Fisetin as a natural senolytic: mechanism, bioaccessibility, and comparison with D+Q — what the evidence demonstrates, what is not ranked, and which hypotheses can be falsified

1. Executive scientific abstract

The available evidence does not support the simple chain “fisetin is a natural senolytic, reaches the ovary as free aglycone, outranks dasatinib + quercetin (D+Q), and extends women's health or lifespan.” Each link names a different object. None closes the complete chain.

In the reproductive window without a gonadotoxic, García et al. 2024 administered fisetin 100 mg/kg oral, three consecutive days per month, in C57BL/6 females from 6 to 10–11 months. Follicles, pregnancy, and litter did not separate from vehicle (P = 0.51–0.91, 0.37 and 0.20). Lipofuscin (P = 0.008) and CD68 macrophages (P = 0.002) fell; Il1b rose. That is neither senolysis nor functional restoration. It is not equivalence either: p>0.05 does not write a region of non-separation.

In young females with doxorubicin, Gao et al. 2023 equated fisetin and D+Q on the broken link: p16, p21, SA-β-gal, Cdkn2a, Cdkn1a, Il6 and Mcp1 fall; γH2AX, follicles, fibrosis, hormones, cyclicity and fertility do not return; TUNEL, Bax and caspase 3 do not separate. Bcl2 mRNA rises with fisetin, not with D+Q. D+Q alone — not fisetin alone — reduces primordial number and proportion. The body writes it: fisetin alone has no evident effect versus vehicle, except that primordial drop of the pair.

The only ovary that moves structure under fisetin is not senolysis. Dong, Zhu, Li, Chang, Yan, Zhang and Niu 2026 (PMID 42193224; not Ding/Zeng) injure 6-week BALB/c with D-galactose and give 10–30 mg/kg daily. Granulosa TUNEL falls and BrdU rises. Compound C occludes EdU in vitro at 5 µM. There is no fisetin arm without injury, nor quercetin, nor D+Q. The E2 ELISA (range 0.31–20 ng/mL) remains under analytical quarantine. The −44.8 kcal/mol docking is rejected. The paper's own introduction concedes that weeks of D-gal do not recapitulate decades.

The comparison with D+Q is not a ranking of two ovarian senolytics. It is flavonol on PI3K–AMPK–mTOR, an elevated apoptotic threshold where the ovary was measured, versus extra damage from the pair (primordials in Gao; adipose and expenditure in Fang 2023, ovary not measured). Fang splits: D+Q harms the female; monthly 100 mg/kg fisetin does not copy weight. That neither clears the ovary nor identifies senolysis.

Bioaccessibility is not a single number either. Krishnakumar et al. 2022 measured free aglycone — without hydrolysis — in 15 volunteers (12 men, 3 women). After 1000 mg unformulated, parent Cmax 9.97 ± 3.97 ng/ml (~0.035 µM), geraldol 24.0 ± 7.69 ng/ml (~0.080 µM), AUC ratio 1.62; parent is quantifiable ~2 h (t½ 1.14 ± 0.09 h, tmax 0.88 ± 0.18 h). FF-20 lifts both Cmax values to ~0.80–0.83 µM. Nobody measured free parent or geraldol in ovary. Shia 2009 shows dominant conjugates after oral dosing in the male rat. The 5 µM bath with which Yousefzadeh 2018 called fisetin senolytic is already three fates: Ercc1 MEFs without death, HUVEC with caspase, D-gal granulosa with EdU↑/TUNEL↓.

Jo et al. 2016 give absolute parent F 7.8% at 100 mg/kg oral and 31.7% at 200 mg/kg, in male ICR mice, with geraldol Cmax/AUC above parent. That F% is not García's ovarian occupancy. MeSH indexes male ICR and protein precipitation. The Cmax table remains paywalled; the 37.032 µM from a 2026 preprint that cites Jo remains barred. Touil 2011 is female C57BL/6, but 223 mg/kg i.p., parent Cmax ~8.7 µM: another route, another exposure.

The ITP (Harrison et al. 2024) starts fisetin at 20 months, 600 ppm, not at 12. Female medians: 871 and 883 versus 876 days. Plasma not reliable. Kidney, brain and liver p16 do not fall. Lifespan transport from Yousefzadeh dies there. In women there is no AMH, cycle, ovary or STRAW+10 under fisetin. Farr, Hickson and Justice are D+Q or another organ. TROFFi and STOP-Sepsis are extraovarian.

Scientific delta of this report (L10-3-REPORT-EN-C1): the English twin freezes the same bilingual factual base. Occupancy cannot share García's phenotype ovary: Krishnakumar leaves unformulated parent quantifiable only ~2 h, so an LC–MS days after the last three-day pulse cannot distinguish “never arrived” from “arrived and washed out.” The first cut is DETECT at T1 = 1.0 h after one PHOSAL pulse, versus the assay's own ovarian LLOQ — not versus 5 µM — with a daily 10 mg/kg × 30-day arm at the same clock. D0-JO-CMAX remains specified and unread. The three project hypotheses are L10-3-AR-H1 v1, L10-3-AR-H2 v1 and L10-3-AR-HT1 v1. L10-3-HG-H1 survives only as a naming restriction on already-measured ovaries: the sign (TUNEL↓, BrdU↑, Bcl2↑, function quiet) is the opposite of senolysis and does not rank against D+Q.

The operational conclusion is negative and actionable. Fisetin is not mature as an ovarian senolytic nor as a female-longevity strategy. The smallest next step is not another D-gal, nor a senolysis factorial, nor an AMH trial. It is not to use the follicle/lipofuscin ovary as a pharmacokinetic time, and not to copy 7.8% as García occupancy.

