Aging Clocks: What They Actually Measure (And Why Most Are Still Wrong)

A single standard deviation increase in the Horvath aging clock score is associated with roughly a 2% increase in mortality risk. The same one-SD shift in DunedinPACE — a third-generation clock — is associated with a 64% increase. Those are not the same instrument measuring the same thing.

That gap is the central tension in Peter Attia’s Episode #386, which takes a rigorous look at what epigenetic aging clocks actually measure, how the science has evolved across three generations of tools, and the question that still doesn’t have a clean answer: does improving your clock score actually improve your health outcomes?

Context

The biological age testing market has grown significantly alongside the broader longevity movement. Dozens of companies now offer at-home epigenetic age tests based on a cheek swab or blood draw — with prices ranging from a few hundred to several thousand dollars. Consumers are being sold the idea that knowing your biological age is actionable information.

The problem is that “aging clock” is not a monolithic category. The tools are built differently, validated differently, and predict different things. Using the wrong clock — or interpreting a score without understanding the underlying methodology — can produce misleading conclusions about your health trajectory.

The Detail

Aging clocks measure DNA methylation: chemical modifications to the genome that change how genes are expressed without altering the underlying sequence. Epigenetic algorithms are trained to read these methylation patterns and output a number that represents some aspect of biological aging.

First-generation clocks, including the widely cited Horvath clock, were trained on chronological age. They are reasonably good at predicting how old someone is, but their predictive value for health outcomes is weak — and they show limited responsiveness to interventions. A 2020 publication by Dr. Jamie Justice found that “no clock was able to satisfy the last requirement: responsive to interventions.” That’s a significant limitation for anyone hoping to use clock scores as feedback on lifestyle changes.

Second and third-generation clocks changed the training target. Rather than predicting chronological age, newer tools like DunedinPACE are trained on longitudinal outcomes — actual aging phenotypes measured over time in real cohorts. DunedinPACE specifically quantifies the pace of aging, not a static age estimate. The mortality predictive gap between Horvath and DunedinPACE — 2% versus 64% per standard deviation — illustrates how consequentially different these tools are.

The most complex current offering is OMICm Age, which integrates over 2,000 proteins, 1,400 metabolites, and 61 clinical variables. That level of signal density pushes predictive accuracy for 10-year mortality to roughly 90%, compared to about 75% for chronological age alone.

Technical precision has also improved substantially. Older methylation arrays showed over 20% variance across replicates. Current instruments achieve ICC precision values above 0.95. The tools are getting sharper — but that doesn’t automatically translate into clinical guidance.

The episode’s most honest question is whether improving a clock score through lifestyle intervention represents genuine biological rejuvenation or simply a change in the biomarker. A 2022 Nature Aging study showed caloric restriction modified DunedinPACE scores — but whether that change predicts better health outcomes decades later remains unresolved.

Source: Peter Attia MD

Content created with AI assistance and reviewed for accuracy.