Longitudinal assessment of the conversion of mild cognitive impairment into Alzheimer's dementia: Observations and mechanisms from neuropsychological testing and electrophysiology

INTRODUCTIONElucidating and better understanding functional biomarkers of Alzheimers disease (AD) is crucial. By analysing a detailed longitudinal dataset, this study aimed to create a model-based toolset to characterise and understand the conversion of mild cognitive impairment (MCI) to AD. METHODSEEG, MRI, and neuropsychological data were collected from participants in San Marino: AD (n = 10), MCI (n = 20), and controls (n = 11). Across two additional years, MCI participants were classified as converters or non-converters. RESULTSWe identified the Stroop Color and Word Test as the largest differentiator for MCI conversion (ROC AUC = 0.795). This was underpinned by disconnectivity in working memory and attention networks. Unsupervised clustering of EEG spectra also differentiated MCI conversion (ROC AUC = 0.710) and was underpinned by reduced excitatory and enhanced inhibitory synaptic efficacy in (prodromal) AD. Combining electrophysiological and neuropsychological assessments increased the accuracy of the differentiation (ROC AUC = 0.880) in comparison to each measure considered individually. CONCLUSIONCombining electrophysiological and neuropsychological assessment with mathematical models can inform the development of non-invasive, low-cost tools for the early diagnosis of AD. HighlightsO_LIWe analysed longitudinal changes in EEG and neuropsychological assessments in MCI C_LIO_LIStroop Color and Word Test error scores were lower in MCI converters C_LIO_LIThe degree of impairment was found to be correlated with functional disconnectivity C_LIO_LIUnsupervised clustering of EEG spectra characterised patterns associated with disease C_LIO_LIMathematical modelling revealed reduced excitatory synaptic efficacy in (prodromal) AD C_LI Research in ContextSystematic review: The authors used PubMed to review the literature on the use of inexpensive modalities, including EEG and neurophysiological testing, for characterising the progression of MCI to AD. Although promising, existing work suggests the full potential of these methods as tools for understanding prodromal AD is still lacking. Interpretation: A novel application of a clustering algorithm to EEG spectra revealed different patient diagnoses could largely be characterised by their cluster assignment. We also found differences in a particular neuropsychological test, the Stroop Color and Word Test. Using mathematical modelling we found there were both network and synaptic mechanisms that underlie these differences. Future directions: Using the methods described herein to build markers for testing MCI to AD conversion on a large independent cohort will be crucial to understanding the full impact and applicability of these approaches. This may ultimately lead to a better characterisation and understanding of the diagnosis and prognosis of AD.