A test tells you that you are «32 biologically» instead of 38. The number is flattering, the presentation is polished. What nobody told you: two very different methods share that name, and the cheaper one is, according to the authors of the study that created both, the more predictive.
Two calculations, one name
In 2018, Morgan Levine's team published in Aging the work that became the reference in the field. They first build a Phenotypic Age: out of 42 candidate markers, a penalised regression keeps 9, to which chronological age is added. Then they train a second model, DNAm PhenoAge, which reproduces that result from 513 DNA methylation sites.
Both output a number expressed in years. They do not measure the same thing, do not cost the same, and do not predict equally well.
| Phenotypic Age | DNAm PhenoAge | |
|---|---|---|
| What it reads | 9 blood markers + age | 513 DNA methylation sites |
| Where to get it | Any medical analysis lab | Dedicated methylation kit |
| Cost | That of an ordinary blood panel | Substantially higher |
| Often repeatable | Yes | Hardly |
| Morbidity and mortality prediction | Better, per Levine et al. | Weaker, per the same authors |
The 9 markers, and what each one tells
| Marker | What it reflects |
|---|---|
| Albumin | Liver function, nutritional status, chronic inflammation |
| Alkaline phosphatase | Liver function and bone remodelling |
| Creatinine | Kidney function |
| Serum glucose | Metabolic regulation |
| CRP (log scale) | Inflammation |
| Lymphocyte percentage | Immunity |
| White blood cell count | Immunity |
| Mean cell volume (MCV) | Red cell line |
| Red cell distribution width (RDW) | Red cell line heterogeneity |
That is the real takeaway of this article: you probably already have those nine values. They sit on a standard complete blood count and biochemistry panel, the kind of panel a doctor prescribes without difficulty.
The absolute number is useless. The gap is not.
«32 biologically» means nothing on its own. What the model produces is a position relative to a reference cohort, translated into years to be readable. The only reading that holds is differential: how many years of gap with your real age, and which way that gap moves between measurements.
A −6 year gap measured once is an anecdote. The same gap moving to −4 then −2 over eighteen months is information, and it is worth far more than the initial number.
« A biological age measured once is a photograph. What you care about is the film. »
What the calculation does not say
- It does not say why. It is an aggregate risk score built on cohorts. It flags a position, never a cause.
- It is sensitive to acute events. CRP is produced by the liver in response to inflammation: an ordinary infection at sampling time is enough to raise it, and the score with it.
- It depends on sampling conditions. Glucose measured after a meal does not mean what fasted glucose means. Changing lab also changes assay techniques.
- Its responsiveness to interventions is poorly documented. The authors themselves note that the score's dynamics after an intervention need more research.
How to make it actually useful
- 1
Get sampled fasted
Glucose is one of the nine markers. Without fasting you are not measuring the same thing twice, and the comparison loses all value.
- 2
Stay with the same lab
Assay techniques and reference intervals vary between labs. Over longitudinal tracking, that variation blends into the signal you are looking for.
- 3
Do not test during an infection
Nor in the days after. CRP needs to come back down, otherwise you are measuring your cold.
- 4
Repeat, and log the conditions
Two points beat one, four beat two. Note the time, the fasting, last night's sleep: those are what will explain the odd readings.
- 5
Track the gap, not the number
And always open the detail. A global score that moves without telling you which marker moved is not actionable.
Why this matters to people who run things
A founder instruments their product, their acquisition, their cash. They refuse to fly on intuition on just about everything, except the one resource that produces all the others. The comparison is a little easy, but it holds: a blood panel twice a year is the cheapest dashboard they will ever buy.
Two of the nine markers, glucose and CRP, are among the most responsive to sleep, food and physical activity. They are also the ones whose drift settles in most quietly during heavy periods. Watching the score without watching those two is like reading revenue without reading margin.
What Helix does
Helix computes the blood version, the one that predicts best, from the nine Levine et al. markers. You enter your panel, the score updates, and the history keeps every measurement with its conditions. The per-marker detail is always open: the global number is only a door. And it is cross-checked with the rest, sleep, stack, nutrition, because a drifting glucose rarely tells its story alone.
Helix is an educational optimisation tool, not a medical device. The calculation is reproducible and its sources are public: that is the minimum you owe someone you are handing an age to.
Want to track the gap rather than the number?
Join the beta