Do Senolytic Supplements Work? The Evidence

Here is what it takes, in 2026, to be reasonably sure a single cell has gone senescent. You stain for senescence-associated beta-galactosidase activity. You run PCR for p16 — except the human CDKN2A locus produces five variant transcripts, two of which encode homologous p16 proteins, and the standard primers for variant 1 cannot avoid also counting variant 5. You check that the cell has stopped dividing, and stopped permanently. You measure the inflammatory secretome it throws off. You count p21-positive cells. Then you cross-reference an algorithm published by the International Cell Senescence Association, alongside a searchable database of senescence-linked genes, both of which exist because no single one of those measures was ever enough on its own.

Then you find out that when a group profiled the transcriptomes and proteomes of fourteen different primary human cell types pushed into senescence more than thirty different ways, the senescent cells did not share a single unique marker. Not one.

That is the state of the target. And in July 2026, a review in Nature Reviews Genetics co-authored by Paul Robbins and Laura Niedernhofer — the two senior authors of the 2018 screen that made fisetin a senolytic candidate in the first place — said so out loud: the field still faces "key challenges in distinguishing senescence from other cell fates" and needs "next-generation biomarkers to capture the varied phenotypes and functions of senescent cells." Which raises an uncomfortable question for anyone holding a bottle. If the scientists who invented this idea cannot yet reliably count senescent cells in a person, what exactly is a senolytic supplement supposed to be doing to yours?

What the new review actually says

The paper is a review, not an experiment. Rahagir Salekeen, Tra L. Kieu, Paul D. Robbins, Laura J. Niedernhofer and Linshan Laux at the Masonic Institute on the Biology of Aging and Metabolism map the genetic pathways that push a cell into senescence and keep it there: the DNA damage response, non-genotoxic stress pathways, chromatin remodelling, and the cell-to-cell signalling that spreads the state to neighbours. It is an authoritative summary of mechanism, and its tone toward the therapeutic future is optimistic. New genetic models, single-cell and spatial multi-omics platforms, and machine-learning approaches, the authors write, promise phenotyping and manipulation of senescent cells "with unprecedented precision and resolution."

Read the verbs. Promise. The precision is the thing being built, not the thing in hand. A companion 2026 effort in Cell described the work of charting human senescence as an exercise in mapping heterogeneity, coining "senotypes" for the tissue-specific varieties, and framing the result as a framework for biomarker discovery rather than a biomarker.

None of this is a knock on the science. It is a description of where the science is. Senescence is not one thing; it is a family of stress states that differ by cell type, by insult, and by how long the cell has been in it. That is precisely why no single validated blood test for senescent cell burden exists outside a research protocol.

Where the senolytic idea came from, and what it was built on

The founding evidence is genuinely striking, and it is almost entirely in mice.

In 2018, a team showed that transplanting relatively small numbers of senescent cells into young mice caused persistent physical dysfunction and spread senescence to the host's own tissues. Fewer cells were needed to do it in older recipients, and their survival dropped. Then the same paper ran the experiment backwards: intermittent oral dasatinib plus quercetin, given to naturally aged mice, alleviated the physical dysfunction and increased post-treatment survival by 36 percent while reducing the mortality hazard to 65 percent.

Later that year, a screen of ten flavonoids for senolytic activity found fisetin the most potent of the panel. Given late in life to wild-type mice, it reduced senescence markers across multiple tissues and extended both median and maximum lifespan. That paper is the origin of every fisetin capsule on the market. Its own authors described the findings as suggesting "the feasibility to translation to human clinical studies" — a sentence about starting trials, not about finishing them.

Eight years later, that translation is still underway, and it has been more complicated than the mouse data implied.

What has actually been tested in people

The honest inventory of human senolytic trials is short, and almost all of it used dasatinib — a prescription tyrosine kinase inhibitor — in combination with quercetin. Not quercetin alone. Not fisetin.

