Longitudinal CDR-Sum of Boxes trajectories in mild cognitive impairment reveal functional heterogeneity beyond cognitive decline: A latent class growth analysis.

Abstract

BackgroundMild cognitive impairment (MCI) is clinically heterogeneous, yet most prognostic studies rely on binary conversion endpoints or expensive biomarkers unavailable in routine practice.ObjectiveTo identify trajectory classes of the Clinical Dementia Rating-Sum of Boxes (CDR-SB) progression in MCI using latent class growth analysis (LCGA) and to determine whether they capture clinically relevant variation not reducible to Mini-Mental State Examination (MMSE) trajectories.MethodsLCGA was applied to longitudinal CDR-SB data from 121 MCI patients (baseline global CDR = 0.5) with two or more serial assessments across up to eight visits at three tertiary hospitals. Models with one through four classes were compared using BIC, AIC, entropy and minimum class size, with annual MMSE change and dementia conversion (global CDR ≥ 1) as external validators.ResultsA four-class model was selected: Stable (n = 20, 16.5%; -0.08 per visit; 0% conversion), Slow (n = 46, 38.0%; +0.41 per visit; 47.8%), Moderate (n = 22, 18.2%; +0.75 per visit; 81.8%), and Fast (n = 33, 27.3%; +2.07 per visit; 100%); conversion-free survival differed across classes (log-rank p < 0.001). Two classes with near-identical MMSE decline differed markedly in CDR-SB slope and conversion. The stability of the Stable class persisted when all classes were restricted to their first two visits.ConclusionsIn this tertiary-care cohort, serial CDR-SB trajectory classification identified clinically meaningful MCI subgroups, including functional heterogeneity invisible to MMSE monitoring alone; pending external validation it may offer practical, low-cost prognostic value where biomarkers are unavailable.