# HSEF Review: Epitalon, Telomerase and Human Healthspan

**HSEF v5.4 evidence review · authored 2026-09-27 · red-team synthesis 2026-09-30**

**Provenance:** 22 source records · 21 registered load-bearing claims · 21 verified · 0 unverified · 0 contradicted · 20 Tier A and 2 Tier B sources · gate passed

> **Adversarial review:** Generated by Sol 6.1; adversarially reviewed by Claude Opus 5.5 (Anthropic) on 2026-10-14. 2 corrections applied (see ADVERSARIAL REVIEW AND APPROVAL). Status: **Approved as revised.**

## AT A GLANCE

| Item | Assessment |
|---|---|
| **HSEF level** | **Γ — Mechanistically plausible, partial system support** |
| **Confidence** | **Low** — small literature, almost all of it from one connected research network, with reproducibility concerns in the assays that carry the claim |
| **Mechanistic role** | **Epiphenomenon** — the classification the framework's derivation reaches at step 2, because no causal dependency has been demonstrated in either direction at the claimed human scope |
| **Aggregation** | Weakest-link, preregistered before adjudication |
| **Rubric routing** | 06-metabolism-aging (primary); 03-epigenetics-gene-regulation (secondary, governing the proposed proximal step). Both return a Γ ceiling |
| **Strongest finding** | Two independent laboratories report telomerase or telomere effects after exposing cultured human cells to the peptide [1–3]; a third independent laboratory reports a telomerase-linked effect in cultured bovine cells [15] |
| **Binding gap** | No study demonstrates telomerase activation after the peptide is administered to a living person, and no trial of the synthetic tetrapeptide measures validated human healthspan [9–11] |
| **Evidence concentration** | 11 of the 12 sources in the original edition originate in one connected research network; no organism-level result has been reproduced by an independent laboratory |
| **Momentum** | Mechanistic only |
| **Stability** | Stable |
| **Divergence** | Severe overstatement: this literature's own titles and conclusions imply roughly Ζ; assessed Γ; gap +3 rungs |

The hypothesis is a conjunction. It requires the same chemically defined synthetic tetrapeptide to activate telomerase in humans, cause durable telomere maintenance, improve health and function during aging, and do so through the proposed telomerase pathway. The evidence reaches the first link in cultured cells. It does not reach the complete chain in people, and at the human scope it does not reach the chain at all.

## LAY SUMMARY

The claim says that the synthetic four-amino-acid peptide Epitalon turns on telomerase, the enzyme that can maintain chromosome ends, and that this molecular effect helps people remain healthy for longer.

The review assigns **Γ**, meaning **mechanistically plausible, partial system support**. Two research groups report telomerase or chromosome-end effects in cultured human cells, a third reports a similar effect in cattle cells, and several animal studies report mixed longevity findings. No study shows telomerase activation after the peptide is given to people, and no trial of the synthetic peptide measures human healthspan.

Confidence in that grade is **low**, which in this framework is a statement about the evidence rather than about the conclusion: the literature is small, nearly all of it comes from one group of connected laboratories, two of the animal reports could only be read in summary, and the one independent laboratory study had all three of its main figures replaced after publication. The missing evidence is a controlled human study linking molecular target engagement to durable functional benefit. The grade describes the present causal evidence and does not determine whether future trials will succeed.

## CLAIM RESTATEMENT AND SCOPE

### Wording submitted

> Epitalon activates telomerase and extends human healthspan.

### Gradeable strong proposition

> In humans, administration of the synthetic tetrapeptide Epitalon (also called Epithalon; sequence Ala-Glu-Asp-Gly, abbreviated AEDG) activates telomerase in somatic cells, produces durable telomere maintenance with net elongation, and thereby slows age-related functional decline sufficiently to extend healthspan.

The identity qualifier is fixed before outcomes: this review concerns the chemically defined synthetic tetrapeptide. It does not silently substitute **Epithalamin**, a heterogeneous pineal extract, or **Thymalin**, a thymic preparation. "Human" is also fixed before outcomes. No subgroup such as "responders" or "susceptible people" is fitted after the result.

One qualification on that qualifier belongs here rather than in a footnote. No study in this corpus reports independent verification of the identity or purity of the material it used. The one independent cell-culture study states that the peptide was purchased from a commercial supplier and reports no identity or purity assay [3]; an independent 2025 overview records that physico-chemical and structural characterisation of the peptide remains limited [17]; and the 2026 regulatory evaluation lists insufficient characterisation among its concerns. "Chemically defined" describes what the molecule is, not what any particular experiment demonstrably contained.

Healthspan means time lived with preserved health and function. A change in telomerase, telomere length, melatonin, clock-gene expression, or another biomarker is not itself healthspan. A disease-specific symptom can be clinically meaningful without becoming a measure of systemic aging.

### Prespecified subclaims

| ID | Subclaim | Scope | Level | Confidence | Binding rule |
|---|---|---|---|---|---|
| **SC1** | Epitalon directly activates telomerase in human somatic cells. | **Cell culture.** Direct enzyme activity in defined human cells; target engagement after administration to a person is a separate scope, carried by SC2. | Γ | Moderate | In-vitro evidence cannot exceed Γ. |
| **SC2** | Epitalon administration durably maintains or lengthens telomeres in living humans. | Longitudinal human tissue measurement beyond assay error and cell-composition shifts. | Γ | Low | Cell-culture and biomarker evidence cannot establish the organism-level inference. |
| **SC3** | Synthetic Epitalon extends human healthspan. | Prespecified function, morbidity, disability-free survival, or a validated multidomain aging endpoint. | Γ | Low | Δ requires an intervention changing lifespan or healthspan; human studies do not measure it. |
| **SC4** | Any human healthspan effect is materially mediated by telomerase and telomeres. | A within-study causal chain from target engagement to functional outcome. | Γ | Low | Cross-system mechanism stitching and biomarker confirmation cannot cross Δ. |
| **SC5** | Epitalon produces a favorable net human healthspan effect. | Systematic long-term benefit and harm ascertainment for verified synthetic Epitalon. | Γ | Low | Short small studies provide partial exposure evidence, not net long-term benefit. |

The preregistered aggregation rule was **weakest-link** because the submitted wording uses "and." All five subclaims land at Γ, so weakest-link, primary-subclaim, and scope-matched aggregation each return Γ. The grade is not an artifact of choosing among those rules.

Two things about that Γ need saying plainly, because a letter that is the same for every subclaim can hide them. First, **at the claimed human scope the evidence is empty**: SC2, SC3, SC4 and SC5 sit at Γ not because human work supports them partially, but because Γ is what the framework assigns when simplified-system evidence exists and endogenous causal dependency has not been shown — and here the simplified systems are cultured cells and three animal species. Second, the case for scoring the human-scoped subclaims lower, at Α, was argued in the red-team pass and rejected: Α describes a claim sustained by analogy rather than causal evidence, and this corpus contains direct interventions with survival endpoints in flies, mice and rats and direct molecular readouts in human cells. What the concentration of that work in one research network changes is the confidence band, not the letter, and that is where this edition moved it.

## COMPETING MECHANISTIC MODELS

| Rank | Model | What it predicts | Fit to the evidence |
|---|---|---|---|
| **1** | **Context-limited cellular telomere response** | The peptide changes telomerase or telomere biology in some proliferating cultured cells, without guaranteed translation to intact humans. | Best fit. It accommodates the positive cell assays, their context dependence, and the absence of human target-engagement or healthspan evidence [1–3,9–11,15]. |
| **2** | **Telomerase-mediated human geroprotection** | Human exposure activates telomerase, preserves telomeres, and slows functional aging. | Plausible but untested at its defining scope. Cell and animal findings support parts of the chain, and telomerase activation by another modality does extend median lifespan in old mice [18], while every human causal link remains missing [1–11]. |
| **3** | **Non-telomerase pleiotropic effect (circadian, neuroendocrine, antioxidant)** | The peptide changes melatonin, clock genes, oxidative-stress responses or senescence markers without telomerase mediating a healthspan effect. | Compatible with the small human biomarker studies [9,10] and strengthened since the prior edition: an independent 2025 study attributes an AEDG effect in injured human retinal pigment epithelium to antioxidant and epithelial–mesenchymal-transition-related action with no telomerase measurement at all [16], and same-network work reports senescence-marker and dendrite effects without telomerase [21]. It lacks validated functional and long-term endpoints. |
| **4** | **Assay or research-program artifact** | Positive findings arise from cell context, analytic flexibility, selective reporting, or correlated bias. | Still viable because most organismal reports come from a connected program, the fly effect was flat across several orders of magnitude of exposure, the most auditable mouse survival result was null on its central summaries, and the independent 2025 paper had all three core data figures replaced by an erratum [3–8]. A complete-artifact account is weakened by two further independent laboratories reporting bounded telomerase-linked effects [3,15]. |

The tournament does not change the HSEF letter. It clarifies what the observed data most directly support. The context-limited model ranks first because it predicts both the positive cell-culture results and the failure of those results, so far, to appear as a demonstrated human causal chain.

