Fisetin, senolytics, and SASP: mechanisms, preclinical evidence, and limits of clinical data

Senescence is not a single type of “old cell,” but a set of durable cell-cycle-arrest states arising after different forms of injury and carrying distinct secretory profiles. In some circumstances, this response limits tumor expansion and supports wound repair; in others, cells that persist chronically alter the tissue environment. A reduction in p16 or an inflammatory cytokine alone therefore does not amount to organ rejuvenation. Fisetin, the dietary flavonol 3,3′,4′,7-tetrahydroxyflavone, is of interest because it reduced the survival of senescent cells in some cellular systems and was accompanied by changes in aging markers and survival in mouse experiments. Translating these findings into a treatment regimen for humans remains a separate task.

It is useful from the outset to distinguish three types of outcome. A senolytic effect means death of previously viable senescent cells, assessed with attention to overall cytotoxicity and the fate of nonsenescent controls. A senomorphic effect changes SASP secretion or related signaling pathways while the cell may remain alive. A decrease in the proportion of SA-β-gal-positive cells, p16INK4a, or an interleukin in a sample is compatible with either scenario but does not distinguish them on its own. Finally, a clinical effect requires comparison of meaningful outcomes in people: function, symptoms, morbidity, or survival. The discussion below keeps these evidence levels separate.

How the senescent state forms

Replicative senescence in culture is associated, among other factors, with the end-replication problem at the ends of linear DNA. With each replication cycle, TTAGGG telomeric repeats may shorten unless the telomerase complex compensates for the loss. The shelterin complex—TRF1, TRF2, RAP1, TIN2, TPP1, and POT1—protects the telomere end, participates in t-loop formation, and limits its mistaken recognition as DNA damage. There is no universal critical length in kilobases: damaged or unprotected telomeres can initiate a response at different measured lengths. An exposed end and other persistent lesions support telomeric DNA-damage response foci involving MRN, ATM/ATR, γH2AX, and 53BP1. ATM activates CHK2; ATR primarily activates CHK1. Through p53 stabilization, this signal increases CDKN1A and p21 expression. This describes a causal sequence, but an individual TIF or γH2AX focus is not a specific diagnostic test for cellular senescence (Hayflick and Moorhead, 1961; Gorgoulis et al., 2019).

A cell can enter senescence without progressive telomere exhaustion. Ionizing radiation, some cytotoxic drugs, HRAS or BRAF hyperactivation, and other forms of replication stress trigger a persistent DNA-damage response. Tumor stress may create a protective barrier to proliferation; this is one reason universal elimination of all senescent cells does not appear to be a safe goal. A distinct variant is mitochondria-induced senescence (MiDAS): respiratory dysfunction, altered NAD⁺/NADH balance, and AMPK–p53 activation may occur with a secretory profile that differs from canonical IL-1-dependent SASP. In MiDAS models, secretion of some IL-1-dependent factors was reduced; this does not establish that IL-6 is absent from every mitochondrially damaged tissue (Wiley et al., 2016).

Two interacting cell-cycle brakes are the p53–p21 and p16INK4a–Rb axes. p21 limits cyclin E/A–CDK2 activity; the CDKN2A product p16INK4a binds CDK4/6 and prevents their action with cyclin D. When Rb is insufficiently phosphorylated, it remains bound to E2F and suppresses transcription of S-phase entry genes, including replication machinery components. Changes in H3K27me3 at the CDKN2A locus, involvement of demethylases such as KDM6B/JMJD3, HDAC recruitment, and formation of H3K9me3-rich senescence-associated heterochromatin foci may stabilize arrest. However, SAHF do not form in every senescent population; their absence does not exclude the state. Durable cell-cycle arrest is also not identical to absolute irreversibility under every experimental manipulation (Campisi and d’Adda di Fagagna, 2007; Gorgoulis et al., 2019).

A damaged cell may survive strong proapoptotic signaling through senescent-cell antiapoptotic pathways (SCAPs). Depending on the model, these include reliance on BCL-2-like proteins BCL-xL, BCL-2, or BCL-w; PI3K–AKT; the peptide inhibitor p21; serpin PAI-1; or receptor tyrosine kinase-related signaling. After recruitment to membrane PIP₃, AKT is phosphorylated by PDK1 at Thr308 and by mTORC2 at Ser473; BAD phosphorylation and other outputs support survival. BH3-only proteins BIM, PUMA, and NOXA may increase after damage, but a “super-high” concentration is not universal across all senescent cells. In the study by Zhu et al., dependence on BCL-xL and other SCAPs varied among preadipocytes, endothelial cells, and fibroblasts. Thus, senolytic selectivity must be established for each cellular model rather than inferred from a general apoptosis-threshold diagram (Zhu et al., 2015).

