Characterizing heterogeneity in cognitive aging through longitudinal trajectory clustering and phenotypic profiling

Cognitive aging is highly heterogeneous: Some individuals maintain high levels of cognitive functioning into late life whereas others experience gradual decline. This inter-individual variability is partly attributable to differences in lifestyle, physical and mental health, as well as genetic variation. A better understanding of the determinants of age-related cognitive change may inform the development of more personalized strategies aimed at promoting healthy cognitive aging and reducing dementia risk. However, previous research has often been constrained by short follow-up periods and the inclusion of only a narrow range of factors within single cohorts, limiting a multi-domain perspective on cognitive aging.

In this context, the present study examined long-term cognitive aging trajectories over an average follow-up of 13 years in a deeply phenotyped cohort. We employed a two-step approach: data-driven clustering to identify subgroups of cognitive aging beyond predefined cut-offs of test scores, followed by a phenome-wide comparison of the identified subgroups to determine the factors associated with distinct trajectories.

Thereby, we included data from 696 dementia-free participants aged 48 years and older, each with at least five records of the German version of the CERAD neuropsychological assessment battery (CERAD-NAB). We applied data-driven clustering to features derived from the participants’ longitudinal records of the CERAD-NAB total scores. These features correspond to the random intercept and random slope obtained from a linear mixed-effects model adjusted for age, sex, and educational years. Subsequently, we compared the identified subgroups across 49 phenotypes spanning domains of mental and neuropsychological health, personality, lifestyle, general physical health, motor function, diet, and genetic and biofluid markers as observed at the phenotypes’ initial records. Multiple testing corrections were performed across all phenotypes within one domain.

Our results reveal distinct trajectories of age-related cognitive change. Using a four-clustering solution, we identified subgroups of i) individuals with high cognitive function at baseline and stable progression across aging, ii) individuals with moderate baseline cognitive function and slow decline, iii) individuals with high baseline cognitive function but rapid decline, and iv) individuals with low baseline cognitive function and progressive decline. Individuals of the high-baseline/stable subgroup compared to those of the low-baseline/declining subgroup showed significantly faster gait speed, higher health-related quality of life, more regular physical and cognitive stimulating activity, and higher plant-based foods intake. While there were fewer phenotypic differences involving the intermediate subgroups, the high-baseline/stable subgroup significantly differed compared to all other subgroups in higher plant-based foods intake.

Together, our study highlights that cognitive decline is not inevitable with aging, as a substantial proportion of individuals maintain cognitive functioning into late life. While our subgroup comparisons pinpoint multidomain factors associated with distinct cognitive aging trajectories, future research is warranted to investigate the underlying biological mechanisms and to explore potential relationships to risk and protective factors for pathological aging. Overall, our study provides a foundation for developing more subgroup-specific interventions aimed at promoting healthy cognitive aging.

References

No references