Longevity drugs: what human trials show

Evidence and US regulatory information through September 22, 2026.

US prescribing information for sirolimus, metformin, semaglutide, and dapagliflozin describes particular diseases or risk groups. Healthspan means years lived in good health, with function and independence preserved. Fewer new diagnoses, heart attacks, kidney failures and disabling symptoms are meaningful outcomes for that goal. The benefit still needs to fit the person’s starting risk and the medicine’s harms. [1] [2] [3] [4]

What has been shown—and what remains open

• Shown: selected medicines reduce important clinical events in specific populations. SELECT found fewer cardiovascular events with injectable semaglutide; DAPA-CKD found fewer major kidney or cardiovascular events with dapagliflozin in people with chronic kidney disease. [7] [8]

• Still uncertain: whether experimental approaches such as rapamycin, dasatinib–quercetin and NR consistently prevent disease or preserve everyday function. Several trials changed biomarkers or found favorable subgroups while missing their main outcome. [5] [9] [10] [25] [27]

• What would change the assessment: well-designed human trials in the relevant population showing sustained improvements in illness, function, or survival, alongside adequate information about harms. The result needs to apply to the actual product and purpose being considered.

A comparison that keeps the outcome visible

These representative studies answer different questions. The study question is separate from an approved indication, and the participant count and follow-up apply to that trial. Here, a primary outcome is the main planned measure of success; secondary outcomes address additional planned questions. Subgroup and exploratory findings usually need confirmation.

A hazard ratio compares the rate at which events occur during follow-up among people still at risk. It is not the percentage of people helped. SELECT's 6.5% and 8.0%, for example, differ by 1.5 percentage points, while its hazard ratio indicates a 20% lower hazard. Both descriptions matter when judging the size of a benefit. In the table, CI means confidence interval, a range expressing statistical uncertainty; eGFR is an estimate of kidney filtration. [7]

Disease prevention is a meaningful healthspan outcome

SELECT and DAPA-CKD measured events that matter directly to patients. Their results support clinical decisions in the populations studied. DAPA-CKD also recorded fewer deaths overall: 4.7% with dapagliflozin and 6.8% with placebo during follow-up. [8] Dismissing such evidence because it is not a trial of “aging itself” would miss a meaningful benefit.

Those benefits still require a safety assessment. Dapagliflozin’s label warns about ketoacidosis (a dangerous buildup of acids in the blood), loss of body fluid, and infections of the urinary or genital tract. In SELECT, adverse events led to permanent treatment discontinuation in 16.6% of semaglutide recipients versus 8.2% with placebo. Benefit and tolerability are both part of the decision. [4] [7]

The limitation concerns transfer to a different person. A participant without diabetes in DAPA-CKD still had chronic kidney disease. A participant without diabetes in SELECT still had cardiovascular disease and overweight or obesity. Neither description is interchangeable with a generally healthy adult considering medication solely because they are getting older.

The same restraint applies across organs. The EVOKE and EVOKE+ trials tested oral semaglutide in 3,808 people with early symptomatic Alzheimer's disease. The 2026 publication reported no significant slowing on the primary clinical progression measure at 104 weeks. [15] This does not erase semaglutide's cardiovascular evidence, nor does a cardiovascular benefit rescue a negative cognition result. Each question requires its own trial.

Metformin and rapamycin show why later evidence matters

Metformin's diabetes-prevention result has stood for decades. Its interpretation changes when the question becomes survival. In roughly 21 years of DPP follow-up, original assignment to metformin did not significantly reduce mortality: the hazard ratio was 0.99, with a 95% confidence interval of 0.79–1.25. [13] Lifestyle assignment also did not significantly reduce mortality in this follow-up. Treatment arrangements changed after the original trial, so this was not 21 years of continuous blinded treatment. [13]

A June 2026 analysis asked another question: did the original intervention delay the accumulation of multiple chronic conditions? Among 1,173 participants with linked Medicare data, the metformin comparison was inconclusive; the lifestyle comparison favored intervention. [14] The subgroup and follow-up design limit causal interpretation. Even so, this is relevant evidence that a diabetes result alone could not supply.

