP components including the N100, P100, N200, P300, P400, P600, a difference wave at 700ms, and also behavioral responses (d' measures). First, each variable was assessed individually in binary logistic regression analyses, and sensitivity and specificity profiles were constructed. Next, the best performers in terms of sensitivity and specificity were entered into regression analyses in pairs, and then in groups of three. Results: Good discrimination was achieved by several combinations of 3 variables, with sensitivity ranging from 76.2% to 77.8%, and specificity from 87.7% to 90.4%. Brain source estimation (GeoSource) showed that the frontal P400, which participated in several good classification outcomes, is generated in the left hippocampus. The combination of multiple cognitive evoked potential components achieved a good sensitivity/specificity profile. Further, the association of hippocampal function with task performance suggests that cognitive evoked potentials may reflect fundamental aspects of the pathophysiology of AD. Taken together, these results support the development of cognitive evoked potentials as functional biomarkers in AD.