The single study that would most clearly separate the first two models is a blinded randomized human trial that measures in-vivo telomerase target engagement, durable tissue-appropriate telomere change, and a prespecified validated healthspan endpoint in the same participants. A mediation analysis would need to show that the functional result tracks the telomerase/telomere effect rather than an alternate peptide action.

## FULL REVIEW

### 1. What must be true for the claim to pass

The wording compresses four evidentiary jumps:

1. a synthetic peptide changes telomerase in cultured cells;
2. a dose given to a person reaches relevant tissues and changes telomerase there;
3. that change durably maintains telomeres in the relevant cells;
4. the molecular change improves age-related function or morbidity enough to extend healthspan.

Each jump has different failure modes. Culture conditions can create exposures that do not occur in a person. Average telomere length can change because cell populations change. Telomere maintenance can occur without a whole-person functional benefit. A clinical effect, if one exists, can arise through circadian, neuroendocrine, immune, stress-response, or other pathways rather than telomerase.

A fifth condition is implicit in the word "extends" and is easy to miss: even a demonstrated telomere-lengthening effect would have to be, on balance, good for a person. That is not a safe assumption about this pathway, and §7 sets out the human genetic evidence that bears on it.

The review therefore does not allow evidence for one link to stand in for another. This is the main reason the submitted claim remains at Γ even though the cell-culture result is more than a bare association.

### 2. Direct telomerase and telomere evidence

#### Foundational fetal-fibroblast study

Khavinson, Bondarev, and Butyugov exposed a telomerase-negative human fetal lung fibroblast culture to the peptide and reported induction of the telomerase catalytic subunit, telomerase enzymatic activity, and telomere elongation [1]. That is the whole of what the accessible record supports.

The prior edition of this review reported specific telomere fluorescence values in arbitrary units and a percentage difference, together with a HeLa positive control. Those specifics are withdrawn here. The primary article is available only as a scanned image on a single-programme website and yielded no machine-readable text when re-checked, no other source reports the values, and the 2026 regulatory evaluation's account of this study likewise reports only that telomere length and telomerase expression increased. Under this project's rule that a number which cannot be verified against its source does not ship, the qualitative finding stays and the figures go. Nothing in the adjudication rests on them.

This is a direct perturbation with a molecular readout, so it supports more than association. Its scope is narrow: one fetal fibroblast culture, with sparse modern detail about replication, statistical structure, and exposure-response behavior. It does not establish pharmacokinetics, target engagement in an intact person, or a functional aging outcome.

A follow-up from the same program reported that primary fetal pulmonary fibroblasts lost proliferative potential at passage 34, while peptide-treated cells with elongated telomeres made ten further divisions, reaching passage 44, and continued dividing [2]. This adds a cellular functional correlate, and it is the only place in the corpus where a telomere change is tied to a cellular phenotype. It is not an independent causal line: same research program, same fetal-fibroblast context.

#### Independent 2025 cell-line study

Al-Dulaimi and colleagues provide the most important independent replication [3]. Reading the corrected full text, the design is asymmetric in a way that matters for interpretation. Two breast-cancer lines were treated daily for four days across a four-point concentration range; normal human fibroblast and mammary epithelial lines were treated daily for three weeks at the top concentration. Normal fibroblasts treated for only four days showed no telomere increase. At three weeks both normal lines showed significantly longer telomeres, raised hTERT mRNA, and raised telomerase activity — approximately fourfold in the fibroblast line and twenty-sixfold in the mammary epithelial line.

The study also found a mechanistic boundary, and reported it with two convergent assays rather than one. In the cancer lines, telomeres lengthened and hTERT mRNA rose (twelvefold and fivefold) with no significant increase in telomerase enzyme activity; C-circle assay showed roughly a tenfold rise in alternative-lengthening-of-telomeres activity in one line and a threefold rise in the other, confirmed by significantly elevated PML bodies in both. The normal lines showed no such ALT activation. The response is therefore not a simple universal "the peptide turns on telomerase" rule across human cells, and the internal negative controls — four days versus three weeks, a telomere decrease at the lowest concentration in one cancer line, and normal-cell telomerase activity that stayed below the level in untreated cancer lines — map where the effect stops. That mapping is what makes this the corpus's most informative experiment.

The article has a publication-status flag. A formal erratum from the same authors states that the wrong figures appeared as Figures 1, 2 and 3 and supplies corrected versions [3]. Those figures contain the telomere-length, hTERT/telomerase, and alternative-lengthening results. This is a figure-handling correction by the authors, not a retraction and not an expression of concern about the data, and the corrected version is the one read here and quoted above. Replacing all three core data figures is nonetheless a reproducibility signal, and it is recorded in the confidence assessment rather than used to discount the numbers.

The article concludes that the findings suggest the peptide could be used safely in healthy individuals to maintain telomeres and influence aging. That inference exceeds the design. A two-dimensional cell-line experiment cannot assess pharmacokinetics, human efficacy, clinical safety, or healthspan. The paper itself identifies its in-vitro design as the main limitation.

#### A third independent laboratory, in another species

An independent Korean laboratory reports that peptide-associated telomerase activation improved bovine oocyte maturation rate and the hatching and implantation potential of post-thaw blastocysts, with supporting reactive-oxygen-species and mitochondrial-membrane-potential measurements [15]. This is cultured non-human mammalian material used for embryo production, not an aging experiment, and it cannot speak to human healthspan. It matters for two narrow purposes: it is a third laboratory outside the originating network reporting a telomerase-linked effect of the peptide, and it therefore weakens a pure assay-artifact account of the cellular findings.

#### The proposed proximal step

Between "peptide in the medium" and "telomerase activity" there is a step the hypothesis needs and the literature has barely tested: how a tetrapeptide reaches and acts on the telomerase gene. The originating proposal is a sequence-complementarity model in which the peptide binds a specific base-pair block, identified as ATTTTC, found repeatedly in the telomerase promoter region [14]; later work from the same programme reports binding to methylated cytosine and to linker histone H1.3 and H1.6, and the 2025 independent study builds its ALT speculation on that reported histone binding [3].

This is an epigenetic gene-regulation claim, and the epigenetics rubric governs it. Under that rubric, sequence-motif occurrence sits in the Weak evidence tier, and the Δ threshold requires an endogenous perturbation showing the proposed feature contributes causally to the claimed output, with specificity demonstrated against global transcriptional collapse and with a matched rescue. None of that has been done for this peptide. The proximal step is the least-studied link in a chain whose distal end is the one the claim advertises.

#### Molecular adjudication

The laboratories make **SC1 biologically plausible with partial system support**. They do not show the molecular effect after administration to a person. The 2025 study also shows that duration and cell identity matter, and that telomere elongation can occur through a different pathway in cancer cells.

SC2 remains at the same ceiling. Average telomere changes in cultured cells do not establish durable telomere maintenance in living human tissues. A valid human test would need prospectively chosen tissues, repeated measurement, adjustment for cell composition, attention to the shortest telomeres rather than only a bulk mean, verified exposure, and direct target engagement.

### 3. Animal lifespan evidence

Animal work supplies organism-level perturbations, but it does not form a clean path to human healthspan.

#### Drosophila: developmental exposure, and a dose-independent effect

The Drosophila study reported an 11–16% increase in adult mean lifespan in the Canton-S strain [4]. Exposure occurred during the egg-to-larval developmental period rather than as an adult or late-life intervention, so the study cannot separate altered development from slowed aging. Under the aging rubric, developmental slowing cannot be counted as a clean geroprotective intervention.

The dose structure deserves more weight than the prior edition gave it. Effective concentrations ran from 0.001 × 10⁻⁶ to 5 × 10⁻⁶ wt.% of culture medium for males and from 0.01 × 10⁻⁶ to 0.1 × 10⁻⁶ wt.% for females, and the paper states explicitly that the increase in lifespan did not depend on dose. A reported effect that is flat across three to four orders of magnitude of exposure is not merely missing a dose-response curve; it is evidence against a saturable, receptor- or promoter-mediated pharmacological mechanism of the kind the hypothesis proposes.

#### Female CBA and senescence-accelerated mice

A female CBA mouse paper reports that treatment from six months of age did not change body weight, food consumption, physical activity or behaviour, but slowed the age-related loss of estrous function, decreased body temperature, decelerated free-radical processes, prolonged lifespan and reduced spontaneous tumour incidence [5]. A study in senescence-accelerated mice reports longer mean and maximum survival among the last 10% of survivors in the senescence-prone substrain, no effect on tumour frequency, and longer survival in tumour-free animals [7]. Only English abstracts were accessible for these Russian-language reports, so this review imports no exact effects, allocation details, censoring rules, or unreported healthspan measures from them.

Both studies arise from the same connected research program as the rest of the organismal literature. Repetition within a program can increase precision and identify context, but it is not an independent causal line capable of excluding the program's shared design and analytic failure modes.