A July 2026 preprint—a report that had not yet undergone peer review—adds a more complicated result: in 145 older adults with glucose intolerance followed for two years, metformin favored a separate index that counts accumulated health problems, but not the main Fried frailty scale once its weight-loss component was excluded. The initial Fried score worsened partly because weight loss counts toward frailty. The differing results make this preliminary report difficult to reduce to a simple claim that metformin prevents frailty. [26]

Rapamycin illustrates a different problem. PEARL's secondary findings attracted attention, including a lean-tissue signal in a subgroup of just eight women receiving the higher study dose. The primary visceral-fat outcome was negative, and a body-composition measurement does not establish improved strength. [5] A separate 13-week trial published in 2026 paired sirolimus or placebo with exercise in 40 older adults. It found no significant benefit on its primary chair-stand comparison; the numerical result favored placebo. [16]

These findings leave questions open, but they also constrain the story. A trial described as investigating healthspan, or years in good health, should be judged by the outcomes it actually measured. Its name cannot supply missing evidence about independence, disease prevention, or survival.

A mechanism is a starting point, not a clinical endpoint

Senolytics aim to remove senescent cells—cells whose normal division has persistently stopped. In the bone study above, the combination of dasatinib and quercetin did not improve the prespecified primary marker of bone breakdown. The bone-formation marker improved at two and four weeks as a planned secondary outcome, but not at 20 weeks. Findings in women with higher markers of cellular senescence came from exploratory subgroup analyses. The trial was open-label, with an untreated control group. [9] These results can guide a subsequent trial; they cannot establish fracture prevention or broad rejuvenation.

The compounds also deserve separate treatment. Dasatinib is a prescription leukemia medicine with substantial safety warnings, including effects on blood-cell counts, bleeding, and fluid retention. [19] Pairing it experimentally with quercetin does not make the combination a routine preventive supplement.

NR is an oral precursor used by the body to make NAD, a molecule involved in cellular metabolism. Its trials offer a particularly clear lesson about what an outcome can tell us. The July 2026 cognition study showed that the intervention reached a biological target: blood NAD rose. It did not show improved cognition. [10] A different trial in 90 people with peripheral artery disease reported a modest improvement in six-minute walking distance after six months. Its estimated NR–placebo difference was 17.6 meters, using a one-sided 90% confidence interval in a small phase 2 design. This was a less stringent statistical threshold than the conventional two-sided 95% interval; confirmation in a larger trial matters. [17] A publisher correction added an omitted trial-registration identifier; it did not change that result. [18]

Walking performance is closer to daily function than a blood biomarker, but the population and uncertainty still matter. These two studies support different conclusions because they tested different outcomes in different people. Neither establishes that oral NMN, an NAD infusion, or an untested formulation produces the same effects.

Biological-age clocks use measurements such as chemical marks on DNA to estimate aspects of aging. A change in the score is not itself evidence of better health. A 2026 metformin pilot randomized 40 older adults with HIV and found no significant benefit on its primary DNA-based biological-age outcome at 96 weeks. A separate semaglutide analysis found favorable changes in several clocks among 84 people with HIV-related central fat accumulation, but it was an exploratory analysis added after the original trial plan, without adjustment for multiple statistical comparisons, and other clocks were unchanged. These exploratory clock findings do not show fewer illnesses or better daily function. [25] [27]

Hormone decisions need their own context

Hormone treatment should be assessed against the symptom, deficiency or health risk being treated, with benefits and harms considered together.

TRAVERSE answered an important cardiovascular safety question about testosterone gel in selected men. Its primary hazard ratio was 0.96, with a 95% confidence interval of 0.78–1.17. That interval met the prespecified noninferiority margin, whose upper boundary was 1.50. A noninferiority trial asks whether a treatment is no worse than a comparator beyond a predefined margin; meeting that margin did not demonstrate a cardiovascular benefit. [11] It does not mean that all harms were absent: atrial fibrillation (an irregular heart rhythm), acute kidney injury, and pulmonary embolism (a blood clot in the lungs) occurred more often in the testosterone group. [11]

Regulatory language has also changed. On February 28, 2025, the US Food and Drug Administration (FDA) announced testosterone labeling changes calling for removal of cardiovascular boxed-warning language and new or strengthened blood-pressure warnings. On June 18, 2026, it requested further revisions, including removal of a limitation concerning age-related low testosterone and changes to prostate-related information. [21] [22] A requested revision and an implemented product label are distinct documents; neither is a general longevity approval.

For menopausal hormone therapy, the overall WHI mortality finding should not be treated as the answer for every age, route, regimen, and treatment goal. The trials used specific oral formulations, and their age-group findings differed. [12] On February 12, 2026, the FDA approved changes to the boxed warnings of six menopausal hormone products. Its notice also emphasized timing of initiation and recognized symptom and bone benefits. [20] The updated-product list identifies the six products individually. [24] A warning change warrants accurate counseling, not a claim that all formulations are risk-free or universally appropriate.