SASP is a secretory program, not an invariant “toxin.” IL-6, IL-1α/β, CXCL8, CCL2, TGF-β, VEGF, matrix metalloproteinases, and extracellular vesicles may appear in different combinations; the profile depends on the stressor, tissue, time since arrest, and immune environment. In some models, IL-1 and NF-κB sustain an inflammatory circuit, JAK/STAT3 transmits signals to neighboring cells, and MMPs alter the extracellular matrix. Paracrine spread of a senescence-like state is possible, but it cannot automatically be equated with clinical propagation of aging through the body. A transient SASP during repair may be useful; a chronic inflammatory program may have different consequences (Coppé et al., 2008; Gorgoulis et al., 2019).

Fisetin’s molecule and the limits of dietary comparisons

Fisetin has the formula C₁₅H₁₀O₆ and a molar mass of 286.24 g/mol. Its C₆–C₃–C₆ scaffold includes a chromone ring with a 4-keto group, 3- and 7-hydroxyl groups, and a 3′,4′-OH catechol moiety on the phenyl ring. The catechol can participate in redox reactions and complex formation, while the conjugated system affects physicochemical properties. Quercetin has an additional 5-OH group, which can form an intramolecular hydrogen bond with the 4-carbonyl. This is a real structural difference, but by itself it cannot calculate affinity for PI3K, BCL-xL, blood–brain barrier transport, or senolytic selectivity. logP varies with method and state and does not replace measurement of free tissue concentration.

In plants, the flavonol scaffold arises through the phenylpropanoid pathway: phenylalanine ammonia-lyase converts L-phenylalanine to cinnamic acid, subsequent reactions lead to 4-coumaroyl-CoA, chalcone synthase condenses this substrate with three malonyl-CoA molecules, chalcone isomerase forms the flavanone scaffold, and hydroxylation and oxidation contribute to flavonol formation. This scheme should not be read as one proven, identical fisetin-biosynthesis route in every plant mentioned. Fisetin and its conjugates are found in smoke tree, some berries, fruits, and vegetables, including strawberries, apples, and persimmons. Food content depends on cultivar, tissue, maturity, and analytical method; exact values from the source table are not sufficiently tied to a stable method and sample to present as universal norms. Dietary intake is not equivalent to an experimental dose of the pure compound. But the reverse conclusion—that a specific gram dose is “necessary” for human senolysis—does not follow either.

In the study by Yousefzadeh et al. (2018), ten flavonoid/polyphenol candidates were compared in cellular models, and fisetin was the most active on the selected measures. A 5 μM dose was used in the initial comparison in Ercc1-deficient fibroblasts; in human IMR90 cells after etoposide, 1–15 μM was assessed over 48 hours. A reduction in viable SA-β-gal-positive cells relative to overall viability is consistent with senolytic activity under these conditions. This is not a ranking of all natural compounds across every cell type. In the same study, responses in selected mouse adipose-tissue populations and human explants were cell-specific; a human explant treated with 20 μM ex vivo is not a trial of fisetin intake in people.

From signaling pathway to cell death: what is known about the mechanism

The phrase “fisetin inhibits PI3K/Akt/mTOR” combines several distinct experimental claims. Class I PI3K uses ATP to phosphorylate membrane PI(4,5)P₂ into PI(3,4,5)P₃; this lipid product recruits the PH domains of AKT and PDK1. PDK1 phosphorylates AKT at Thr308, while mTORC2 phosphorylates it at Ser473. Active AKT acts on BAD, TSC2, FOXO, and other substrates, while mTORC1 regulates growth and protein synthesis through S6K and 4E-BP1. Reduced phosphorylation of these outputs after fisetin treatment has been observed in different cellular studies; however, such an immunoblot by itself does not prove that fisetin occupies the ATP pocket of PI3Kα or mTOR, nor does it establish the same mechanism across all senescent cell types. A direct binding claim requires biochemical affinity, competition with ATP, structural confirmation, and comparable cellular concentrations. Docking models of individual PI3Kα and mTOR residues and the IC₅₀ values reported for them establish neither direct binding nor a mechanism of senolysis.

