Epigenetic age test explained

TL;DR

An epigenetic age test estimates a person’s biological age by analyzing DNA methylation patterns at specific genomic sites, using mathematical models called epigenetic clocks.

• The major clocks have evolved through three generations. The first generation (Horvath, Hannum) predicts chronological age. The second generation (PhenoAge, GrimAge) predicts mortality and disease. Third generation (DunedinPACE) measures the current pace of aging.
• Different clocks answer different questions. DunedinPACE responds to short-term interventions within months. First-generation clocks shift only over 6 to 12 months and are less clinically useful.
• Epigenetic age tests have real reliability concerns. Early clocks varied 3 to 5 years on the same sample tested twice. Newer clocks like DunedinPACE achieve intraclass correlation above 0.90.
• Inside a longevity clinic workflow, epigenetic age tests work alongside blood biomarker panels and functional measures. No single test captures the full aging picture.

A Brief History of Epigenetic Age Testing

In 2013, Steve Horvath at UCLA built the first DNA methylation age clock. It uses 353 CpG sites, chemical tags on the DNA at specific locations across the genome, to estimate a person’s age from a blood or saliva sample with a correlation of 0.96 to chronological age (Horvath, Genome Biology 2013).

Thirteen years later, the field has moved through three distinct generations of clocks. Each measures something different from the last. An epigenetic age test in 2026 is not the same instrument that Horvath built. Knowing what version of the technology a test actually uses, and what that version is good for, is the part that determines whether the result means anything inside a clinical workflow.

This article covers the following:

• What an epigenetic age test measures
• How the three generations of clocks differ
• What DunedinPACE changed about the field
• The reliability questions critics raise
• Where Systems Age fits, and
• What these tests do and do not tell you in clinical practice.

What an epigenetic age test actually measures

DNA methylation is the attachment of methyl groups, small chemical tags, to specific cytosine bases in DNA, typically at CpG sites where a cytosine sits next to a guanine in the sequence. Methylation does not change the DNA sequence itself. It changes whether and how strongly the gene at that location gets expressed.

Methylation patterns shift with age in highly consistent ways. Some sites become more methylated. Others become less. The pattern across hundreds or thousands of sites carries a signal about biological aging.

An epigenetic age test runs the following workflow. Sample collection (blood, saliva, or cheek swab). DNA extraction. Methylation array analysis, typically using Illumina methylation chips that read methylation levels at hundreds of thousands of CpG sites. The methylation data is fed into a mathematical model, the epigenetic clock algorithm. The output is a biological age estimate or a pace-of-aging score.

The format of the result depends on the clock used.

• First-generation results are an age in years (for example, biological age 47).
• Second-generation results may include disease risk scores.
• Third-generation results, specifically DunedinPACE, give a rate (for example, 1.05, meaning aging 5 percent faster than calendar time).

The science was established in the early 2010s. Hannum and colleagues and Horvath both published foundational papers in 2013. Bird’s 2002 review in Genes and Development is the deeper background on methylation biology itself. The clocks built since 2013 are refinements and reinterpretations of what those original papers demonstrated.

The three generations of epigenetic age testing clocks

First-generation clocks (Horvath 2013, Hannum 2013):

They were built to predict chronological age from methylation patterns. Horvath’s clock uses 353 CpG sites and validates across 51 different human tissue types. The correlation with chronological age sits at 0.96. Hannum’s contemporaneous clock was designed specifically for blood samples. Both did what they were designed to do well.

The conceptual limitation was inherent in the training target. A clock trained to predict the number on a passport carries limited information about how fast someone is actually aging. The clinical utility is limited. These clocks detect large deviations from expected aging but lack the resolution to track most lifestyle interventions.

Second-generation clocks:

PhenoAge (Levine et al. 2018, Aging) was trained on a phenotypic age estimate derived from nine clinical biomarkers and mortality risk. GrimAge (Lu et al. 2019, Aging) was trained directly on mortality, incorporating DNA methylation surrogates for plasma proteins associated with mortality.

