Senolytics

Senolytics

Cellular senescence is a durable arrest with a secretory programme, a clearance problem, and both protective and maladaptive roles. Senolysis kills a subset of those cells. Senomorphics try to quiet what they secrete. The mouse record is large. The human record is small, short, and mostly surrogate. This article grades that gap. It is not a protocol.

Compiled by South Beach Longevity · 20 August 2026 Copyright 2026 Series SBL-41 / SP-SENOLYTICS · Register A scientific article Sources peer-reviewed cell, animal, and human studies; trial-registry records; labelled reviews · verified NCBI records Constraint This document describes published research. It is not medical advice. No human use, dose, route or schedule is recommended anywhere in this document.

How to read this document Every finding is labelled, in the sentence that reports it, by the kind of study that produced it. A genetically engineered mouse is not a patient. A fall in p16 transcript, SA-β-galactosidase, or a circulating SASP cytokine is not a hospitalization, a fracture, or a death. A licensed kinase-inhibitor label is not a senolytic approval. Amounts and schedules appear only as reported experimental or registry parameters, always with the population attached. Nothing here is a recommendation. Findings are graded in place as established, strongly supported, emerging, plausible, or speculative. Two further labels mark careful absences rather than verdicts: not established, where the evidence is too thin to place a claim on the ladder at all — untested or insufficient, an absence of proof rather than disproof; and not supported, where the weight of evidence leans against a claim but stops short of a formal refutation. Preclinical senolysis, human biomarker studies, and clinical outcomes are kept in separate sentences.

01 The cell state, and four things it is not

Cellular senescence is a durable form of proliferative arrest, usually accompanied by macromolecular damage, a remodelled secretome, and resistance to apoptosis. Hayflick and Moorhead described the finite replicative life of human diploid fibroblasts in culture; Hayflick then named the limit (Hayflick and Moorhead, 1961; Hayflick, 1965). That culture observation is established. It is not a theory of organismal ageing by itself, and it is not a licence to treat every non-dividing cell in an old tissue as senescent.

Four things follow immediately and will be enforced for the rest of this document.

First, senescence is not quiescence, terminal differentiation, or death. A quiescent cell can re-enter cycle. A differentiated cell has a programme. A dead cell is gone. A senescent cell remains metabolically active, often enlarged, and frequently secretory. Campisi and d’Adda di Fagagna stated that distinction as the field’s working definition (Campisi and d’Adda di Fagagna, 2007). Gorgoulis and a multi-laboratory consensus later insisted that no single marker is sufficient and that a multi-parameter path is required (Gorgoulis et al., 2019). Both papers are syntheses. They are strongly supported as maps of a large experimental literature. They are not a bedside assay.

Second, senescence is not only an ageing defect. Muñoz-Espín and Serrano reviewed beneficial and pathological roles in the same cell state (Muñoz-Espín and Serrano, 2014). He and Sharpless did the same from a medical angle (He and Sharpless, 2017). van Deursen argued that senescent cells are causal in selected ageing phenotypes in mice, not merely correlative passengers (van Deursen, 2014). Those claims must be graded separately for development, wound healing, tumour suppression, and late-life accumulation.

Third, a senescent cell is not a universal drug target. Childs and colleagues treated senescent cells as an emerging target for diseases of ageing, not as a completed pharmacology (Childs et al., 2017). Chaib, Tchkonia, and Kirkland later wrote the translational path as a path, not as an arrival (Chaib, Tchkonia, and Kirkland, 2022). Grosse and colleagues showed that some p16-high cell types are indispensable for mouse healthspan (Grosse et al., 2020). That paper is a genetic experiment, not a press release, and it is load-bearing for Part Five.

Fourth, this document is not medical advice and not a senolytic protocol. Reported milligrams, cycles, and combinations are experimental or registry parameters attached to named populations. They are not instructions for any reader.

02 Replicative, stress-induced, and oncogene-induced routes

The field uses three entry routes. They share an arrest. They do not share a cause.

Replicative senescence is the Hayflick limit: successive divisions erode telomeres until a DNA-damage checkpoint fires. d’Adda di Fagagna and colleagues showed that telomere-initiated senescence in human fibroblasts is a DNA-damage-checkpoint response (d’Adda di Fagagna et al., 2003). That cell-culture result is established. It does not prove that every p16-positive cell in an aged human organ arrived there by telomere erosion.

Stress-induced premature senescence is arrest after genotoxic, oxidative, metabolic, or oncogenic stress without requiring the Hayflick clock. Rodier and colleagues showed that persistent DNA-damage signalling triggers senescence-associated inflammatory cytokine secretion (Rodier et al., 2009). Wiley and related mitochondrial-dysfunction work sits in that family of mechanisms; the present harvest treats the DNA-damage paper as the load-bearing citation.

