Among 5,844 postmenopausal women tracked from the mid-1990s, every standard deviation increase in epigenetic age acceleration (EAA) reduced odds of surviving to 90 by 7–18% for first-generation clocks. Third-generation clocks performed markedly better: AgeAccelGrim2 showed a 34% reduction in odds (OR=0.66), PCGrimAge 36% (OR=0.64), and PCPhenoAge 27% (OR=0.73) per SD increase in EAA. Crucially, none of the 15 clocks distinguished women who reached 90 with intact cognition from those who arrived there cognitively impaired—both groups looked identical through the epigenetic lens.
This finding lands at the heart of a fundamental gap in longevity science: biological clocks calibrated on mortality endpoints systematically ignore the brain. The result is practically significant. Current marquee clocks—GrimAge, PhenoAge, DunedinPACE—were trained on blood-based systemic biomarkers and all-cause mortality, meaning they capture cardiovascular and metabolic aging well but are essentially blind to neurodegeneration trajectories. For adults using commercial epigenetic testing to assess health trajectory, this is a calibration warning: a favorable clock reading says nothing about cognitive fate at extreme age.
The study's strengths include cohort size, long follow-up, and simultaneous comparison of 15 clocks. Its limitations are single-sex design (women only), a single baseline blood draw, and observational architecture preventing causal inference. Scientifically, this is confirmatory of clock utility for mortality while paradigm-shifting in exposing their cognitive blind spot—pointing urgently toward neuro-specific epigenetic clock development.
