UBX1325 (Foselutoclax): The Senolytic That Got Closest to Proving It — and What It Still Didn’t Prove

Every few months somebody sends me a screenshot claiming senolytics are “proven in humans now.” The molecule they’re usually pointing at is UBX1325, also called foselutoclax. I want to walk through what that program actually did, because it’s the single most rigorous senolytic dataset in humans — and it’s also a case study in how far “most rigorous” still is from “proven.”

What UBX1325 Is

UBX1325 is a small-molecule inhibitor of BCL-xL, developed by UNITY Biotechnology. BCL-xL is a member of the BCL-2 family — the proteins that decide whether a cell commits apoptosis or keeps living. Senescent cells are unusual: they’ve stopped dividing, they refuse to die, and they lean hard on anti-apoptotic proteins like BCL-xL to stay alive. Block that protein and the senescent cell’s own death program finally executes.

Crucially, UBX1325 was never a pill. It’s an intravitreal injection — a needle into the vitreous of the eye, delivered by a retina specialist. That delivery route is not a detail. It is the entire reason the program existed at all.

The Senescent Cell and SASP Thesis, In One Section

The idea is simple enough that it spread fast. Cells that hit replicative or stress-induced senescence stop dividing but don’t clear out. They accumulate with age. And they secrete — a mix of inflammatory cytokines, chemokines, proteases and growth factors collectively called the senescence-associated secretory phenotype, or SASP. The claim is that SASP is a meaningful driver of chronic low-grade inflammation, tissue dysfunction, and age-related disease.

The animal evidence for this is genuinely strong. In a 2016 Nature paper, clearing p16Ink4a-positive cells from naturally aging mice extended median lifespan and attenuated age-related deterioration in kidney, heart and fat. That’s not a mouse engineered to age badly — that’s normal mice, cleared genetically, living longer and better. I covered the broader framing of this in my piece on senolytic therapy and so-called zombie cells.

The gap is the translation. Mouse genetics clears senescent cells cleanly, everywhere, on command. Drugs do not.

Why Systemic BCL-xL Senolytics Kept Failing

The most potent senolytic class we have — BCL-xL inhibitors — has a hard biological ceiling in humans, and the ceiling is your platelets.

Platelets are anucleate. They have no nucleus, no transcription, no way to mount a stress response. Their survival clock is set almost entirely by BCL-xL. Megakaryocyte-restricted deletion of Bcl-x in mice collapses platelet lifespan from roughly five days to roughly five hours, and platelet counts fall to about 2% of normal. Pharmacologic BCL-xL inhibition with navitoclax (ABT-263) triggers BAK/BAX-mediated platelet apoptosis and causes dose-dependent thrombocytopenia in mice, dogs and humans.

That’s why navitoclax has dose-limiting thrombocytopenia. You cannot reach senolytic exposure systemically without dropping platelets. That’s not a formulation problem you fix with a better capsule — it’s on-target, on-mechanism toxicity. Researchers have tried engineering around it, most notably by converting navitoclax into a PROTAC degrader (PZ15227) recruiting the cereblon E3 ligase, which is poorly expressed in platelets. That reduced platelet toxicity preclinically while preserving senolytic potency — but it remains preclinical.

UNITY’s answer was different and, honestly, smarter: don’t go systemic at all. Inject a tiny dose — 10 μg — into a closed compartment where senescent endothelial cells are doing measurable local damage, and where systemic exposure stays negligible. The eye. That’s the whole strategic insight.

What the Human Data Actually Shows

BEHOLD (Phase 2, versus sham)

BEHOLD enrolled 65 patients with diabetic macular edema who were already on anti-VEGF therapy and still had a visual acuity deficit and residual retinal fluid — a hard-to-treat group. They were randomized 1:1 to a single 10 μg intravitreal injection of UBX1325 or sham, and followed to 48 weeks.

A single injection produced a gain of +6.2 ETDRS letters from baseline, a difference of +5.6 letters versus sham, sustained through 48 weeks. Retinal thickness stayed stable while sham worsened. Roughly 53% of UBX1325 patients needed no anti-VEGF rescue over 48 weeks, versus 22% of sham. No intraocular inflammation, no retinal artery occlusion, no endophthalmitis, no vasculitis. It was published in NEJM Evidence in April 2025.

That is a real result. One shot, a year of durability, in eyes that had already failed standard therapy.

