Although episodic memory impairment is usually the earliest sign of Alzheimer's disease (AD), there are up to 15% of patients presenting with early impairment in non-memory cognitive functions (i.e., atypical AD). Stratifying patients with AD may aid clinical trials. Previous studies divided patients by cognitive profile, focusing on cross-sectional analyses without testing stability of clusters over time. We used principal component analysis followed by cluster analyses in 127 patients with AD based on 24 cognitive scores at 0, 6, 12, and 24 months follow-up. We investigated the definition of clusters and their stability over time as well as interactions of cluster assignment and disease severity. At each time point, six distinct factors and four distinct clusters were extracted that did not differ substantially between time points. Clusters were defined by cognitive profile rather than disease severity. 85% of patients fell into the same cluster twice, 42% three times, and 17% four times. Subjects with focal semantic impairment progressed significantly faster than the other cluster. Longitudinally, focal deficits increased relatively rather than tending toward average disease severity. The observed similar cluster definitions at each time point indicate the validity of the approach. Cluster-specific longitudinal increases of focal impairments and significant between-cluster differences in disease progression make this approach useful for stratified inclusion into clinical trials.
Keywords: Alzheimer's disease; atypical AD; cluster analysis; clustering; subgroups; subtypes; typical AD.