Oncogene-induced senescence is the arrest that follows an oncogenic insult in a still-intact checkpoint. Serrano, Lin, McCurrach, Beach, and Lowe showed that oncogenic Ras provokes premature senescence with accumulation of p53 and p16INK4a (Serrano et al., 1997). Kuilman and colleagues later showed that the arrest can be relayed by an interleukin-dependent inflammatory network (Kuilman et al., 2008). Acosta and colleagues showed that a complex secretory programme, orchestrated in part by the inflammasome, can impose paracrine senescence on neighbours (Acosta et al., 2013). Those are cell and mosaic-tissue experiments. They are strongly supported as mechanisms. They are not human outcome trials.

Campisi described four faces of the state — tumour suppression, ageing, tissue repair, and a secretory programme that can do harm at a distance (Rodier and Campisi, 2011; Campisi, 2013). The article uses that split. It does not collapse the four faces into one slogan.

03 DNA damage, p16, p21, and the arrest decision

The arrest is enforced by cyclin-dependent-kinase inhibitors. Two of them dominate the literature.

p21 (CDKN1A / WAF1 / CIP1 / SDI1) was cloned as a p53-inducible mediator of tumour suppression and as a Cdk-interacting inhibitor of G1 cyclin-dependent kinases, and again from senescent-cell expression screens (el-Deiry et al., 1993; Harper et al., 1993; Noda et al., 1994). In many acute-damage models it is the early brake.

p16INK4a (CDKN2A) was identified as a specific inhibitor of cyclin D/CDK4 (Serrano, Hannon, and Beach, 1993). It accumulates with time in several murine and human tissues and became the most widely used in-vivo marker of a senescent-like state (Krishnamurthy et al., 2004; Sharpless and Sherr, 2015). Beauséjour and colleagues showed that human fibroblast senescence can be reversed when p53 is inactivated if p16 has not yet locked the arrest, and is far harder to reverse once p16 is established (Beauséjour et al., 2003). That is a culture result. It is strongly supported as a description of two cooperating but non-identical brakes. It is not a statement that every p16-high cell in a human biopsy is permanently senescent.

d’Adda di Fagagna reviewed senescence as a DNA-damage response (d’Adda di Fagagna, 2008). Hernandez-Segura, Nehme, and Demaria listed hallmarks that travel together more often than any one of them travels alone: arrest, SA-β-galactosidase, SASP, morphological change, and selected chromatin and lysosomal features (Hernandez-Segura, Nehme, and Demaria, 2018). Sharpless and Sherr warned that forging a signature of in-vivo senescence is harder than staining a dish (Sharpless and Sherr, 2015). That warning is the hinge of section 08.

04 SASP

The senescence-associated secretory phenotype is the reason a non-dividing cell can still change a tissue. Coppé, Patil, Rodier, and Campisi showed that senescent cells secrete a suite of cytokines, chemokines, growth factors, and proteases, and that the composition depends on the trigger and on p53 status (Coppé et al., 2008). Coppé and colleagues later named the dark side of that programme for tumour suppression: the same secretome that can recruit clearance can also promote inflammation, angiogenesis, and a pro-tumorigenic microenvironment (Coppé et al., 2010).

Laberge and colleagues showed that mTOR regulates a pro-tumorigenic SASP in part by promoting IL-1A translation (Laberge et al., 2015). Xu and colleagues showed that JAK inhibition alleviates SASP and frailty measures in old mice (Xu et al., 2015). Moiseeva and colleagues showed that metformin can inhibit SASP by interfering with IKK/NF-κB activation in cell systems (Moiseeva et al., 2013). Those three papers are the mechanistic basis for calling some interventions senomorphic rather than senolytic. They are not human outcome trials, and metformin’s licensed indications are not senomorphic approvals.

SASP is established as a cell-biological object. It is emerging as a pharmacologic target. It is speculative as a stand-alone explanation of any named human disease until the disease trial is produced.

05 Immune clearance and tissue-specific burden

A senescent cell that is cleared is a different epidemiologic object from a senescent cell that persists. Kang and colleagues showed that senescence surveillance of pre-malignant hepatocytes limits liver-cancer development in mice (Kang et al., 2011). Xue and colleagues showed that restoring p53 in murine liver carcinomas can trigger senescence and tumour clearance (Xue et al., 2007). Krizhanovsky and colleagues showed that senescence of activated stellate cells can limit liver fibrosis (Krizhanovsky et al., 2008). Those are murine experiments. They are strongly supported as proofs that immune recognition of senescent cells can be beneficial.

The opposite paper is also required. Ovadya and colleagues showed that impaired immune surveillance accelerates accumulation of senescent cells and ageing phenotypes in mice (Ovadya et al., 2018). Pereira and colleagues showed that senescent cells can evade clearance via HLA-E-mediated inhibition of NK and CD8 T cells (Pereira et al., 2019). Prata, Tchkonia, and Kirkland reviewed immune clearance as a physiological senolytic (the review is used here only as a map; primary claims rest on the mouse and human-cell papers).

