Background: The recently proposed latent variable δ is a new tool for dementia case finding. It is built in a structural equation modeling framework of cognitive and functional data and constitutes a novel endophenotype for Alzheimer's disease (AD) research and clinical trials.
Objective: To investigate the association of δ with AD biomarkers and to compare the prediction of δ with established scales for conversion to dementia in patients with mild cognitive impairment (MCI).
Methods: Using data from a multicenter memory clinic study, we examined the external associations of the latent variable δ and compared δ with well-established cognitive and functional scales and cognitive-functional composite scores. For that purpose, logistic regressions with cerebrospinal fluid (CSF) biomarkers and conversion to dementia as dependent variables were performed with the investigated scores. The models were tested for significant differences.
Results: In patients with MCI, δ based on a broad range of cognitive scales (including the ADAS-cog, the MMSE, and the CERAD neuropsychological battery) predicted an abnormal CSF Aβ42/tau ratio indicative of AD (n = 340, AUC = 0.78, p < 0.001), and predicted incident dementia within 1-3 years of follow-up (n = 525, AUC = 0.84, p < 0.001). These associations were generally stronger than for any other scale or cognitive-functional composite examined. Homologs of δ based on reduced test batteries yielded somewhat lower effects.
Conclusion: These findings support the interpretation of δ as a construct capturing the disease-related "essence" of cognitive and functional impairments in patients with MCI and dementia, and suggest that δ might become an analytical tool for dementia research.
Keywords: CSF AD biomarker signature; delta model; incident dementia; mild cognitive impairment.