What the evidence says about longevity drugs and biological-age tests

What does the evidence on longevity drugs actually say?

Longetic keeps lab reports and a medication list, and this article is about what belongs in them. Across the documents read here, one compound has a repeated animal result, one human trial of it did not move its primary outcome, and on the regulator pages checked no medicine is licensed for ageing at all. Anything taken for longevity is therefore taken off-label, which is exactly why the written record matters.

Three different kinds of claim get mixed together in the same conversation: what happened in mice, what happened in people, and what a test sold online returns. They carry very different amounts of evidence, and the sources below are the primary documents rather than coverage of them.

One note on what this article does not do. It gives no doses, no targets and no protocol, and it tells nobody to start or stop anything. Where two respectable sources print different wording for the same result, both are named.

What happened when rapamycin was tested in mice?

The strongest animal result in this field is a 2009 paper in Nature. Mice fed rapamycin late in life lived longer, and later cohorts in the same testing programme recorded the result again. The widely quoted figures of 14% and 9% are not median lifespan: the paper attaches them to age at 90% mortality, which is a different measure.

Harrison and colleagues, published on 8 July 2009, report that rapamycin "extends median and maximal lifespan of both male and female mice when fed beginning at 600 days of age." Six hundred days is late in a mouse life, and that closing condition is part of the finding rather than a detail attached to it.

Those two figures arrive with their own frame. The paper states: "On the basis of age at 90% mortality, rapamycin led to an increase of 14% for females and 9% for males." Age at 90% mortality is a maximal-lifespan measure. Anyone who repeats those two numbers as median lifespan has quietly changed what the paper said.

The 2009 paper came out of the Interventions Testing Program, which the National Institute on Aging describes as "designed to identify agents that extend lifespan and healthspan in mice", with testing at three sites. Its results table lists rapamycin again and again: started at 20 months in a 2005 cohort, at 9 months in 2006, at three doses in 2009 and in several 2015 arms, with increases in median and maximum lifespan recorded for both sexes in most of them.

What happened when metformin was tested the same way?

In the same programme, metformin on its own produced no increase in median or maximum lifespan. The only arm containing metformin that did extend lifespan also contained rapamycin, so it cannot show what metformin contributed. That is a narrower statement than metformin doing nothing in animals generally: it was one dose, one cohort and one starting age.

The programme publishes its results as a table, last updated on 13 April 2026, headed "Increase in Median (Med) and Maximum (Max) Lifespan". The metformin row records a 2011 cohort at 1000 ppm, started at 9 months, and its result cell reads "none".

That third row is why the blunt version of this claim will not hold. Anyone who opens the table finds a positive metformin arm, so the honest sentence names the combination and says plainly that a combination cannot attribute the effect to either half.

What did the one-year human trial of rapamycin find?

One randomised placebo-controlled trial gave healthy adults low-dose intermittent rapamycin for 48 weeks, and its primary outcome, visceral fat measured on a DXA scan, did not change significantly. The authors' own conclusion is about safety and about secondary findings in women. The lead sponsor is a company that sells rapamycin prescriptions, which belongs in any reading of it.

The PEARL trial was published in Aging on 4 April 2025. The paper describes "This 48-week decentralized, double-blinded, randomized, placebo-controlled trial" of the intervention, registered as NCT04488601, and states that "The primary outcome measure was visceral adiposity (by DXA scan)".

On that outcome the paper says: "Visceral adiposity did not change significantly", and prints p = 0.942 beside it. The authors do not turn that into a verdict on the trial, and this article does not do it for them. What they do write is a conclusion about safety: "Low-dose, intermittent rapamycin administration over 48 weeks is relatively safe in healthy, normative-aging adults", and, as a separate clause, "and was associated with significant improvements in lean tissue mass and pain in women."

Two numbers are worth keeping straight. The paper reports that "A total of 114 participants completed the study and were included in data analysis." The trial registry records 129 enrolled. And 48 weeks is not twelve months, however a headline rounds it.

The lead sponsor on the registry record is a telehealth company that sells rapamycin prescriptions. That does not make the result wrong. It does mean the trial is not an independent one, and a reader deciding what to make of it should know whose trial it was.