Tissue distribution is not uniform. Krishnamurthy and colleagues reported Ink4a/Arf as an ageing biomarker across several murine tissues (Krishnamurthy et al., 2004). Tuttle and colleagues systematically reviewed cellular-senescence markers against chronological age in human tissues and found the association real, heterogeneous, and assay-dependent (Tuttle et al., 2020). Idda and colleagues surveyed senescent-cell markers with age in human tissues and likewise found no single universal stain (Idda et al., 2020). Waaijer and colleagues showed that the number of p16INK4a-positive cells in human skin reflects biological age more tightly than chronological age in that sample (Waaijer et al., 2012). Those human observational papers are strongly supported as evidence that a senescent-like burden rises with age in some tissues. They do not identify a universal targetable cell.

06 Beneficial roles: development, wound healing, tumour suppression

Three beneficial roles are established in animals and plausible as constraints on indiscriminate senolysis in humans.

Development. Muñoz-Espín and colleagues and Storer and colleagues independently showed programmed senescence during mammalian embryonic development and patterning (Muñoz-Espín et al., 2013; Storer et al., 2013). Those are developmental-biology papers. They do not say that an adult senolytic will abort a pregnancy. They do say that senescence is a physiological tool, not a synonym for decay.

Wound healing. Demaria and colleagues showed that senescent cells are required for optimal wound healing in mice through secretion of PDGF-AA (Demaria et al., 2014). Transient senescence at a wound edge is therefore not automatically a disease. A drug that clears those cells at the wrong time is a different experiment from a drug that clears chronically accumulated cells in an old kidney.

Tumour suppression. Oncogene-induced senescence, p53 restoration, and immune surveillance papers already cited make the tumour-suppressive face strongly supported in mice. Campisi’s reviews collect that case (Campisi, 2013). The adversarial question in section 24 is what happens if a senolytic removes a barrier that was still working.

07 Maladaptive accumulation with age

The causal mouse case for maladaptive accumulation is the INK-ATTAC series. Baker and colleagues showed that clearance of p16Ink4a-positive senescent cells delays ageing-associated disorders in a progeroid background (Baker et al., 2011). Baker and colleagues later showed that naturally occurring p16Ink4a-positive cells shorten healthy lifespan in non-progeroid mice (Baker et al., 2016). Those two Nature papers are strongly supported as genetic evidence that a p16-high population can be causal in mice. They are not human lifespan trials.

Disease-specific mouse papers then attached named pathologies to that population: senescent intimal foam cells at all stages of atherosclerosis (Childs et al., 2016); fibrotic pulmonary disease (Schafer et al., 2017); age-dependent hepatic steatosis (Ogrodnik et al., 2017); post-traumatic osteoarthritis (Jeon et al., 2017); age-related bone loss (Farr et al., 2017); vasomotor dysfunction (Roos et al., 2016); tau-dependent pathology after clearance of senescent glia (Bussian et al., 2018); obesity-induced metabolic dysfunction (Palmer et al., 2019); obesity-induced anxiety and impaired neurogenesis (Ogrodnik et al., 2019); worse recovery after myocardial infarction in aged mice (Walaszczyk et al., 2019); and senescent beta cells in a type-1-diabetes model (Thompson et al., 2019). Xu and colleagues showed that transplanting senescent cells into young mice induces an osteoarthritis-like condition, which is the gain-of-function complement to clearance (Xu et al., 2017). Xu and colleagues later reported that senolytics improve physical function and increase remaining lifespan when senescent cells are transplanted or when old mice are treated (Xu et al., 2018).

That catalogue is strongly supported as a mouse atlas. It is emerging as a human therapeutic hypothesis. It is not a list of approved indications.

Grosse and colleagues remain the required counterweight: deleting some defined p16-high cell types in mice impaired healthspan (Grosse et al., 2020). Cohn, Gasek, Kuchel, and Xu reviewed single-cell heterogeneity as the reason one marker cannot carry a drug programme (Cohn et al., 2023). Di Micco, Krizhanovsky, Baker, and d’Adda di Fagagna reviewed mechanisms-to-therapy as a still-open translation (Di Micco et al., 2021).

08 Why senescent cells are hard to identify in humans

This is the measurement problem, not a footnote.

Dimri and colleagues introduced SA-β-galactosidase as a biomarker that identifies senescent human cells in culture and in ageing skin (Dimri et al., 1995). Lee and colleagues later showed that the activity is lysosomal β-galactosidase, detectable when lysosomal mass rises, not a unique senescence enzyme (Lee et al., 2006). The stain is therefore useful and non-specific. It is established as a culture companion marker. It is not a stand-alone in-vivo identity.

