Identifying cognitive trajectories and predicting rapid decline of cognitive function in early Alzheimer’s disease

Abstract

AbstractBackgroundThe clinical manifestations and progression of cognitive function vary between patients with Alzheimer disease (AD) and mild cognitive impairment (MCI). Predicting disease course is important for families and clinicians because most patients experience progressive and irreversible cognitive decline despite variations.MethodsThis study sought to identify groups of patients following a similar trajectory of cognitive decline over time, and then developed a prediction model to identify patients at higher risk for a speedy cognitive decline. This study was conducted on a clinic‐based cohort of 233 early AD and MCI patients who converted to AD within 2 years in Taiwan. A group‐based trajectory modeling technique characterized the trajectory groups of Mini‐Mental State Examination (MMSE) scores over time. Feature‐selection strategy was then used to investigate predictors of the trajectory of rapid cognitive decline. Finally, multivariable logistic regression (LR) was applied to predict the trajectory group membership from the predictors.ResultsThe best‐fitting model includes two meaningful trajectories: Group 1 shows a rapid decline in the MMSE scores over time (decrease of 5.4±3.9 points within 2 years); Group 2 shows a slow progression (decrease of 1.2±3.9 points within 2 years). The baseline MMSE score, instrumental activities daily living (IADL) total score, and APOE ε4 status are significant predictors in the prediction model for classifying the trajectory of cognitive decline. After bootstrapping validation, the performance of the LR model using these predictors to predict trajectory groups shows 85.3% of area under curve (AUC), 79.0% of sensitivity, 73.1% of specificity, and 75.4% of predictive accuracy.ConclusionsAmong various biomarkers and neuropsychological battles for patients with AD or MCI, baseline MMSE score, IADL total score and APOE ε4 status are the ones important and unchallenging to obtain in dementia clinics. The LR model for the prediction model is easy to understand and apply in the clinical practice. This model gives clinicians and patients a glimpse of future cognitive progression and aids further healthcare planning.