If AKT activity falls, phosphorylated BAD may be released from 14-3-3 proteins and increase the availability of proapoptotic factors at the mitochondrial outer membrane. The outcome nevertheless depends on the balance among BIM, PUMA, NOXA, antiapoptotic BCL-xL, BCL-2, MCL-1, and effector BAX/BAK; no single Ser136 residue predicts the outcome on its own. Classical senolytics exploit cell-specific dependence on these networks. The proposed direct BH3-mimetic occupation of the BCL-xL groove by fisetin, with contacts at Tyr101 and Arg139, is based on a structural interpretation for which the available primary evidence provides no direct confirmation. It is especially important to distinguish docking, altered BCL2L1 expression, and displacement of BIM from a purified complex: these are different evidence levels. Fisetin should not be called pharmacologically equivalent to navitoclax or venetoclax without comparative biochemistry.

The execution phase of the mitochondrial pathway is well described independently of fisetin when the proapoptotic signal is sufficient. Activating BH3-only proteins promote conformational change and oligomerization of BAX/BAK in the mitochondrial outer membrane. Mitochondrial outer-membrane permeabilization (MOMP) allows cytochrome c to leave the intermembrane space for the cytosol; it is not identical to an immediate, complete loss of potential across the inner membrane. Cytosolic cytochrome c and ATP/dATP support assembly of the APAF1 complex, which recruits procaspase-9; dimerization and proteolysis activate caspase-9, followed by executioner caspases-3/7. Smac/DIABLO weakens caspase inhibition by IAP proteins. Cleavage of ICAD releases CAD for DNA fragmentation, while activation of XKR8 and reduced retention of phosphatidylserine on the inner membrane leaflet promote a signal for phagocytes. Apoptosis may end in efferocytosis without secondary necrosis, but this depends on tissue state and the phagocytic system; completely “immunologically silent” death is not guaranteed.

Fisetin selectivity is not a binary molecular property. In Yousefzadeh et al. (2018), viability of SA-β-gal-positive cells in certain cultures fell more than overall viability, while responses among adipose-tissue cell subtypes differed. Other cells may differ in exposure, injury severity, polyphenol metabolism, and SCAP dependence. This is why findings in Ercc1-deficient embryonic fibroblasts, etoposide-treated IMR90 cells, and mouse adipose tissue should be reported with the model specified. The concept of “apoptotic priming” helps formulate a hypothesis, but it does not establish a universal safe threshold of 5–15 μM for normal stem cells or human organs.

SASP can be attenuated without eliminating the cell

A senomorphic effect requires a different reading of the data. During canonical NF-κB activation, the IKKα/IKKβ/NEMO complex phosphorylates IκBα at Ser32/Ser36; SCF^β-TrCP then marks the inhibitor for proteasomal degradation, allowing RelA/p65–p50 to translocate to the nucleus. These steps support transcription of some SASP genes, but not the entire secretome or every type of senescence. If fisetin reduces IκBα phosphorylation, nuclear p65, or secreted cytokines in culture, that supports a change in signaling output in that model. The specific fisetin contacts with Cys99/Asp166 of IKKβ and an IC₅₀ of 8–12 μM stated in the original manuscript should not be accepted as a proven direct enzyme mechanism without separate analysis of purified kinase.

A second possible control layer is mRNA lifetime. The MKK3/6–p38 MAPK–MK2 axis affects phosphorylation of the RNA-binding protein TTP/ZFP36; phosphorylated TTP binds 14-3-3 and is less able to promote decay of AU-rich cytokine transcripts, whereas active TTP may recruit CCR4–NOT and RNA degradation. Ser52/Ser178 are often cited for mouse TTP; residue numbers should not be mechanically transferred to the human protein. This is a plausible architecture for SASP regulation, but the available primary evidence does not establish that fisetin binding to Met109 of p38α is the necessary cause of every observed decrease in IL-6. Measured mRNA half-lives depend on the model and are not universal constants.