Sharpless and Sherr, Hernandez-Segura and colleagues, González-Gualda and colleagues, and the 2019 Cell consensus all require multi-parameter identification: arrest, a CDK inhibitor, a damage or chromatin mark, a SASP panel, and a tissue context (Sharpless and Sherr, 2015; Hernandez-Segura, Nehme, and Demaria, 2018; Gorgoulis et al., 2019; González-Gualda et al., 2021). SenNet recommendations and the 2024 MINCS guidelines restated that rule for in-situ detection and for in-vivo experimentation (Suryadevara et al., 2024; Ogrodnik et al., 2024). Those consensus documents are strongly supported as methods law. They are the reason a consumer “senolytic panel” that reports one cytokine is not a senescence assay.

Tuttle’s systematic review and Idda’s tissue survey show that human marker–age correlations exist and are inconsistent across organs and methods (Tuttle et al., 2020; Idda et al., 2020). Waaijer’s skin p16 counts track biological age in that tissue (Waaijer et al., 2012). Blood p16 in T cells has been used as a systemic ageing biomarker (Krishnamurthy et al., 2004, in the murine founding paper; human blood applications are younger and more variable). None of these is a companion diagnostic for senolysis.

The honest sentence is therefore: senescence can be measured in humans, but not yet reliably enough, or specifically enough, to support a claim that a drug cleared the right cells in the right tissue. That sentence is established as a methods conclusion. It governs every human table in Part Four.

09 Senolysis versus senomorphics

Senolysis is the selective killing of senescent cells. Senomorphics are agents that suppress or reprogramme SASP, or otherwise modify senescent-cell behaviour, without a requirement that the cell die. Kirkland and Tchkonia used that split in their translational reviews (Kirkland and Tchkonia, 2020; Chaib, Tchkonia, and Kirkland, 2022). Robbins and colleagues reviewed senolytic drugs as agents that reduce senescent-cell viability to extend healthspan (Robbins et al., 2021). The split is conceptual and pharmacologic. It is not a marketing taxonomy.

Xu’s JAK-inhibitor paper and Laberge’s mTOR–IL-1A paper are senomorphic experiments (Xu et al., 2015; Laberge et al., 2015). Zhu’s dasatinib-plus-quercetin paper and Yosef’s and Chang’s BCL-2-family papers are senolytic experiments (Zhu et al., 2015; Yosef et al., 2016; Chang et al., 2016). A flavonoid that is senolytic in a dish at a concentration never reached in human plasma after a food-like dose is still a senolytic experiment. It is not a dietary protocol.

10 Agent and mechanism classes

The classes that matter mechanistically, with the founding papers attached, are these.

Dasatinib plus quercetin. Zhu and colleagues identified senescent-cell anti-apoptotic pathways and reported that dasatinib plus quercetin preferentially reduced senescent-cell viability (Zhu et al., 2015). Dasatinib is a licensed BCR-ABL/SRC-family kinase inhibitor for Philadelphia-chromosome-positive leukaemias (Kantarjian et al., 2010; Cortes et al., 2016). That licence is established. It is not a senolytic indication. Quercetin is a dietary flavonol with low and variable human oral bioavailability (Graefe, Derendorf, and Veit, 1999; Manach et al., 2004). Combining a prescription kinase inhibitor with a poorly bioavailable flavonoid is a specific experimental object. It is not “a natural senolytic stack.”

Fisetin. Zhu and colleagues later reported fisetin, with BCL-XL inhibitors, as agents that target senescent cells (Zhu et al., 2017). Yousefzadeh and colleagues reported that fisetin is senotherapeutic and extends health and lifespan in mice (Yousefzadeh et al., 2018). Touil and colleagues described fisetin disposition and metabolism in mice, including geraldol as an active metabolite (Touil et al., 2011). Those are mouse and cell papers. Human intermittent fisetin regimens sold as senolytics are ahead of the human outcome record.

Navitoclax and BCL-2-family approaches. Zhu and colleagues identified navitoclax as a senolytic targeting BCL-2-family anti-apoptotic factors (Zhu et al., 2016). Yosef and colleagues showed directed elimination of senescent cells by inhibition of BCL-W and BCL-XL (Yosef et al., 2016). Chang and colleagues showed that ABT-263 (navitoclax) clears senescent cells and rejuvenates aged haematopoietic stem cells in mice (Chang et al., 2016). In people, navitoclax is an oncology drug whose dose-limiting toxicity is thrombocytopenia (Wilson et al., 2010; Rudin et al., 2012). That human harm is established in cancer trials. It is the reason BCL-XL remains a difficult systemic senolytic target.

FOXO4-related experimental approaches. Baar and colleagues reported that a FOXO4-directed peptide induced targeted apoptosis of senescent cells and restored tissue homeostasis after chemotoxicity and in ageing mice (Baar et al., 2017). That is a landmark mouse paper. It is emerging mechanistically and not a human medicine. No completed FOXO4-DRI registration trial was returned by ClinicalTrials.gov in the reviewed record.

