Comparing predictors of conversion and decline in mild cognitive impairment

Abstract

Objective: A variety of measurements have been individually linked to decline in mild cognitive impairment (MCI), but the identification of optimal markers for predicting disease progression remains unresolved. The goal of this study was to evaluate the prognostic ability of genetic, CSF, neuroimaging, and cognitive measurements obtained in the same participants.

Methods: APOE epsilon4 allele frequency, CSF proteins (Abeta(1-42), total tau, hyperphosphorylated tau [p-tau(181p)]), glucose metabolism (FDG-PET), hippocampal volume, and episodic memory performance were evaluated at baseline in patients with amnestic MCI (n = 85), using data from a large multisite study (Alzheimer's Disease Neuroimaging Initiative). Patients were classified as normal or abnormal on each predictor variable based on externally derived cutoffs, and then variables were evaluated as predictors of subsequent conversion to Alzheimer disease (AD) and cognitive decline (Alzheimer's Disease Assessment Scale-Cognitive Subscale) during a variable follow-up period (1.9 +/- 0.4 years).

Results: Patients with MCI converted to AD at an annual rate of 17.2%. Subjects with MCI who had abnormal results on both FDG-PET and episodic memory were 11.7 times more likely to convert to AD than subjects who had normal results on both measures (p <or= 0.02). In addition, the CSF ratio p-tau(181p)/Abeta(1-42) (beta = 1.10 +/- 0.53; p = 0.04) and, marginally, FDG-PET predicted cognitive decline.

Conclusions: Baseline FDG-PET and episodic memory predict conversion to AD, whereas p-tau(181p)/Abeta(1-42) and, marginally, FDG-PET predict longitudinal cognitive decline. Complementary information provided by these biomarkers may aid in future selection of patients for clinical trials or identification of patients likely to benefit from a therapeutic intervention.

Figure FDG-PET and AVLT survival curves show increased conversion over time for abnormal relative to normal subjects Predicted survival curves based on Cox proportional hazards models (table 3) illustrate the univariate results for (A) FDG-PET and (B) AVLT, which were the 2 variables that remained significant in the multivariate model. Both curves show that for each variable, a higher proportion of AD− subjects (solid black line) remained dementia-free over time compared to AD+ subjects (dotted line). Age, education, and sex were included as control covariates. Proportion of subjects remaining dementia-free is shown on the y-axis.