In partnership with
2 minutes. Your URL. A customer profile worth using.
Most founders can describe their product. They can't describe their customer. Not in a way that actually changes how they sell.
HubSpot for Startups built a free tool to fix that. Paste in your URL, answer a few quick questions, and it generates a structured profile of your best-fit customer. Firmographics, buying triggers, the works.
Takes 2 minutes. No spreadsheet required.
LONGEVITY LATEST ISSUE 27 · 9 SEPTEMBER 2026
LONGEVITY LATEST · DEEP DIVE
Before You Buy a Biological Age Test
What the number means, why it can move, and when it should change a decision.
By Christian Thomsen · Companion to Issue 27 · 9 September 2026 · ~6-minute read
Imagine opening a report at 54 and seeing a biological age of 46. It feels like good news. Another model analyses that sample and gives you 57. The temptation is to decide one must be wrong, preferably the second.
This example is illustrative. It shows why a familiar word, “age”, can make different statistical outputs look interchangeable. We understand a birthday. We have to learn what each model means.
My buying rule: explain the measurement, its uncertainty and its intended use before telling me whether the result is flattering.
Before paying, ask about six things: the model, who it was validated in, its uncertainty, retesting, the decision it changes and control of your data. The full questions are on the final page. First, here is how to judge the answers.
Three claims that need separate evidence
Claim | Evidence needed |
The score is repeatable. | The same material produces sufficiently similar results under stated laboratory conditions. |
The score predicts health. | It predicts relevant future outcomes in people like the intended customer. |
Acting on the score improves health. | A test-guided strategy improves meaningful outcomes compared with appropriate usual care. |
A company may have evidence for the first two and still lack the third. Predicting risk and proving benefit from changing a score are different jobs.
FDA guidance makes the distinction for surrogate endpoints: using a marker in place of a clinical outcome requires substantial evidence that it represents clinical benefit in the relevant setting. A correlation alone does not complete that validation.
For the buyer: when you hear “scientifically validated”, ask which of these three claims was tested.
Read the label before the number
PhenoAge: the name has two meanings
Clinical Phenotypic Age combines chronological age and nine routine blood biomarkers, including glucose, albumin and C-reactive protein. DNAm PhenoAge uses methylation to estimate that clinical construct. The original work examined associations with mortality and other ageing-related outcomes.
Before buying: ask whether the service uses blood chemistry or DNA methylation. Entering familiar blood-test results into a calculator does not produce the same measurement as a methylation laboratory. Neither replaces interpretation of an abnormal clinical result.
GrimAge: risk expressed in years
The original GrimAge combined methylation-based estimates of smoking exposure and selected proteins with age and sex to predict mortality risk, then expressed its output on an age scale. It is not a countdown.
Before buying: ask what health decision the risk estimate supports. The “years” cannot tell you how much life a treatment would add.
DunedinPACE: a rate rather than an age
DunedinPACE used two decades of repeated physiological measurements in the Dunedin cohort to develop a pace measure, then estimated it from blood methylation. Around 1 represents the reference pace. A value of 0.90 is below that reference, not “ten years younger”.
The original study reported strong technical reliability and associations with morbidity, disability and mortality. Those findings support its research role. They do not show that moving your own result from 1.00 to 0.90 adds 10% to your lifespan.
Before buying: check the units. A pace result and an age estimate cannot be compared as if they were the same thing.
Why the answers differ
Models use different inputs and targets. Think of a credit score and a monthly cash-flow statement: both say something about finances, but they answer different questions.
Stop shopping for the lowest age. Decide what you want to learn, then ask whether this test can tell you.
What does a lower follow-up result prove?
Suppose an illustrative report reads 56 at baseline and 53 three months later. You started exercising, lost weight and added several supplements. The score fell by three years. You still do not know what would have happened without those changes, or which change caused the difference.
Noise: could another test tell another story?
Higgins-Chen and colleagues found substantial differences between repeat measurements of the same material, then improved agreement using principal-component versions of the clocks. A published model is therefore only part of the story; the complete service must perform reliably.
Ask: “If you tested my sample twice, how different might the answers be?” A company may show that its test ranks people consistently while the repeated numbers for one person still differ enough to confuse a follow-up.
Then ask how large a change between visits must be before they regard it as interpretable. Both measurements contribute uncertainty. If the provider cannot explain that in ordinary language, do not let a small change drive a large purchase.
Why a trial average is different
CALERIE found a favourable DunedinPACE result alongside non-significant PhenoAge and GrimAge findings. Reading all the results gives a more useful picture than selecting the lowest-looking clock.
A controlled trial can detect a small group effect that would be difficult to identify in one customer. That does not invalidate the trial, or guarantee that your own retest can detect the same benefit.
Five checks on an age-reversal headline
Endpoint: was the clock or health outcome chosen before the results were known?
Comparator: was there an appropriate control group measured over the same period?
Full result: how many clocks were tested, and which did not move?
Magnitude: is this change within one group, or the difference between groups?
Meaning: did researchers measure disease, function or survival, or only a biomarker?
Use those checks on a diet, supplement or prescription-drug claim. They help you decide what the study supports before you decide what to buy.
The six questions to ask before paying
Keep this list beside the sales page. A useful answer should be specific enough to compare with another provider’s answer.
Which test? Give me the exact model, version, specimen type and published validation paper. Explain any changes from the published method.
Whose validation? Was it evaluated in people of my age and background, and does that evidence apply to the sample type you collect?
How much uncertainty? Show repeat-measurement differences and explain how large a follow-up change must be before you regard it as interpretable.
What happens on retesting? How are laboratory batches and algorithm updates handled? Can earlier samples be recalculated if the model changes?
Which decision? What action follows from an older score? Is there evidence that making that decision because of the test improves outcomes?
Who controls the data? Explain storage, reuse, sharing and deletion, including what happens to the physical sample.
These are questions, not assumptions about every provider. If an answer is missing, keep that uncertainty in the buying decision.
When curiosity is enough
Buying out of curiosity is a legitimate choice. Call it that, set a budget and decide beforehand how you will handle an unwelcome result. Otherwise one purchase can become a cycle of retests and products bought to improve an uncertain number.
For routine health decisions, I would prioritise clinically interpreted risk factors, symptoms, function and established preventive care. A young-looking clock should not excuse an abnormal conventional result. An older-looking one should not frighten you into an unproven treatment.
The science is interesting. The purchase needs a purpose. Make the test explain how it helps with the decision in front of you.
Sources and further reading
Evidence reviewed through 6 September 2026. Numerical examples are hypothetical. Educational content only; do not use biological age scores independently to diagnose illness or change treatment.
© 2026 FrontWave Media Ltd · Longevity Latest
Pique’s new X·E Fountain energy drink delivers clean, caffeine-free energy without the jitters, crash, or dehydration. Vitamin B Complex, magnesium, and electrolytes help support steady focus, while crisp cucumber-celery-lime makes every sip ridiculously refreshing. Get 15% off for life.