Biological Age Tests: What the Number Can and Cannot Tell You

A patient of fifty-two puts a handsomely designed report on the desk. The first line gives his biological age as 46.3 — almost six years younger than the date on his identity card. He paid a good deal for the test, and the question he asks is exactly the right one: what now. The same appointment turns up a blood pressure of 148 over 92, no treadmill in a decade, and an ApoB nobody has ever measured.

Two different things hide under one name. The first is sound biology: people of the same chronological age grow old at markedly different rates, and the difference is detectable and predicts outcomes. The second is a consumer product: one number, to one decimal place, sold as a personal verdict. The first deserves serious attention. The second runs several good years ahead of the evidence, and that is the distinction worth keeping in mind throughout.

The idea itself is right — ageing does not run at one speed

Take ten sixty-five-year-olds from the same neighbourhood. One rides eighty kilometres on a Saturday morning; another struggles to get out of a low chair without pushing off with his hands. The difference between them is not merely a matter of lifestyle, and it is not a matter of appearance. It predicts hospital admissions, fractures, disability and death better than anything printed on an identity card. Geriatricians have worked with it for decades under the name frailty, assessing it with remarkably dull instruments — gait speed, grip strength, unintentional weight loss, self-reported exhaustion.

That chronological age is a crude proxy for the rate of ageing is therefore neither controversial nor new. The open question is narrower: whether that rate can be compressed into one number, from one tube of blood, precisely enough to justify an individual decision. That is where the argument sits, not in the idea.

Four families of biological age test

The most discussed method is the epigenetic clock. The DNA sequence stays largely fixed across a lifetime, but the methylation pattern laid over it — methyl groups attaching to CpG sites and shaping gene expression — changes with age in a fairly orderly fashion. An algorithm trained on hundreds of those sites returns a number. What decides the value of that number is not the biochemistry but a single question: what the algorithm was trained on.

First-generation clocks, the best known published in 2013, were trained to predict chronological age itself — a strange target, since a perfect clock of that kind tells you nothing your identity card does not, and its clinical value lies entirely where it disagrees with the card. The second generation, PhenoAge and GrimAge, was trained on clinical markers and on mortality instead, and is indeed more strongly related to health outcomes. A third family, the pace clocks, led by DunedinPACE, came out of long-term follow-up of a birth cohort and reports not an age but a rate: how many years of biological change you accumulate in each calendar year.

Methylation is not the only route. Glycan-based measures examine the sugar chains on IgG antibodies, whose patterns shift with age and with inflammatory state. Proteomic panels scan thousands of plasma proteins and try to estimate the ageing of different organs separately — one of the more interesting directions in the field, and one of the youngest clinically. The fourth family, least glamorous, is simply the functional measures: VO₂max, grip strength, gait speed, time to rise from a chair.

What a biological age test can actually tell you

In large cohorts, second-generation clocks are associated with all-cause mortality and cardiovascular disease, and in some work with cancer incidence and cognitive decline — even after adjustment for chronological age. That is a real finding, and not trivial. It is also a population-level one: several thousand people whose clocks run ahead will, as a group, experience more events than a comparable group. It does not say, and cannot say, that a particular person running three years ahead is on a worse trajectory than his neighbour whose result matches his age.

The gap between a correlation at group level and a decision for one person is not peculiar to this field. ApoB also predicts events at population level. But ApoB has decision thresholds, an available intervention, and randomised trials showing that lowering it reduces events. The biological age number has none of the three: no professional society guideline says what to do with a result, and no evidence shows that lowering the number lowers the risk. It is not yet an established surrogate measure — and medicine is full of surrogates that looked convincing and did not hold up.

The methodological problems that do not appear on the report

Reliability between samples. This is the problem people seldom discuss, and a particularly hard one to ignore. Studies that examined it found that repeat samples from the same person — sometimes a split of one and the same blood draw — produce differences of years in the number returned. Versions built on principal components rather than individual CpG sites are considerably more stable, but not everything sold to consumers uses them, and you will usually not know which was used.

Cell composition in the sample. Methylation is measured from a mixture of white blood cells, each type with a pattern of its own. A shift in the ratio of neutrophils to lymphocytes — acute illness, an infection that resolved a fortnight ago, sometimes an unusual physical load — moves the result without your rate of ageing having changed at all.