PAI-1, adrenomedullin, and lifetime smoking pack-years are among the components. The clinical utility is meaningfully higher. GrimAge acceleration independently predicts cardiovascular disease, cancer, and all-cause mortality after adjusting for known risk factors. The clock measures something clinically actionable.

Third-generation clocks:

They measure pace of aging rather than static age. DunedinPACE (Belsky et al. 2022, eLife) is the canonical example. It was trained from the Dunedin longitudinal cohort, where researchers tracked rate of decline in 19 organ-function biomarkers in the same individuals from age 3 to age 51. The result is a methylation signature associated with faster or slower decline. A DunedinPACE score of 1.0 means aging one year per calendar year. 0.85 means 15 percent slower than average. 1.2 means 20 percent faster.

The clinical utility is the highest of any clock. Test-retest reliability (intraclass correlation) is above 0.90. The Waziry CALERIE caloric restriction trial documented an 11 percent pace-of-aging reduction over two years, the first randomized-trial evidence that any clock responds to a validated longevity intervention.

How to select the right epigenetic clock

From operator experience, clock selection often defaults to whichever clock comes packaged in a commercial test kit. Frequently a first-generation Horvath or its derivative. This is the wrong starting point. The right question is which clinical question the clinic is trying to answer. The clock follows from that question.

“Aging biology is inherently multi system and probabilistic. AI’s maturity in the form of technology today is predictive/probabilistic (not yet handling causaulity).”

Samir Mitra, Founder and CEO of Reya.ai. LinkedIn Post 11 Addendum, 2024.

How DunedinPACE changed what epigenetic testing measures

A static biological age estimate tells you the cumulative state. A pace-of-aging measurement tells you current trajectory. These are different questions with different clinical implications.

A 40-year-old with a biological age of 35 might still be aging at 1.15 years per calendar year right now. The static number is reassuring. The pace might be the actual problem. The reverse is also possible. A biological age of 45 with a pace of 0.85 means the cumulative state is concerning but the recent trajectory is good. The clinical interpretation differs in each case.

A clinic running DunedinPACE quarterly can detect intervention responses within a single quarter. A clinic running first-generation Horvath has effectively no short-term signal. Methylation age shifts on six-to-twelve-month timescales, too slow for most intervention tracking.

The mortality data matters at the population level. Belsky 2022 documented that fast agers (DunedinPACE above 1.0 by one standard deviation) had 56 percent higher seven-year mortality and 54 percent higher seven-year chronic disease risk than slow agers. Small differences in pace translate to large differences in outcomes.

The CALERIE validation is the strongest randomized-trial evidence for any clock’s responsiveness. 11 percent pace-of-aging reduction over two years of caloric restriction. The first time a clock demonstrated responsiveness to a controlled longevity intervention.

In Reya’s customer base, pace of aging is the metric clinics actually use for quarterly tracking. First-generation clocks become an annual snapshot at best. The clock that runs the workflow is the one that responds to the workflow’s tempo.

Reliability, precision, and the questions critics raise

Epigenetic age tests have real limitations.

Coverage of methylation variability in PNAS has documented that biological age can fluctuate by up to five years across a single day depending on factors like sample collection time, recent meal, recent illness, hydration, and acute stress. It means single-point biological age testing is more variable than the test marketing suggests.

The reliability problem is even larger. First-generation clocks have intraclass correlation values in the 0.4 to 0.7 range. The same sample tested twice can produce results varying by three to five years. This is a serious limitation for using these clocks to track intervention response.

Solution to the limitations above

Higgins-Chen and colleagues offered a solution in 2022 (Nature Aging). Their principal-component-based computational method substantially improved clock reliability when applied to existing first and second-generation clocks. The newer iteration of GrimAge (GrimAge2, Lu et al. 2022) incorporates similar reliability improvements.

DunedinPACE was designed for this problem from the start. It achieves an intraclass correlation above 0.90, the highest of any clock. This is why DunedinPACE is the clock recommended for short-interval retesting.

A 2025 piece in The Conversation argued that biological age tests are useful for researchers but less useful for individual consumers chasing a single number. The critique is partially fair. A standalone result for an individual member can be misleading without longitudinal context. The same test, used inside a continuous clinical workflow tracking intervention response over time, carries more signal than the single-point result the critique addresses.

