Senolytics Finished Real Trials. Here's What They Showed.

The full story

Senolytics have the best origin story in geroscience. In mice, a drug-inducible genetic switch that eliminated p16Ink4a-positive senescent cells delayed fat, muscle and eye pathology in a progeroid strain, and clearing those cells late in life attenuated disorders that had already set in - a 2011 result in Nature, and still one of the cleanest proofs of principle in the field. Then a 2018 study in Nat Med showed that transplanting even small numbers of senescent cells into mice caused persistent physical dysfunction, and that intermittent oral dasatinib plus quercetin in naturally aged mice increased post-treatment survival by 36% while cutting the mortality hazard to 65%.

Both of those are mouse experiments, and the 2011 one is a genetic tool rather than a drug. What matters is that the drugs then went into people. Six human senolytic studies have reported. Every one of them is small, several are open-label with no control arm, and the ones that were randomized with a prespecified efficacy endpoint have missed it. That is not the same as failure. It is also not what the supplement aisle is selling.

The first human studies were pilots, and said so

The first senolytic trial in humans, published in 2019 in EBioMedicine, was open-label with no control arm in 14 patients with idiopathic pulmonary fibrosis. Six-minute walk distance, four-metre gait speed and chair-stand time all improved significantly. Pulmonary function, clinical chemistries, frailty index and self-reported health did not change at all, and effects on circulating inflammatory factors were inconclusive. In 14 unblinded patients who knew they were being treated, physical-performance tests are exactly the measures you would expect to move first, for reasons that have nothing to do with senescent cells.

A second 2019 report in EBioMedicine was more interesting because it was more modest: an open-label phase 1 pilot with no control group in nine people with diabetic kidney disease, mean age 68.7 years, eGFR 27, only two of them women. Three days of dasatinib 100 mg plus quercetin 1000 mg reduced p16-positive and p21-positive cells in fat, lowered senescence-associated beta-galactosidase activity and adipose macrophages, and dropped circulating IL-1alpha, IL-6, MMP-9 and MMP-12 within 11 days. No clinical outcome was measured, and none was claimed. What that study established is target engagement - the drugs do in humans what they are supposed to do to senescent cells. That is a real and necessary result. It is also the beginning of the argument, not the end.

Then the controlled trials reported

The randomized follow-up to the open-label fibrosis pilot appeared in 2023 in EBioMedicine as a phase 1, single-blind, single-centre, placebo-controlled trial - with 12 participants, six per arm, and feasibility and tolerability as its stated objective. Frailty, pulmonary and physical function, in the authors' words, "do not appear to differ meaningfully between groups," and the study was explicitly under-powered for efficacy. One asymmetry did show up: sleep disturbance and anxiety were disproportionately reported in the dasatinib-plus-quercetin arm, four of six versus none of six.

The biggest randomized academic readout came in 2024 in Nat Med: a phase 2 trial of intermittent dasatinib plus quercetin in postmenopausal women, 60 analysed, which missed its primary endpoint. The 20-week change in the bone-resorption marker CTx did not differ between senolytic and control arms - a median of -4.1% versus -7.7%, P=0.611. A bone-formation signal on P1NP at two and four weeks had evaporated by 20 weeks, at -9% and P=0.149. Any skeletal benefit turned up only in an exploratory analysis of the highest tertile of T-cell p16. Worth noting from the registry: this trial was itself open-label and randomized, with percent change in serum CTx at 20 weeks as its single prespecified primary outcome, and 74 enrolled. There is no ambiguity about what it was built to test or how it came out.

The most expensive miss was not academic. On 17 August 2020, UNITY Biotechnology reported that its senolytic UBX0101 failed in 183 patients with painful knee osteoarthritis: "There was no statistically significant difference between any arm of UBX0101 and placebo at the 12-week endpoint for change from baseline in WOMAC-A," with p values of 0.5222, 0.8069 and 0.9870 across the three doses. The company dropped the programme. A randomized placebo-controlled trial in 183 patients is larger than any academic senolytic trial that has reported, and it is the result the category discusses least.

The Alzheimer's trial asked a different question than you think

The 2023 senolytic Alzheimer's study in Nat Med was an open-label, proof-of-concept phase 1 trial in five participants, and its registry record is unusually clarifying: the prespecified primary outcomes were brain penetrance of dasatinib and of quercetin measured in cerebrospinal fluid. Not cognition. Not disease progression. On that question it returned a genuinely useful answer - dasatinib reached cerebrospinal fluid in four of five participants, and quercetin was not detected at all. Cognitive and neuroimaging measures did not differ from baseline. Cerebrospinal-fluid interleukin-6 and GFAP went up, at P=.008 and P=.028.

Read that as a pharmacology result and it is valuable: half the famous drug combination may not be reaching the organ everyone wants it to reach. Read it as an Alzheimer's efficacy result and you have invented a finding the trial was not designed to produce.

What the evidence does not show

No human senolytic trial has shown improved survival, slowed disease progression, or reduced hard clinical events. Nothing has demonstrated an effect on human aging, because no trial has tested one. The mouse lifespan results are mouse lifespan results, and the fisetin work - most potent of ten flavonoids screened, life-extending in wild-type mice, reported in 2018 in EBioMedicine - also found that fisetin cleared senescence in only a subset of cells in mouse and human fat, meaning the drug is selective in ways nobody has mapped in a person.

Then there is dosing, which is where consumer products quietly fall apart. A registered phase 2 sepsis protocol published in 2024 in Trials, targeting 220 patients over 65, doses fisetin at 20 mg per kilogram - roughly 1.4 grams for a 70-kilogram adult, far above what over-the-counter fisetin capsules contain. No results have been reported; it is a protocol paper, so it establishes intended dosing and nothing else. A bottle labelled "senolytic" is not delivering a trial dose, and if it were, that would raise a different question rather than settle this one.

The people who discovered this drug class put it plainly in their own 2020 review in J Intern Med:

until such studies are done, it is too early for senolytics to be used outside of clinical trials

That is the Mayo Clinic group whose mouse work created the field, declining to endorse consumer use of dasatinib, quercetin or fisetin. When the founders are the most cautious voices in a category, the marketing is running ahead of the science by definition.

How to read the next headline about this

• Open the registry entry, not the press release. Look for the single prespecified primary outcome and check whether the reported number is that outcome. Brain penetrance, tolerability and feasibility are legitimate endpoints - and they are not efficacy.
• Look for a control arm and blinding. Physical-performance tests in unblinded patients move for reasons unrelated to the drug. Open-label plus no controls means the result cannot separate biology from expectation.
• Treat "exploratory subgroup" as a flag, not a finding. A benefit confined to one tertile of a biomarker after the primary endpoint missed is a hypothesis for the next trial.
• Compare the dose to the trial dose. If a registered protocol uses grams and the capsule uses milligrams, they are not the same intervention.

Senolytics remain one of the few longevity ideas with a coherent mechanism, animal evidence in the right direction, and drugs that demonstrably hit their target in humans. That is genuinely more than most of this beat can claim. What is missing is the boring part - adequately powered randomized trials with clinical endpoints that get met rather than missed. The most defensible statement available is that senolytics do something measurable to senescent cells in people and have not yet been shown to make anyone healthier. The interesting readouts are still ahead. So is the possibility that they disappoint.