The absence of a standard. Every laboratory has its own chip array, its own normalisation and its own proprietary algorithm. Results cannot be compared between laboratories, and sometimes not with one from the same laboratory two years ago, if the algorithm has since been updated. There is no agreed unit here of the kind millimoles per litre provides.

And the conceptual problem, the gravest of them. The number is not the source of the information but a product of it. GrimAge, for example, draws much of its predictive power from surrogate measures of smoking and of plasma proteins related to inflammation and to metabolic function. In other words, it reconstructs from the methylation pattern things that can be measured directly, cheaply and without algorithmic uncertainty. If blood pressure, ApoB, HbA1c, fitness and body composition have already been measured, what does a number summarising them add, when it cannot say which of them to work on?

A number that can move by years between two samples from the same person, that cannot be compared across laboratories, and for which no clinical guideline says what to do with it, is not a diagnosis. At best it is a measure of trend.

What can be measured today, and improved as well

The measures most strongly related to healthy ageing are available already, can be measured repeatedly and reliably, and are mostly modifiable. These are what we measure before clocks enter the conversation:

• VO₂max — aerobic capacity. Of the single measures, its association with mortality is among the strongest recorded, it comes from an objective measurement, and it improves with structured training at almost any age.
• Grip strength and strength generally — an inexpensive proxy for whole-body strength and neuromuscular integrity, and a long-standing predictor of disability and death.
• Body composition, and visceral fat in particular — a measure that explains much of what weight alone misses, and responds to diet and training within months.
• The metabolic profile — ApoB, HbA1c, fasting insulin, triglycerides. They have thresholds, interventions and trials.
• Blood pressure — the least impressive parameter in preventive medicine and among the most consequential, and still often taken in haste, without repetition and without home readings.

Each of these passes tests that the epigenetic clocks still do not: reliable repeat measurement, thresholds derived from clinical outcomes, and a known way to move it. The limitation is worth stating honestly here too — for blood pressure and ApoB the evidence that moving the marker moves the risk comes from randomised trials; for fitness, strength and body composition it is largely observational, though highly consistent across decades and populations.

The Israeli context

In Israel, biological age tests are almost always sold privately, mostly at lifestyle clinics and private institutes. As far as we are aware they are not included in the national health basket — the publicly funded list of covered services — and the arrangements under the supplementary insurance plans offered by the health funds (kupot holim) vary between funds and are revised periodically. Worth checking directly with your own fund rather than relying on an assurance from whoever is selling the test.

Meanwhile, much of what carries the real value is already within reach. Blood pressure, a lipid profile, HbA1c and renal function are generally available through routine follow-up with your family physician, though the scope, the frequency and the co-payment conditions vary between funds and with your risk factors — worth checking what is covered for you at yours. Executive health screenings, a common workplace benefit, usually include an exercise test, from which a rough estimate of fitness can be derived. The gap between what people pay for and what changes outcomes is unusually visible here, and many of the people we meet have spent money on the first without completing the second.

Where we stand, and how to use this in practice

We offer biological age panels — glycan and proteomic — selectively, as an addition within the membership, and always in the same frame: a tool for trend, not a verdict. For someone already measuring the fundamentals and working on them systematically, successive measurements can add a layer of information. For someone who arrives with the number before the fundamentals, it draws attention away from where it is needed. If you choose to do it anyway, these are the conditions that make a result useful rather than misleading:

• Same laboratory, same algorithm, over time. A single result is not information; it is a point.
• Same sampling conditions: morning, fasting, not during acute illness and not in the weeks after an infection.
• At least a year between measurements. A change measured after three months is mostly noise.
• Read direction and consistency, not an absolute value and not a decimal place.
• The test does not replace a screening test, and does not replace a risk factor measured directly.

Two practical conclusions follow. If the result comes back younger than your age, it is no licence to postpone colorectal cancer screening, ignore borderline blood pressure, or forgo direct measurement of fitness and body composition. If it comes back older, the first thing to check is not another panel but blood pressure, ApoB, HbA1c, fitness, sleep and waist circumference — and there, almost always, the explanation is already waiting.

Biological age is one of the good ideas in preventive medicine. Its measurement has simply not caught up with it. Until it does, the numbers we already measure well — and that you can genuinely change — are worth more than any number claiming to summarise them.