Is the semaglutide result an ageing result?

No, and the licence says so in its own words: the use approved after the large cardiovascular trial is reducing the risk of major cardiovascular events in a tightly defined group of adults. The trial behind it excluded anyone with a history of diabetes. It is a cardiovascular result in a cardiovascular population, not a finding about ageing.

The United States label, revised in February 2026, carries the indication "to reduce the risk of major adverse cardiovascular (CV) events", defines those as "(CV death, non-fatal myocardial infarction, or non-fatal stroke)", and scopes them "in adults with established CV disease and either obesity or overweight." That scope is doing real work: established disease, and then either of two weight categories, rather than weight on its own.

The label's own account of the trial records that "Patients with a history of type 1 or type 2 diabetes were excluded." and that "The estimated hazard ratio (95% CI) was 0.80 (0.72, 0.90)". That is a genuine effect on a hard endpoint. It is also an effect on that endpoint in that population, which is what the licence says and the whole of what it says.

If a medicine of this kind has been prescribed to you, what belongs in your own record is covered in what to log if one of these is prescribed to you rather than repeated here.

Is any medicine licensed for ageing in Britain or Spain?

Not on the regulator pages read for this article. The two compounds discussed most often are authorised in Europe for other things entirely: one for preventing transplant rejection and for a rare lung condition, the other for type 2 diabetes. Taking either for longevity is off-label use, and that phrase is the reason a written record matters.

The European listing for sirolimus states that "Rapamune is indicated for the prophylaxis of organ rejection in adult patients at low to moderate immunological risk". It carries a second indication, "Rapamune is indicated for the treatment of patients with sporadic lymphangioleiomyomatosis with moderate lung disease", which the page completes in a separate clause with "or declining lung function." The authorisation dates from 2001, and the product information was last updated on 4 June 2026.

For metformin, the European referral page states that "Metformin is a medicine used on its own or in combination with other medicines for the treatment of type 2 diabetes." The British regulator, in a safety update published on 20 June 2022, writes that "Metformin is a medicine authorised to treat type 2 diabetes mellitus" and, separately, "to help prevent type 2 diabetes in patients at high risk of developing it."

A limit on that finding, stated plainly. This is what the regulator pages read for this article say. It is not an audit of every entry in every register, and the Spanish medicines agency was not read on this particular question, so nothing here should be taken as a statement about what Spain has or has not authorised.

What is a biological-age test actually returning?

A number produced by a model, not a reading taken off a sample. The published descriptions are explicit about it: these are machine learning tools fitted to data, and the age they return is an estimate. What the different families of them set out to measure is a separate question with its own answer.

A 2025 review in Nature Reviews Genetics puts it directly: "epigenetic clocks have emerged as powerful machine learning tools", "not only to estimate chronological and biological age". Elsewhere the same review calls one "a cohort-based machine learning algorithm".

A 2024 systematic review in Aging is blunter about the state of the field. It notes that "Epigenetic clocks generally follow the standard protocol for regression problems.", that "Despite a recent surge in biological clock research, best practices are still empirical.", and that "their ability to provide insights into mortality and disease has been shown to be moderate".

What the separate families of these tests set out to measure, and why three of them can disagree with each other, is the subject of a separate article on what these tests measure and is not repeated here.

It is worth saying that an age-shaped score is not in itself a commercial trick. The NHS publishes one. Its heart age tool, last reviewed on 20 January 2026, tells the reader that "Your heart age gives you an idea of how healthy your heart is." and states outright that "The NHS heart age calculations are generated by JavaScript using the JBS3 Cardiovascular Risk Assessment." A named model, and a claim pitched at giving the reader an idea rather than at measuring anything.

How reproducible is the number?

Less reproducible than a single printed figure suggests, though the honest version is narrower than the one usually repeated. In one published study, running the same samples twice produced differences of up to nine years across six well known clocks. Up to is a maximum rather than a typical gap, and the same authors published a retrained version precisely to shrink it.

That study, published in Nature Aging in 2022, states that "Epigenetic clocks are widely used aging biomarkers calculated from DNA methylation data," and then, in the next clause, "but this data can be surprisingly unreliable." Its headline finding is that "technical noise produces deviations up to 9 years between replicates for six prominent epigenetic clocks". The figure legends record the design: "ICCs are quantified across 36 samples with 2 technical replicates each."