2. Scientific question and relevance

The question is which mechanisms connect fisetin — mechanism, bioaccessibility, and comparison with D+Q — to women's health and longevity, and which new hypotheses can be falsified. Answering it requires separating objects that the title fuses:

1. Is the causal agent fisetin, geraldol, a conjugate, quercetin, or D+Q?
2. Does the affected cell meet a durable arrest distinguishable from luteinization, atresia, reversible quiescence, and proliferation (Q0-STATE)?
3. Is there selective apoptosis, or do markers fall while TUNEL is null or reversed?
4. Does an independent ovarian function follow the cellular change?
5. Is the exposure food matrix, unformulated bolus, formulation, monthly gavage, a three-day pulse, or daily — and which free species occupies the compartment (Q0-EXPOSURE)?
6. Do D+Q and fisetin break the same ovarian link, or diverge by extra damage from the pair?
7. Does that trajectory mediate an organ event in women, with competing risks?

The relevance is high because “natural senolytic” is a commercially useful and scientifically lazy epithet. The ovary is especially vulnerable to identity errors: p16, p21, SA-β-gal, lipofuscin, CD68, IL-6, AMH, follicles and litter are not interchangeable. Neither are acute oxidative injury in a 6-week BALB/c, doxorubicin in an 8-week C57BL/6J, and reserve decline toward 6–9 months. A longer reproductive life can associate with skeletal or metabolic benefit and, at the same time, with greater exposure to hormone-dependent risks (L10-1-K10). The correct unit is an organ event with mediation and competing risks, not a reproductive surrogate.

3. Scope, population, and life stage

The primary biological estimand is the adult female in natural reproductive aging, without chemotherapy: the object of García experiment 2. For human translation, the relevant population is women classified by STRAW+10: late reproductive (−3b/−3a), transition (−2/−1) and early postmenopause (+1c), analyzed separately. Chronological age alone does not substitute for stage.

The doxorubicin (Gao), D-galactose (Dong/Zhu/Niu) and VCD (Ávila, abstract) models are retained as injury contrasts. They inform cytoprotection, marker–structure dissociation, or a null of induced marker; they are not averaged with the natural transition. PCOS, ischemia–reperfusion, cyclophosphamide POF at 100 ng/kg (Lin 2020), porcine granulosa, ovarian cancer and bird remain outside the estimand. Trials in adipose, lung, bone, 6MWD or SOFA are extraovarian evidence.

Prescription, dose recommendation, senolytic ranking, combination with NAD+ precursors, product proposals, user data, identification of García's Cdkn1/Serpin1/Ccxl5 as Cdkn1a/SERPINE1/Cxcl5, and any inference from fertility or menopause to longevity without an organ event are excluded. SERPINE1 remains barred as an analyte. CD38 remains parked (L9-4-K6). COMT is reopened only as the pharmacokinetic enzyme of the circulating species, not as ovarian occlusion (H65).

4. Background knowledge and mechanistic map

4.1 Operational definitions

• Fisetin: 3,3′,4′,7-tetrahydroxyflavone. Chemical identity, not a senolytic finding. It lacks quercetin's 5-OH; that difference is a chemical assumption under test, not ovarian evidence.
• Geraldol: 3′-O-methyl fisetin. A distinct species. In unformulated human plasma it exceeds parent. Touil makes it more cytotoxic than parent in tumor and endothelium, at extraovarian micromolar concentrations.
• D+Q: a pharmacological combination. It is not a single mechanism until neither component is shown to suffice. In L10-2, H0 remains as ovarian senolysis.
• Senolysis: selective death of a cell that already met durable arrest plus an orthogonal readout, with survival of non-senescent sisters. Lowering p16, p21, SA-β-gal or lipofuscin is not senolysis. “Natural” is not a mechanism.
• Bioaccessibility: the fraction released from the matrix. It is not Cmax. It is not occupancy.
• Free species: unconjugated aglycone in the compartment where the effect is claimed. Total, glucuronide, sulfate and hydrolyzed do not substitute (L9-4-K7 transports the rule, not 2.48 µM quercetin).
• DETECT versus OCCUPY: exceeding the ovarian LLOQ of free parent or geraldol is not occupying AMPK, mTOR or SCAP.
• Ovarian function: an endpoint independent of the marker. The marker cannot be the success.
• Healthspan/longevity: years free of an organ event or survival, with mediation and competing risks. Not AMH, follicles, cycle, pregnancy, litter, or age at menopause.

4.2 Favored chain that the literature would have to demonstrate

Matrix or bolus called fisetin
  → nominated free species (parent, geraldol or conjugate) in ovary
  → occupancy of a target written before the phenotype
  → selective death of a qualified durable arrest (Q0-STATE)
  → change in an independent ovarian function
  → endocrine or systemic trajectory
  → organ event in women
  → net effect under competing risks

The current literature breaks before the second link. There is no in-vivo ovarian measurement of free aglycone. Where the ovary was measured, death is null or reversed. The bridge to an organ was not trialed.

4.3 Competing mechanisms

A — marker/infiltrate drop without senolysis (H2). García: lipofuscin and CD68 fall, function quiet, Il1b rises. Gao: marker down, TUNEL null. Yousefzadeh: MEFs at 5 µM without death.

B — cytoprotection / cycle re-entry (H4 / L10-3-EV-R1). Dong: TUNEL↓ BrdU↑ via AMPK/mitophagy. Khan 2012: PI3K/Akt/mTOR↓ and p-AMPKα↑ at 5–20 µM in NSCLC. Li 2021: quercetin alone restrains PI3K–AKT–mTOR/FOXO3A in cyclophosphamide.

C — species / schedule failure (split H5). Oral is not the 5 µM bath. Geraldol and conjugates can dominate. A García null can be washout, not biology. A Dong rescue can be another exposure (daily, another vehicle, D-gal).

D — shared failure under doxorubicin, extra damage from the pair (H3/H6). Gao equates the marker and splits primordials. Fang splits female adipose. Neither is demonstrated ovarian senolysis.

E — selective fisetin senolysis (scope H1). No primary of the estimand supports it. HUVEC is not ovary.

5. Evidence method

The complete artifacts of scoping, map, verification, synthesis, computational decision, hypotheses, adversarial review and experimental design were integrated. Claims were kept by layer: human (H), animal (A), in vitro/ex vivo (V), extraovarian (X), computational (C) and inferred (I).