#### SHR mice: null central summaries, positive tail

The most auditable mouse survival report randomised female outbred SHR mice 54 per group and treated them from three months of age until natural death [6]; four mice per group were euthanised at twelve months for the cytogenetic study, so the paper reports its lifespan and tumour tables on 50 per group, which is the denominator behind the percentages below. Mean lifespan was 455 ± 31 days with the peptide against 456 ± 29 days in controls. The median was **lower** in the treated group: 487 days against 512. The favourable signals were confined to the upper tail — mean lifespan of the last 10% of survivors 803 ± 15.0 days against 709 ± 10.8, a 13.3% increase (P < 0.01), and maximum lifespan 830 days against 739, a 12.3% increase.

The prior edition reported the median as though it favoured treatment. It did not, and the corrected direction strengthens rather than weakens the review's conclusion: both central summaries of the survival distribution were null or slightly unfavourable while a secondary region was positive.

Two further features of this study belong in the record. Treated mice **consumed more food** than controls from the fifth to the sixteenth month of life, with no body-weight difference at any point of observation — the article's own abstract states that treatment did not influence food consumption, and its table is the finer-grained record, a discrepancy worth naming because the abstract is what most readers see. Caloric intake is a listed confounder in this field, and a treatment associated with higher intake and unchanged weight is a metabolic observation in its own right.

The favourable non-survival findings should also be on the record, since this is the only rodent study in the corpus with systematically collected harm-relevant data. Total spontaneous tumour incidence was 32% (16/50) with the peptide against 36% (18/50) in controls; leukaemia occurred in 1 treated animal against 6 controls, a sixfold reduction (P < 0.01); bone-marrow chromosome aberrations at 12 months were 6.8 ± 0.21% against 8.2 ± 0.41%, a 17.1% decrease (P < 0.05). The authors conclude that their data suggest geroprotector activity and the safety of long-term administration in mice. None of this is a healthspan endpoint, none of it has been independently reproduced, and none of it transfers to people — but omitting it while reporting the null mean would be asymmetric scrutiny.

#### Rats: an effect that appears only under disrupted lighting

In male rats treated from four months of age under three lighting regimens — standard 12-hour light, the natural illumination of north-western Russia, and permanent illumination — mean lifespan was virtually unchanged; the peptide was associated with significant normalisation of the population aging rate and of mortality-rate doubling time **in the naturally-lit and constantly-lit groups**, and it significantly inhibited spontaneous tumours, chiefly testicular leydigomas and leukaemias, under every regimen [8]. The companion study in female rats from the same programme is explicit about the pattern: no lifespan change under the standard day/night regimen, maximum lifespan prolonged by 95 and 24 days under natural and constant light respectively, and mean lifespan of the last 10% prolonged by 137 and 43 days [13].

The prior edition described the maximum-lifespan extension as "limited to constant darkness." Neither survival paper has a constant-darkness arm; four-condition designs including light deprivation exist elsewhere in this rat programme but not here, which is the likely origin of the error. The corrected reading is a sharper argument than the original made. The apparent survival benefit appears in animals whose circadian environment has been disrupted and is absent in animals on a standard light cycle — which reads as partial mitigation of an induced pathology rather than as an effect on normal aging. Treatment also began in young adulthood rather than late life, and neither study measured telomerase or telomeres.

#### The comparison the hypothesis needs, and does not have

The proposed mechanism has been tested directly in old animals, by a different modality, and the result was positive. Adeno-associated-virus delivery of mouse TERT to 1- and 2-year-old mice improved insulin sensitivity, osteoporosis, neuromuscular coordination and molecular aging biomarkers, increased median lifespan by 24% and 13% respectively, and did not increase cancer relative to control littermates; a catalytically inactive TERT produced none of the benefits [18].

That study is the benchmark this literature should be read against, and it cuts three ways. It shows the mechanism class is not implausible at the organism level, which is the strongest available case for the nominal model. It is also the template for the specificity test the Epitalon literature has never run — the inactive-enzyme control is what makes the mouse result mechanism-specific. And it does not help the submitted claim: the intervention is a vector-delivered gene, not a peptide; universal cap 3 caps overexpression evidence at Γ for claims about the endogenous pathway; and a mouse result cannot carry a human-scoped conjunction.

#### Animal adjudication

The animal literature contains positive signals, null central summaries, developmental exposure, a dose-independent effect, strain and sex restriction, an effect confined to disrupted lighting environments, and strong clustering within one research network. It does not provide the rubric's gold-standard pattern of late-life intervention, replicated lifespan plus healthspan improvement across independent laboratories or backgrounds, and mechanistic rescue.

Even a robust animal lifespan effect would not by itself establish the submitted human claim. Here, the animal result is less robust than that and does not measure the proposed telomerase causal chain.

### 4. Human intervention evidence

#### Pineal-peptide melatonin study in older adults

Korkushko and colleagues report that in old monkeys and elderly people, pineal peptide preparations restored night release of endogenous melatonin and normalised its circadian rhythm, with elderly people who had pineal functional insufficiency showing an increase in night melatonin; the report also asserts that the preparations have no side effects [9].

Two limits on how far this source can be used. First, the study reports **two different preparations together** — Epithalamin, the heterogeneous pineal extract, and the synthetic tetrapeptide — and the accessible record does not attribute the human melatonin result to one rather than the other. Since this review excludes a long-term mortality report on identity grounds (§5), consistency requires saying so here rather than describing the study as a clean synthetic-tetrapeptide result. Second, the design specifics asserted in the prior edition — randomisation, a placebo arm, a ten-day course — are withdrawn. The accessible abstract states none of them; the primary is an image-only scan; and the 2026 regulatory evaluation records for this study that route, dosing, duration, number of subjects, demographics and methods of safety data collection "were not provided in the abstract" and that the full article was unavailable in English.

What remains is that products described as pineal peptides, including this one, have been given to elderly people with a reported biomarker change. That is not evidence of telomerase activation, telomere maintenance, or healthspan.

#### Night-shift circadian-gene study

Ivko and colleagues report that AEDG increased urinary 6-sulfatoxymelatonin excretion about 1.7-fold in middle-aged people and normalised hyper-expression of the Clock and Csnk1e genes in leukocytes by roughly 1.9–2.1-fold while increasing hypo-expressed Cry2 about twofold in people with reduced melatonin-producing function [10].

The design, as summarised in the 2026 regulatory evaluation rather than in the primary Russian text, was 75 healthy women aged 40–50 working mainly night shifts: 35 with age-appropriate melatonin-metabolite excretion as a comparison group, and 40 with low excretion allocated 20 to peptide and 20 to placebo for twenty days. That synthesis records that the study did not specify blinding. It also records a secondary observation that the peptide group showed positive dynamics of psycho-emotional and general physical condition in 83% of cases against 25% on placebo, with no description of how either was assessed — which is why this review does not treat that figure as a functional outcome. A second paper reports clock-gene findings from the same cohort; it is one study reported more than once, not a second independent clinical line.

Neither report measures telomerase activity, telomere length, validated multidomain healthspan, disability-free survival, or long-term morbidity.

#### Retinitis pigmentosa

The retinitis-pigmentosa report combines an animal component in Campbell rats with a clinical component in which the peptide was administered locally to the eye in patients with degenerative retinal lesions, reporting a positive clinical effect in 90% of cases [11]. PubMed classes it as a controlled clinical trial; randomisation and blinding are not described in the accessible record, and the comparator received conventional treatment. Even if the visual result is accepted as reported, a local ophthalmic outcome in a retinal disease is not evidence that the intervention slows systemic aging. The study measured neither telomerase nor telomeres. An independent 2025 cell study of AEDG in injured human retinal pigment epithelium is consistent with a retinal effect arising through antioxidant and epithelial–mesenchymal-transition-related routes rather than telomere biology [16].

#### What the human evidence does and does not establish

The human studies demonstrate that people have received products described as Epitalon or AEDG and that short-term physiological or disease-specific signals have been reported. Across the identified primary reports, none measures telomerase activity after administration, none establishes longitudinal telomere maintenance, and none uses a validated healthspan endpoint [9–11].

One qualification bounds that absence claim. A 2024 review from the originating institute reports that the peptide regulates circadian genes and markers including TERF-1, prohibitin, SIRT1 and SIRT6 in buccal epithelium [21]. TERF-1 is a telomere-associated protein. Whether the underlying studies administered the peptide to people or applied it to cultured buccal cells cannot be determined from the accessible Russian-language abstract. The absence claim is therefore stated precisely: no identified study measures telomerase enzyme activity or telomere length after administration to a person, and this open question is recorded for a future pass.

Exact-name searches of ClinicalTrials.gov on 2026-09-30 for Epitalon, Epithalon, Epithalone, Ala-Glu-Asp-Gly and AEDG peptide returned no registered study; searches for Epithalamin and Thymalin likewise returned none. The FDA's evaluation of epitalon-related bulk drug substances of 12 May 2026, prepared for the Pharmacy Compounding Advisory Committee meeting of 23–24 July 2026, states that no pharmacokinetic data were found for epitalon free base or epitalon acetate, that there are no clinical data supporting their safety when used in humans, and that there is a lack of evidence supporting effectiveness for the proposed use — which was insomnia, not healthspan; it concludes that the evaluation criteria weigh against listing, citing insufficient characterisation, absent documentation of historical compounding use, no clinical efficacy evidence, and safety concerns including potential immunogenicity. A peer-reviewed 2026 analysis of the committee's recommendations records the same peptide among synthetic peptides promoted for healthy aging and the concerns raised in review [22]. These search and regulatory artifacts corroborate the gap; they are not causal evidence, and the registry and FDA items are not counted among the numbered peer-reviewed ledger sources.

