Epigenetic Age Results — What the Number Actually Tells You

A patient came in this week with a one-page report. Chronological age 55, biological age 48, pace of aging 1.12. And a line saying their biological age fell four years in twelve months on a supplement stack. They want to know if the stack works.

That report is measuring something real. The epigenome — the chemical marks sitting on top of the DNA sequence that control which genes get read — genuinely changes with age in reproducible ways. The test is real. The interpretation on that page is not.

What specifically is being measured in a blood draw?

DNA methylation — a methyl group attached to a cytosine at a CpG site, meaning a cytosine-guanine pair in the sequence. When those marks accumulate in a gene's promoter, the gene goes silent. The pattern shifts with age, illness, smoking, adiposity. The clock reads that drift.

So what is the clock itself?

A regression model. It takes methylation values at a few hundred to a few thousand CpG sites and returns a weighted sum fitted to match whatever outcome the builders chose. First-generation clocks, like Horvath 2013, were fitted to chronological age. They don't predict mortality particularly well because they weren't trained to.

And newer ones?

PhenoAge and GrimAge were trained on biological endpoints — mortality, disease incidence. GrimAge in particular has prospective data linking higher clock values to earlier all-cause mortality. That's trial-level epidemiological evidence, not mechanism only. But those are research instruments validated in populations, not in individuals retested over twelve months.

That's the key thing here, isn't it — the four-year change claim. How reproducible is a single person's result?

Poorly. Test-retest variability on the same sample across commercial labs can be several years. Within a lab, batch effects and normalization choices shift the number. A four-year change in one person over one year is well within that noise range for most commercially sold versions. The companies selling the test and the supplement together have an obvious conflict there.

Is there any trial evidence that any intervention actually moves a validated clock?

Some. The Horvath group's TRIIM trial administered growth hormone, metformin, and DHEA and reported reversal of DNAm age on the Horvath clock — that's published, but it's a small pilot, no clinical endpoints, and the regimen is not approved for this purpose. Rapamycin has clock data in animals. Caloric restriction trials in humans show modest signals. None of this justifies the supplement claim on that report.

So what do I actually tell the patient?

Three things. The layer being measured is real and matters to aging biology. The instrument was built as a population research tool and is being retailed as a personal one — that gap is not bridged yet. And the four-year drop is not distinguishable from noise in a commercially sold test with one retesting point.

Should they stop the test entirely?

I wouldn't tell them it's worthless — the science will mature. I'd say don't make financial or medical decisions based on a single-vendor number that also sells the intervention. If they want to track it, use the same lab, same conditions, multiple time points, and interpret trends not single values.

What's the one thing you'd want me to hold onto from this?

The biology is ahead of the product. Epigenetic clocks are genuine research tools with real predictive signal at the population level. They are not yet validated as individual diagnostic instruments. Until that gap closes, the number is hypothesis-generating, not actionable.