“Biological age” is one of the most oversold numbers in consumer health. It is also, done carefully, genuinely useful. The difference is entirely in the method — so here is ours, in full, including the parts that are unflattering.

The method

We use the Klemera–Doubal method, published in 2006 and still the most defensible approach in the literature. The idea is simple. Take a large reference population. For each blood marker, work out how it changes with age across that population. Then, for one person, ask: given this set of results, what age does this bloodwork most resemble?

It is a statistical estimate, not a measurement. Nothing in your blood knows how old you are.

Fitted on national data, not on marketing

We fitted our model on the National Health and Nutrition Examination Survey — the CDC programme that has been measuring the health of a representative sample of Americans for decades. Our reference population is 5,142 adults aged 20 to 79. The data is public, which means anybody can check our work.

The result correlates with chronological age at 0.95 in the validation, and mean biological age tracks real age within about a year in every decade from the twenties to the seventies. That last part matters: a model that says everyone is 45 will look accurate on average and be useless for individuals.

The markers that carry the signal

Ranked by how strongly each moves with age in the national data:

HbA1c — average blood sugar. The strongest single age marker we found.
Ferritin — iron storage, which climbs with age.
Creatinine — reflects kidney filtration.
Alkaline phosphatase — liver and bone turnover.
Triglycerides — fat metabolism.
Albumin — the one that falls rather than rises.
Uric acid, total cholesterol and hsCRP — weaker, but contributing.

The finding that surprised us

HDL and LDL cholesterol carry almost no age signal at all. In the national data, HDL correlates with age at 0.045 and LDL at 0.030 — statistically indistinguishable from nothing. The two numbers most people think of as their “health score” tell you close to zero about biological ageing.

They are still worth measuring. They are meaningful cardiovascular markers and they belong on a panel. They simply cannot carry a biological age, and any product implying otherwise is not being straight with you.

Two markers we deliberately refuse to use

eGFR and free testosterone are both excluded, because both are calculated from age or from other markers already in the model. Feeding an age-derived number back into a model that predicts age is circular. It would make our correlation look better and mean less.

And one we removed on purpose

Vitamin D was the second-strongest signal in our first model. We took it out. In the national data, older adults have higher vitamin D — which is almost certainly because older adults are more likely to take a supplement. That is a measure of behaviour, not of ageing.

Removing it cost us real precision. It was the honest thing to do anyway. If a number can be improved by taking a pill that does not change how you are ageing, it is not measuring what it claims to measure.

How to read your own result

A few years either side of your real age is completely ordinary.
It is not a diagnosis, not a life-expectancy figure, and not a medical test.
One reading is a point. Two readings a year apart are a direction, which is far more useful.
If a single marker is out of range, that individual result matters more than the composite — take it to a clinician.

We show you which markers went into your number and the formula behind anything calculated, because a score you cannot interrogate is a score you should not trust.

What a Biological Age Result Actually Measures — And What It Does Not