ASPIRE (Phase 2b, versus aflibercept)

This is the one people cite as “beat a real drug.” Here’s what happened. ASPIRE (NCT06011798) randomized 52 patients 1:1 to 10 μg UBX1325 or 2 mg aflibercept every eight weeks for six months. The primary endpoint was non-inferiority to aflibercept on mean BCVA change from baseline to the average of weeks 20 and 24.

Endpoint	Result
Primary (avg. weeks 20 & 24)	Not met — non-inferiority margin to aflibercept not reached
BCVA change, baseline to week 24	+5.2 ETDRS letters (UBX1325 arm)
BCVA change, baseline to week 36	+5.5 ETDRS letters (UBX1325 arm)
Week 36	Statistically non-inferior to aflibercept
Prespecified subgroup, baseline CST <400 μm	Numerically favored UBX1325 (exploratory)
Weeks 16–20	CST increase in UBX1325 arm; some patients required supplemental anti-VEGF

So: it missed. Then it caught up. “Non-inferior at week 36” means “we could not statistically distinguish it from aflibercept twelve weeks after the point where we said we’d measure.” That is meaningfully different from beating a real drug, and anyone telling you otherwise is selling something.

The subgroup where it looked better — patients entering with central subfield thickness under 400 microns — was prespecified, which is better than data dredging. It is still a subgroup of 52 people, and subgroups of 52 people generate hypotheses, not conclusions.

The part nobody quotes

In May 2025, UNITY announced complete 36-week ASPIRE data and simultaneously announced a reduction in force affecting its entire workforce, closure of the ASPIRE study, and exploration of strategic alternatives. The company filed its certificate of dissolution in Delaware in September 2025. UBX1325 currently has no sponsor advancing it.

Also worth remembering: this same company’s earlier senolytic, UBX0101 (an MDM2/p53 inhibitor injected into the knee), showed nothing versus placebo on WOMAC-A pain at 12 weeks in Phase 2 and was discontinued in 2020. And UBX1325 itself failed to hit non-inferiority against aflibercept through 24 weeks in the ENVISION wet AMD study. The senolytic scoreboard in humans is not a winning streak.

What This Does NOT Prove

• It does not prove senolytics extend lifespan. No human trial has ever measured that for any senolytic.
• It does not prove systemic senolysis is safe or effective. The entire design premise was avoiding systemic exposure.
• It does not prove the drug worked by killing senescent cells in humans. Selective senescent-cell clearance was demonstrated preclinically. In the human eye, we observed a functional outcome — letters gained — not a verified cell count.
• It does not generalize across tissues. The retina is immune-privileged, closed, small, and directly measurable. Your liver, brain and vasculature are none of those things.
• It does not survive the commercial test. A drug nobody will fund is a drug that never gets a Phase 3.

The Honest Comparison: Fisetin, Quercetin, and D+Q

Here’s where I want to be blunt, because the supplement world took a 10 μg needle in one eye and turned it into a marketing claim for capsules.

Let’s look at what the oral senolytic human evidence actually consists of:

• D+Q in idiopathic pulmonary fibrosis (2019): open-label, no control group, n=14, three weeks of intermittent dosing. Physical function measures improved. Feasibility study. Not efficacy.
• D+Q in diabetic kidney disease (2019): n=9, a three-day course, open-label. Adipose and skin senescent-cell markers went down. Nine people. Three days. No clinical outcome.
• D+Q in IPF, randomized (2023): phase I, single-blind, single-center, placebo-controlled — designed and powered for feasibility and tolerability, not efficacy.
• D+Q in postmenopausal bone (2024, Nature Medicine): the best-designed oral senolytic trial we have. Phase 2, n=60, randomized. The primary endpoint failed. Change in the bone resorption marker CTx at 20 weeks did not differ from control (−4.1% vs −7.7%, P=0.611). A bone formation marker rose transiently at 2 and 4 weeks and was gone by 20. An exploratory subgroup with the highest T-cell p16 burden looked better. No serious adverse events.
• Fisetin: multiple registered trials, results still sparse in peer-reviewed print. A randomized knee osteoarthritis trial reported at OARSI in 2025 (34 fisetin vs 40 placebo) found MRI T2 cartilage values similar between groups and no between-group difference in self-reported pain.

Add it up. Total randomized, placebo-controlled, adequately powered oral senolytic trials with a positive primary clinical endpoint: zero. Small n. Short durations. Surrogate endpoints — marker changes, not disease outcomes. That’s the state of the field. I’ve written elsewhere about how people actually run dasatinib plus quercetin protocols, and nothing here changes my view that it’s an experiment, not a therapy.