The ovarian search (PubMed, 2026-09-27) identified 46 records for fisetin AND (ovary OR ovarian OR follicle OR oocyte OR granulosa). Primaries were audited in PMC BioC or HTML when they responded. Gao 2023, Dong/Zhu/Niu 2026, Krishnakumar 2022, Harrison/ITP 2024, Yousefzadeh 2018 and Lin 2020 were reread in the body. García and Fang are inherited verified from L10-1/L10-2. Ávila 2024 remains at abstract; n and dose are not invented. Jo 2016: abstract and MeSH; Cmax table not opened on 2026-09-28 (Europe PMC isOpenAccess=N). A preprint that cites 37.032 µM does not enter.

p>0.05 was never interpreted as equivalence. Sections, fields, wells and mating rounds were not promoted to female count. 2.48 µM quercetin, 5 µM well, 9.97 ng/ml, Lin's 100 ng/kg, −44.8 kcal/mol, 8.7 µM i.p., 7.8% F and 0.81 mg/kg HED are not copied as ovarian thresholds.

No computation was executed (decision: skip). NCBI GEO on 2026-09-27 returned zero ovarian fisetin series. GSE202601 does not choose among H2, H4 and H5. BioNeMo does not reorder.

In this turn the HTML of PMC13203114 was reconfirmed: Dong/Zhu/Niu authorship, E2 kit 0.31–20 ng/mL, BrdU “three mice per group,” and the introduction's caution of weeks versus decades. Krishnakumar PMC9574875 remains the human free-aglycone PK.

6. Evidence map

7. Contradictory evidence and null results

7.1 The marker does not predict structure or function

García is the null closest to the transition: lipofuscin and CD68 fall; follicles, pregnancy and litter do not separate; Il1b rises, incompatible with “clean SASP.” Gao adds the strongest dissociation under injury: the marker falls and the organ does not return, with TUNEL null. Ávila, at abstract layer, states that neither D+Q nor fisetin lower VCD-induced ovarian senescence. Those observations are not equivalences, but they do prevent treating any p16/p21/lipofuscin drop as ovarian restoration.

7.2 Death changes sign across objects

• Yousefzadeh MEFs: SA-β-gal↓, without death.
• Yousefzadeh HUVEC: caspase 3/7↑ — endothelium, not ovary.
• Gao: TUNEL “no obvious change”; Bcl2↑ with fisetin.
• Dong: TUNEL↓ and BrdU↑ in granulosa that was already dying under D-gal (L10-3-AR-C2). Restoring nucleoside is Saul's anti-senotype (L10-1-K8), not a senolytic that “worked.”
• L10-2 already had TUNEL down (Du, cisplatin+D+Q) and up (Su). They are not averaged. None pass Q0-STATE.

7.3 Three ovaries are not one phenotype

García, Gao and Dong differ in strain, age, injury, schedule, vehicle and operational n. Fusing them into an “elevated apoptotic threshold” lets the hypothesis live if any of the three signs appears (L10-3-AR-C1). The laboratory already withdrew that disjunction in Fang. Here it is the same trick.

7.4 Jo is not García, and F% is not the bath

Sharing 100 mg/kg oral does not equalize exposure (L10-3-AR-C3). The 7.8 → 31.7 nonlinearity on doubling the dose kills the pairing inside Jo. Touil, the only female C57BL/6 PK in this chain, is micromolar i.p. Daily Dong 10–30 mg/kg cannot share a single species-failure with monthly García (L10-3-AR-C4).

7.5 The life null is not an ovarian-senolysis null — and not the test either

The ITP does not extend pooled median. It also does not prove that García's oral was empty: it is 600 ppm in UM-HET3 diet from 20 months, with plasma not reliable. A lifespan null without PK does not kill marker H2 nor save H1.

7.6 The human gap is more precise than “women are missing”

Zero ovary, AMH, cycle or STRAW+10 under fisetin. Three women in a single-dose PK. Farr randomized 60 postmenopausal women to D+Q, with a null skeletal primary. Expanding TROFFi does not fill the ovary. El Sayed 2026 is not cited as translation evidence: its own abstract concedes that reproductive clinical evidence is absent.

8. Multiscale mechanistic synthesis

8.1 Protein

Khan places fisetin 5–20 µM on PI3K/Akt/mTOR↓ and p-AMPKα↑, with mTOR docking at −7 to −8 kcal/mol — the credible range, not Dong's −44.8. Li places quercetin alone on the same axis in cyclophosphamide. Dong occludes AMPK in vitro when granulosa stops dying and re-enters cycle. That names flavonol class, not senotype. Bcl2 rises where H1 would require the apoptotic threshold to fall. p21 does not rescue: in ovary it marks luteal exit and reversible quiescence by AMH (L10-1-K5). CDKN2A does not choose p16INK4a versus p14ARF (L10-1-K4).

8.2 Cell

5 µM bath (Yousefzadeh / Dong in vitro)
  ├─ Ercc1 MEFs ── SA-β-gal↓, without death
  ├─ HUVEC     ── caspase 3/7↑   (endothelium; Zhu flavonol arm)
  └─ D-gal granulosa ── EdU↑ TUNEL↓ via AMPK (Compound C occludes)
Oral called fisetin
  ├─ García 3 d/month, 100 mg/kg, PHOSAL, adult C57BL/6
  ├─ Gao every 2 d × 3 wk, 100 mg/kg, PEG400, young + Dox
  ├─ Dong daily 10–30 mg/kg, another vehicle, BALB/c + D-gal
  ├─ Jo 100 mg/kg, male ICR, F 7.8% ── not García
  └─ Touil 223 mg/kg i.p., female C57BL/6, ~8.7 µM ── not the gavage
Human UF plasma ── parent ~0.035 µM, geraldol ~0.080 µM, t½ ~1.1 h
Ovarian Q0-STATE lineage ── does not exist (L10-1)

Fisetin is not used to decide which nuclei “were” senescent. Porcine granulosa lowers P4 at 1–50 µM: adverse sign retained.

