Epigenetic clocks, quantifying biological age through DNA methylation (DNAmAge), have emerged as potential indicators of brain aging. As the variety of DNAmAge algorithms grows, consensus on their efficacy in predicting age-related changes is lacking. This study aimed to explore the intricate relationship between diverse DNAmAge algorithms and structural and cognitive markers of brain aging. Within a cohort of 796 elderly patients (mean age, 65.8 ± 7.9 years), we scrutinized 11 DNAmAge algorithms, including Horvath, Hannum, Zhang's clocks, PhenoAge, GrimAge, DunedinPACE, and principal component (PC)-based PCHorvath, PCHannum, PCPhenoAge, and PCGrimAge. We evaluated their association with baseline cognition and cognitive decline, assessed through follow-up evaluations at three (T1) and six (T2) years postbaseline. Additionally, we examined their relationship with structural magnetic resonance imaging markers of brain aging, including white matter. Zhang's clock was the best predictor of decline in memory (β = -0.04) and global cognition (β = -0.03), whereas PCGrimAge was the best predictor of speed decline (β = -0.17). The DNAmAge algorithms were the second-best predictors in explaining cognitive variability after education in memory and global cognition (R2 partial = 1.66% to 2.82%) and the best predictors for speed decline (R2 partial = 2.13%). PC-trained DNAmAge algorithms outperformed their respective original version. DNAmAge algorithms are strong and independent predictors of cognitive decline in the normal elderly population and explain additional variability in cognitive decline beyond that accounted for by conventional risk factors.
Read full abstract