SIRT1 is an NAD⁺-dependent deacetylase: during deacetylation it consumes NAD⁺, producing nicotinamide and O-acetyl-ADP-ribose. Deacetylation of RelA Lys310 may attenuate the NF-κB transcriptional response. AMPK responds to cellular energy status and, when activated at Thr172, changes the balance between anabolism and catabolism; its connection with NAMPT may affect NAD⁺ salvage. A change in SIRT1/AMPK activity after fisetin in a model cannot automatically be called proven direct allosteric binding of fisetin to SIRT1. Clinical reductions in hs-CRP, IL-6, or TNF-α after fisetin intake are likewise not confirmed by the available primary studies. For broader context on NAD⁺ metabolism and CD38, see the material on apigenin and CD38; it does not establish a therapeutic synergy between two supplements.

The same boundary applies to “restoration of stem-cell niches.” Chronic SASP can disrupt the environment of Pax7-positive muscle satellite cells, hematopoietic stem cells, and mesenchymal stromal cells. Reducing inflammatory secretion or removing senescent cells in an animal model may change niche conditions. But the specific claim that fisetin restores MyoD/myogenin, repopulates tissue with young cells, and restores clinical regeneration in humans requires direct analysis of the relevant model and outcomes. It cannot be inferred from a general understanding of SASP.

Organ phenotypes: outcomes from different interventions cannot be combined into a single fisetin effect

The study by Yousefzadeh et al. (2018) used several regimens that should be distinguished. In Ercc1-deficient progeroid mice, cellular and tissue markers were assessed after fisetin was incorporated into the diet; in part of the experiment, 500 ppm feed was given intermittently. In 22–24-month-old mice, brief administration of 100 mg/kg for five days reduced the proportion of SA-β-gal-positive cells in inguinal adipose tissue, and CyTOF showed heterogeneous responses across cell populations. Human adipose-tissue samples were treated with 20 μM ex vivo: this demonstrates a biological response in human tissue in a dish, not an achievable concentration or benefit after oral intake.

A separate cohort of ordinary mice began receiving 500 ppm fisetin-containing feed at 85 weeks of age; this was a late-life, long-term feeding study, not evidence for a clinical two-day “pulse.” In a small group of approximately 8–9 animals per arm, the authors reported improvements in median and maximum survival, geropathology scores, and some SASP markers. The magnitude of the survival effect cannot be transferred across sexes and experiments without specifying the calculation method and uncertainty of the estimate. Small groups, strain, and long-term exposure limit translation to humans. A decrease in Cdkn2a/Cdkn1a in tissue may also reflect changes in cell composition and expression rather than an exact count of senescent cells eliminated.

For bone and joints, mechanisms were studied under different experimental interventions. Jeon et al. (2017) showed that local clearance of senescent cells after joint injury altered osteoarthritis features and the regenerative microenvironment; this was not a clinical trial of oral fisetin. Farr et al. (2017) studied targeted action on senescent cells in a mouse model of age-related bone loss; the result does not verify a specific claim about fisetin, RANKL, and restoration of bone mineral density. The biochemical sequence MMP-13/ADAMTS-5 → degradation of collagen II and aggrecan is worth describing as a cartilage-catabolism model, but safranin O staining or an increase in BV/TV cannot be attributed to fisetin without the corresponding experiment. The physiological target and a demonstrated drug effect therefore remain distinct.

Neuronal evidence is also heterogeneous. In studies by Currais et al., fisetin altered inflammatory and behavioral measures in transgenic and rapidly aging mouse models; changes in p25/CDK5, tau phosphorylation, and behavior in these models do not establish senolytic clearance of glia. The study by Bussian et al. (2018) compellingly linked genetic removal of senescent glia with attenuation of tau-dependent pathology in mice; it did not study fisetin treatment. The molecule’s lipophilicity alone does not prove free exposure in the hippocampus, a specific human concentration, or restoration of long-term potentiation. “Complete prevention of dementia” is not an outcome in the cited evidence base.

In the vascular system, senescent endothelial and smooth-muscle cells may alter NO signaling, inflammatory secretion, matrix, and osteogenic programs such as RUNX2. Roos et al. (2016) showed attenuation of vascular dysfunction in aged or obese mice after long-term senolytic intervention, but this cannot be rewritten as direct evidence that fisetin reduces pulse-wave velocity, reverses calcification, and stabilizes plaques. Similarly, improvements in HOMA-IR, steatosis, and leptin have not been established in a specifically identified primary fisetin experiment with measured outcomes. Each organ requires its own evidence chain, including intervention, model, and control, rather than reference to the general possibility of senolysis.