8.3 Tissue

Four ovaries, the same compound name, signs that cannot be averaged. The toxicant and the calendar command. L7-4 (loss of luteal exposure) is not filled: Dong does not move P4; porcine granulosa lowers it.

8.4 System

Three exposures the title calls “natural”: food (not quantified as tissue dose), unformulated bolus sub-µM geraldol-dominant, and formulation ~0.83 µM. Conjugates invisible in Krishnakumar. Fang: D+Q harms the female; fisetin does not copy; ovary absent. The “better profile” can be absence of dasatinib, not clearance of an ovarian senotype.

8.5 Woman and organ event

The chain cuts at protein (no ovarian occupancy), at the cell (no Q0-STATE), at the ITP (pooled null, plasma not reliable) and at the human ovarian absence. Later menopause splits bone and cancer without uniform lifespan. Even if a reserve rescue arrived, it would not be longevity.

9. Computational layer

Decision: skip (L10-3-CD-C1). Counting BCL2, PRKAA1/PRKAA2, PIK3CD and FOXO3 in GSE202601 is computable and does not choose among marker drop, flavonol class and aglycone failure: those symbols are constitutive follicular physiology, Dong already named granulosa when occluding AMPK, and an snRNA zero is not absence (L10-1-K4). GEO returned zero ovarian fisetin series. Dong's n=3 RNA-seq is not indexed. Extraovarian series (Alport kidney, aorta, TB lung, eSC, pancreas) do not bring the fisetin / quercetin / D+Q factorial. GSE264327 is nano D+Q without fisetin. ChEMBL does not write the ovarian threshold. BioNeMo does not measure occupancy at 0.035 µM or at 5 µM. The L10-1 SenMayo module is not reopened to authorize a senolytic.

10. Primary hypothesis

L10-3-AR-H1 v1 — García's bolus does not inherit Jo PK, and F% is not occupancy

Role: project primary after adversarial review. It replaces the milligram identity of L10-3-HG-H2 v1. It does not claim that García's ovary was occupied. It claims that nobody measured that occupancy in that female, and that Jo does not substitute for it.

Falsifiable statement. In adult C57BL/6 females, under the frozen oral schedule of García experiment 2 — 100 mg/kg, three consecutive days, once a month, PHOSAL/PEG/ethanol vehicle — Cmax or AUC of free fisetin aglycone and free geraldol in the ovary, measured in a satellite at T1 = 1.0 h after the last gavage of one pulse and against thresholds written before the phenotype, are not inferred from the 7.8% absolute F nor from Jo 2016's geraldol>parent ratio in male ICR. If, on opening Jo's table, parent Cmax at 100 mg/kg oral in that male sits below the written well threshold, that is male ICR PK, not García. If it sits above that threshold, H2 dies as “murine oral does not reach 5 µM” and still does not occupy García's ovary until LC–MS of that female.

Proposed mechanism. Strain, sex, vehicle, route, hydrolysis and duty cycle alter the species that arrives. An F% is dose-corrected oral/i.v. AUC. It is not ovarian concentration. Krishnakumar leaves unformulated parent below the analytical floor at two hours: García's follicle ovary, harvested days or a month after the last pulse, does not vote occupancy.

Evidence for. Jo MeSH and abstract (male ICR, nonlinear F, precipitation). Touil (female C57BL/6, i.p., ~8.7 µM). García: ovary measured, free species not. Krishnakumar: another species, another dose, 3 women, t½ 1.14 h.

Adversarial evidence. Jo's table was not read. A micromolar Cmax at 100 mg/kg in ICR would weaken “poor parent” and would not touch strain/sex. gmGUS (L9-4-Q9) could occupy the follicle with poor plasma: it requires temporal aglycone and loss on blocking the enzyme; it is not granted. Daily Dong could occupy at lower milligrams: that is AR-H2, not a Jo transport.

Predictions.

1. Extracting Jo's table, without a new animal, gives parent and geraldol Cmax/AUC at 100 and 200 mg/kg, reported as male ICR.
2. Ovarian LC–MS at T1 = 1.0 h on one PHOSAL pulse is the only DETECT close for the García object.
3. Touil is not used as García Cmax.
4. FF-20 (~0.83 µM human plasma) is not averaged with unformulated oral nor with Jo.
5. None of this measures senolysis. Q0-STATE remains prior for anyone who wants the name.

Kill criteria. García is declared to inherit 7.8% F or Jo's geraldol ratio: the call is invalid. Ovarian DETECT on García's schedule, two replicates, threshold written before: free parent or geraldol exceeds LLOQ; then this primary dies as “unassignable occupancy” and the phenotype moves to HG-H1(i) or to SCOPE-H1 according to the direction of death. 37.032 µM, 5 µM or 9.97 ng/ml is used as ovarian threshold: the call is invalid.

Status: proposed.
Maturity: literature H1 for Jo/García non-identity. H0 for ovarian occupancy.
Confidence: 0.84 that Jo is not García PK; 0.31 that García's ovary, once measured at T1, sits below LLOQ. The experiment was not run; the figures do not change.

11. Competing hypothesis

L10-3-AR-H2 v1 — the monthly null and the daily rescue are not one species failure

Role: competing. Contrary to the conjunction of HG-H1 and contrary to reading Dong as proof that “oral arrives” without naming milligrams, vehicle or strain.

Falsifiable statement. In females, at a nominated free species (parent, geraldol, conjugate pool) and LLOQ written before, García's schedule (100 mg/kg, three consecutive days per month, PHOSAL, adult C57BL/6, no gonadotoxic) and a daily 10 mg/kg × 30-day schedule in the same PHOSAL, same colony, same T1 = 1.0 h, do not coincide in ovarian DETECT, or coincide in DETECT and do not coincide in the TUNEL/nucleoside sign when phenotype is later opened. D-gal, if it enters, is a separate factorial (injury × schedule), not the natural estimand. BALB/c is not substituted by C57BL/6 after seeing the result. Dong's vehicle is not mixed.