HSP90 approaches. Fuhrmann-Stroissnigg and colleagues identified HSP90 inhibitors as a senolytic class in a screen and in mice (Fuhrmann-Stroissnigg et al., 2017). HSP90 inhibitors have their own oncology toxicity history. The senolytic claim is preclinical.

Other experimental senolytics. Cardiac glycosides were reported as broad-spectrum senolytics (Guerrero et al., 2019; Triana-Martínez et al., 2019). USP7 inhibition, BET-family degraders, glutaminolysis inhibition, and synthetic-lethal metabolic targeting have each produced senolysis in experimental systems (He et al., 2020; Wakita et al., 2020; Johmura et al., 2021; Dörr et al., 2013). Those are plausible class expansions. None is a longevity indication.

Immune-mediated strategies. Amor and colleagues engineered uPAR-targeted CAR T cells that reverse senescence-associated pathologies in mice (Amor et al., 2020) and later reported prophylactic and lasting metabolic benefit in mice (Amor et al., 2024). That is emerging immunotherapy, not a clinic-ready senolytic.

Targeted delivery. β-galactosidase-targeted prodrugs and an antibody–drug conjugate against a senescent-cell membrane marker have been reported in mice (Cai et al., 2020; Poblocka et al., 2021). Gasek and colleagues reviewed targeting strategies, including local delivery (Gasek et al., 2021). Local delivery is the design logic behind intra-articular UBX0101 and intravitreal UBX1325 (foselutoclax): keep a BCL-family poison out of platelets. The logic is plausible. The human outcomes are in Part Four.

11 What a valid human target would require

A valid human senolytic indication would need four things at once: (1) a tissue in which senescent cells are shown, by multi-parameter methods, to be causal or at least necessary for the phenotype; (2) an agent that engages that population with an acceptable off-target profile; (3) a pre-specified clinical outcome, not only a biomarker; and (4) a harm profile that does not erase the benefit by impairing healing, haemostasis, or tumour suppression. Chaib, Tchkonia, and Kirkland stated a version of that translational path (Chaib, Tchkonia, and Kirkland, 2022). Kirkland and Tchkonia’s earlier “from discovery to translation” review is the same claim in an earlier tense (Kirkland and Tchkonia, 2020). The present document treats those reviews as maps. The verdicts below are from primary trials and registry records.

12 Genetic clearance is not a drug

INK-ATTAC and related p16-driven suicide systems ask a clean question: if a p16-high cell is removed, which phenotypes move? Baker 2011 and Baker 2016 answered “several ageing-associated disorders, and healthy lifespan” in mice (Baker et al., 2011; Baker et al., 2016). Grosse 2020 answered “not those cells, or not all of them, and not in every tissue” (Grosse et al., 2020). Both answers can be true because p16-high is a marker class, not a cell type. Pharmacologic senolysis is a cruder instrument than a tissue-restricted suicide gene. Mouse drug papers that cite INK-ATTAC as if the drug inherited the genetic specificity are over-claiming.

13 Animal lifespan and healthspan

Xu and colleagues reported that senolytics improved physical function and increased remaining lifespan in old mice and in mice transplanted with senescent cells (Xu et al., 2018). Yousefzadeh and colleagues reported fisetin extension of health and lifespan in mice (Yousefzadeh et al., 2018). Those two papers are the usual sources for “senolytics extend lifespan.” They are mouse papers. They are strongly supported as mouse results. They are not human lifespan evidence.

Disease-model healthspan effects, already listed in section 07, are collected in Matrix C. The pattern is broad and impressive in rodents: vessels, bone, fat, liver, lung, joint, brain, heart, and pancreas have each moved in at least one clearance or senolytic experiment. The pattern is also selected: laboratories publish the models that work. Nulls are thinner in the reviewed record than positives. That publication shape is itself a finding.

Anderson and colleagues showed length-independent telomere damage driving post-mitotic cardiomyocyte senescence (Anderson et al., 2019). Minamino and colleagues showed a crucial role for adipose-tissue p53 in insulin resistance (Minamino et al., 2009). Those papers widen the target beyond fibroblasts. They also widen the risk that a senolytic will hit a cell that was doing a job.

14 Transferability limits

Mice are small, short-lived, inbred or genetically engineered, and housed in specific-pathogen-free rooms. Their senescent-cell burden, immune clearance, and SASP composition are not human by default. Di Micco and colleagues and Chaib and colleagues both treat translation as the unsolved step (Di Micco et al., 2021; Chaib, Tchkonia, and Kirkland, 2022). The present grade is mechanical: a mouse healthspan effect is preclinical. It becomes emerging in humans only after a controlled trial with a clinical outcome. It becomes established only after replication and an acceptable harm profile. No senolytic has crossed that last line as of the compilation date.

