Fifteen Epigenetic Clocks Can Tell You If You Will Reach 90, But Not Whether You Will Still Know Your Own Name

#1

Researchers measured 15 blood-based epigenetic clocks in 5,844 women enrolled in the Women’s Health Initiative Memory Study between 1996 and 1999, then followed them for up to 28 years to see who reached age 90. Every clock predicted survival to 90, with the third-generation and second-generation clocks (AgeAccelGrim2, PCGrimAge, PCPhenoAge, DunedinPACE) roughly two to three times stronger than the original Horvath and Hannum clocks. The headline result is the negative one. Not a single clock, including the newly published Intrinsic Capacity clock that was partly trained on cognitive testing, was better at predicting survival to 90 with intact cognition than survival to 90 with dementia or mild cognitive impairment. Among women who did reach 90, baseline clock values had essentially no association with whether cognitive impairment developed along the way.

The promise attached to epigenetic clocks has always been bigger than survival. Anyone can measure death. What people actually want from a biological age test is a readout on the quality of the years ahead, and specifically on whether the brain will still be working. This study, the largest of its kind to follow women all the way to age 90, is the first to put that question to a serious test. The answer is no.

The design is straightforward. Blood was drawn from nearly 6,000 women in the late 1990s, when they were on average 70 years old and cognitively normal. Fifteen clocks were computed from that single blood sample. The women were then tracked, with annual cognitive assessment and adjudicated dementia diagnoses, until February 2024. By then 3,233 had reached 90, of whom 1,726 still had preserved cognition and 956 did not. The remaining 2,611 died before 90.

On plain survival, the clocks worked. Faster biological aging on AgeAccelGrim2 or PCGrimAge cut the odds of reaching 90 by around a third per standard deviation, and the pattern held for every clock tested. Clocks built to capture longer telomeres or higher intrinsic capacity ran the other way, as expected.

Then the researchers split the survivors. If a clock measures healthspan rather than lifespan, it should track more tightly to the women who arrived at 90 with their minds intact. None did. GrimAge2 predicted survival with dementia (odds ratio 0.67) just as strongly as survival with preserved cognition (0.64). The confidence intervals overlap almost completely. Where small differences appeared, they usually leaned the wrong way.

The most likely explanation is structural. GrimAge2, the best performer, is not really a measure of methylation age at all. It is a stack of DNA methylation proxies for inflammatory and metabolic proteins in plasma, plus an estimate of lifetime smoking exposure. That combination kills people through hearts, lungs, kidneys and vasculature, and it does so on a schedule the clock reads well. Neurodegeneration runs on a partly separate track, through cell types with their own methylation landscapes, and a profile taken from circulating white blood cells does not see it.

For a field increasingly willing to sell methylation panels as a healthspan dashboard, this is an inconvenient and useful result. The clocks predict when you die. They do not yet predict how you arrive.

Actionable Insights

Nothing here is an intervention. This is an observational study of a blood test, and it cannot tell you that changing your clock changes your fate. What it gives you is a calibrated sense of what a clock reading is worth.

In this cohort, 55.3 percent of women reached 90. A woman one standard deviation worse than average on GrimAge2 fell to about 45 percent, one standard deviation better rose to about 65 percent. So the full spread across that range is roughly 20 percentage points, which matters at population scale and much less for any single person.

As a standardized effect size, the strongest clock reaches Cohen’s d of about 0.25. Convention calls anything below 0.2 negligible and 0.2 to 0.5 small. In discrimination terms, GrimAge2 sorts future survivors from future decedents at about 56 percent accuracy once age, smoking and disease history are already in the model. A coin flip is 50 percent. That is the honest size of the signal.

Three takeaways. A poor clock score is a risk nudge, not a prognosis. If you buy a test, the second and third generation clocks (GrimAge2, PCGrimAge, PCPhenoAge, DunedinPACE) carry roughly three times the signal of Horvath or Hannum. And do not read any current clock as a measure of your future brain, because on this evidence it is not one.

Context and Source

  • Open Access Papaer: Epigenetic Clocks of Biological Aging and Cognitively Healthy Longevity: The Women’s Health Initiative Memory Study
  • Authors: LaCroix AZ, Zhang B, Nguyen S, Resnick SM, Ferrucci L, Horvath S, Lu AT, Espeland MA, Rapp SR, Beckman K, Nievergelt C, Maihofer AX, McEvoy LK, Shadyab AH
  • Lead institution: Herbert Wertheim School of Public Health and Human Longevity Science, University of California San Diego, La Jolla, California
  • Country: United States
  • Journal: The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences (Medical Sciences section), Oxford University Press.
  • Impact evaluation: The impact score of this journal is 4.6 (2025 Journal Impact Factor; CiteScore 8.2), evaluated against a typical high-end range of 0 to 60+ for top general science, therefore this is a Medium impact journal.

Related Reading:

1 Like