Data-driven subtypes of subjective cognitive decline: neuropsychological profiles, Alzheimer's disease biomarkers, and clinical trajectories.

Objectives: Subjective Cognitive Decline (SCD) is a heterogeneous condition recognized as the earliest manifestation of Alzheimer's disease (AD). We hypothesized that the heterogeneity of SCD may be synthesized in distinct subtypes.

Methods: We analyzed data from the AD Neuroimaging Initiative (ADNI) database. For all participants, demographic variables, cognitive measures, APOE genotype, CSF biomarkers, brain MRI, and FDG-PET data were available. Participants underwent follow-up visits every 6 or 12 months.

Results: 542 cognitively normal (CN), 346 SCD, and 423 early mild cognitive impairment (E-MCI) individuals were included. A data-driven approach based on cognitive measures identified three SCD clusters (k1, k2, k3) that performed differently in verbal memory (k2 outperformed all the groups and k3 showed the poorest performance, p < 0.001) and in executive function (k1 had the lowest scores, p = 0.006). Regarding CSF biomarkers, k2 exhibited lower p-tau (20.6 ± 9.2 vs. 24.2 ± 13.6, p = 0.03) and k3 had higher Aβ₄₂ levels (1131.3 ± 379.8 vs. 942.87 ± 355.3, p = 0.01) compared to the E-MCI group, while there were no differences between k1 and E-MCI. Regarding brain FDG-PET, k1 demonstrated reduced uptake compared to CN, k2, and k3 (p < 0.001). During follow-up, k1 exhibited a higher rate of progression to MCI or dementia compared to k2 and k3 (Log-rank χ² = 18.18, p = 0.0002) and a steeper decline in general cognition and long-term verbal memory compared to k2.

Interpretation: We proposed a three-subgroup system classification for SCD, reflecting different cognitive profiles and longitudinal trajectories. Classifying individuals with SCD may enhance diagnostic pathways and inform personalized interventions.