Proposed mechanism. Duty cycle and vehicle are part of exposure (L10-2-K8 applied to fisetin). A daily that DETECT-passes and a monthly that DETECT-fails convert García's null into relative underexposure, not into “antiapoptotic flavonol.” If both fail, HG-H2 keeps both schedules at T1 and Dong's rescue is not attributed to ovarian free aglycone in C57BL/6+PHOSAL.

Evidence for. The two schedules, the two vehicles and the two strains/ages, already verified. Dong without fisetin in uninjured; BrdU n=3. García function without separation on a much less dense schedule.

Adversarial evidence. Nobody has measured ovarian free species on either. Both can be empty or both full. Changing strain and injury together is a confounder. A nonspecific antioxidant in D-gal can move TUNEL/BrdU without occupied flavonol. 10 mg/kg in PHOSAL is not Dong's paper.

Kill criteria. Both schedules, same sex/strain/vehicle/clock, produce indistinguishable DETECT (both above or both below LLOQ) and — if phenotype is opened in uninjured — the same TUNEL/nucleoside sign. Then the schedule split dies. Adjudication with Jo, with F% or with 5 µM: the call is invalid. García, Gao and Dong are averaged to manufacture a phenotype: the call is invalid.

Status: proposed.
Maturity: literature H1 that the schedules are not one. H0 for the DETECT split.
Confidence: 0.78 that García and Dong are not the same experiment; 0.27 that daily will DETECT-pass and monthly will not.

L10-3-HG-H1 v1 — retained only as a naming restriction

Role: hypothesis of already-measured ovaries, no longer the primary. Status: weakened.

What survives. In García, marker/infiltrate without demonstrated independent function. In Gao, Bcl2↑ with fisetin, late TUNEL null, PMF↓ only with D+Q alone. In Dong, death runs opposite to senolysis. SCOPE-H1 remains H0 on those objects.

What is withdrawn. The conjunction as a single apoptotic threshold. Cycle re-entry of a granulosa that would have passed Q0-STATE. The clause that fuses Fang and Gao. The 0.29 confidence in the complete natural factorial.

Maturity: H1 for “not senolysis” on the measured objects. H0 for the transition female with Q alone and free species.
Revised confidence: 0.58 that none of the three ovaries demonstrates senolysis; 0.18 that a single natural factorial will show quiet function, non-Q0 nucleoside and PMF/adipose loaded by the pair.

L10-3-HG-H2 v1 — retained as human plasma and ovarian gap

Role: hypothesis of another object. Status: weakened.

What survives. Krishnakumar UF, free geraldol > parent, AUC ratio 1.62, 12 men and 3 women, no ovary, no conjugates. Shia: conjugates after oral in the male rat. ITP: plasma not reliable.

What is withdrawn. Jo as García/Fang PK. “7.8% is already not the 5 µM bath.” The 0.22 confidence that García's null dissolves on occupying tissue.

Maturity: H1 for human UF plasma. H0 for ovary and for male ICR as female estimand.
Revised confidence: 0.70 that unformulated human oral is not the 5 µM parent bath; 0.16 that that fact explains García's null.

12. Translational hypothesis

L10-3-AR-HT1 v1 — the rule of not substituting the organ still holds; circulating is not the cell

Role: translational. Splits L10-3-HG-HT1 v1. It is not a biomarker. It is not a product. HUMAN_QA_REQUIRED.

Falsifiable statement. In adult women classified by STRAW+10, outside any user cohort, a parameter made of AMH, follicles, menstrual regularity, age at menopause, plasma Cmax/AUC of fisetin or geraldol, adipose p16/p21, 6MWD, SOFA, or Fang's systemic direction will not be accepted as mediator of a single-organ event. The only parameter that could reach that study is one that has passed Q0-STATE and Q0-EXPOSURE in ovary, and the AR-H1/AR-H2 cut in the female, and that then beats stage and a prespecified out-of-sample systemic panel, with competing risks. Until then mediation is not measured. rs4680, if recorded, is an exposure stratifier, not a senolysis predictor.

Predictions.

1. Breaking the rule invalidates the study. An event trial is not needed for the rule to hold.
2. An AMH RCT, or expanding TROFFi, neither kills nor saves the empirical form.
3. Krishnakumar is not reanalyzed as a STRAW interaction (female n = 3).
4. Later menopause still splits bone and cancer without uniform lifespan (L10-1-K10). It is not attributed to fisetin.
5. El Sayed 2026 is not cited as translation evidence.

Evidence for. Zero ovary, AMH, cycle or STRAW+10 under fisetin. PK n=3 women. ITP pooled null, plasma not reliable. Already consolidated rule: AMH, follicles and menopause are not the event.

Adversarial evidence. The empirical form is not measured. It can die if a parameter that does not yet exist beats the panel. An extraovarian event at ~0.80 µM FF-20 would kill ovarian mediation and leave pleiotropy. The 0.44 confidence of HG-HT1 is withdrawn: it described the gap, not a prediction result.

Kill criteria. It is marked H5 or biomarker: not permitted. Plasma Cmax, Fang, Hickson or 6MWD is used as if it were the ovarian effect in transitioning women. Q0-STATE, ovarian Q0-EXPOSURE and AR-H2 are passed, and the parameter predicts an organ better than stage and the panel, out of sample, with competing risks: then the empirical form dies. The rule of not using the surrogates does not die with it.

Status: proposed, parked for human execution.
Maturity: H0. HUMAN_QA_REQUIRED.
Confidence: 0.66 in the design rule; 0.10 in the complete empirical form.

13. Falsifiable predictions and kill criteria

SCOPE-H1 remains H0 because it has not been trialed with identity, exposure and factorial. It is not considered refuted for a cell that has not yet been identified; it does remain outside what existing studies demonstrate.

14. Discriminating experiment

This is research. It is not a regimen for a woman. These quantities are not copied as a recommendation. There is no gavage at this stage. The order is not reopened.

