Can you really measure your biological age? The evidence on epigenetic clocks

For decades, chronological age has been the primary metric used to assess disease risk and mortality. But chronological age is a blunt instrument. Two 60-year-olds can have vastly different health profiles, cellular function, and remaining lifespans. The quest to quantify this difference has led to the development of "epigenetic clocks"—algorithms that estimate biological age by analyzing chemical modifications on our DNA.[1][7]

In recent years, these clocks have transitioned from niche research tools to a booming consumer and clinical market. The global epigenetic biological age testing market reached $850 million in 2025 and is projected to grow 17.5% annually. Driven by a 40% reduction in DNA methylation assay costs, tests that once cost thousands of dollars are now accessible to consumers for under $200.[6]

But as the market expands, so does the confusion. Not all epigenetic clocks measure the same thing. To understand their utility, we must look at the underlying mechanism. Epigenetic clocks work by analyzing DNA methylation—a process where chemical tags called methyl groups attach to specific regions of DNA known as CpG sites.[1][7]

These methyl groups do not change the underlying genetic code, but they act as volume dials, turning gene expression up or down. As we age, certain CpG sites predictably gain or lose methylation. By using machine learning to analyze these patterns across thousands of individuals, scientists have built algorithms that calculate how old a person's cells appear to be at a molecular level.[1][7]

The field is currently divided into three distinct "generations" of clocks, each with different clinical value. First-generation models, such as the widely known Horvath and Hannum clocks developed around 2013, were trained specifically to predict chronological age. They do this remarkably well, with a median error of just 3.6 years.[1]

However, predicting chronological age is not the same as predicting health. If a clock perfectly guesses that you are 50 years old, it provides no insight into whether you are aging exceptionally well or exceptionally poorly. This limitation spurred the development of second-generation clocks.[1][7]

Second-generation models, most notably PhenoAge and GrimAge, were trained not on chronological age, but on clinical biomarkers and mortality data. GrimAge, in particular, has emerged as a powerhouse diagnostic tool. A January 2026 review in Frontiers in Molecular Biosciences synthesized multiple cohorts and concluded that GrimAge epigenetic age acceleration is arguably the strongest methylation-based predictor of cardiovascular and all-cause mortality currently available.[1][3]

The predictive power is substantial. A one-year increase in GrimAge acceleration—meaning your biological age is running one year ahead of your chronological age—is associated with roughly a 10% increase in mortality risk in some cohorts. A five-year biological age acceleration correlates with an approximate 30% increase in mortality risk, independent of traditional clinical risk factors.[1][6]

Beyond mortality, these clocks capture organ-specific vulnerabilities. A February 2026 systematic review in eBioMedicine analyzed 13 studies and found that individuals with accelerated epigenetic aging were consistently more likely to suffer a stroke. The association was stronger for first-ever strokes than recurrent events, suggesting the clocks detect vascular vulnerability before it becomes clinically apparent.[2]

The latest evolution in the field is the third-generation clock, exemplified by DunedinPACE (Pace of Aging Calculated from the Epigenome). While GrimAge acts as an odometer measuring total accumulated damage, DunedinPACE acts as a speedometer, measuring how fast you are aging right now.[4][5]

A score of 1.0 on DunedinPACE means you are aging at a standard rate of one biological year per chronological year. A score of 1.2 means you are aging 20% faster. Research indicates that scores above 1.0 correlate with a 56% increased mortality risk and a 54% higher chronic disease risk over a seven-year period.[5]

Crucially, a landmark 2026 study in Nature Aging confirmed that longitudinal changes in these clocks predict mortality independently of where a patient starts. This means a patient whose DunedinPACE score is worsening year over year carries a significantly different risk profile than one whose pace is stable, even if their absolute biological ages are identical.[4]

This brings up the most critical question for consumers and clinicians: Are these clocks modifiable? The evidence suggests they are. A December 2025 multi-cohort study reported that smoking, high BMI, and elevated glucose accelerate aging as measured by DunedinPACE, while physical activity and a healthier diet slow it down.[2]

However, a major layer of uncertainty remains. Most epigenetic clocks still derive their predictive features from correlation rather than proven causation. It remains an open question whether DNA methylation changes actively drive organ decline and aging, or if they are merely downstream passengers accompanying the damage.[2][7]

Furthermore, testing methodology matters. While consumer saliva tests are popular, blood-based epigenetic tests remain the gold standard for accuracy. Saliva samples can yield biological age estimates that vary wildly, whereas blood samples provide highly reproducible results.[5]

As the science matures, epigenetic clocks are poised to shift from longevity enthusiast novelties to standard clinical endpoints. By providing a quantifiable metric for biological decline, they offer a way to test preventative interventions in months rather than waiting decades to observe mortality outcomes.[2][7]