### 5. Epitalon is not Epithalamin

The substance distinction is load-bearing. Epitalon is the defined tetrapeptide Ala-Glu-Asp-Gly, synthesised on the basis of the amino-acid composition of Epithalamin [17]. Epithalamin is a heterogeneous preparation derived from bovine pineal tissue. The names are similar, and both appear in work from the same research tradition, but they are not interchangeable interventions.

The long-term human report titled "Peptides of pineal gland and thymus prolong human life" administered Epithalamin and Thymalin to 266 elderly people over six to eight years and reports mortality reductions of 1.6–1.8-fold in the Epithalamin group, 2.0–2.1-fold with Thymalin, 2.5-fold with both, and 4.1-fold in a group retreated annually for six years [12]. Whatever the merits of that study on its own terms, it cannot establish efficacy, mechanism, pharmacokinetics, or safety for the synthetic tetrapeptide. Other long-term reports using Epithalamin face the same identity boundary.

This exclusion is not semantic bookkeeping. If long-term survival evidence for one preparation is assigned to another, the apparent human evidence base changes from "no healthspan trial" to "long-term mortality studies." The latter description is factually wrong for the synthetic tetrapeptide assessed here. The rule cuts both ways, which is why §4 states that the principal human biomarker source reports both preparations and does not separate their results.

### 6. Telomerase mediation is untested

The nominal mechanism predicts more than parallel positive findings. It predicts that the peptide's organismal benefit depends materially on telomerase and telomere maintenance.

No identified study blocks telomerase during exposure to the peptide and tests whether an organismal benefit disappears. No matched rescue restores the proposed pathway. No human study measures target engagement, a telomere intermediate, and a functional endpoint in one longitudinal design. No mediation analysis distinguishes the telomerase route from circadian, neuroendocrine or antioxidant actions — and the alternative routes are not hypothetical: an independent 2025 study reports a functional cellular effect of AEDG with no telomerase involvement measured [16], and the peptide's reported binding to DNA and linker histone would be expected to have consequences beyond one promoter [3,14].

What the missing experiment looks like is not a mystery, because the field has already run it for the mechanism class. In the mouse gene-therapy study, a catalytically inactive TERT abolished every benefit, which is what made the result mechanism-specific rather than merely concurrent [18]. Nothing equivalent exists for this peptide in any system.

The cell and organismal studies therefore cannot be stitched into one causal experiment. The cell papers show what the peptide can do under selected culture conditions. The animal and human papers report other outcomes in different systems. Similar direction across separate systems is hypothesis-generating; it is not mechanism-specific causal mediation.

### 7. Net benefit and safety

The claim says "extends healthspan," which is inherently a net outcome. A transient biomarker improvement does not establish more healthy life if the intervention causes offsetting morbidity or other long-term harm.

The identified human studies are too small, short, and incompletely reported to establish long-term safety. They do not provide human pharmacokinetic characterization, systematic serious-adverse-event ascertainment over an aging-relevant horizon, or a healthspan benefit against which harms could be weighed [9–11]. The 2026 regulatory evaluation states in terms that no pharmacokinetic data and no clinical safety data exist for the substances it assessed.

Where the prior edition left the benefit-and-harm discussion as an unsourced generality, four specific pieces of evidence bear on it, and none of them can move the level.

**In the cells.** The one independent cell study found that the peptide raised alternative-lengthening-of-telomeres activity roughly tenfold in one breast-cancer line and threefold in another, with elevated PML bodies in both, while telomerase activity rose only in the normal lines [3]. The authors read the normal-cell specificity of ALT activation as reassuring, and note that telomerase activity in treated normal cells stayed below the level in untreated cancer lines, which argues against acquired immortality in a single step. Both readings are available from the same data; what is not available from a two-dimensional cell-line experiment is a conclusion about human safety in either direction.

**In the mice.** Telomerase activation by gene therapy in 1- and 2-year-old mice did not increase cancer incidence relative to control littermates [18]. That is the most direct animal evidence on the specific worry, and it is reassuring for that modality, in that species, at that age of onset.

**In the rodent tumour data for this peptide.** Total tumour incidence was unchanged with a sixfold reduction in leukaemia in SHR mice [6], tumour frequency was unaffected in senescence-accelerated mice [7], spontaneous tumours were inhibited under all photoregimens in male rats [8], and tumour incidence fell in CBA mice [5]. All from one network, none designed to settle human safety.

**In human genetics.** The proposed mediator, unlike the intervention, is well instrumented. A Mendelian randomisation study across 35 cancers and 48 non-neoplastic diseases, with 420,081 cases and 1,093,105 controls, found that per standard deviation of genetically higher telomere length the odds of several site-specific cancers rose — glioma 5.27 (95% CI 3.15–8.81), serous low-malignant-potential ovarian cancer 4.35 (2.39–7.94), lung adenocarcinoma 3.19 (2.40–4.22), neuroblastoma 2.98 (1.92–4.62), bladder 2.19 (1.32–3.66), melanoma 1.87 (1.55–2.26), testicular 1.76 (1.02–3.04), kidney 1.55 (1.08–2.23), endometrial 1.31 (1.07–1.61) — while the odds of some non-neoplastic diseases fell, including coronary heart disease 0.78 (0.67–0.90), abdominal aortic aneurysm 0.63 (0.49–0.81), coeliac disease 0.42 (0.28–0.61) and interstitial lung disease 0.09 (0.05–0.15) [19]. A polygenic analysis in 472,174 UK Biobank participants estimated that at age 40, people whose leukocyte telomere length was more than one standard deviation below the population mean had a 2.5-year lower life expectancy than those at least one standard deviation above [20].

Two scope limits keep that evidence in its place. It concerns lifelong, germline-determined differences in telomere length, not pharmacological lengthening in an adult, and the two are not interchangeable exposures. And it is association-based: universal cap 2 bars it from raising any subclaim above Β on its own. What it does establish is that "longer telomeres" is not a monotonic good in humans, so the hypothesis's final step — from telomere maintenance to net healthspan gain — is a substantive empirical claim with evidence pointing in both directions, not a corollary.

SC5 stays at Γ because limited short-term exposure evidence exists in humans and partial-system harm data exist in animals, while the net long-term human effect is untested. Telomerase biology remains a reason to measure outcomes, not grounds to declare the compound either safe or dangerous.

### 8. Boundary check and cap application

The metabolism-and-aging rubric separates biomarker change, lifespan, healthspan, stress resistance, and development. Its Δ threshold requires at least one intervention changing lifespan or healthspan in a way consistent with the hypothesis, and requires that developmental effects be separated from true aging effects. Biomarker change alone does not satisfy it.

Five Γ ceilings bind:

1. **SC1:** the direct telomerase evidence is in vitro.
2. **SC2:** the telomere evidence is confined to cultured cells and is a biomarker for the human claim.
3. **SC3:** the human interventions do not measure healthspan.
4. **SC4:** the proposed mediation chain is assembled across different systems without necessity or rescue.
5. **SC5:** the human studies do not establish long-term net benefit.

A sixth ceiling comes from the second rubric. The proposed proximal step — peptide binding to DNA or linker histone, activating transcription of the telomerase catalytic subunit — falls under the epigenetics and gene-regulation rubric, whose Δ threshold requires an endogenous perturbation with demonstrated specificity against global transcriptional or cell-cycle collapse and a matched rescue, and whose Weak tier is where sequence-motif evidence sits. That rubric therefore also returns Γ. Under the cross-field rule the lowest ceiling governs; here both are Γ, so routing does not change the letter, but the second ceiling is now on the record rather than assumed.

Animal pharmacologic interventions can cross a causal threshold for a narrow animal subclaim. That does not create a contradiction with the Γ cap. The threshold applies within the system in which the intervention was run; the cap governs inference from that system to human healthspan. Developmental fly exposure, conditional rodent survival signals, and one-network replication are not a field-valid human equivalent.

Upper-ladder equivalents were also considered, and one entry from the prior edition needs a correction of emphasis rather than of fact. There is no Mendelian-randomization instrument for **exposure to the synthetic peptide**, no natural experiment in that exposure, no independent drug-class convergence on a peptide-specific effect, no human population dose-response, and no therapeutic hard aging endpoint. Instruments do exist for the proposed **mediator**, telomere length, and they are reported in §7; they cannot carry this claim because they concern a different exposure (lifelong germline variation, not administration of a compound), because they are association-based, and because they point in opposite directions for different outcomes. The low grade does not result from demanding an impossible randomized mechanism trial while ignoring equivalent evidence; the equivalents for the intervention are absent, and the equivalents for the mediator do not transfer.