Clinical translation: a registered trial is not a result

AFFIRM and other registered fisetin studies show that tolerability, inflammatory measures, and physical function have been posed as questions for human research. Published primary results confirming reductions in cytokines and improved mobility in these programs are not available in the cited evidence. NCT03430037 and NCT03675724 should be treated as protocol identifiers, not confirmation of efficacy. A regimen reported in a particular study protocol and population should not be recast as a general treatment standard or a self-prescribing protocol. Primary functional and biomarker endpoints, participant counts, withdrawals, and adverse events should be assessed from the specific study results if published before conclusions are drawn.

COVID-FISETIN and the experiment by Camell et al. (2021) illustrate the same boundary. In aged mice, senolytic interventions altered outcomes after coronavirus infection and supported the hypothesis that senescent cells contribute to the inflammatory response. They do not demonstrate prevention of pulmonary fibrosis, reductions in D-dimer, or lower mortality in people after fisetin. Evidence from studies of diabetic kidney disease and peripheral artery disease likewise should not be described as proven clearance of podocytes, improved eGFR, or longer walking distance: the paper by Hickson et al. (2019) concerns dasatinib plus quercetin in diabetic kidney disease, not fisetin. The open-label pilot by Justice et al. (2019) in idiopathic pulmonary fibrosis also studied D+Q. The molecules, design, and population are different, which is fundamental to interpretation.

Assessing senescence in studies requires a combination of markers and tissue context. p16INK4a/CDKN2A in peripheral-blood T lymphocytes reflects the state of that population and may correlate with age, but it is not a validated quantitative counter of all senescent cells in heart, joint, and brain. SA-β-gal at approximately pH 6 and the fluorogenic substrate C₁₂FDG are used in specific cell protocols; lysosomal activity and staining conditions limit specificity. Plasma IL-6, CCL2, MMP-3, activin A, and PAI-1 are not unique to SASP. cfDNA, miR-34a, and miR-146a may have many sources. There is no generally accepted clinical threshold by which such a panel proves senolytic action of fisetin in a particular patient.

Exposure, conjugation, and the problem of comparing concentrations

Fisetin is poorly soluble in water; dissolution, food, vehicle, and intestinal transport affect aglycone availability. In enterocytes and the liver, the flavonol undergoes phase II reactions: UDP-glucuronosyltransferases transfer a glucuronyl group from UDP-glucuronic acid to hydroxyls, sulfotransferases use PAPS as a sulfate donor, and catechol-O-methyltransferase transfers a methyl group from SAM to the catechol moiety, producing geraldol among other metabolites. These reactions change polarity, distribution, and the form measured analytically. Specific UGT isoforms, positions of individual conjugates, and the affinity of each metabolite for BCL-xL require analytical or biochemical confirmation. An “absolute bioavailability of 5–10%” likewise cannot be asserted for humans without an intravenous comparator and an appropriate design.

The key translational comparison is the free concentration of unchanged aglycone in the relevant tissue versus the concentration at which an effect was observed in the cellular system. Total plasma AUC, including conjugates or measured after sample hydrolysis, does not answer this question without analyte identification. With low solubility, first-pass metabolism, protein binding, and tissue heterogeneity, 5–20 μM in culture medium cannot be equated to a capsule dose. Uncertainty also applies to the brain: systemic exposure does not guarantee a concentration beyond the blood–brain barrier.

In a small randomized, double-blind, crossover pharmacokinetic study, Krishnakumar et al. (2022) enrolled 15 healthy volunteers. They compared ordinary fisetin with FF-20, a fenugreek galactomannan-based formulation, and reported substantially greater plasma exposure to fisetin and the metabolite geraldol with the tested formulation. The primary publication reported an AUC ratio of approximately 26.9 under the conditions of that study; it must be interpreted with the analyte matrix, dose, and method specified, not as “27-fold stronger senolysis.” The authors are affiliated with the ingredient developer and a commercial partner; this does not invalidate the pharmacokinetic observation but increases the value of independent replication. The study did not measure clearance of senescent cells, function, or lifespan. Increased AUC does not justify deriving a dose of 500–1000 mg or arithmetically dividing a presumed “senolytic course.”