The 2024 bone trial deserves the most attention, because it is the largest randomised test of an oral senolytic in ordinary ageing people rather than in a specific lung or kidney disease, and because it did not work as hoped. Sixty postmenopausal women received intermittent dasatinib plus quercetin or control. The primary endpoint was the change at 20 weeks in CTx, a marker of bone resorption. It did not differ between groups: a median of −4.1 percent on the senolytic against −7.7 percent on control, with a P value of 0.611. The bone formation marker P1NP rose 16 percent relative to control at two weeks and again at four weeks, then was no different at twenty. No serious adverse events occurred.

Then the interesting part. In exploratory analysis, the women in the highest tertile for T-cell p16 messenger RNA — the ones with the most measurable senescent burden to begin with — did respond: P1NP up 34 percent and CTx down 11 percent at two weeks, and radius bone mineral density up 2.7 percent at twenty. A follow-up paper narrowed that further, reporting that the specific variant-5 transcript predicted the skeletal response better than the standard variant 1+5 assay, and that a plasma panel of secreted factors performed about as well.

That is a real and important signal. It is also, precisely, a hypothesis generated after the fact from a trial that missed its primary endpoint, in a subgroup defined by an assay whose stated purpose is selecting participants for trials. It says the next trial should enrol people with high senescent burden. It does not say the drug works in people who have not been measured.

The gap between the trials and the supplement shelf

Every oral result above involves dasatinib — the one exception, in the eye, used a different senolytic molecule entirely. The supplement aisle sells the other half of the oral combination, plus fisetin, and inherits the association.

Quercetin on its own has never been shown to clear senescent cells in a human being. In the Alzheimer's trial, quercetin was not detectable in cerebrospinal fluid at all.

It does have a separate and more modest body of human evidence, and it is worth taking on its own terms rather than as a senolytic claim. In a 2007 crossover trial, 730 mg a day for 28 days left blood pressure unchanged in prehypertensive participants but lowered systolic pressure by 7 mmHg and diastolic by 5 mmHg in those with stage 1 hypertension. A 2009 crossover trial of 150 mg a day in 93 overweight adults with metabolic syndrome traits found systolic pressure down 2.6 mmHg overall, and 2.9 mmHg in the hypertensive subgroup. A 2015 trial of 162 mg a day from onion skin extract in 70 overweight-to-obese patients found no significant effect on 24-hour ambulatory or office blood pressure in the group as a whole, with a 3.6 mmHg systolic reduction confined again to the hypertensive subgroup. That is a small, condition-dependent blood-pressure effect in people who already have high blood pressure. It is not a senescence claim, and nobody should read it as one.

Fisetin's problem is different and more basic: absorption. In the first human pharmacokinetic study of fisetin, fifteen healthy volunteers took a single 1,000 mg dose of unformulated fisetin, and the twelve-hour area under the curve was 12.67 — against 341.4 for an encapsulated formulation delivering only 192 mg of actual fisetin. That is a 26.9-fold difference in exposure from under a fifth of the dose. Peak plasma concentration was 9.97 ng/ml for the plain powder against 238.2 ng/ml for the formulated version. Whatever a plain fisetin capsule is doing, most of it is not reaching your bloodstream.

Dose is the other quiet gap. The one randomised phase 2 trial designed to test whether oral fisetin improves physical function in humans — in postmenopausal breast cancer survivors, with six-minute walk distance as the primary endpoint — uses 20 mg per kilogram per day for three days of each fourteen-day cycle. For a 70 kg participant that is about 1,400 mg a day, on three days out of fourteen. The two fisetin trials in ordinary adults used 200 mg a day and 100 mg a day. Those are three very different interventions wearing the same molecule's name, and the 20 mg/kg trial has published its design without yet publishing results.

What fisetin has produced in humans so far is thin. A twelve-week randomised trial in 44 men with obesity found that 200 mg a day lowered TNF-alpha, and lowered IL-6 in the fisetin-only arm, alongside a training programme that did most of the metabolic work. A trial of 100 mg a day against a circulating inflammation marker in adults over fifty has published a protocol, and no results. And the widely shared 2025 paper whose title promises that intermittent fisetin supplementation "Improves Physical Function and Decreases Cellular Senescence in Skeletal Muscle With Aging" is, read past the title, a study in old mice.

