PhenoAge explained: the clinical-biomarker and epigenetic versions

Prepared by: Research Desk. Sources are linked for verification.

A "PhenoAge" figure on a biological-age report could mean one of two things. The name covers a blood-chemistry score called Phenotypic Age and an epigenetic clock called DNAm PhenoAge. Both come from the same 2018 paper by Morgan Levine and colleagues [1], and both are expressed in years.

Both rest on the same idea: a useful aging clock should track differences in mortality and healthspan, not just the number of birthdays. But they are built from different raw data, and they are not interchangeable.

This explainer covers how each version was constructed, what the original study reported, and where PhenoAge fits among other widely cited aging clocks.

Why PhenoAge was built differently

Steve Horvath published the first widely used multi-tissue epigenetic clock in 2013. It was trained to predict chronological age from DNA methylation across many human tissues and cell types [2]. That design produces a clock that estimates calendar age very accurately. But calendar age is already known.

Levine and colleagues argued that a clock trained only on chronological age would miss part of what separates people of the same age who differ in lifespan and healthspan [1]. Their alternative was a two-step design:

• Step one: build a composite measure of "phenotypic age" from clinical measures that capture mortality risk.

• Step two: train a methylation clock to predict that composite, rather than chronological age [1].

Step one: Phenotypic Age from blood chemistry

The first step uses no DNA at all. Phenotypic Age combines chronological age with nine blood-based measures, selected for their association with mortality in a large U.S. population survey dataset [1]:

• albumin

• creatinine

• glucose

• C-reactive protein (an inflammation marker)

• lymphocyte percentage

• mean red blood cell volume

• red cell distribution width

• alkaline phosphatase

• white blood cell count

Together these markers span liver and kidney function, glucose metabolism, inflammation, and immune and blood-cell characteristics. The model weights them into a single mortality-risk score, then rescales that score into years so the output reads as an age.

A Phenotypic Age above chronological age means something specific. In the training data, that blood-chemistry profile resembled the profiles of older people with higher mortality risk.

All nine inputs are standard laboratory measurements. That is why Phenotypic Age is the version people often calculate themselves from existing blood results, using the published formula.

That accessibility comes with practical caveats. These are general observations about working with any weighted formula, not findings reported in the original paper:

• A formula built on specific units will produce a misleading output if a marker is entered in different units.

• Because each input carries a fixed weight, one unusual reading can move the result. A short-lived rise in an inflammation marker, for example, would feed straight into the score.

Step two: DNAm PhenoAge, the epigenetic version

In the second step, the researchers used Phenotypic Age as the training target. They then searched whole-blood DNA methylation data for the combination of sites that best predicted it [1]. The resulting clock, DNAm PhenoAge, uses 513 CpG sites.

The output is again in years. But it now estimates the blood-chemistry Phenotypic Age from methylation data rather than measuring blood chemistry directly.

That distinction matters. DNAm PhenoAge is a methylation-based proxy for a phenotype, not a copy of it. The two versions are related, but they can disagree for the same person, since each draws on different biological signals and different sources of measurement noise.

What the 2018 study reported

The original paper reported that DNAm PhenoAge outperformed earlier epigenetic measures at predicting a range of aging outcomes [1]. These included:

• all-cause mortality

• cancers

• healthspan

• physical functioning

• Alzheimer's disease

These were associations observed in human cohort data. They show that the score carries predictive information. They do not show that changing the score changes those outcomes.

Two further findings shape how the clock is discussed today:

1. Tissue reach. DNAm PhenoAge was developed in whole blood. Even so, the authors reported that it correlated strongly with age in every tissue and cell type they tested [1].

2. Biology behind the signal. In a transcriptional analysis, higher epigenetic age relative to chronological age was associated with more activity in pro-inflammatory and interferon pathways. It was also associated with less activity in transcriptional and translational machinery, DNA damage response and mitochondrial signatures [1].

The inflammatory link fits a wider research theme. A 2024 review describes the senescence-associated secretory phenotype (SASP): senescent cells release inflammatory and other signalling factors, with both physiological and pathological implications [5].

That does not make PhenoAge a direct measure of cellular senescence. It does offer context for two things: why an inflammation marker appears in the blood-chemistry version, and why inflammatory signatures show up alongside the epigenetic one.

Where PhenoAge sits among aging clocks

Like GrimAge, PhenoAge was trained on health-related outcomes rather than on calendar age alone [1][3]. The later clocks took that idea in different directions:

• GrimAge moves further toward mortality. It combines methylation-based surrogates of plasma proteins and smoking pack-years, and it was trained on time-to-death [3].

• DunedinPACE takes another route. It was trained on the pace of change in multiple biomarkers, measured repeatedly over two decades in a single birth cohort. It reports a rate rather than an age [4].

Our biological age test guide has a fuller side-by-side of these clocks and explains how methylation arrays are read.

