European Journal of Neurology | June 27, 2021

Santangelo R, Agosta F, Masi F, Spinelli EG, Cecchetti G, Caso F, Mandelli A, Cardamone R, Barbieri A, Furlan R, Magnani G and Filippi M

European journal of neurology. 2021

DOI: https://doi.org/10.1111/ene.14999

Abstract

Background

Alzheimer’s disease (AD) is characterized by a heterogeneous course. Predicting a fast rather than a slow decline over time is crucial to both provide a reliable prognosis and elaborate stricter enrolment criteria in clinical trials. Here we searched for independent predictors of cognitive decline rate to assess the risk of fast disease progression already at baseline.

Methods

Fifty-three subjects with an “in-vivo biomarker confirmed” diagnosis of AD were included. Neuropsychological assessment, plasma neurofilaments (NfL) concentrations and, in a subsample of 23 patients, brain magnetic resonance imaging were available. Patients were labelled FAST or SLOW depending on the Mini-Mental State Examination (MMSE) points lost per year (FAST if more than 3 points). We adopted single logistic regression models to search for independent predictors of FAST progression.

Results

At baseline no differences were found between FAST and SLOW subgroups in demographics, MMSE scores, vascular burden and medial temporal lobe atrophy measurements. Higher plasma NfL concentrations and worse scores at semantic verbal fluency (SVF) and clock drawing test (CDT) were independent predictors of FAST decline, after controlling for age, education, sex and baseline disease severity stage. The regression model combining all the predictors correctly classified 80% of patients overall. The risk of FAST decline was 81.2% if all the three predictors were abnormal (i.e., SVF ≤21.5, CDT ≤5.5, NfL ≥22.19).

Conclusions

An easily applicable algorithm, including plasma NfL measurement and two neuropsychological tests worldwide adopted in clinical practice (SVF and CDT), may allow clinicians to reliably stratify AD patients in relation to the risk of fast cognitive decline.

This study was performed using the Quanteix HD-1 Analyzer.