15 Human biomarker studies, kept separate from outcomes

Hickson and colleagues published a preliminary report that dasatinib plus quercetin decreased senescent-cell markers in adipose tissue and skin from individuals with diabetic kidney disease, with a later corrigendum (Hickson et al., 2019; Hickson et al., 2020). The parent registry record is NCT02848131, a Mayo Phase 2 study whose primary outcome is change in the proportion of senescent cells (ClinicalTrials.gov, NCT02848131, retrieved 20 August 2026). That paper is a human biomarker study. It is emerging as evidence that an intermittent D+Q course can move tissue markers in a small open-label sample. It is not a demonstration that diabetic kidney disease improved as a clinical outcome.

Justice and colleagues’ first-in-human IPF study reported feasibility and some physical-function signals in an open-label, fourteen-participant design (Justice et al., 2019). Function here is closer to an outcome than a cytokine, but the design cannot separate placebo, training, or regression effects. The grade is emerging for feasibility and not established for disease modification.

Farr and colleagues’ Phase 2 randomized trial in postmenopausal women measured bone-metabolism markers after intermittent senolytic therapy (Farr et al., 2024). Registry NCT04313634 lists change in CTX as the primary outcome, actual enrollment 74, status completed (ClinicalTrials.gov, NCT04313634). A randomized bone-turnover trial is stronger than an open-label pilot. It remains a biomarker-primary skeletal study unless a fracture or BMD endpoint is pre-specified and met. The published paper is graded emerging for bone-remodelling effects and not as an osteoporosis indication.

Gonzales and colleagues published the SToMP-AD pilot protocol (Gonzales et al., 2022). The completed Phase 1/2 registry record NCT04063124 enrolled five participants; primary outcomes were brain penetrance of dasatinib and of quercetin (ClinicalTrials.gov, NCT04063124). A 2023 Research Square preprint reported first clinical-trial outcomes of senolytic therapy for Alzheimer’s disease (Gonzales et al., 2023). A preprint is labelled as a preprint. It is not treated here as a peer-reviewed outcome paper.

UNITY’s intra-articular UBX0101 Phase 2 knee-osteoarthritis trial (NCT04129944) enrolled 183 participants; the programme did not become an approved OA medicine. Intravitreal foselutoclax (UBX1325) has a completed Phase 2a DME study (BEHOLD, NCT04857996, n=65), a Phase 2 wet-AMD study (ENVISION, NCT05275205, n=51), and a Phase 2b DME study (ASPIRE, NCT06011798, n=52) (ClinicalTrials.gov records, retrieved 20 August 2026). Klier and colleagues reported safety and efficacy of UBX1325 in diabetic macular oedema in NEJM Evidence (Klier et al., 2025). Chew published an accompanying commentary (Chew, 2025). Macha and colleagues reported multifocal electroretinography changes after UBX1325 in neovascular AMD (Macha et al., 2024). Those ophthalmology papers are the closest the field currently comes to a clinical-outcome senolytic programme, in a local compartment, against an active anti-VEGF control in later studies. They are emerging. They are not a systemic longevity claim, and they are not an approved medicine as of compilation.

16 Completed trials: populations, endpoints, adverse events

Nambiar and colleagues reported a Phase I, single-blind, single-centre, randomized, placebo-controlled pilot of dasatinib plus quercetin in idiopathic pulmonary fibrosis, designed for feasibility and tolerability (Nambiar et al., 2023). That paper is the correct object for “IPF senolytic RCT.” It is not a Phase 3 antifibrotic trial. Enrollment is small. The honest reading is: intermittent D+Q was feasible in that sample; disease modification is unproven.

Dasatinib’s adverse-event profile in leukaemia — cytopenias, pleural effusion, pulmonary arterial hypertension, bleeding, QTc effects — is established on the DASISION programme and the licensed label (Kantarjian et al., 2010; Cortes et al., 2016). Those harms were characterized at oncologic exposures and chronic daily dosing. Intermittent senolytic schedules are a different exposure. They do not erase the pharmacology. Quercetin’s human limitation is absorption and variability, not a leukaemia-style toxicity database (Graefe, Derendorf, and Veit, 1999; Manach et al., 2004). Navitoclax’s thrombocytopenia is established (Wilson et al., 2010; Rudin et al., 2012). FOXO4-DRI has no adequate human safety database in the reviewed record. Fisetin’s human safety database as a senolytic is small and schedule-specific; food-use history is not a senolytic safety file.

Matrix E collects those harms. The rule is: report the source-indication toxicity beside the senolytic hypothesis, and do not treat intermittent use as a demonstrated detoxification.