The complication nobody markets

The pitch treats senescent cells as pure waste. The biology does not.

Senescence is a tumour suppressor mechanism — that is what it evolved to be, a way of permanently benching a cell whose DNA is too damaged to trust with another division. And in 2026, researchers described a rapid-onset senescence response in skin that appears within minutes to hours of a wound, running not through new transcription but through pre-existing messenger RNA already sitting in the cell. Those newly senescent cells at the wound edge secrete factors that pull the repair forward. Suppress that early response, genetically or pharmacologically, and wound closure is markedly delayed. Suppress it later in repair, and nothing happens.

So the target is not "senescent cells." It is a particular subset, in particular tissues, at particular times, and the field's own view is that we cannot yet reliably tell them apart. That is the argument for intermittent "hit-and-run" dosing rather than continuous suppression — and it is also an argument against a daily capsule taken indefinitely on the theory that less senescence is always better.

Safety is not a settled question either. A 2025 study cultured primary human neurons and astrocytes to test two senolytic regimens that had worked in Alzheimer's mouse models. Dasatinib plus quercetin looked safe in human neurons. Navitoclax, which works by targeting the Bcl-2 family of anti-apoptotic proteins, showed non-selective neurotoxicity. Different senolytics are not interchangeable, and potency in a dish is not a virtue by itself.

What a careful reader should take from this

The people who built this field have been unusually direct about where it stands. A 2020 review by the senolytics group put it plainly: until the trials are done, "it is too early for senolytics to be used outside of clinical trials." A 2025 translational review of geroscience in JAMA reached the same place from the outside — senescent cells are more abundant with age, and greater abundance is associated with more physical impairment and increased mortality, but "potential health benefits of reducing senescent cells in humans remain unclear."

Five years separate those two sentences. The second is not more confident than the first.

What that leaves, practically, is modest. Fisetin is described in the literature as a flavonoid from strawberries and apples, and quercetin as a flavonol found in apples, berries and onions. Eating those foods is a defensible thing to do for reasons that have nothing to do with senescence, and the amounts they deliver are nowhere near the doses these trials used. If you are already taking a fisetin or quercetin capsule, the accurate description of what you are doing is a personal experiment with an unmeasured endpoint, not a treatment — and the two supplements are not the same bet, since quercetin at least has direct randomised human data on a hard-ish outcome in one specific group, while fisetin's published human data are largely absorption studies and one small inflammation trial.

Quercetin absorption, at least, is somewhat controllable. A 2025 systematic review and meta-analysis of 31 human intervention studies found that the chemical form matters enormously — oligoglucoside forms were roughly twentyfold more bioavailable than the aglycone — and that taking it with dietary fat and fibre roughly doubled bioavailability. If a label does not say which form it contains, that is information.

And if this line of research is what interests you, the useful move is not a bottle. It is a trial. A directory of actively recruiting aging studies, with each study team's own eligibility criteria and contact details, is kept here. Enrolling puts you on the side of the question where the p16 assay actually gets run.

How you would know if this changed

Three specific things would move senolytics from plausible to established, and none of them has happened yet:

A validated biomarker you can order. Right now the best candidate is a T-cell p16 variant-5 assay used to select participants for trials, with a supporting plasma panel of secreted factors. Watch for that, or something like the circulating IL-23R signal reported in 2025, becoming a standardised clinical test. Until then, no validated clinical test measures senescent cell burden outside a research protocol.

A randomised trial with a hard endpoint that hits its primary outcome. Not feasibility, not tolerability, not a marker at two weeks in a subgroup. Fracture, function, disease progression, prespecified.

Human evidence for the specific molecule in the specific bottle. Dasatinib plus quercetin is not fisetin. A trial of one is not evidence for the other, and neither is a mouse.

The Nature Reviews Genetics review is a good paper. It is worth reading it the way its authors wrote it: as a map of a mechanism that is finally becoming legible, and an explicit statement that the tools to act on it precisely are still being built. The supplement industry read the same literature and shipped the product first. That is the gap.

Educational, not medical advice.