From operator experience, the reliability differential is the actual clinical decision point. Clinics that take this seriously order DunedinPACE specifically or use the Higgins-Chen-improved composites. The clinics that order first-generation Horvath as an annual snapshot are leaving most of the clinical signals on the table.

“The opportunity is not just more biomarkers. It is better integration, systems that connect biology, data, and clinical reasoning in ways that actually extending health span.”

Samir Mitra, Founder and CEO of Reya.ai. LinkedIn Post 11 Addendum, 2024.

Where Systems Age and multi-system measurement fit

The newest direction in epigenetic age testing is multi-system. Rather than producing a single biological age number, the 2025 Sehgal and Levine paper in Nature Aging introduced Systems Age, an 11-system aging score from a single blood methylation test.

The 11 systems are Heart, Lung, Kidney, Liver, Brain, Immune, Inflammatory, Blood, Musculoskeletal, Hormone, and Metabolic.

The framework matters because aging does not happen uniformly. Two 50-year-olds with the same global biological age might have very different patterns. One has a 40-year-old metabolic system and a 60-year-old cardiovascular system. The other has the reverse. They need different clinical protocols. A single biological age number obscures the difference.

Sehgal and Levine’s framework outperformed existing global clocks for predicting system-specific diseases. It is the first multi-system clock that translates into actionable clinical differentiation between members.

The direction is clear. By 2027 and 2028, multi-system aging profiles will likely become the clinical standard. Single-number biological age testing will become a summary metric rather than the primary clinical output. Longevity clinics that adopt this framework early have a workflow advantage that compounds.

“Immune shifts, metabolic changes, and epigenetic drift interact over time. Yet healthcare often treats them in isolation.”

Samir Mitra, Founder and CEO of Reya.ai. LinkedIn Post 11 Addendum, 2024.

What changes epigenetic age

Epigenetic age responds to specific interventions. The evidence base is concrete enough to name the studies rather than speak in generalities.

• Caloric restriction has the strongest randomized-trial evidence. The Waziry CALERIE trial documented an 11 percent DunedinPACE reduction over two years.
• Exercise produces measurable epigenetic age reductions in multiple intervention trials. Sustained aerobic and resistance training combined show the largest effects.
• DNA methylation maintenance happens during deep sleep. Chronic short sleep accelerates epigenetic aging within days.
• Inflammation reduction translates to slower pace of aging on second-generation clocks because GrimAge and PhenoAge incorporate inflammatory surrogates directly. Lower hsCRP through diet, omega-3 intake, and exercise produces measurable biological age effects.
• Smoking is the strongest single accelerator. Dose-dependent across every epigenetic clock. Cessation produces measurable reversal over time (Klopack et al. 2022 Clinical Epigenetics).
• Mindfulness, social connection, and time in nature each carry measurable epigenetic signatures in intervention trials.

For a deeper treatment of the interventions and their evidence base, see Reya’s piece on what biological age actually is and how to change it.

What separates members who actually see epigenetic age improvements from members who do not is sustained behavior change over years, not the choice of intervention.

Clinical Realities of Epigenetic Age Testing

What they tell you?

Whether biological aging is running ahead of or behind chronological age. Whether recent interventions are producing detectable cellular-level change. Mortality and disease risk stratification (with second-generation clocks specifically). System-specific aging patterns (with Systems Age, when available).

What they do not tell you?

The cause of accelerated aging. Only that it is happening. Causal interpretation requires the full clinical picture. They do not capture acute health changes. Epigenetic clocks are slow. For acute concerns, blood biomarkers are a better tool. And they do not produce actionable information without context. A biological age number without medical history, medication list, recent illness, and stress context can mislead more than inform.

Inside a longevity clinic workflow, epigenetic testing is one layer alongside quarterly blood biomarker panels, wearable data, functional measures, and lifestyle assessments. The clock alone is a summary metric. We need the architecture around the clock to turn the data into outcomes.

Coordinating these data streams across appointments, between members, and over years is the workflow problem Reya was built to solve.

