Diagnostic Evidence: Cardiovascular & Biological Age Testing · Review 04
Epigenetic and Biological Age Clocks: Correlative Research Tool or Clinical Test?
Biological age clocks correlate with disease risk across large populations, but no methylation clock has been validated as a test that tells one person, reliably, whether they are aging faster or slower, or whether a specific intervention changed that.
- PROCEDURE STATUS
- Measuring DNA methylation itself is an established, reproducible laboratory technique. The composite "biological age clocks" built from those measurements are not an established clinical test.
- INDICATION EVIDENCE
- Experimental for individual clinical use
- REVIEW STATUS
- Clinical review required
- NEXT REVIEW
- January 2027
In short
DNA methylation can be measured reliably, and population studies show these "biological age clocks" track disease risk and mortality across large groups. But at the individual level, no clock has been validated as a diagnostic test, and the best available randomized trial evidence shows an inconsistent, modest signal that different clocks do not even agree on. A clock score changing after an intervention is not the same as proving that intervention slows aging or improves health, and no physician should make a treatment decision based on it alone.
The clinical question
Can DNA-methylation-based "biological age" clocks be used at the individual level to guide clinical decisions or to prove that an intervention slows aging, or are they presently valid only for population-level research?
What is established
Measuring DNA methylation, the chemical marks added to DNA that change with age and exposure, is a mature and reproducible laboratory technique. Methylation arrays and sequencing methods can detect these marks at specific genomic sites with good technical reliability. That part is not in dispute.
What is built on top of that raw measurement is different. Researchers have trained statistical models, commonly called epigenetic or biological age clocks, to combine methylation values at selected sites into a single score. Some of these models, including GrimAge, were built specifically to predict mortality and disease risk rather than chronological age, and in large validation cohorts GrimAge outperformed earlier clocks at predicting time-to-death and time-to-disease at the population level.[1]
What the evidence for individual use actually shows
The GrimAge validation work is a strong result for what it is: a demonstration that, averaged across large cohorts, people with a higher GrimAge score tend to die sooner and develop disease sooner than people with a lower score. That is a population-level prediction study. It shows the score carries real information in aggregate. It does not, by itself, tell a single patient how accurately their own number reflects their own future risk.[1]
The strongest test of whether an intervention can change biological age comes from CALERIE-2, a randomized trial of two years of caloric restriction in healthy, non-obese adults. A secondary analysis applied several methylation clocks to blood samples from that trial. The result was a small slowing of one clock, DunedinPACE, in the caloric restriction group relative to controls. Other clocks measured in the same trial, on the same participants, did not show a statistically significant change.[2]
The trap: a clock moving is not a proven outcome
CALERIE-2 is a well-designed, well-powered randomized trial, which makes its mixed result more informative, not less. If biological age were a single, well-established property that any competent clock could detect, the different clocks applied to the same blood samples from the same people should have agreed. They did not. That disagreement is the clearest evidence available that these scores are still immature research instruments, not interchangeable readouts of one true biological quantity.[2]
A number changing is not the same as a benefit proven
Even where a clock moves in the hoped-for direction, that is a change in a statistical model's output, not a demonstrated reduction in disease, disability or death. CALERIE-2 did not report those hard outcomes for its epigenetic sub-analysis, and no trial to date has shown that deliberately changing a clock score extends life or prevents disease.
What remains uncertain
A 2025 methodological review examined epigenetic clocks against the standard criteria expected of any clinical biomarker, including test-retest reliability and a clear signal-to-noise ratio at the individual level. Its conclusion was blunt: current clocks do not meet those criteria, and using them to guide an individual's clinical decisions, rather than to study populations, can be uninformative and potentially harmful.[3]
This is a critical review rather than a new dataset, and it does not overturn the population-level associations shown by GrimAge or the modest CALERIE-2 signal. What it does is directly counter the marketing claim that a single biological age test result is ready to tell one person, with clinical confidence, how fast they are aging or whether a program is working.[3][1][2]
What this means for you
AION does not treat a biological age clock result as proof that a patient is aging faster or slower than their peers, and does not use a change in that score to claim that a program has worked. Individually validated diagnostics, cardiovascular markers, metabolic panels and imaging, remain the primary basis for physician decisions.
Where an epigenetic or biological age test is offered, the physician frames it to the patient as exploratory context rather than a diagnosis, explains that different clocks can disagree, and does not attach a fixed protocol to the result. Whether the test is ordered at all, and how any result is interpreted alongside a patient's actual clinical picture, remains a physician decision made case by case.[3]
Source register
Every material source used in this review, with the study design and the limitation that matters when interpreting it.
- [1]Lu AT, Quach A, Wilson JG, et al. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging (Albany NY). 2019;11(2):303-327.
Aging (Albany NY) · 2019 · Biomarker development and validation cohort study
PMID 30669119 · DOI 10.18632/aging.101684
- WHAT IT ADDS
- GrimAge outperformed earlier DNA-methylation clocks in predicting time-to-death and time-to-disease in large validation cohorts.
- LIMITATION
- Association and prediction study only, at the level of aggregate cohorts. Does not establish that the score is a reliable individual-level clinical test.
- [2]Waziry R, Ryan CP, Corcoran DL, et al. Effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial. Nat Aging. 2023;3(3):248-257.
Nature Aging · 2023 · Secondary analysis of a randomized controlled trial (CALERIE-2, n=220)
PMID 37118425 · DOI 10.1038/s43587-022-00357-y
- WHAT IT ADDS
- Two years of caloric restriction produced a small slowing of DunedinPACE relative to controls; other clocks measured in the same trial did not show significant changes.
- LIMITATION
- Small, clock-inconsistent effect. The trial was not powered around epigenetic endpoints, and no hard clinical outcome, such as disease incidence or mortality, was measured for this sub-analysis.
- [3]Apsley AT, Etzel L, Ye Q, Shalev I. From population science to the clinic? Limits of epigenetic clocks as personal biomarkers. Epigenomics. 2025;17(18):1447-1461.
Epigenomics · 2025 · Critical and methodological review
PMID 41403206 · DOI 10.1080/17501911.2025.2603880
- WHAT IT ADDS
- Argues epigenetic clocks fail to meet standard criteria for clinical biomarkers at the individual level, and that individual-level clinical use can be uninformative and potentially harmful.
- LIMITATION
- Narrative and critical review rather than new primary data, though it directly addresses commercial claims of clinical readiness for these tests.
Research changes the question.
A physician owns the answer.
This review is educational and remains marked for clinical review. It does not determine whether any therapy is appropriate for an individual patient.