The same paper offers a fix rather than a verdict. The authors write that "Our retrained PC versions of six clocks show agreement between most replicates within 1.5 years". Most, not all, and of their own revised versions rather than of the ones sold to the public.

There is an older measurement paper often pressed into service here, and it does not say what it is usually quoted as saying. Published in Patterns on 23 April 2020, it confirms that its duplicates were genuinely the same blood: "DNA was sourced from a single blood draw via a single extraction". It then reports that individual measurements are often poor: "Probe reliabilities were skewed toward zero, with a mean of 0.21 (median = 0.09)." and "The median reliability in our EPIC-EPIC comparison was 0.26."

But those same authors defend the clocks against their own background finding. They write that "Each aging algorithm had median probe reliabilities higher than that of the background distribution." and that "externally validated DNA methylation algorithms are generally composed of reliable probes". Citing that paper as proof these tests are unreliable reverses its conclusion.

The practical consequence is small and specific. A difference between two of these results is not automatically a difference in you, and the same discipline that makes two ordinary lab results worth comparing applies here too. That discipline is the subject of keeping two results comparable.

What do British and Spanish sources say about buying these tests?

Neither names them: Britain warns about commercially sold tests as a class, and a government report calls consumer genomic tests for non-medical traits largely unregulated. Spain legislates on how self-tests may be sold rather than on what they return. Both countries address the channel, and neither addresses the product itself.

The UK National Screening Committee, in guidance published on 21 November 2023, states flatly that "False positive results are inevitable in screening." and "False negative results are also inevitable in screening." It adds that "Companies may offer screening that is not evidence-based." and that "Often there is no quality assurance, meaning the tests might not be the same everywhere."

The national patient page on screening, last reviewed on 30 July 2024, puts the same point more mildly: "Some of the tests on offer are not recommended by the UK National Screening Committee (UK NSC)".

A Government Office for Science report, published on 26 January 2022 and so already nearly five years old, is the closest thing to a British statement on tests sold straight to the public. It says that "DTC genomic tests for non-medical traits are also largely unregulated", continuing in a separate clause "relying instead on voluntary self-regulation by the test providers", and notes that "Many of these companies are also based outside of the UK". The scope matters and is printed in the source itself: non-medical traits.

Spain legislated on the channel. A royal decree in force since 24 October 2025 covers in vitro diagnostic products, and the agency's own summary describes the change as "eliminando la necesidad de prescripción para los productos de autodiagnóstico" while "manteniendo su venta exclusivamente a través de las oficinas de farmacia". The prescription requirement went; the pharmacy-only sale stayed. Reporting either half on its own misstates what happened.

What the decree does not do is say whether a test of this kind falls inside that definition at all. The agency's own page on the area, last updated on 20 May 2026, describes a framework still in motion: "La legislación sobre productos sanitarios de diagnóstico in vitro está actualmente en fase de transición".

Across the official pages read for this article in both countries, none mentions biological age, epigenetic age or DNA methylation. Catalonia and Andalusia could not be read for this article and are therefore unchecked rather than silent.

What to write down and bring to your doctor

Three ordinary things: the name and strength of anything taken for longevity, exactly as printed on the box, and every test bought privately, with its date and the laboratory's own reference range kept attached to the value. The third is the question you want answered, written before the appointment rather than remembered during it.

Anything taken off-label is still a medicine as far as the rest of your care is concerned. It belongs on the same list as everything else, written as printed rather than as remembered, so that the person treating you is not working from a gap.

A result bought privately only exists in your care if you bring it. Keep the date, the name of the laboratory, and the range the laboratory itself printed beside the value, because that range belongs to the laboratory rather than to the number.

Longetic holds reports from any laboratory in one line and shows each value next to that laboratory's own reference range, and it holds a medication list with reminders and a dose schedule. It does not interpret results, does not convert units, does not check interactions, and it does not calculate a biological age. A stored report stays a stored report.

Going into the appointment, a written list beats a remembered one. What to write down before an appointment covers the rest of it.

Frequently asked questions