The exact arithmetic behind blood-based biological age: the nine PhenoAge markers, where the coefficients come from, and what your number does and doesn't mean.
Got a standard blood panel lying around? Good — that's all the raw material a biological age needs. No quiz, no guessing app. Just published arithmetic. Here's what actually goes into the number.
A blood-based biological age needs these nine values, all routine on standard panels:
Plus two things that are not blood: your chronological age and your sex. In the PhenoAge model — published by Morgan Levine and colleagues in Aging, 2018 ("An epigenetic biomarker of aging for lifespan and healthspan", Aging 10(4):573–591) — these eleven inputs go into a two-step formula: first a linear combination of the markers produces an intermediate score, then a Gompertz-style transformation converts it to years.
None of this is a secret. The coefficients sit there in the paper, printed. The BioAge R package (Kwon & Belsky, 2021) implements them as `phenoage_calc`. Any tool that refuses to show you the math is asking you to trust a black box instead.
The original cohort behind the formula was NHANES IV — more than 11,000 participants. In that population, higher blood-age relative to calendar age tracked with higher all-cause mortality. That is a statistical association across thousands of people, measured over years.
It is not:
What it is good for: tracking. One draw gives you a position. A second draw a year later gives you a direction. The trend is the information — especially when something in your routine changes.
Unit errors kill quietly. Glucose in mg/dL is about 18× the same reading in mmol/L. Creatinine in mg/dL is 88.4× µmol/L. A calculator that accepts both without complaint will hand you a nonsense age sooner or later. Delta rejects unrecognized units rather than converting them — a panel that cannot be read safely is a panel that should not be scored.
Missing markers get imputed somewhere. If a lab panel skips alkaline phosphatase, some calculators fill in a population average and still print a confident number. That number is now partly about the average person, not you. The honest move is refusing to score, or saying plainly how much of the input was imputed.
One draw is noise. Inflammation markers move with a cold. Creatinine moves with dehydration. A single reading is a snapshot with error bars; the response between readings is the signal worth watching.
Draw once. Wait a year. Draw again. The change in the number — not the number — is what tells you whether what you are doing is moving anything. If you want a faster read, the DELTA protocol brackets a defined 48-hour fast with three draws and reads the response curve, with a hard safety screen first: no fasts for anyone under 21, type 1 diabetes, pregnancy or breastfeeding, a history of disordered eating, advanced kidney/liver/cardiac disease, or concurrent SGLT2 inhibitors, insulin, or sulfonylureas.
The measurement is honest arithmetic. The interpretation is a conversation with your clinician — it was never supposed to be anything else.