### 9. Falsification ledger

Five prespecified patterns are met:

1. **Human biomarkers without healthspan.** Melatonin and clock-gene changes are reported, but no validated healthspan endpoint was measured [9,10].
2. **Development is not separated from aging.** Fly exposure occurred only from egg through larval development, and the reported effect was dose-independent across orders of magnitude [4].
3. **Central survival summaries are null or unfavourable.** SHR mice showed no mean-lifespan benefit and a slightly lower median in the treated group, with the favourable signal confined to the upper tail [6]; male rats showed virtually unchanged mean lifespan [8].
4. **The effect tracks an induced pathology rather than aging.** Rat survival signals appear in animals under disrupted lighting and are absent under a standard light cycle [8,13].
5. **Mechanistic necessity is untested.** No telomerase blockade, matched rescue, inactive-analogue control, or mediation test shows that an organismal outcome requires the proposed pathway, although the design for such a test exists in the mouse gene-therapy literature [18].

These observations do not prove that the peptide cannot affect aging. They show that the experiments required to establish the submitted human causal claim have not been completed.

### 10. Steelman and final adjudication

The strongest case for **Δ**, one rung higher, is stronger than the prior edition allowed. The peptide is a direct intervention; several animal studies report survival effects; two laboratories report telomerase-related effects in human cells and a third in bovine cells; and, decisively for the mechanism class, telomerase activation by gene therapy in one- and two-year-old mice produced healthspan improvements and a 24% and 13% median-lifespan increase with no excess cancer, an effect abolished by a catalytically inactive enzyme [18]. So the pathway the hypothesis names is not a speculative one, and it has been shown to be modifiable with benefit in a mammal.

That case is nonetheless rejected. The aging rubric requires a field-valid intervention changing the claimed lifespan or healthspan endpoint, and it requires developmental effects to be separated from aging effects. No human healthspan intervention with this peptide exists; the rodent record for this peptide is null on its central summaries, developmental in the fly, and confined to circadian-disrupted groups in the rat; and the mouse gene-therapy result is a different intervention in a different species, capped at Γ for endogenous-pathway claims by universal cap 3, with no evidence that the peptide engages that pathway in any living animal.

The strongest case for **Β**, one rung lower, is that the organismal findings are concentrated in one research network, the independent cell paper had all three core data figures replaced, the fly effect was dose-independent, the most auditable mouse result was null on mean and median, and the complete human chain has never been observed. That case is also rejected. Β is defined as reproducible correlation or co-occurrence without demonstrated causal dependency, and this corpus is not that: defined agents were applied to human cells under controlled conditions and direct molecular readouts were taken, with internal negative controls mapping where the effect stops. That is causal perturbation in a simplified system, which is Γ's definition, and cap 1 makes Γ the ceiling it cannot pass.

The final assessment is therefore:

> **Γ — Mechanistically plausible, partial system support · Low confidence**

The level and the confidence band answer different questions, and this edition separates them where the prior one did not. The level is stable: an extensive primary-source search finds no human target-engagement or healthspan study, and every accessible line of evidence stops at the same scope boundary, so Γ is where the ceiling sits and it is unlikely to move without new experiments. The confidence band is low because confidence in this framework measures evidence density and reproducibility, not the reviewer's certainty about the verdict: eleven of the twelve sources in the prior edition originate in one connected research network, no organism-level finding has been reproduced outside it, and the single independent cell-culture study required replacement of all three core data figures. The band rests on those evidence-side facts alone; that two rodent survival reports are available only as English abstracts and two primaries are unindexed scans is an access limitation recorded in REVIEWER NOTES, not a determinant of confidence. A framework that recorded "High" here would be recording confidence in its own reasoning rather than in the literature.

### 11. Role of the mechanism

The classification is **epiphenomenon**, reached at **step 2** of the framework's derivation procedure.

The procedure is answered in order and stops at the first match. Step 1 asks whether the mechanism is produced by the phenotype rather than producing it; nothing in the corpus addresses that, so it does not match. Step 2 asks whether there is no demonstrated causal dependency in either direction, only co-occurrence. At the scope this classification attaches to — human aging trajectories and healthspan — that is exactly the position: no human study measures the mechanism at all, no necessity or rescue test exists in any system, and no mediation analysis links the mechanism to a functional outcome. Step 2 matches, and the derivation stops there.

The prior edition recorded `reinforcing / stabilizing factor`, which is step 5 — the residual branch, reached only when the evidence shows that the mechanism amplifies, accelerates or maintains a phenotype initiated upstream. Its own derivation conceded that this "organism-level role remains unproven" and labelled the classification provisional. A role that the review states is unproven has not been derived, and the framework's guardrail is explicit that missing evidence does not earn benefit of the doubt.

Three things this classification does **not** assert. It does not say the peptide has no biological activity: a bounded causal dependency is demonstrated in cell culture, where exposure was followed by raised hTERT expression, raised telomerase activity, longer telomeres and additional population doublings, and that is recorded at its own scope under SC1. It does not rank the competing models; the tournament is a tiebreaker only and was not needed here. And it does not change the level — role classification never does.

### 12. Momentum, stability, and divergence

**Momentum — Mechanistic only.** The causal output on this peptide since the prior scored literature is mechanistic and in vitro: the corrected 2025 independent cell-line study [3], an independent bovine embryo-production study [15], an independent study in injured human retinal pigment epithelium that measures no telomerase [16], same-network work on senescence markers in induced neurons [21], and two reviews [17,22]. The regulatory record advanced in 2026 without any new clinical data. No registered human healthspan trial was found on 2026-09-30. Nothing identified addresses the boundary check.

**Stability — Stable.** No named pending result could be identified that measures both human telomerase target engagement and a healthspan endpoint. Under the framework, absence of a specific moving trigger makes the score Stable by definition. A registered trial of the kind described in §"Proposed decisive experiments" would make it Rising.

**Divergence — Severe overstatement.** The implied level is computed from this literature's own artifacts rather than from its tone: "Epithalon peptide induces telomerase activity and telomere elongation in human somatic cells" [1]; "Peptide promotes overcoming of the division limit in human somatic cell", concluding that the peptide "prolonged the vital cycle of normal human cells due to overcoming the Heyflick limit" [2]; "Peptides of pineal gland and thymus prolong human life" [12]; "Geroprotective effect of ala-glu-asp-gly peptide…" [8]; and, in the independent 2025 paper, "this would suggest that epitalon can be safely used in healthy individuals to maintain telomeres and thereby influence the aging process" [3]. A competent reader would infer replicated organism-level causal benefit with safety established — roughly Ζ. The assessed level is Γ. Γ → Δ → Ε → Ζ is a three-rung gap. This judgement applies to this peptide's own literature and its commercial-style downstream framing, not to geroscience as a field.

## PROPOSED DECISIVE EXPERIMENTS

### 1. Human target-engagement and healthspan trial

Run a preregistered, double-blind randomized trial of chemically verified synthetic Epitalon in an older population, with identity and purity of the material documented before the trial begins. Fix enrollment strata, tissues, endpoints, and analysis before outcomes. Measure direct telomerase activity in prospectively selected tissues, tissue-appropriate telomere dynamics, a validated multidomain healthspan or disability-free-survival endpoint, morbidity, and systematic adverse outcomes, including cancer incidence over a follow-up long enough to be informative.

The design needs placebo, blinded allocation and analysis, batch identity, pharmacokinetics, cell-composition adjustment, multiplicity control, and follow-up long enough to distinguish transient biomarkers from durable function. Adequate exposure without target engagement, or target engagement without functional benefit, would falsify the corresponding link. Replicated target engagement, telomere preservation, and clinically meaningful healthspan improvement would cross the current boundary.

### 2. Telomerase-mediation experiment

In matched human primary-cell and tissue systems, combine peptide exposure with selective TERT or telomerase-pathway inhibition and a pathway-specific rescue. Measure telomerase activity, shortest-telomere burden, alternative-lengthening-of-telomeres activity, replicative function, senescence-independent cell fitness, and off-target transcriptomic responses.

The controls that make this decisive are already established in the telomerase literature: an exposure-matched vehicle, inhibitor alone, a rescue that restores telomerase without reproducing unrelated peptide effects, and — the analogue of the inactive-TERT arm that made the mouse gene-therapy result mechanism-specific [18] — a sequence-scrambled or otherwise inactive peptide control, together with blinded replication in an independent laboratory. If the cellular effect persists when telomerase is selectively blocked, the nominal mediator is falsified. If blockade abolishes the effect and pathway-specific rescue restores it, the mechanism is materially narrowed. Because the peptide is reported to bind DNA and linker histone [3,14], the same design should test whether the proximal chromatin step is required, with specificity against global transcriptional disruption.

### 3. Independent late-life preclinical aging study

Before interpreting early-life rodent or developmental fly results as geroprotection, independent laboratories should administer verified synthetic Epitalon beginning late in life in at least two mammalian backgrounds and both sexes, under a standard light cycle and with a circadian-disrupted arm, since the existing rat signal appears only under disrupted lighting. Prespecify a primary endpoint and measure survival, frailty, physical and cognitive function, morbidity, food intake, body weight, tumor incidence, pharmacokinetics, tissue telomerase activity, and telomere dynamics.