“Longevity clinics need software that supports personalized, preventive, predictive and participatory (4P) care in order to deliver meaningful results to their customers. This contrasts with software designed for sick-care, which is episodic, reactive, and notes-focussed to drive billing.”

Samir Mitra, Founder and CEO of Reya.ai. Longevity.Technology interview, February 2025.

Integrating epigenetic age tests into practice

Epigenetic age testing has officially transitioned from an experimental novelty into a legitimate clinical tool, provided practitioners select the right clock for the right question. The stark reliability differences between early and modern generations make this choice critical for patient outcomes.

As the field rapidly advances toward continuous monitoring and multi-system profiles rather than single-number annual snapshots, clinics must adapt their workflows to keep pace.

Forward-thinking practices are leaving rigid legacy software behind and partnering with Reya. By deploying Reya’s custom agentic AI layer directly on top of your traditional EMR, you can effortlessly coordinate complex biomarkers and deliver true preventative care. Discover how Reya can optimize your clinical workflows by scheduling a consultation.

Frequently Asked Questions

References

1. Horvath, S. (2013). DNA methylation age of human tissues and cell types. Genome Biology, 14(10), R115. https://doi.org/10.1186/gb-2013-14-10-r115

2. Hannum, G., Guinney, J., Zhao, L., Zhang, L., Hughes, G., Sadda, S., … Zhang, K. (2013). Genome-wide methylation profiles reveal quantitative views of human aging rates. Molecular Cell, 49(2), 359–367. https://doi.org/10.1016/j.molcel.2012.10.016

3. Bird, A. (2002). DNA methylation patterns and epigenetic memory. Genes & Development, 16(1), 6–21. https://doi.org/10.1101/gad.947102

4. Levine, M. E., Lu, A. T., Quach, A., Chen, B. H., Assimes, T. L., Bandinelli, S., … Horvath, S. (2018). An epigenetic biomarker of aging for lifespan and healthspan. Aging, 10(4), 573–591. https://doi.org/10.18632/aging.101414

5. Lu, A. T., Quach, A., Wilson, J. G., Reiner, A. P., Aviv, A., Raj, K., … Horvath, S. (2019). DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging, 11(2), 303–327. https://doi.org/10.18632/aging.101684

6. Belsky, D. W., Caspi, A., Corcoran, D. L., Sugden, K., Poulton, R., Arseneault, L., … Moffitt, T. E. (2022). DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife, 11, e73420. https://doi.org/10.7554/eLife.73420

7. Waziry, R., Ryan, C. P., Corcoran, D. L., Huffman, K. M., Kobor, M. S., Kothari, M., … Belsky, D. W. (2023). Effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial. Nature Aging, 3(3), 248–257. https://doi.org/10.1038/s43587-022-00357-y

8. Higgins-Chen, A. T., Thrush, K. L., Wang, Y., Minteer, C. J., Kuo, P.-L., Wang, M., … Levine, M. E. (2022). A computational solution for bolstering the reliability of epigenetic clocks: Implications for clinical trials and longitudinal tracking. Nature Aging, 2(7), 644–661. https://doi.org/10.1038/s43587-022-00248-2

9. Lu, A. T., Binder, A. M., Zhang, J., Yan, Q., Reiner, A. P., Cox, S. R., … Horvath, S. (2022). DNA methylation GrimAge version 2. Aging, 14(23), 9484–9549. https://doi.org/10.18632/aging.204434

10. Sehgal, R., Markov, Y., Qin, C., Meer, M., Hadley, C., Shadyab, A. H., … Levine, M. E. (2025). Systems Age: A single blood methylation test to quantify aging heterogeneity across 11 physiological systems. Nature Aging, 5, 1880–1896. https://doi.org/10.1038/s43587-025-00958-3

11. Klopack, E. T., Carroll, J. E., Cole, S. W., Seeman, T. E., & Crimmins, E. M. (2022). Lifetime exposure to smoking, epigenetic aging, and morbidity and mortality in older adults. Clinical Epigenetics, 14(1), 72. https://doi.org/10.1186/s13148-022-01286-8