Fast Declining Prediction in Alzheimer's Disease from Early Clinical Assessment.

Intoduction: The heterogenicity in Alzheimer's Disease (AD) progression hinders individual prognosis. The present work is an observational 2-year longitudinal study in patients with mild cognitive impairment due to AD (n= 52, with positive CSF biomarkers). The aim of this study is to predict which patients are at risk of fast progression. For this, 3 neuropsychological tests based on different domains (clinical dementia, cognition, delayed memory) and the sum of them were used.

Method: The tests were performed at diagnosis time (T1) and two years after the diagnosis time (T2). Then, the corresponding progression models were developed using each individual test and their sum as a variable response.

Results: As a result, the model based on cognition status to predict fast decline (differences in the Z score (T2-T1) <1.5 were considered fast declining) provided satisfactory performance (AUC 0.74, 83.3% of sensibility and 70.2% of specificity); the models based on clinical dementia and delayed memory to predict fast declining showed low AUC and sensitivity. Nevertheless, the model based on the sum of the 3 tests showed the highest AUC (0.79), low sensitivity (63.6%), and high specificity.

Conclusion: The developed progression models could provide useful information to clinicians and AD patients regarding their fast/normal decline in general or specific domains.