Pair-feeding or intake adjustment is essential given the higher food consumption observed in treated mice, as are blinded outcome scoring, harmonized husbandry, and full-cohort survival analysis reporting mean and median rather than the longest-lived tail. A benefit confined to one background, sex, lighting regime, or upper tail would preserve the present context-limited interpretation. Concordant late-life healthspan benefit with target engagement and no major trade-off would materially strengthen the case.

## KNOWLEDGE GAPS

- **Human target engagement:** It is unknown whether administered synthetic Epitalon reaches relevant human tissues and activates telomerase at tolerable exposure. No pharmacokinetic data exist.
- **Human telomere consequence:** No longitudinal study establishes durable tissue-specific telomere preservation beyond measurement error and cell-composition change.
- **Validated healthspan outcome:** No study of the synthetic peptide measures multidomain function, morbidity, or disability-free survival.
- **Mechanistic mediation:** No experiment shows that the cellular, circadian, or organismal findings require telomerase, and no inactive-analogue control has been reported in any system.
- **The proximal step:** How a tetrapeptide reaches and acts on the telomerase gene rests on a sequence-complementarity model and reported DNA and histone binding; independent structural and binding work is absent, and under the epigenetics rubric this is the weakest-tier evidence in the chain.
- **Whether telomere lengthening is a net good in humans:** Human genetic evidence links longer telomeres to higher risk of several cancers and lower risk of some non-neoplastic diseases, so the final step of the hypothesis is a substantive empirical question rather than a corollary — and it concerns lifelong germline variation, not adult pharmacological lengthening, so even that evidence is scope-limited.
- **Product identity and pharmacology:** Epitalon and Epithalamin are repeatedly discussed together despite being different preparations; no cited study reports verification of the identity or purity of the material used; and human pharmacokinetics for the defined tetrapeptide are absent from the identified literature.
- **Long-term safety:** Systematic human follow-up over a healthspan-relevant horizon is absent, and no clinical safety data exist.
- **Independent reproduction:** The animal and human literature comes almost entirely from a connected research network; independent organism-level reproduction is needed, and the three independent studies that exist are all in cell culture.

## REVIEWER NOTES

### Source access

Complete full text was read for the corrected 2025 cell paper and its erratum [3], the 2003 SHR-mouse paper [6], and the 2026 FDA evaluation. The 2000 Drosophila paper [4] was read at abstract and partial-full-text depth. The remaining records were assessed from primary abstracts and metadata: [1], [2], [5], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [19], [20], [21], [22].

Two corrections to the prior edition's access record belong here. The prior edition reported complete full text for the 2003 fetal-fibroblast paper [1], the 2007 human melatonin paper [9], the 2008 male-rat paper [8] and the retinitis-pigmentosa paper [11] from scanned PDFs on a single-programme website. Those files are image-only and yielded no machine-readable text when re-checked in the red-team pass, so this edition treats those four sources at abstract-and-metadata depth and withdraws every specific that only the scans supported — the arbitrary-unit telomere values in [1], the randomisation, placebo and ten-day course in [9], and the constant-darkness attribution in [8]. Where the prior edition's numbers could be confirmed from an independently hosted primary PDF, as for [6], they are retained and one was corrected.

No exact numerical claim from an abstract-only record exceeds what that abstract supports, with one declared exception: the design of the night-shift study [10] is taken from the 2026 FDA evaluation's synthesis of the primary Russian text, and is attributed to that synthesis in the prose. Citation-context counts are omitted rather than entered as zeros.

### Search artifacts

ClinicalTrials.gov was searched on 2026-09-30 for Epitalon, Epithalon, Epithalone, Ala-Glu-Asp-Gly and AEDG peptide, and separately for Epithalamin and Thymalin: no registered study. A PubMed sweep of the peptide's names returned 167 records, of which the 2019–2026 window was examined in full for missed evidence. The FDA's evaluation of epitalon-related bulk drug substances was read as an evidence-gap and product-identity cross-check. These artifacts carry no causal grade and are not counted among the 22 numbered peer-reviewed ledger sources.

### Calibration

No prior scored dossier on this hypothesis was found and no sealed expected band was supplied, so this review is not a calibration entry. `hypothesis-index.md` carries no entry for this hypothesis; the change log recommends a stub.

### What validator passage means

The validator checks schema presence, canonical descriptor use, scope, aggregation arithmetic, cap ceilings, source identifiers, status flags, claim-to-source links, provenance counts, and required prose blocks. A pass confirms structural and internal consistency. It is not an independent replication of the biological studies or an external scientific endorsement, and it cannot detect a claim that misdescribes its source: the prior edition passed 34 of 34 checks while reporting a reversed median, a lighting regimen absent from the cited paper, and a confidence band derived from the wrong definition.

## RED-TEAM SYNTHESIS NOTE

This edition is the published output of a cross-model red-team pass. The review was authored on 2026-09-27 by a different model family; the red team and synthesis were performed on 2026-09-30 by Claude (Anthropic) against the same framework packet.

**The level did not change.** Γ survived arguments in both directions, each rejected on rubric text: Β because this corpus contains controlled perturbation with direct molecular readouts rather than correlation, and Δ because no intervention changes a claimed lifespan or healthspan endpoint at the claimed scope.

**Two score elements changed.** Confidence moved from High to Low, because the framework defines confidence as evidence density and reproducibility and every criterion of its Low band is met here. The role of mechanism moved from `reinforcing / stabilizing factor` to `epiphenomenon`, because the framework's five-step derivation matches at step 2 and the prior classification was reached at step 5 without the evidence step 5 requires.

**Every reference was re-resolved, and every identifier in the prior edition was correct** — no confabulated PMID or DOI was found, and no source was removed. Ten verified references were added. Four factual corrections were made: a reversed median lifespan in the mouse survival study, a lighting regimen attributed to a paper that does not contain it, arbitrary-unit values from a scan that cannot be re-verified, and human study design specifics the accessible record does not support. Material omitted in both directions was restored: favourable rodent tumour and genomic-stability findings, the alternative-lengthening result in cancer cells, the mouse telomerase gene-therapy benchmark, and the human genetic evidence on telomere length. A second rubric was added to the routing.

Full findings, dispositions and the citation-verification table are in `redteam/REDTEAM_REPORT.md`; the before-and-after score table, the itemised change list and the open issues are in `redteam/SYNTHESIS_CHANGE_LOG.md`.

## ADVERSARIAL REVIEW AND APPROVAL

**Authoring model:** Sol 6.1 (recorded in this document's own provenance as authored 2026-09-27 by OpenAI Codex under HSEF v5.4 and then red-teamed and rewritten by Claude on 2026-09-30; see RED-TEAM SYNTHESIS NOTE). **Adversarial reviewer:** Claude Opus 5.5 (Anthropic). **Date:** 2026-10-14. **Edition reviewed:** the 9 October 2026 revised edition, as it now stands.

**Scope of checks.** All 22 numbered PMIDs plus the erratum PMID 41240216 were resolved against PubMed and compared on first author, title, journal, year, volume, issue, pages and DOI: 23/23 resolved, every field matching, no retracted publication and no expression of concern. Thirty-one lettered sub-claims across eight sources were spot-checked (PMIDs 14501183, 11087911, 28241208, 34611362, 22585399, 40493162, 40908429, 39788414), and the whole SHR mouse paragraph — the review's densest numeric passage, and the one whose figures this edition says it corrected — was re-verified line by line against the primary PDF the reference links. Scoring logic was re-derived from `references/hsef-framework.md`, `references/evidence-rubrics/06-metabolism-aging.md` and `references/evidence-rubrics/03-epigenetics-gene-regulation.md`; `hypothesis-index.md` was confirmed to carry no entry for this hypothesis, as the review states.

### Findings

| ID | Severity | Location | Original to corrected | Evidence |
|---|---|---|---|---|
| F1 | Minor | §3, "SHR mice: null central summaries, positive tail" | "used female outbred SHR mice, 54 per group, treated from three months of age until natural death" to a sentence that also records that four mice per group were euthanised at twelve months for the cytogenetic study, so the paper's lifespan and tumour tables report 50 per group | Both numbers are correct in the source — the abstract and Methods say 54 per group, while Tables 5, 6 and 7 all state "Number of mice 50 50" — but the paragraph then quoted tumour percentages over a denominator of 50 without explaining the shift, so a checking reader meets 54 and then 16/50 with nothing to reconcile them. The percentages behind the 50 are exactly right: Table 7 prints "Number of tumor-bearing mice 18 (36%)" for saline against "16 (32%)" for Epitalon. |
| F2 | Minor | §10 "Steelman and final adjudication"; score card `final_score.confidence_rationale` (3 fields) | The confidence derivation no longer lists access depth among its determinants: the two items that were access facts — two rodent survival reports available only as English abstracts, two primaries image-only scans — are now recorded as an access limitation rather than as reasons for the band | Lead policy for this round: how much of a paper the reviewer could read is neither evidence density nor reproducibility and never justifies a band. Step two of that test was applied and the **band stays Low**, because what remains is sufficient on its own and is Low-band text verbatim — a small literature, eleven of twelve prior-edition sources from one connected research network, no organism-level result reproduced outside it, and all three core data figures of the one independent cell study replaced by erratum. Striking the invalid reason could only have raised the band, and here it does not: the valid determinants are themselves Low criteria, and the corpus is nowhere near "extensive independent replication across labs, models, and assay modalities". |