Step 1 — D0-JO-CMAX: zero animals

From Jo 2016, in this order, and in writing before converting to µM: numeric table of Cmax, tmax, AUC and t½ of parent and geraldol at 2 mg/kg i.v. and at 100/200 mg/kg oral; if there is no table, two independent readings of the same figure; sex, strain, vehicle, n and hydrolysis if the body writes them. The figure is reported as male ICR. It is not transported to García. 37.032 µM is not used. If it is not estimable after one retry, one continues without declaring that murine oral sits below 5 µM. 7.8%/31.7% F does not substitute for Cmax. The animal facility is not opened to “resolve” an unreadable table.

Step 2 — Q0-EXPOSURE-PULSE: satellite, not the coprimary ovary

C57BL/6 females, ~6 months, no gonadotoxic. One pulse: three consecutive days, 100 mg/kg oral, PHOSAL/PEG/ethanol. T1 = 1.0 h after the third gavage (rounding of Krishnakumar's human oral tmax; not Jo tmax; not Touil's 15 min). Arms: vehicle and fisetin. Unit: the female. Floor of 4 per arm (df = 6, t = 2.446911851144979). Three species: free parent, free geraldol (no hydrolysis) and conjugate pool (separate aliquot, enzymatic hydrolysis, reported as increment). LLOQ from a matrix-matched curve in vehicle ovary homogenate, written before. Plasma from the same T1 is archived; it does not substitute for ovary.

DETECT: female positive if ovarian concentration ≥ LLOQ of that free species. Arm positive if ≥3 of 4. Vehicle ≥ LLOQ invalidates the run. High conjugate does not rescue aglycone DETECT. DETECT is not AMPK/mTOR OCCUPY.

If PULSE DETECT fails for parent and geraldol: HG-H2 keeps that pulse; biology is not declared; AR-H1 survives the first cut. If it passes: AR-H1 dies as unassignable occupancy; the chronic phenotype is still not called biology of that species. Lipofuscin, CD68, follicles, TUNEL and BrdU do not vote on this satellite.

Step 3 — Q0-EXPOSURE-DAILY: schedule split, same clock, same vehicle

10 mg/kg oral, once daily, 30 days, PHOSAL, same colony, T1 = 1.0 h after the last gavage. Not Dong's DMSO/PEG300/Tween. Not BALB/c. Not D-gal. 20 or 30 mg/kg is not chosen after seeing DETECT. Joint interpretation with PULSE (table in experiment-design.md §5): daily passes / monthly fails → AR-H2 survives; both pass → the occupancy split dies; both fail → HG-H2 keeps both schedules; monthly passes / daily fails → unexpected, archived, not relabeled senolysis.

Step 4 — closed from this design

Chronic García phenotype: other females, only if PULSE DETECT passed. Q0-STATE, 2×2 D-gal, vehicle / F / Q / D+Q factorial, G8 and empirical HT1 are not opened from this stage. The expensive error would be LC–MS of the lipofuscin ovary and calling that H5.

15. Biomarkers and stratification

There is no validated biomarker. The following are research readouts, not biomarkers:

• Free parent, free geraldol and conjugate pool in the compartment, at a PK time written before: exposure gate (DETECT ≠ OCCUPY).
• Durable arrest plus an orthogonal readout that fails in luteum, atresia and nucleoside-incorporating cells: identity gate.
• Selective death versus sisters: senolysis gate.
• Independent ovarian function: endpoint, not marker.
• STRAW+10: minimum human stratification, not a criterion of a trial this stage does not authorize.
• rs4680: stratifier of the geraldol/parent ratio, not a predictor of ovarian senescence.

p16/p21, SA-β-gal, lipofuscin, CD68, AMH, follicles, menstrual regularity, age at menopause, plasma Cmax, 6MWD and SOFA cannot be presented as a single biomarker of senolysis, health or longevity. Farr's lymphocyte p16 is exploratory stratification of skeletal response under D+Q, not an ovarian measure of fisetin.

16. Individual variability

Expected heterogeneity includes reproductive stage, basal reserve, reproductive history, toxicant and time since injury, tissue composition, luteal state, adiposity, renal/hepatic function, free exposure, vehicle, microbiota, administration calendar, and — as PK, not as senotype — COMT activity and local deconjugation. Human studies must distinguish:

• natural transition versus iatrogenic ovarian insufficiency or PCOS;
• ovary with follicles versus late postmenopause;
• local action versus systemic disease;
• organ-specific response versus a global index;
• sex: Fang already showed dimorphism under D+Q; Jo is male ICR; Krishnakumar is 12 men and 3 women.

Women's participation is not equivalent to estimating modification by stage (L9-3-K9). Three volunteers from a single-dose PK do not estimate STRAW+10.

17. Pharma relevance and maturity

Pharma relevance is negative derisking, not readiness (L10-3-ED-C6). There is no nominated target. There is no modality. There is no indication in women's health. There is no patient stratum this stage authorizes. Maturity does not reach H5. HUMAN_QA_REQUIRED before any partnership conversation.

1. The object of this line is not a natural ovarian senolytic ready to evaluate. Fisetin, geraldol, conjugates, quercetin and D+Q remain distinct exposures. The only work authorized now does not even measure senolysis: it measures whether a García pulse and a 10 mg/kg daily are detectable at 1 h in C57BL/6 female ovary.
2. The direction already measured in ovary runs opposite to a senolytic. Dong lowers TUNEL and raises BrdU. Gao raises Bcl2 with fisetin and leaves TUNEL null. García moves lipofuscin/CD68 without function. A development that treats that direction as “senescent clearance” repeats the use-desire the laboratory closed.
3. Geraldol is not a development shortcut. Touil makes it more cytotoxic than parent in tumor and endothelium. Krishnakumar finds it dominant in unformulated human plasma. That is an object problem, not an ovarian active-metabolite advantage.
4. Dasatinib has an approved use outside this estimand. That existence is not a development signal in ovarian aging. This report does not evaluate that indication and does not transfer 5 mg/kg or 100 mg/kg to a person.
5. There is no mechanistic target this stage leaves standing. Dong's AMPK is occluded in vitro at 5 µM. Khan is NSCLC. Jin is not reopened. COMT is a PK enzyme. CD38 remains parked.
6. What would have to exist before a human evaluation is, in this order, a Jo table read as ICR, ovarian DETECT at T1 on García's pulse, the daily split at the same clock, a chronic phenotype only if DETECT passed, Q0-STATE only if someone still says senolysis, and an organ parameter that beats stage and the panel with competing risks. All of that is H0. An AMH RCT, expanding TROFFi or citing El Sayed 2026 does not fill G7.

