Data-driven neuropsychological phenotypes in the Baltimore Longitudinal Study of Aging.

INTRODUCTION: This study aimed to identify phenotypes of subtle variation in multidomain cognitive performance and examine their longitudinal associations with Alzheimer's disease and related dementias (AD/ADRD) biomarkers and cognitive outcomes. METHODS: Among 1192 cognitively unimpaired (CU) older adults from the Baltimore Longitudinal Study of Aging, latent profile analysis (LPA) identified phenotypes based on baseline patterns of neuropsychological test performance. Mixed-effects and Cox models examined longitudinal differences in cognitive status and AD/ADRD biomarkers (phosphorylated tau-181 [pTau181], amyloid-beta 42/40 ratio [Aβ42/Aβ40], neurofilament light [NfL], and glial fibrillary acidic protein [GFAP]) across phenotypes. RESULTS: LPA identified the following cognitive phenotypes: Overall Low Average, Dysexecutive (n = 112); Overall Average, Low Memory (n = 284); Overall Average, High Memory (n = 449); High Executive, Relatively Low Memory (n = 214); and Overall High Performing (n = 133). Phenotypes differed in longitudinal rates of cognitive decline and increases in NfL. DISCUSSION: Subtle variations in neuropsychological performance among CU older adults have implications for long-term cognitive health and may help inform Alzheimer's disease and related dementias diagnosis and disease monitoring. HIGHLIGHTS: Significant cognitive heterogeneity exists in CU older adults.LPA identified phenotypes based on cognitive performance.Personality and psychosocial characteristics differed by cognitive phenotype.Cognition over time and risk of cognitive impairment differed by cognitive phenotypes.The phenotype with greatest executive dysfunction had the fastest increase in NfL.