Stage vs. Subtype Hypothesis in Alzheimer's Disease: A Multi-Cohort and Longitudinal Bayesian Clustering Study

Background: The heterogeneity in Alzheimer’s disease (AD) has recently become a topic of intensive research due to its implications in understanding the underlying pathophysiological mechanisms of AD. We aimed to assess whether AD atrophy subtypes reflect different disease stages or biologically distinct subtypes. Methods We used longitudinal magnetic resonance images of 1196 individuals (891 AD dementia patients and 305 cognitively unimpaired individuals) from ADNI, AIBL, J-ADNI, and AddNeuroMed cohorts. AD patients were divided in discovery and prediction datasets. In the discovery dataset we applied a novel longitudinal clustering model that estimates differential atrophy patterns from the age of clinical disease onset. We calculated cluster-specific cognitive trajectories and validated our model in the prediction dataset. Results We discovered five longitudinal atrophy patterns in 320 amyloid-β positive AD patients over 8 years, with different atrophy and cognitive trajectories. Our findings replicated the previously reported AD subtypes (typical, limbic predominant, hippocampal sparing, minimal atrophy), but featured only two distinct longitudinal neurodegeneration pathways (mediotemporal vs. cortical). AD subtypes with distinct atrophy trajectories converged in late disease stages. We found a substantial agreement between our model estimates and the prediction dataset’s (571 AD) atrophy levels. Conclusion: Two neurodegeneration pathways were defined, suggesting that some of the previously reported atrophy subtypes may reflect different disease stages rather than distinct subtypes. However, substantial heterogeneity exists within the two pathways with different rates of atrophy and cognitive decline, potentially caused by a complex combination of protective/risk factors and concomitant non-AD brain pathologies. Clearly distinct atrophy trajectories may converge in late disease stages. By shifting from the cross-sectional understanding of AD subtypes to the new perspective brought by longitudinal clustering, we set the ground to unravel the heterogeneity in AD, which may provide a platform for developing disease-modifying treatments in the future. Funding: No funding is declared. Declaration of Interest: None to declare. Ethical Approval: All participants provided written informed consent in accordance with the Helsinki declaration and approval for the studies was obtained by the local ethical committees.