In 2013, Steve Horvath published a paper in Genome Biology that changed how we think about aging. He trained a machine-learning algorithm on DNA methylation data from 8,000 samples across 51 cell types and tissues and produced the first "epigenetic clock" — a set of 353 CpG sites whose methylation status could predict chronological age with a median error of 3.6 years. The insight was not that methylation changes with age. That was already known. The insight was that the same sites change in the same direction across virtually every tissue in the body, suggesting a shared biological program.

Horvath's clock was a first-generation tool. It predicted chronological age, which is useful for forensics but tells you less about health. The second generation arrived when researchers asked a different question: instead of predicting how old you are, can we predict how soon you will die? Lu et al. answered this in 2019 with GrimAge, an epigenetic clock trained on plasma protein surrogates and smoking pack-years as intermediate outcomes, then calibrated against time to death. GrimAge acceleration — the gap between your GrimAge and your chronological age — predicts all-cause mortality, cardiovascular disease, cancer incidence, and physical decline more strongly than any prior clock.

GrimAge v2, published by Lu et al. in 2022, added new CpG-based plasma protein surrogates and improved prediction further. In independent cohorts, a one-year increase in GrimAge v2 acceleration is associated with a 10–15% increase in mortality risk over the following decade. This is not a theoretical biomarker. Insurance actuaries, clinical trialists, and pharmaceutical companies are now using GrimAge v2 as a surrogate endpoint for interventions that claim to slow aging. The FDA has not formally validated it as a primary endpoint, but it appears in the statistical analysis plans of multiple aging intervention trials.

Daniel Belsky at Columbia took a fundamentally different approach with DunedinPACE, published in eLife in 2022. Rather than estimating total biological age from a single blood draw, DunedinPACE measures the current pace of aging — how many years of biological aging are occurring per calendar year. It was trained on longitudinal data from the Dunedin Study, a birth cohort of 1,037 individuals followed from birth to age 45 with repeated organ-system measurements. A DunedinPACE of 1.0 means you are aging at the expected rate. Below 1.0 means slower. Above 1.0 means faster.

The distinction matters clinically. GrimAge tells you where you are — your cumulative biological age debt. DunedinPACE tells you how fast you are currently accruing that debt. You might have a high GrimAge but a low DunedinPACE if you changed your lifestyle recently. The combination of both clocks gives a more complete picture than either alone.

Consumer testing has made these tools accessible. TruDiagnostic's TruAge test ($229–$499) reports both GrimAge v2 acceleration and DunedinPACE from a blood sample collected at home. Elysium Health's Index test reports a proprietary biological age estimate and DunedinPACE. Both require a blood draw kit, which is mailed to a CLIA-certified lab. Results arrive in 4–6 weeks. The scientific foundation is solid — both tests use Illumina EPIC methylation arrays — but individual measurements have noise. A single test can vary by 1–2 years between draws taken weeks apart.

What moves the clocks? The CALERIE trial showed that two years of 25% caloric restriction slowed DunedinPACE by 2–3% compared to controls. Exercise, sleep quality, Mediterranean diet patterns, and smoking cessation have all been associated with slower epigenetic aging. The SELECT trial's post-hoc analysis found semaglutide reduced GrimAge acceleration by 1.4 years independent of weight change. These are early but consistent signals that epigenetic aging responds to intervention.

For patients considering testing, the practical advice is straightforward. Establish a baseline, ideally with both GrimAge and DunedinPACE. Make intervention changes — exercise, diet, sleep, stress management. Retest after 6–12 months. Look at the trend across multiple measurements rather than fixating on any single number. One test is a snapshot. Three tests over 18 months is a trajectory. The trajectory is what matters.