You run a blood-based biological age calculator and get one number. A few months later, you send in a DNA methylation kit and get a different number — sometimes a couple of years off, sometimes more. Neither reading is a mistake. "Biological age" isn't one measured quantity that different tools happen to read differently, like two thermometers checking the same room. It's a summary score, and different tools build that summary from different raw material.
What "biological age" actually means
There's no single biological structure anyone can measure that directly outputs "biological age" the way a thermometer outputs temperature. Every biological age tool starts by picking a set of measurable biomarkers, then trains a statistical model — usually against real-world outcomes like time-to-death in a large cohort — to combine those biomarkers into one score, calibrated to read in years. The choice of which biomarkers go in, and what outcome the model is trained against, is what actually defines the score. Change either one, and you get a legitimately different measurement, not a more or less accurate version of the same one.
The blood-based version: what a score like PhenoAge is built from
The Longevity Lab's biological age estimate follows the same approach as PhenoAge, one of the more widely validated blood-based aging scores. It's built from nine routine clinical measures — albumin, creatinine, glucose, C-reactive protein, lymphocyte percentage, mean cell volume, red cell distribution width, alkaline phosphatase, and white blood cell count — combined with chronological age into a single score. It was trained directly against mortality data from a large national health survey (NHANES III), so its "years" are calibrated against how those exact biomarker patterns predicted actual survival in that population (Levine et al., 2018). It's a snapshot of organ-system function and inflammation, using markers a standard blood panel already captures.
The epigenetic version: a different measurement entirely
DNA methylation clocks measure something biologically distinct: the pattern of methyl groups attached at specific locations along your DNA, at specific CpG sites that shift in predictable ways with age. Different epigenetic clocks are built for different jobs — some (like the original Horvath and Hannum clocks) were trained to predict chronological age itself, while others were trained to predict something else entirely and only incidentally correlate with age (Field et al., 2018). One clock in this second category, "DNAm PhenoAge," was built specifically to approximate the blood-based PhenoAge score using only methylation data — it's a methylation-based stand-in for the blood score, trained to predict it, not a second independent measurement of the same underlying biology (Levine et al., 2018).
Why they don't always agree
Because these are different measurement classes trained against different targets, disagreement isn't a flaw — it's expected. A 2021 analysis looked at the genetics underlying two different blood-based aging measures and found they were associated with meaningfully different sets of genetic variants, pointing to at least partially distinct underlying biology even between two scores built from similar blood chemistry (Kuo et al., 2021). A 2026 analysis went further, comparing five widely used epigenetic clocks side by side and finding they tracked distinct biological pathways and gene-expression signatures of aging, rather than converging on one shared "true" aging process each clock approximates with more or less noise (Arpawong et al., 2026).
| Measurement type | Built from | Trained to predict |
|---|---|---|
| Blood-based (e.g., PhenoAge) | 9 routine clinical biomarkers + chronological age | Mortality risk in a real-world cohort (NHANES III) |
| Epigenetic — chronological-age clocks (e.g., Horvath, Hannum) | DNA methylation at specific CpG sites | Chronological age itself |
| Epigenetic — outcome-trained clocks (e.g., DNAm PhenoAge, GrimAge) | DNA methylation at specific CpG sites | A different target score or outcome (e.g., approximating PhenoAge, or mortality directly) |
Why this is easy to miss
Consumer framing tends to present "biological age" as if it were one objective fact about a person's body, since it's easier to market a single number than to explain a scoring methodology. That framing makes disagreement between two legitimate tools look like one of them must be wrong. In practice, the more useful question isn't "which number is correct" — it's "what is this particular score actually built from, and does that map onto something I can act on."
How to actually use this
- Don't directly compare a score from one method to a score from a different method — a PhenoAge-style blood score and a DNA methylation clock aren't measuring the same thing, even when both are called "biological age."
- Track trend within the same method over time. A blood-based score moving in one direction across a year is more informative than comparing that score, once, to a differently-built epigenetic number.
- Look at what's actually driving the number. A blood-based score is only as informative as its underlying biomarkers — if inflammation (CRP) or kidney function (creatinine) moved, that's the real, actionable signal, and the summary score is just a convenient way of packaging it.
- Treat any single biological age number, from any method, as a starting point for a conversation about specific biomarkers — not a verdict.
When to actually check in with someone
No biological age score — blood-based or epigenetic — is a diagnostic tool, and none of the research above supports using one to self-diagnose a health condition. If a score (or one of its underlying biomarkers) looks meaningfully off to you, that's worth raising directly with a doctor or qualified healthcare provider, who can interpret it alongside your actual medical history rather than as an isolated number.