Then there’s the pharmacology problem. UBX1325 delivered a precise dose into a compartment where the target cells live. Oral fisetin has poor bioavailability and rapid conjugation. Quercetin is worse. Whether these compounds ever reach a senolytic concentration in a human tissue that matters is not established. Reduce inflammation? Plausible. Selectively execute senescent cells in your kidney? Not shown.

What Would Actually Have to Be True

For systemic senolytics to work as advertised, several things need to hold simultaneously:

1. Senescent cells must be causal in the specific disease, not just correlated with it.
2. The drug must reach a senolytic concentration in the relevant tissue — verified, not assumed.
3. Selectivity must hold in humans, sparing the cells that legitimately need BCL-xL (platelets, neurons, others).
4. Clearance must produce a functional benefit, and that benefit must persist between intermittent doses.
5. The benefit has to show up on a hard clinical endpoint, in an adequately powered randomized trial, over a duration long enough to matter.

UBX1325 arguably satisfied 1 through 4 in one tissue, in one disease, with a needle. Nothing in the oral senolytic literature has cleanly satisfied even the second item. That’s the gap, and it’s wide.

My practical position hasn’t changed: measure things. If you’re going to run experiments on yourself, at minimum know your baseline — the bloodwork markers that actually matter tell you more about whether an intervention is helping than any mechanism story does. And the boring inputs — sleep, load management, structured recovery — still have better evidence behind them than any senolytic on the market.

Where That Leaves Us

UBX1325 is the strongest thing the senolytic field has produced in humans: one intravitreal injection, roughly five and a half letters of vision held for a year against sham, a clean local safety profile, published in a serious journal. That’s real and deserves credit. It also missed its primary endpoint in the head-to-head trial, and its sponsor dissolved.

If the best-case senolytic result in human history is “worked locally in one eye disease, tied the standard of care three months late, and then the company folded,” then the honest read is that the senescent-cell hypothesis is promising and unfinished. It is not a validated basis for buying capsules. Anybody who tells you otherwise skipped the part where the primary endpoint failed.

FAQ

Did UBX1325 actually beat aflibercept in the ASPIRE trial?

No. ASPIRE missed its primary endpoint, which was non-inferiority to aflibercept on visual acuity change at the average of weeks 20 and 24. UBX1325 reached statistical non-inferiority at week 36, meaning it could not be distinguished from aflibercept at that later timepoint. Non-inferiority is not superiority, and it was achieved after the prespecified primary analysis had already failed.

Why was the eye chosen as the first target tissue for a BCL-xL senolytic?

Because systemic BCL-xL inhibition destroys platelets. Platelets have no nucleus and depend almost entirely on BCL-xL for survival, so drugs like navitoclax cause dose-dependent thrombocytopenia in mice, dogs and humans. Injecting a 10 microgram dose directly into the vitreous of the eye delivers drug to the target tissue while keeping systemic exposure negligible, which sidesteps the platelet toxicity that limits systemic senolytics.

What happened to UNITY Biotechnology and UBX1325?

In May 2025 the company announced complete 36-week ASPIRE results alongside a workforce reduction affecting all employees, closure of the ASPIRE study, and a search for strategic alternatives. It filed its certificate of dissolution in Delaware in September 2025. UBX1325 currently has no sponsor advancing it into further trials.

Does this trial prove that fisetin or quercetin work as senolytics in people?

No. UBX1325 was a precisely dosed injection into a closed, immune-privileged compartment. Fisetin and quercetin are orally dosed compounds with poor bioavailability, and it has not been established that they reach senolytic concentrations in human tissues. The results from an intraocular injection say nothing about what an oral flavonoid does systemically.

What is the strongest randomized evidence for oral senolytics so far?

A 2024 phase 2 randomized controlled trial in Nature Medicine tested intermittent dasatinib plus quercetin in 60 postmenopausal women. The primary endpoint, change in the bone resorption marker CTx at 20 weeks, did not differ between groups. A bone formation marker rose transiently at 2 and 4 weeks but not at 20 weeks. No serious adverse events occurred. To date, no adequately powered randomized trial of an oral senolytic has met a positive primary clinical endpoint.

Medical disclaimer: This article is for informational purposes only and is not medical advice; UBX1325 is an investigational intraocular injection not approved for any use, and you should consult a qualified physician before making any health decisions.