17 Ongoing trials

ClinicalTrials.gov returned 71 unique studies on the senolytic-related queries run for this compilation (20 August 2026). They are not 71 independent, adequate, outcome-driven tests of senolysis. Many are small, open-label, biomarker-primary, dietary-supplement, or adjacent (for example Qualia Senolytic NCT06953518, an open-label cytokine pilot). Material ongoing or recently active registry objects include SENIOR (NCT06018467, D+Q ± nicotinamide riboside, bone-resorption CTX, n=120 estimated, Denmark); Cedars-Sinai adipose single-nuclei mapping with D+Q (NCT05653258, n=160 estimated); SEN-SURVIVORS in adult survivors of childhood cancer (NCT04733534); ALSENLITE (NCT04785300, open-label D+Q in AD/MCI, n=20 estimated); SToMP-AD Phase 2 (NCT04685590, n=48 estimated, primary: adverse events versus placebo); fisetin vascular function in older adults (NCT06133634, n=70 estimated); and D+Q for accelerated ageing in mental disorders (NCT05838560, n=40 estimated). FOXO4-DRI returned no registry hits. Navitoclax-plus-senescence returned none as a senolytic programme.

Registry status is a snapshot. It is established as a description of what was registered. It is not a result.

18 Commercial protocols versus the trial record

Commercial intermittent fisetin or quercetin protocols typically inherit their rhetoric from Zhu 2015–2017 and Yousefzadeh 2018, sometimes from Hickson 2019. They do not inherit the mouse strain, the combination with dasatinib, the tissue assay, or the trial population. Quercetin’s human oral bioavailability is low and formulation-dependent (Graefe, Derendorf, and Veit, 1999; Manach et al., 2004). Fisetin’s published disposition paper in the reviewed record is murine (Touil et al., 2011). A food-like flavonoid exposure is not a demonstrated senolytic exposure. The grade for commercial “senolytic” supplement protocols as clinical senolysis is speculative, and the practice is ahead of the evidence. That sentence is a scientific finding, not a consumer recommendation in the other direction.

19 Are senescent cells valid universal targets?

Challenge. If senescence is heterogeneous, tissue-specific, and sometimes beneficial, a universal senolytic target is a category error.

Evidence. Gorgoulis 2019, Sharpless and Sherr 2015, Cohn 2023, Tuttle 2020, Idda 2020, and Grosse 2020.

Resolution. They are not valid universal targets. They are valid conditional targets in tissues where a defined senescent population is shown to drive a phenotype and where clearance can be confined. The INK-ATTAC positives and the Grosse indispensable-cell negatives are both required. Verdict: the universal-target claim fails. The tissue-and-context claim remains open.

20 Could indiscriminate senolysis impair healing or tumour suppression?

Challenge. Demaria 2014, Muñoz-Espín 2013, Storer 2013, Xue 2007, Kang 2011, and Krizhanovsky 2008 show useful senescence. Clearing the wrong cell at the wrong time could impair wounds or remove a tumour-suppressive barrier.

Resolution. The risk is plausible and not theoretical only in mice. Human senolytic trials to date are too small and too short to measure wound-failure or incident-cancer signals with any power. Local delivery (joint, vitreous) is a rational attempt to limit that risk. Intermittent scheduling is a hypothesis, not a demonstration that the risk is gone. Verdict: indiscriminate systemic senolysis carries a real, unquantified human risk to repair and to tumour suppression. The risk is a reason for indication discipline, not a reason to deny all research.

21 Are mouse effects transferable to humans?

Challenge. Xu 2018 and Yousefzadeh 2018 are easy to over-read.

Resolution. Transfer is not a property of a pathway cartoon. It is a property of a trial. Haematopoietic, immune, telomere, and SASP differences are large enough that mouse healthspan is a discovery engine, not a surrogate. Verdict: mouse effects are not transferable by analogy. They are transferable only by human outcome trials that have not yet been produced for systemic senolysis.

22 Are current human trials too small and too short?

Challenge. n=5 to n=74 for most systemic programmes; weeks to months of follow-up; biomarker primaries.

Resolution. Yes. Nambiar 2023 is a feasibility RCT. Hickson 2019 is a preliminary biomarker report. Justice 2019 is open-label. Farr 2024 is larger and randomized but bone-turnover-primary. SToMP-AD Phase 1 enrolled five. UBX1325 ophthalmology trials are the only completed programmes with tens of participants and clinical-function endpoints, and they are local, not systemic. Verdict: current systemic human trials are too small and too short to establish clinical benefit. That is a fact about the record, not a slur on the investigators.

23 Are commercial fisetin and quercetin protocols ahead of the evidence?

Resolution. Yes. The founding senolytic papers used specific combinations, concentrations, and assays (Zhu et al., 2015; Zhu et al., 2017; Yousefzadeh et al., 2018). Human flavonoid pharmacokinetics do not reproduce those conditions by default (Graefe, Derendorf, and Veit, 1999; Manach et al., 2004; Touil et al., 2011). No adequate randomized human outcome trial supports a commercial intermittent flavonoid regimen as senolysis. Verdict: commercial protocols are ahead of the evidence.

24 Does reduction in senescence biomarkers demonstrate clinical benefit?

Resolution. No. Hickson 2019 moved tissue markers. That is a pharmacodynamic observation. HF-ACTION and Look AHEAD, in a sibling SBL title, already taught the series that a better surrogate is not an event. The same law applies here. Verdict: biomarker reduction does not demonstrate clinical benefit. It can support target engagement if the marker is valid, which in this field it often is not.