Reading a PhenoAge result carefully

Several practical points follow from how the two versions were built. The last two are general measurement considerations rather than findings from the cited studies:

• Identify the version. A "PhenoAge" line on a methylation report is DNAm PhenoAge. A number computed from a blood panel is Phenotypic Age. Tracking one against the other over time mixes two different instruments.

• One reading is one data point. Both versions were developed to predict outcomes across populations [1]. A single result also reflects measurement noise and short-term biological variation, so a difference between two tests is not automatically a real change.

• Method consistency. Laboratory methods and array platforms are not guaranteed to produce identical methylation readings. For any trend, keeping the lab and method constant gives a more like-for-like comparison than switching providers.

• Regulatory status. Many laboratory-run tests, including some consumer methylation tests, fall under the laboratory-developed test (LDT) category.

• FDA's LDT page describes the May 2024 rule that would have phased LDTs into device oversight, and its vacatur by a federal court in 2025 [6].

• A September 2025 Federal Register notice implemented that vacatur [7].

• CLIA certification, administered through CMS, concerns laboratory testing quality standards [8]. It does not mean an aging score has been validated for interpreting any one person's result.

Our page on whether biological age tests are FDA approved covers that regulatory history in detail.

PhenoAge in intervention research

Epigenetic clocks have drawn interest as research endpoints because human lifespan studies take decades. A 2018 review described DNA methylation clocks as an emerging way to evaluate anti-aging strategies much faster than waiting for mortality outcomes [9]. Because it was trained against a mortality-linked composite, PhenoAge is one of the clocks that could be considered for such a role.

The caveat is the one noted above. The 2018 PhenoAge study established predictive associations. It did not establish that moving the score moves health outcomes [1].

When a study reports a shift in DNAm PhenoAge, it is reporting a shift in a biomarker. Whether that shift corresponds to any change in mortality or healthspan is a separate question, and a biomarker result alone cannot answer it.

Our approach to sourcing, and how the site is funded, are described on the about page.

Beyond methylation: other candidate aging markers

Methylation is not the only readout researchers are exploring:

• A 2020 study built a proteomic atlas of senescence-associated secretomes, largely from cultured human cells, as a resource for aging biomarker development [10].

• A 2025 human study mapped the proteomic landscape of multiple aging phenotypes [11].

These lines of work echo the PhenoAge logic: anchor a biomarker to phenotypes that matter rather than to age alone. The difference is that they use proteins as the measured signal. For now, they remain research tools rather than consumer products.

Further reading

A good aging-biology book is a practical way to understand the science behind clock numbers before comparing test reports. Useful neutral criteria for choosing one:

• Recency. Methylation clocks changed rapidly after 2013. Titles published before 2018 predate PhenoAge and the later outcome-trained clocks.

• Sourcing. Titles that cite primary studies let claims be traced back to the data.

• Balance. The strongest books clearly separate what studies measured from what authors speculate.

• Depth and format. Textbook-style overviews suit researchers, while narrative science books suit general readers. Sample chapters show the level before purchase.

Whatever the format, a book is most useful read alongside the original papers listed under Sources below.

Frequently asked questions

Is PhenoAge the same as Phenotypic Age?

Not exactly. Phenotypic Age is the blood-chemistry score built from nine clinical markers plus chronological age. DNAm PhenoAge is the epigenetic clock trained to estimate it from 513 methylation sites [1]. They come from the same paper but use different samples and can give different numbers.

Can PhenoAge be calculated from a standard blood panel?

The blood-chemistry version uses common laboratory markers, so it can be computed from existing results with the published formula [1]. As with any weighted formula, entering a marker in the wrong units or including one atypical value can shift the output. The epigenetic version requires a DNA methylation test.

How is PhenoAge different from GrimAge and DunedinPACE?

PhenoAge is trained on a clinical-marker composite of mortality risk. GrimAge was trained on time-to-death using protein and smoking surrogates [3]. DunedinPACE estimates a pace of aging from repeated biomarker measurements [4]. Each clock answers a slightly different question, which is one reason reports listing several clocks rarely agree.

Last updated: 19 October 2026.

Related on this site

• BioAge Peptide home

• Are biological age tests FDA approved? How aging tests are regulated

• Biological age tests: how epigenetic clocks work and what they mean

• GrimAge explained: how the mortality-trained epigenetic clock works

• Biological age test accuracy: reliability, noise and repeat results

• How epigenetic clocks work: DNA methylation and age estimation

• First, second and third generation epigenetic clocks, explained

• The Horvath clock: the 2013 multi-tissue epigenetic age estimator

Not medical advice. This page reports published research and regulatory records for information only.

Research-only notice: Content on this site is provided for research and educational purposes only and is not intended for human use. It is not medical advice.

Translation notice: Some content on this site may have been machine-translated using Google Translate and may contain errors.