Some Epigenetic Clocks Correlate with Risk of Dementia

Key Takeaways

• AgeAccelGrim2 predicts dementia risk modestly.
• DunedinPACE shows non‑significant risk increase.
• First‑generation clocks lack predictive power.
• Study involved 6,069 women, 9.3‑year follow‑up.
• No subgroup differences observed.

Summary

A recent analysis of 6,069 cognitively normal women examined whether epigenetic aging clocks predict incident mild cognitive impairment or dementia. Second‑ and third‑generation clocks (AgeAccelPheno, AgeAccelGrim2, DunedinPACE) were compared with first‑generation Horvath and Hannum measures. Only the AgeAccelGrim2 clock showed a statistically significant hazard ratio (1.11 per SD) after Bonferroni correction, while DunedinPACE and AgeAccelPheno were non‑significant. No meaningful effect modification was observed across age, race, hormone therapy, menopause type, or APOE ε4 status.

Pulse Analysis

Epigenetic clocks translate DNA methylation patterns into a single age estimate, offering a molecular gauge of biological aging. First‑generation models such as Horvath and Hannum were built on broad tissue datasets, while newer second‑ and third‑generation clocks incorporate biomarkers linked to morbidity and mortality, aiming for tighter health relevance. As the population ages, researchers are probing whether these clocks can flag individuals at heightened risk for neurodegenerative conditions, where early intervention could be transformative.

In the Women’s Health Initiative Memory Study, researchers applied three advanced clocks—AgeAccelPheno, AgeAccelGrim2, and DunedinPACE—to blood samples from women averaging 70 years old. Over a median 9.3‑year follow‑up, 1,307 participants developed mild cognitive impairment or probable dementia. Adjusted Cox models revealed a 11% increase in hazard per standard‑deviation rise in AgeAccelGrim2, surviving stringent Bonferroni correction, whereas DunedinPACE (7% increase) and AgeAccelPheno (1% increase) did not reach significance. First‑generation clocks showed no association, suggesting that the inclusion of mortality‑related biomarkers in newer algorithms may capture disease‑relevant aging processes more effectively.

These results carry practical implications for biotech firms and clinicians seeking biomarkers to enrich dementia risk stratification. The modest yet significant link of AgeAccelGrim2 suggests it could serve as a component in multi‑modal predictive models, but the lack of consistency across clocks cautions against wholesale adoption. Future work must validate these findings in diverse cohorts, explore longitudinal changes in clock readings, and integrate them with neuroimaging and cognitive testing to refine precision‑medicine approaches for Alzheimer’s and related dementias.