Maturity:

• Fisetin ovarian senolysis (SCOPE-H1): H0.
• “Not senolysis” on already-measured ovaries (restricted HG-H1): literature H1, without formal equivalence.
• Jo/García non-identity (AR-H1): methods H1; ovarian occupancy H0.
• Schedule split (AR-H2): H1 that the schedules are not one; DETECT H0.
• Geraldol-dominant human UF plasma (restricted HG-H2): H1 for plasma; ovary H0.
• AMPK cytoprotection in D-gal: literature H1 for that object; transition H0.
• Translational event hypothesis: H0, HUMAN_QA_REQUIRED.
• Healthspan / G8: closed.

18. Limitations

1. A systematic review with a PRISMA protocol was not performed; the search was directed at the estimand.
2. Several studies have small n, ambiguous unit or counts by section. Dong: BrdU n=3, pie embryos n=3, RNA-seq n=3. Gao: qPCR n=3.
3. Ávila 2024 remains at abstract layer; n and dose were not invented.
4. There is no ovarian free exposure nor target occupancy. DETECT is designed, not run.
5. No in-vivo ovarian study splits vehicle, fisetin, quercetin and D+Q on coprimaries of selective death and function.
6. None meets the required cellular-identity gate.
7. Calendars, vehicles, strains and toxicants are heterogeneous and are not averaged.
8. p>0.05 does not establish equivalence; several “nulls” may be imprecise.
9. Jo's table was not read. 7.8% F remains not a concentration. García was not reread on 2026-09-28 (PMC HTTP 500 / captcha in the design turn); experiment 2 methods are inherited.
10. T1 = 1.0 h is a proxy of human UF tmax, not the C57BL/6 female tmax in PHOSAL.
11. Four females do not estimate an EC50. That is why DETECT is a detectability floor, not target occupancy.
12. Daily 10 mg/kg in PHOSAL is not Dong. A negative daily DETECT does not kill the D-gal paper in BALB/c.
13. The 1–1000 ng/ml curve is human plasma. The ovarian LLOQ can be worse.
14. Dong's E2 ELISA (ng/mL) and the −44.8 kcal/mol docking remain rejected.
15. Animal data do not authorize dose or intervention in people.
16. The bridge to longevity has no measured event, mediation or competing risks.
17. PubMed returned captcha in this turn; reconfirmation of Dong and Krishnakumar was against open PMC, not against PubMed esummary.

19. Conclusions

Fisetin can lower a lysosomal marker or an infiltrate in an adult female ovary without separating function; it can share with D+Q the marker–structure dissociation under doxorubicin, with Bcl2 up and primordials down only in the pair; and it can, in D-gal of a young BALB/c, lower TUNEL and return nucleoside through AMPK. That does not demonstrate ovarian senolysis, does not rank a natural senolytic against D+Q, and does not connect to an organ event in women.

The title's bioaccessibility is three exposures and three species. Unformulated human oral is geraldol-dominant, sub-µM and short half-life. García's 100 mg/kg milligram does not inherit an ICR male's F. The phenotype ovary is not a PK time.

The scientific priority is neither to expand enthusiasm nor to repeat flavonol versus vehicle in another D-gal. It is to separate species, schedule and organ. The first discriminant can cost zero animals if Jo is left to be read as ICR. The first animal, if opened, asks DETECT at 1.0 h, not senolysis. Women's longevity is not at stake in either of those two endpoints.

