Is a PhenoAge below my real age good news?

It means your nine markers look like those of a younger person in the reference cohort. That is associated with lower mortality across large populations. It is not a measurement of your cells, and it says nothing certain about you as an individual — the standard error on the estimate is several years wide.

Why does it need nine blood tests?

Because those are the nine the equation was fitted on. Dropping one and substituting another produces a different number that is no longer the published, validated model. If you are missing a marker, the honest answer is that we cannot compute it.

Which marker moves the result most?

Red cell distribution width and albumin carry the largest weights per unit of typical variation, which surprises most people who expect glucose or cholesterol to dominate. That is one reason PhenoAge is a poor tool for judging whether a supplement worked.

Can I compare my result with an epigenetic age test?

Not directly. PhenoAge from blood chemistry and DNA-methylation clocks are different models with different inputs and only moderate agreement. Two "biological ages" that disagree by a decade is a normal finding, not an error.

Biological Age Calculator (PhenoAge)

Enter your numbers

10 questions. Questions that only apply to some people appear once your earlier answers decide that they do.

From your health data

10 of these questions can be answered from lab results and device data you have already connected: Chronological age, Albumin, Creatinine, Fasting glucose, C-reactive protein, Lymphocytes, Mean cell volume, Red cell distribution width, Alkaline phosphatase, White blood cell count. Connect them once in your member account and they arrive here pre-filled and labelled as coming from your health data — and you can always type over the value.

Set up health data in your member account

This public page never reads your health data. Pre-filling happens inside the signed-in app, which is a separate, access-controlled surface.

Chronological ageFrom your health data — connect to fill this automatically

Albumin(g/L)From your health data — connect to fill this automatically

Typical adult range 35-50 g/L. A US lab reporting 4.3 g/dL is 43 g/L.

Creatinine(mg/dL)From your health data — connect to fill this automatically

Typical adult range 50-110 umol/L. A US lab reporting 0.9 mg/dL is about 80 umol/L. Enter this in mg/dL. We convert it to umol/L{creatinine} for the formula.

Fasting glucose(mg/dL)From your health data — connect to fill this automatically

Typical fasting range 3.9-5.5 mmol/L. A US lab reporting 95 mg/dL is about 5.3 mmol/L. Enter this in mg/dL. We convert it to mmol/L{glucose} for the formula.

C-reactive protein(mg/L)From your health data — connect to fill this automatically

High-sensitivity CRP in mg/L. Values at or above 10 mg/L usually mean acute illness.

Lymphocytes(%)From your health data — connect to fill this automatically

Lymphocytes as a percentage of white cells, not the absolute count. Typical range 20-45%.

Mean cell volume(fL)From your health data — connect to fill this automatically

Red cell distribution width(%)From your health data — connect to fill this automatically

Typical range 11.5-14.5%. This is the single most influential input.

Alkaline phosphatase(U/L)From your health data — connect to fill this automatically

White blood cell count(×10³/µL)From your health data — connect to fill this automatically

In 10^3/uL, which is numerically the same as 10^9/L. Typical range 4.0-10.0.

Nothing you type is stored unless you are signed in and ask us to save it.

How this is calculated

Formula version phenoage-levine2018-1.0.0

Nine blood markers and chronological age are combined with the published Levine PhenoAge coefficients into a Gompertz linear predictor. That predictor gives a modelled 10-year all-cause mortality risk, which is then mapped back onto an age scale by inverting an age-only mortality model: phenotypic age is the age at which the average person in the reference cohort carried the same risk. Inputs are taken in SI units because the coefficients were fitted on SI values; CRP enters as the natural logarithm of the concentration in mg/dL.

What it cannot do

Fitted on NHANES III (US, 1988-1994) adults; absolute phenotypic age carries that population and era.
No individual-level prediction interval is published, so no confidence interval is shown.
We could not locate a journal-published worked example (full input vector plus expected output) to pin a regression test against. The coefficients are cited individually and the back-transformation constants are proven mutually consistent in the test suite. See GAPS.md.
The CRP term is the natural log of mg/dL. This follows the BioAge reference implementation, which divides an mg/L value by ten before taking the log. A ten-fold error here biases every result by about 2.4 years.
Extremely disordered inputs can drive the modelled mortality risk to 1, at which point the back-transformation is undefined. Such inputs are rejected by range validation rather than clamped.
Race, medication use and diagnoses are not inputs; two people with identical panels get identical results.

Sources

Levine ME et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY) 2018;10(4):573-591DOI 10.18632/aging.101414 (opens in a new tab)PMID 29676998 (opens in a new tab)
Liu Z et al. A new aging measure captures morbidity and mortality risk across diverse subpopulations from NHANES IV. PLoS Med 2018;15(12):e1002718DOI 10.1371/journal.pmed.1002718 (opens in a new tab)PMID 30596641 (opens in a new tab)
Kwon D, Belsky DW. A toolkit for quantification of biological age from blood chemistry and organ function test data: BioAge. GeroScience 2021;43:2795-2808DOI 10.1007/s11357-021-00480-5 (opens in a new tab)PMID 34725754 (opens in a new tab)

The formula version is pinned so that a result you saved or shared can be reproduced. When we change a coefficient, the version changes and old results stay attributed to the version that produced them.

Questions people ask

Is a PhenoAge below my real age good news?: It means your nine markers look like those of a younger person in the reference cohort. That is associated with lower mortality across large populations. It is not a measurement of your cells, and it says nothing certain about you as an individual — the standard error on the estimate is several years wide.
Why does it need nine blood tests?: Because those are the nine the equation was fitted on. Dropping one and substituting another produces a different number that is no longer the published, validated model. If you are missing a marker, the honest answer is that we cannot compute it.
Which marker moves the result most?: Red cell distribution width and albumin carry the largest weights per unit of typical variation, which surprises most people who expect glucose or cholesterol to dominate. That is one reason PhenoAge is a poor tool for judging whether a supplement worked.
Can I compare my result with an epigenetic age test?: Not directly. PhenoAge from blood chemistry and DNA-methylation clocks are different models with different inputs and only moderate agreement. Two "biological ages" that disagree by a decade is a normal finding, not an error.