Meta-analysis preprint: of five DNA methylation clocks, only GrimAge was associated with future dementia

Meta-analysis preprint: of five DNA methylation clocks, only GrimAge was associated with future dementia

A systematic review preprint published on medRxiv on 21 October 2026 included 37 observational studies covering 46 365 people and seven additional studies examining genetic associations. The authors investigated whether age estimates derived from chemical marks on DNA in readily available samples help predict brain diseases.

Epigenetic clocks are models that estimate age from chemical marks on DNA. Blood can be sampled repeatedly for this analysis, whereas changes in the brain are much harder to measure in living people. Different clocks were trained for different tasks: early models predicted chronological age, while GrimAge predicted mortality risk.

The authors separately pooled studies that measured methylation before diagnosis and then followed participants over time. In the fully adjusted analysis, a higher GrimAge value, by an amount corresponding to the typical spread of values in the group, was associated with an approximately 11% higher risk of a subsequent dementia diagnosis. For Horvath1, Hannum, PhenoAge and DunedinPACE, the authors found insufficient statistical evidence of an association with future dementia.

GrimAge was developed using methylation features associated with smoking and levels of certain blood proteins. The authors suggest that this may allow it to reflect aspects of the body's condition associated with vascular, metabolic, inflammatory and behavioral risk factors for dementia.

Most included studies measured methylation in blood, although neurodegeneration occurs in the central nervous system. The authors propose evaluating models tailored to specific diseases and cell types by testing whether they predict disease onset and progression.