20. References

1. Yousefzadeh MJ, Zhu Y, McGowan SJ, et al. Fisetin is a senotherapeutic that extends health and lifespan. EBioMedicine. 2018;36:18–28. DOI: 10.1016/j.ebiom.2018.09.015. PMID: 30279143. PMC6197652.
2. Zhu Y, Tchkonia T, Pirtskhalava T, et al. The Achilles' heel of senescent cells: from transcriptome to senolytic drugs. Aging Cell. 2015;14:644–658. DOI: 10.1111/acel.12344. PMID: 25754370.
3. Khan N, Afaq F, Khusro FH, Mustafa Adhami V, Suh Y, Mukhtar H. Dual inhibition of phosphatidylinositol 3-kinase/Akt and mammalian target of rapamycin signaling in human nonsmall cell lung cancer cells by a dietary flavonoid fisetin. Int J Cancer. 2012;130:1695–1705. DOI: 10.1002/ijc.26178. PMID: 21618507. PMC3267899.
4. Gao Y, et al. Increased cellular senescence in doxorubicin-induced murine ovarian injury: effect of senolytics. GeroScience. 2023;45:1775–1790. DOI: 10.1007/s11357-023-00728-2. PMID: 36648735. PMC10400526.
5. García DN, et al. Study of D+Q/fisetin, ovarian reserve and fertility in C57BL/6 females. GeroScience. 2024. DOI: 10.1007/s11357-024-01089-0. PMID: 38358579. PMC11009191.
6. Dong J, Zhu Y, Li Z, Chang Y, Yan X, Zhang C, Niu D. Fisetin Attenuates D-Gal-Induced Ovarian Aging by Modulating Mitophagy via the AMPK/mTOR Pathway. Antioxidants. 2026. DOI: 10.3390/antiox15050602. PMID: 42193224. PMC13203114. Not Ding/Zeng.
7. Ávila et al. Study of D+Q/fisetin after VCD-induced chemical estropause. Life Sciences. 2024;123073. DOI: 10.1016/j.lfs.2024.123073. PMID: 39307182. Abstract; n and dose not extracted.
8. Fang Y, et al. Comparative study of D+Q and fisetin in male and female C57BL/6 mice. GeroScience. 2023. DOI: 10.1007/s11357-023-00843-0. PMID: 37296266. PMC10643448.
9. Li J, et al. Quercetin prevents primordial follicle loss via suppression of PI3K/Akt/Foxo3a pathway activation in cyclophosphamide-treated mice. Reprod Biol Endocrinol. 2021. DOI: 10.1186/s12958-021-00743-y. PMID: 33892727.
10. Krishnakumar IM, Jaja-Chimedza A, Joseph A, Balakrishnan A, Maliakel B, Swick A. Enhanced bioavailability and pharmacokinetics of a novel hybrid-hydrogel formulation of fisetin orally administered in healthy individuals. J Nutr Sci. 2022;11:e74. DOI: 10.1017/jns.2022.72. PMID: 36304817. PMC9574875.
11. Jo JH, Jo JJ, Lee JM, Lee S. PK study of fisetin and geraldol in mouse. J Chromatogr B. 2016. DOI: 10.1016/j.jchromb.2016.10.034. PMID: 27810278. Abstract/MeSH; table not opened.
12. Touil YS, Auzeil N, Boulinguez F, et al. Fisetin disposition and metabolism in mice: Identification of geraldol as an active metabolite. Biochem Pharmacol. 2011;82:1731–1739. DOI: 10.1016/j.bcp.2011.07.097. PMID: 21840301.
13. Shia CS, Tsai SY, Kuo SC, Hou YC, Chao PD. Metabolism and pharmacokinetics of 3,3',4',7-tetrahydroxyflavone (fisetin), 5-hydroxyflavone, and 7-hydroxyflavone and antihemolysis effects of the metabolites of 3,3',4',7-tetrahydroxyflavone. J Agric Food Chem. 2009. DOI: 10.1021/jf802378q. PMID: 19090755.
14. Harrison DE, Strong R, Reifsnyder P, et al. Astaxanthin and meclizine extend lifespan in UM-HET3 male mice; fisetin, SG1002 (hydrogen sulfide donor), dimethyl fumarate, mycophenolic acid, and 4-phenylbutyrate do not significantly affect lifespan. Geroscience. 2024. DOI: 10.1007/s11357-023-01011-0. PMID: 38041783. PMC10828146. Fisetin from 20 months, not 12.
15. Farr JN, Atkinson EJ, Achenbach SJ, et al. Effects of intermittent senolytic therapy on bone metabolism in postmenopausal women: a phase 2 randomized controlled trial. Nat Med. 2024;30:2605–2612. DOI: 10.1038/s41591-024-03096-2. PMID: 38956196. D+Q, not fisetin.
16. Hickson LJ, Langhi Prata LGP, Bobart SA, et al. Senolytics decrease senescent cells in humans: Preliminary report from a clinical trial of Dasatinib plus Quercetin in individuals with diabetic kidney disease. EBioMedicine. 2019. DOI: 10.1016/j.ebiom.2019.08.069. PMID: 31542391.
17. Justice JN, Nambiar AM, Tchkonia T, et al. Senolytics in idiopathic pulmonary fibrosis: Results from a first-in-human, open-label, pilot study. EBioMedicine. 2019. DOI: 10.1016/j.ebiom.2018.12.052. PMID: 30616998.
18. Lin J, Nie X, Xiong Y, et al. Study of cyclophosphamide POF and fisetin 100 ng/kg. Am J Transl Res. 2020;12(1):203–247. PMID: 32051749. PMC7013230. No DOI.
19. Bujnakova Mlynarcikova A, Scsukova S. Study of fisetin and steroidogenesis in porcine granulosa. Environ Toxicol Pharmacol. 2018. DOI: 10.1016/j.etap.2018.03.001. PMID: 29544186.
20. Jin et al. Single-cell/nuclear atlas of human ovarian aging. Nat Aging. 2025. DOI: 10.1038/s43587-024-00762-5. PMID: 39578560. PMC11839473. GSE202601.
21. Moustafa et al. Fisetin in letrozole PCOS. Naunyn Schmiedebergs Arch Pharmacol. 2024. DOI: 10.1007/s00210-024-03124-5. PMID: 38789632. Outside the estimand; not Alenezi.
22. Chahal, Kabra. Fisetin vs metformin in mifepristone PCOS; dorsomorphin reverses. Naunyn Schmiedebergs Arch Pharmacol. 2024. DOI: 10.1007/s00210-024-03257-7. PMID: 38963551. Outside the estimand; not Ali.
23. El Sayed et al. Review of senotherapeutics in reproductive health. Nutrients. 2026. DOI: 10.3390/nu18030393. PMID: 41683217. Frame rejected; concedes absent reproductive clinical evidence.
24. Saul D, et al. Senescence atlas; does not score senotype until Ki67+, CD45-high and S phase are withdrawn. 2025. Inherited via L10-1-K8. The ovary is not in that atlas.
25. Coppé JP, Rodier F, Patil CK, et al. Tumor suppressor and aging biomarker p16INK4a induces cellular senescence without the associated inflammatory secretory phenotype. J Biol Chem. 2011. DOI: 10.1074/jbc.M111.257071. PMID: 21903580. CDKI arrest without SASP in fibroblast; not ovary.

Claims added in REPORT_EN: L10-3-REPORT-EN-C1.

Hypotheses touched: L10-3-AR-H1 v1, L10-3-AR-H2 v1, L10-3-AR-HT1 v1, L10-3-HG-H1 v1, L10-3-HG-H2 v1, L10-3-HG-HT1 v1, L10-3-SCOPE-H1 v1, L10-3-SCOPE-H5 v1.

This is a scientific report and does not constitute medical advice, a prescription, or authorization to use fisetin, geraldol, quercetin, dasatinib or their combination. It is not written into memoria agentes/Lua Labs/reportes/. Promotion of the final report occurs only after QA.