### Rejected findings (checked, no change)

- **Every number in the SHR mouse paragraph is correct, including the one this edition says it corrected.** Table 6 of the primary prints mean life span 456 ± 29 days (saline) against 455 ± 31 (Epitalon); **median 512 against 487**, so the median is indeed lower in the treated group, as this edition states and the prior edition did not; mean life span of the last 10% of survivors 709 ± 10.8 against 803 ± 15.0 with +13.3% and P < 0.01; maximum life span 739 against 830 with +12.3%. Table 7 gives tumour-bearing mice 18 (36%) against 16 (32%) and leukaemia 6 against 1. The Results text gives bone-marrow chromosome aberrations at twelve months as 8.2 ± 0.41% (saline) against 6.8 ± 0.21 (Epitalon), −17.1%, P < 0.05. Table 2 shows daily food consumption significantly higher in the treated group at 5, 7, 9, 11, 13 and 16 months while Table 1 shows no body-weight difference at any point — so the review's observation that the abstract's "did not influence food consumption" is contradicted by the paper's own table is correct, and is the kind of finding an abstract-only pass would miss.
- **Headline letter Γ and weakest-link aggregation.** Re-derived independently. The submitted wording is a conjunction, which is what weakest-link is for; all five subclaims sit at Γ, so the three rules genuinely converge and the review says so rather than hiding the dependency. The Β case and the Δ case are both argued and both rejected on rubric text, in that order, which is the v5 requirement that the one-rung-above case be written first. No index entry prescribes a different rule, and `index_rule_deviation` is correctly empty.
- **Confidence band Low.** The band itself was re-derived under the lead policy for this round and **stands**. Two of the six items the derivation listed were access facts rather than evidence facts, and those were struck (F2); what remains is a small literature, eleven of twelve prior-edition sources from one connected research network, no organism-level result reproduced outside it, and all three core data figures of the one independent cell study replaced by erratum — which is the Low band’s own text, several criteria over. Striking an invalid downward reason can only raise a band, and here it cannot: the corpus is nowhere near "extensive independent replication across labs, models, and assay modalities". The review already declines High on the correct ground, that High would record confidence in the reviewer’s reasoning rather than in the literature.
- **Role `epiphenomenon` at step 2.** The derivation is answered in order and the stopping point is justified: at the scope the classification attaches to — human aging and healthspan — no study measures the mechanism, no necessity or rescue test exists in any system, and no mediation analysis links mechanism to outcome. The cell-culture causal dependency is real and is recorded at SC1's own scope rather than being used to reach step 5. The previous `reinforcing / stabilizing factor` was reached at step 5 while its own derivation called the organism-level role unproven, which is not a derivation.
- **Momentum "Mechanistic only", Stability "Stable", Divergence "Severe overstatement".** Each carries the artifact the framework requires. Momentum names five completed mechanistic items and states that nothing identified addresses the boundary check, which is the definition of the value — not the "Decisive work underway" boilerplate this round was told to look for. Stability names the absence of a moving trigger and cites the framework's own "Stable by definition" rule. Divergence names the implied level from five quoted artifacts, states the arithmetic (Γ to Ζ is three rungs) and bounds the judgement to this peptide's literature.
- **Cap applications.** Cap 1 holds SC1 at Γ for in-vitro evidence; cap 2 bars the Mendelian-randomisation and polygenic telomere evidence from raising any subclaim above Β; cap 3 caps the AAV-TERT mouse result at Γ for endogenous-pathway claims. The second rubric's Γ ceiling on the proximal chromatin step is recorded rather than assumed, and the cross-field lowest-ceiling rule is applied explicitly.
- **"The earlier adversarial approval refers to its original edition".** Verified rather than assumed, because this clause was spliced in unconditionally across the 9 October revisions and is false in some cases. Here it is **true**: `prepublication-enrichment-2026-10-05/ADVERSARIAL_REVIEW.md` records this case as "Approved as reviewed" by Claude Opus 5.5 on 2026-10-06, with 8/8 identifiers verified and the grade unchanged. That approval's object was the dated research addendum rather than this review edition, which is what "refers to its original edition" says, so the clause stands unaltered.
- **No Consensus self-contradiction.** REVIEWER NOTES and the Search-artifacts subsection describe the PubMed sweep, the ClinicalTrials.gov searches and the FDA evaluation; neither denies provider use, so the spliced provenance block contradicts nothing. The splice itself already declines to credit a shared template sentence about an amyloid-specific follow-up as a case-specific run for this topic, which is the right disposition for that wording.
- **"Best-supported rival" renderer bug.** Not applicable: this review has no such field, and in the competing-models table the rank-1 model is the context-limited cellular response while the nominal telomerase-mediated geroprotection model sits at rank 2, with the prose naming rank 1 correctly.
- **PMID 28241208 collective author.** PubMed's author array for the Haycock Mendelian-randomisation paper begins with an empty entry standing for the Telomeres Mendelian Randomization Collaboration. Reference 19 preserves the group name ("Haycock PC, Burgess S, Nounu A, et al; Telomeres Mendelian Randomization Collaboration"), so the collective-author builder defect did not fire.
- **PubMed title brackets.** References 5, 7, 10 and 21 are Russian-language articles whose PubMed titles are bracketed; the reference lines drop the brackets. Cosmetic, and covered by the standing convention on PubMed title formatting. Reference 17's "2691" is MDPI's article number rather than a page, which PubMed confirms via the pii.
- **Validator.** PASS, 34 of 34 checks, with no failure and no warning — including the `v5.lay_summary` check that the v5.3 reviews in this round cannot satisfy. The review's own §"What validator passage means" correctly notes that a pass cannot detect a claim that misdescribes its source, which is the gap this pass closed by re-reading the SHR primary.

### Residual limitations

Two sources carry the only remaining soft spots. The Al-Dulaimi cell paper's treatment durations, the telomerase-activity dissociation in the cancer lines and the C-circle and PML assays are body-level details not present in its abstract; the review declares complete full-text access to the corrected article, and this pass confirmed that the abstract contradicts none of them, but did not re-read the corrected figures — which, given that all three core data figures were replaced by erratum, is where any further check should go. Separately, the SHR primary is internally inconsistent about the leukaemia P value: the Results text and abstract give a sixfold reduction at P < 0.01 while Table 7's footnote marks the same comparison P < 0.05. The review follows the Results text, which is defensible; the discrepancy is recorded here rather than resolved.

**Approval:** Following adversarial review and the corrections listed above, the revised document is approved. — Claude Opus 5.5 (Anthropic), 2026-10-14.