Funding and interests are part of the appraisal

Commercial involvement varies across these studies. SELECT was funded by Novo Nordisk, DAPA-CKD by AstraZeneca, and TRAVERSE by AbbVie and other sponsors. PEARL's authors disclosed employment and shareholdings in AgelessRx. The senolytic study disclosed patent-related interests among several authors. In contrast, the 2026 NR cognition study reported funding from the US National Institutes of Health (NIH), company-supplied capsules and no author conflicts. [5] [7] [8] [9] [10] [11]

These disclosures do not decide whether a result is valid. They help readers assess who designed and reported the work, alongside how treatment was assigned, missing data, outcomes specified in advance, and complete findings.

What to ask when a drug is presented as a longevity treatment

Start with the proposed benefit. “Better metabolic health” should be translated into something assessable: fewer diagnoses, fewer hospitalizations, improved walking, reduced symptoms, or a laboratory change. Ask whether that was the main planned outcome or a finding discovered among many analyses.

Then ask whether the participants resemble you. Existing disease, starting risk, age, and sex can change both the likely benefit and the harms. Ask about absolute results, follow-up length, adverse effects, and what evidence supports the exact formulation being discussed.

Finally, establish its status. Approval for one condition, an off-label prescription, a registered trial, and an online product listing answer different questions. At the September 22, 2026 evidence date, the FDA stated that retatrutide is not a component of an approved drug; promising trial results had not changed that US status. [23] An investigational treatment's scientific interest does not establish routine access or product quality.

How Healthy Longevity Clinic experts evaluate the evidence

For a reader hoping to stay independent and avoid serious illness, Healthy Longevity Clinic’s interpretation starts with the outcome that matters and the risk they already face. A reduction in heart attacks or kidney failure in a matching patient group is directly useful evidence. A change in NAD, body composition, or a biological-age clock answers a narrower question. These findings should not receive the same clinical meaning simply because all are discussed under “longevity.” [5] [7] [8] [10] [25] [27]

SELECT shows how to make the comparison concrete: 6.5% versus 8.0% had a major cardiovascular event over a mean 39.8 months, while adverse effects led more people to stop semaglutide. That supports an individualized discussion for someone with established cardiovascular disease and overweight or obesity, without diabetes; it does not provide the same answer for a low-risk person. In DPP, intensive lifestyle intervention reduced diabetes incidence more than metformin during the original trial, so an evidence-based conversation should include the established non-drug option as well. [6] [7]

Negative primary outcomes also help decisions. The NR cognition trial showed that doubling blood NAD did not translate into measured cognitive improvement over 12 weeks. PEARL’s body-composition signals did not demonstrate better strength. For preventive use in generally healthy adults, the assessment would change with reproducible gains in function or fewer diseases in that population, with enough follow-up to weigh harms. A promising mechanism or a single favorable clock result cannot supply that evidence. [5] [10] [25] [27]

Three questions for a clinician

1. Is the proposed goal fewer serious illnesses, better function or symptoms, or a change in a laboratory marker—and which outcome actually improved in the trial?

2. How closely do I match the study’s disease status, starting risk, age, and treatment formulation?

3. What are the absolute benefits, adverse effects, and follow-up limits, and how do established options such as the DPP lifestyle intervention compare for my actual health goal?

Common questions

Is there an approved drug that slows aging in healthy people?

The US labels discussed here cover specific diseases and risk groups. There is no general healthy-aging indication in those labels. Their disease-specific benefits can still support healthspan when the patient’s risks match the evidence. [1] [2] [3] [4]

If a study enrolled people without diabetes, were they healthy?

Not necessarily. SELECT participants had cardiovascular disease and overweight or obesity. DAPA-CKD participants had chronic kidney disease. “Without diabetes” does not mean “without disease.” [7] [8]

Does a better biological-age score mean better health?

These trials did not establish that. The metformin pilot’s main clock outcome was not significantly improved, while the semaglutide analysis found mixed clock results in a specific HIV population. Neither clock analysis demonstrated fewer illnesses or improved everyday function. [25] [27]

Does a negative trial mean the medicine never helps?

It narrows what can be claimed for that product, population, and outcome. Oral semaglutide did not significantly slow clinical progression in EVOKE and EVOKE+, while injectable semaglutide reduced cardiovascular events in SELECT. The separate findings can both be true. [7] [15]