Liposomes, phospholipid complexes, and self-emulsifying systems remain formulation approaches. Their ability to improve dissolution does not prove lymphatic absorption, liver bypass, or clinical efficacy for all products; composition, particle size, stability, and measurement of free aglycone are needed for comparison. Comparing formulations only by nominal milligrams, without bioavailability and composition, is methodologically unsound.

Intermittent dosing is a research hypothesis, not a routine protocol

The hit-and-run idea arose from the hypothesis that after a brief elimination of some senescent cells, the biological effect may persist longer than the senolytic itself. Model experiments support this for some interventions; for fisetin, the specific regimens ranged from five-day gavage to periods of feeding and late-life long-term dietary administration. Thus, neither “daily intake is categorically prohibited” nor “two days each month are enough” is a proven general law. Moreover, transient senescence may participate in repair, and progenitor cells require PI3K–AKT signaling for survival and proliferation. These considerations justify specific studies of safety, timing, and context, but do not permit a universal “niche recovery interval.”

Regimens in individual research protocols are not approved treatments for aging, osteoarthritis, or COVID-19, and conversion to another body weight establishes neither benefit nor safety. Plasma AUC differences between FF-20 and ordinary powder cannot be converted into an “equivalent” course: the relationship between AUC and senolytic effect in humans has not been measured, and the fraction of free active aglycone in an organ is unknown. For a specialist, appropriate wording is to specify the protocol, formulation, population, inclusion criteria, primary outcome, and adverse events of a particular trial; until those data are available, the intervention should not be presented as standard care.

These data do not support ranking the “best senolytic,” since the molecules, doses, tissues, and endpoints have not been compared in an adequate head-to-head clinical trial. Cell-screen comparisons of fisetin with quercetin, curcumin, and EGCG also do not support ready-made fisetin “stacks.” The article on polyphenols and their safety is useful for distinguishing dietary exposure from concentrated formulations, but does not justify combining them for senolysis. Spermidine and apigenin affect autophagy/EP300 and NAD⁺/CD38 metabolism, respectively, in other experimental contexts. There are no clinical grounds for recommending combined or sequential regimens with fisetin.

Safety cannot be inferred from how common fisetin is in foods. Comparing food amounts with concentrated experimental regimens means comparing different exposures and possible interactions. Findings on inhibition of CYP1A2, CYP2C9, CYP3A4 and effects on platelet aggregation should be treated as signals for pharmacological evaluation, especially with narrow-therapeutic-index medicines and hemostatic disorders; cellular IC₅₀ values do not predict the magnitude of a drug interaction in humans. A specific discontinuation interval before surgery and a universal list of absolute contraindications for high-dose fisetin have not been established in clinical studies. Pregnancy, lactation, active treatment of serious disease, bleeding, and imminent surgery require individualized medical assessment because adequate data on high-dose fisetin are lacking. The same applies to liver and kidney function. Complete blood count, ALT/AST, creatinine, IL-6, or hs-CRP may be tracked under a specific protocol, but the combination of measures and their repeat frequency do not constitute a validated algorithm for assessing senolysis.

The clinical conclusion remains limited to experimental evidence. Fisetin is an important tool for testing whether the fate of selected senescent populations and tissue SASP can be altered. In humans, controlled exposure to active forms, representative tissue markers, benefit on a prespecified outcome, and an acceptable adverse-event profile must be demonstrated. Until these links are established, signs of asthenia, osteoarthritis, or elevated hs-CRP alone are not indications for a high-dose “rejuvenation” course. For researchers, the central next step is trials with transparent selection of population and formulation, separation of senolytic clearance from senomorphic SASP suppression, and publication of negative results alongside positive ones.

How this review was prepared

This narrative review of fisetin, cellular senescence, and SASP draws on a selection of 19 sources. In the 4 October 2026 editorial version, material discrepancies among cellular, animal, and pharmacokinetic evidence were checked directly. Cases in which findings from other approaches to senescent-cell removal had been attributed to fisetin were identified. No exhaustive search, meta-analysis, formal risk-of-bias assessment, or independent peer review was conducted; full texts that were not accessed are not treated as independently verified.

Cell culture, human-tissue explants, continuous feeding in mice, and a brief pharmacokinetic study in volunteers are not combined into evidence for clinical senolytic therapy. Trial registration identifies a research question, not a confirmed effect. Regimens used in models are not treated as universal recommendations for people.