## REFERENCES

1. Khavinson VKh, Bondarev IE, Butyugov AA. Epithalon peptide induces telomerase activity and telomere elongation in human somatic cells. *Bulletin of Experimental Biology and Medicine*. 2003;135(6):590–592. [PMID 12937682](https://pubmed.ncbi.nlm.nih.gov/12937682/) · [DOI 10.1023/A:1025493705728](https://doi.org/10.1023/A:1025493705728)
2. Khavinson VKh, Bondarev IE, Butyugov AA, Smirnova TD. Peptide promotes overcoming of the division limit in human somatic cell. *Bulletin of Experimental Biology and Medicine*. 2004;137(5):503–506. [PMID 15455129](https://pubmed.ncbi.nlm.nih.gov/15455129/) · [DOI 10.1023/B:BEBM.0000038164.49947.8C](https://doi.org/10.1023/B:BEBM.0000038164.49947.8C)
3. Al-Dulaimi S, Thomas R, Matta S, Roberts T. Epitalon increases telomere length in human cell lines through telomerase upregulation or ALT activity. *Biogerontology*. 2025;26(5):178. [PMID 40908429](https://pubmed.ncbi.nlm.nih.gov/40908429/) · [PMC full text](https://pmc.ncbi.nlm.nih.gov/articles/PMC12411320/) · [DOI 10.1007/s10522-025-10315-x](https://doi.org/10.1007/s10522-025-10315-x). Published erratum replacing Figures 1–3: [PMID 41240216](https://pubmed.ncbi.nlm.nih.gov/41240216/) · [DOI 10.1007/s10522-025-10326-8](https://doi.org/10.1007/s10522-025-10326-8).
4. Khavinson VK, Izmaylov DM, Obukhova LK, Malinin VV. Effect of epitalon on the lifespan increase in *Drosophila melanogaster*. *Mechanisms of Ageing and Development*. 2000;120(1–3):141–149. [PMID 11087911](https://pubmed.ncbi.nlm.nih.gov/11087911/) · [DOI 10.1016/S0047-6374(00)00217-7](https://doi.org/10.1016/S0047-6374(00)00217-7)
5. Anisimov VN, Khavinson VKh, Zavarzina NIu, et al. Effect of pineal peptide on parameters of the biological age and life span in mice. *Rossiiskii Fiziologicheskii Zhurnal imeni I. M. Sechenova*. 2001;87(1):125–136. [PMID 11227856](https://pubmed.ncbi.nlm.nih.gov/11227856/)
6. Anisimov VN, Khavinson VKh, Popovich IG, et al. Effect of Epitalon on biomarkers of aging, life span and spontaneous tumor incidence in female Swiss-derived SHR mice. *Biogerontology*. 2003;4(4):193–202. [PMID 14501183](https://pubmed.ncbi.nlm.nih.gov/14501183/) · [DOI 10.1023/A:1025114230714](https://doi.org/10.1023/A:1025114230714) · [Primary PDF](https://www.demogr.mpg.de/publications/files/1513_1189524145_1_BG-Anis-Epitalon-SHR.pdf)
7. Anisimov VN, Popovich IG, Zabezhinskiĭ MA, et al. Effect of epitalon and melatonin on life span and spontaneous carcinogenesis in senescence accelerated mice (SAM). *Voprosy Onkologii*. 2005;51(1):93–98. [PMID 15909815](https://pubmed.ncbi.nlm.nih.gov/15909815/)
8. Vinogradova IA, Bukalev AV, Zabezhinski MA, Semenchenko AV, Khavinson VKh, Anisimov VN. Geroprotective effect of Ala-Glu-Asp-Gly peptide in male rats exposed to different illumination regimens. *Bulletin of Experimental Biology and Medicine*. 2008;145(4):472–477. [PMID 19110597](https://pubmed.ncbi.nlm.nih.gov/19110597/) · [DOI 10.1007/s10517-008-0121-7](https://doi.org/10.1007/s10517-008-0121-7)
9. Korkushko OV, Lapin BA, Goncharova ND, et al. Normalizing effect of the pineal gland peptides on the daily melatonin rhythm in old monkeys and elderly people. *Advances in Gerontology*. 2007;20(1):74–85. [PMID 17969590](https://pubmed.ncbi.nlm.nih.gov/17969590/)
10. Ivko OM, Linkova NS, Ilina AR, Sharova AA, Ryzhak GA. AEDG peptide regulates human circadian rhythms genes expression during pineal gland accelerated aging. *Advances in Gerontology*. 2020;33(3):429–435. [PMID 33280326](https://pubmed.ncbi.nlm.nih.gov/33280326/)
11. Khavinson V, Razumovsky M, Trofimova S, Grigorian R, Razumovskaya A. Pineal-regulating tetrapeptide epitalon improves eye retina condition in retinitis pigmentosa. *Neuro Endocrinology Letters*. 2002;23(4):365–368. [PMID 12195242](https://pubmed.ncbi.nlm.nih.gov/12195242/)
12. Khavinson VKh, Morozov VG. Peptides of pineal gland and thymus prolong human life. *Neuro Endocrinology Letters*. 2003;24(3–4):233–240. [PMID 14523363](https://pubmed.ncbi.nlm.nih.gov/14523363/)
13. Vinogradova IA, Bukalev AV, Zabezhinski MA, Semenchenko AV, Khavinson VKh, Anisimov VN. Effect of Ala-Glu-Asp-Gly peptide on life span and development of spontaneous tumors in female rats exposed to different illumination regimes. *Bulletin of Experimental Biology and Medicine*. 2007;144(6):825–830. [PMID 18856211](https://pubmed.ncbi.nlm.nih.gov/18856211/) · [DOI 10.1007/s10517-007-0441-z](https://doi.org/10.1007/s10517-007-0441-z)
14. Khavinson V, Shataeva L, Chernova A. DNA double-helix binds regulatory peptides similarly to transcription factors. *Neuro Endocrinology Letters*. 2005;26(3):237–241. [PMID 15990728](https://pubmed.ncbi.nlm.nih.gov/15990728/)
15. Ullah S, Haider Z, Perera CD, Lee SH, Idrees M, Park S, Kong IK. Epitalon-activated telomerase enhance bovine oocyte maturation rate and post-thawed embryo development. *Life Sciences*. 2025;362:123381. [PMID 39788414](https://pubmed.ncbi.nlm.nih.gov/39788414/) · [DOI 10.1016/j.lfs.2025.123381](https://doi.org/10.1016/j.lfs.2025.123381)
16. Gatta M, Dovizio M, Milillo C, et al. The antioxidant tetrapeptide Epitalon enhances delayed wound healing in an in vitro model of diabetic retinopathy. *Stem Cell Reviews and Reports*. 2025;21(6):1822–1834. [PMID 40493162](https://pubmed.ncbi.nlm.nih.gov/40493162/) · [DOI 10.1007/s12015-025-10911-x](https://doi.org/10.1007/s12015-025-10911-x)
17. Araj SK, Brzezik J, Mądra-Gackowska K, Szeleszczuk Ł. Overview of Epitalon — highly bioactive pineal tetrapeptide with promising properties. *International Journal of Molecular Sciences*. 2025;26(6):2691. [PMID 40141333](https://pubmed.ncbi.nlm.nih.gov/40141333/) · [DOI 10.3390/ijms26062691](https://doi.org/10.3390/ijms26062691)
18. Bernardes de Jesus B, Vera E, Schneeberger K, Tejera AM, Ayuso E, Bosch F, Blasco MA. Telomerase gene therapy in adult and old mice delays aging and increases longevity without increasing cancer. *EMBO Molecular Medicine*. 2012;4(8):691–704. [PMID 22585399](https://pubmed.ncbi.nlm.nih.gov/22585399/) · [DOI 10.1002/emmm.201200245](https://doi.org/10.1002/emmm.201200245)
19. Haycock PC, Burgess S, Nounu A, et al; Telomeres Mendelian Randomization Collaboration. Association between telomere length and risk of cancer and non-neoplastic diseases: a Mendelian randomization study. *JAMA Oncology*. 2017;3(5):636–651. [PMID 28241208](https://pubmed.ncbi.nlm.nih.gov/28241208/) · [DOI 10.1001/jamaoncol.2016.5945](https://doi.org/10.1001/jamaoncol.2016.5945)
20. Codd V, Wang Q, Allara E, et al. Polygenic basis and biomedical consequences of telomere length variation. *Nature Genetics*. 2021;53(10):1425–1433. [PMID 34611362](https://pubmed.ncbi.nlm.nih.gov/34611362/) · [DOI 10.1038/s41588-021-00944-6](https://doi.org/10.1038/s41588-021-00944-6)
21. Ivko OM, Trofimova SV, Trofimov AV, Sharkovich Z, Mogilev VA. Peptidergic regulation of expression of cellular aging marker proteins in buccal epithelium. *Advances in Gerontology*. 2024;37(5):516–524. [PMID 39742404](https://pubmed.ncbi.nlm.nih.gov/39742404/)
22. Willis J, Vu N. Synthetic peptides in pharmacy compounding: analysis of PCAC recommendations and industry safety standards. *International Journal of Pharmaceutical Compounding*. 2026;30(4):329–336. [PMID 42752426](https://pubmed.ncbi.nlm.nih.gov/42752426/)

**Unnumbered context artifacts:** [FDA Evaluation of Epitalon-Related Bulk Drug Substances, 12 May 2026, prepared for the Pharmacy Compounding Advisory Committee meeting of 23–24 July 2026](https://www.fda.gov/media/193345/download) · [ClinicalTrials.gov search, 2026-09-30](https://clinicaltrials.gov/search?term=Epitalon). These were used for evidence-gap, product-identity and registry checks, not as causal evidence.

<!-- consensus-consistency-2026-10-09 -->
### Consensus provenance reconciliation — 9 October 2026

Case-specific Consensus discovery records were located and linked to this final edition. This supplements the review's provenance; it leaves the original assessed claim, grades and source counts intact. Provider summaries and full-text badges do not establish examination of original primary passages by this reviewer. Source access and incorporation retain the limits stated in the original methods and any dated research update.

Recorded evidence artifacts:

- `deliverables/prepublication-enrichment-2026-10-05/website-series/epitalon-telomerase/consensus-browser-links.json`
- `deliverables/prepublication-enrichment-2026-10-05/website-series/epitalon-telomerase/consensus-browser.txt`
- `deliverables/prepublication-enrichment-2026-10-05/website-series/epitalon-telomerase/consensus-targeted-check.txt`

The located artifacts do not state a reliable source-comparison count. No count is inferred.

**Existing source reconciliation:** Yes; 8 identifiers; selected Methods/Results passage and correction PMID 41240216 noted; corrected figures not re-quantified. Targeted paper check covers specified cell lines/results only; correction means image-based quantitative claims need corrected-figure appraisal.

Matched primary identifiers in the preserved dated ledger: PMID 11087911, PMID 12937682, PMID 14501183, PMID 14523363, PMID 15455129, PMID 28241208, PMID 40493162, PMID 40908429. These are historical checks, not newly credited complete-paper examinations.

Shared template wording about an amyloid-specific Deep follow-up is not credited as a case-specific run for this topic. The actual raw case records above establish provider use.

The earlier adversarial approval refers to its original edition; this provenance supplement is not a new independent or blinded adjudication. No score or confidence change is made by this reconciliation.
<!-- end-consensus-consistency -->