25 Can senescence be measured reliably in humans?

Resolution. Partially, in research settings, with multi-parameter panels and tissue context (Gorgoulis et al., 2019; Suryadevara et al., 2024; Ogrodnik et al., 2024). Not reliably enough for a single blood test to adjudicate a senolytic claim. Verdict: measurement is the field’s unsolved companion-diagnostic problem.

26 Dedicated skeptical review — assume narrower clinical utility than expected

Assume, as required, that senolytics will ultimately have narrower clinical utility than the longevity field currently expects.

Under that assumption the surviving uses look like this. First, local, high-burden, low-repair compartments — a joint, a vitreous cavity, a fibrotic niche — where a senescent population can be reached without chronic systemic BCL-XL or kinase inhibition. UBX1325 is the existence proof that industry already made that bet; UBX0101 is the existence proof that the bet can fail. Second, short, intermittent courses in a named disease with a pre-specified outcome, not a consumer longevity stack. Third, senomorphics as a quieter, possibly more tractable path in diseases where SASP is the driver and the cell still has a job. Fourth, immune and targeted-delivery strategies that inherit oncology’s lesson: specificity is the drug.

What falls away under the assumption is the idea of a general “senolytic” as a longevity vitamin; the idea that p16 is a universal kill-switch; the idea that a flavonoid capsule is translationally equivalent to INK-ATTAC; and the idea that a two-week biomarker study can carry a healthspan claim. Kirkland’s own translational reviews, read without the keynote cadence, already contain most of those restrictions (Kirkland and Tchkonia, 2020; Chaib, Tchkonia, and Kirkland, 2022). Grosse 2020 and the measurement consensus are the rest.

Formal resolution of the skeptical brief. The brief is accepted as the prior. The evidence did not overturn it. Promising geroscience remains. Established human medicine, for systemic senolysis as a class, does not.

27 Translational-evidence hierarchy

The hierarchy used in Matrix F is ordered, not diplomatic.

1. Established human medicine — licensed indication, adequate trials, known harm. Example: dasatinib for CML (Kantarjian et al., 2010). Not a senolytic claim.
2. Emerging local clinical programmes — UBX1325 ophthalmology (Klier et al., 2025). Disease-adjacent, not longevity.
3. Human feasibility and biomarker trials — Justice 2019, Hickson 2019, Nambiar 2023, Farr 2024, SToMP-AD n=5.
4. Genetic mouse causality — Baker 2011, Baker 2016, with Grosse 2020 as a restriction.
5. Pharmacologic mouse healthspan — Xu 2018, Yousefzadeh 2018, and the disease-model atlas.
6. Cell-system mechanism — Zhu 2015 and the class papers.
7. Reviews, keynotes, and commercial protocols — hypothesis generators, not evidence.

A claim may move up only by producing the object on the rung above. It may not borrow prestige from a lower rung.

Research matrices

Matrix A — Senescence markers

Matrix B — Agent and mechanism

Matrix C — Animal lifespan and healthspan

Matrix D — Human trial matrix

Registry facts: ClinicalTrials.gov API v2, 20 August 2026. Published facts: NCBI records cited in the table.

Matrix E — Safety and adverse events

Matrix F — Translational-evidence hierarchy

28 Formal resolutions

The seven required challenges are resolved in sections 19–25. The dedicated skeptical brief is accepted in section 26. The hierarchy in section 27 is the operational summary.

Promising geroscience: the cell state is real; SASP is real; mice can be made healthier by removing some senescent cells; several pharmacologic classes exist; local BCL-family senolysis has reached controlled human ophthalmology trials.

Established human medicine: dasatinib for CML is medicine. Navitoclax thrombocytopenia is medicine’s harm. Senolysis as a longevity or multi-morbidity therapy is not, as of 20 August 2026, established human medicine.

Standing constraint This document describes published research. It is not medical advice. It does not recommend human use of any compound, food, or method and specifies no dose, route, or schedule for any person. Investigational and prescription agents named here are discussed as scientific objects. They are not sourced, titrated, or combined for any reader.

Evidence handling

Study type is labelled in the reporting sentence. Animal and in-vitro results are never phrased as human outcomes. Conflicting evidence is presented as conflict: Baker 2016 and Grosse 2020 are both kept. Reviews are maps. Registry records are snapshots dated 20 August 2026. Preprints are labelled. Project 06 was queried read-only (senolytic title hits were present in the live catalog and the integration layer; they were used for discovery, not as citation authority). Project 07 news was discovery only and is not cited as evidence. the NCBI record is the citation authority for published papers.

Literature through August 2026. Claims that a later trial overturns a grade should be re-adjudicated against the primary paper and the registry, not against this paragraph.

References