Version 1.0
The grading rubric, in full
Every tier badge on this site links here. A grade whose derivation you cannot check in one click is an assertion, not a rating.
Last updated
The seven rungs
Ordered strongest first. A claim sits on the highest rung for which at least one record qualifies.
human RCT
A prospective trial in human participants with randomised allocation and a control arm, reporting the outcome this claim is about.
Qualifies
- Allocation to groups is randomised and made before the outcome is measured.
- There is a control arm — placebo, an active comparator, or documented standard of care.
- The outcome supporting this specific claim is a reported endpoint, not an incidental observation.
- Participant number and effect size are reported, with a measure of uncertainty.
- Trial registration is recorded, or its absence is stated on the record.
Does not qualify
- A single-arm study, however large, and however the paper describes itself.
- Allocation decided by clinician or participant preference — that is observational, rung 2.
- A trial of a different route, formulation or salt than the claim describes.
- A conference abstract with no published methods. It caps at rung 2 until the full paper exists.
- Post-hoc subgroup findings presented as trial endpoints.
What this rung does not license you to say
A rung 1 claim about one outcome says nothing about any other outcome. A trial establishing glycaemic control does not grade a claim about hair growth, muscle retention, or anything else the same molecule is sold for. This is the single most common misreading in this category.
For example Semaglutide and body-weight reduction: multiple randomised, placebo-controlled trials with pre-registered weight endpoints.
human observational
A study of humans in which exposure was not assigned by the investigator, but a comparison is available — between groups, or against a documented baseline.
Qualifies
- A defined human population with stated inclusion criteria.
- A comparison group, or a within-subject baseline measured before exposure.
- Follow-up duration and loss to follow-up reported.
- Confounders named, whether or not they were adjusted for.
Does not qualify
- No comparison of any kind — that is a case series, rung 3.
- Self-reported use with no verification of what was actually taken. Grey-market compounds frequently are not what the label says.
- Surveys run by a seller on their own customers, at any sample size.
- Aggregated forum or social-media posts. See 'What is not evidence' below.
What this rung does not license you to say
Association is not effect. A rung 2 claim may never be phrased causally on this site — not 'improves', not 'reduces', not 'causes'. The wording must survive the possibility that the association is entirely confounded.
For example A registry cohort reporting outcomes among people prescribed a GLP-1 agonist, compared with matched non-users.
human case series
One or more documented human cases reported in the literature, with individual-level detail and no control group.
Qualifies
- Clinician-documented and published, with case-level detail.
- Dose, route and duration reported for each case.
- Outcome assessment described, even if not blinded.
Does not qualify
- Anecdote without clinical documentation, wherever it appears.
- Testimonials, influencer accounts, and vendor-published 'results'.
- Community discussion threads, wherever they are hosted. Attributed discussion is not clinical documentation.
What this rung does not license you to say
No frequency estimate may be derived from a case series. There is no denominator, so a rate cannot exist. If you see a percentage attached to a rung 3 claim anywhere on this site, it is a bug — report it.
For example A published series of four patients treated off-label, reporting individual dose, duration and outcome.
animal in vivo
A study in live non-human animals, with a control group, reporting the outcome this claim is about.
Qualifies
- Species, strain, group sizes and sex reported.
- A control or sham group.
- Dose reported in mg/kg with the route stated.
- Exposure duration reported.
Does not qualify
- Fewer than three animals per group.
- No control or sham group.
- Dose given only as a total mass with no body weight, so it cannot be scaled or compared.
What this rung does not license you to say
AN ANIMAL DOSE IS NOT A HUMAN DOSE. Allometric scaling to a human-equivalent dose produces a plausible starting point for designing a trial — not a recommendation, and not a safe dose. Rodent healing and regeneration models in particular have a long record of not translating to humans. Where this site shows a human dose range extrapolated from animal data, it says so on the record and it is not a rung 1, 2 or 3 claim.
For example Tendon-to-bone healing in a rat model, treated versus sham, with histological scoring.
in vitro
A study in cells, tissue or a cell-free preparation, outside a living organism.
Qualifies
- Concentration reported in molar or mass units, with the assay named.
- A control condition.
- Cell line or tissue source identified.
Does not qualify
- Concentrations that could not be reached in a living body, unless that limitation is stated on the record alongside the finding.
- No control condition.
What this rung does not license you to say
No dose, no route, and no bioavailability may be inferred. A concentration in a dish says nothing about what a vial delivers to a tissue. In-vitro activity is a reason to run a study, not a reason to take something.
For example Receptor-binding affinity measured in a cell-free assay, with a reference ligand as control.
theoretical
An inference from established biology to a claim about this compound, where no study has measured the claimed outcome for this compound.
Qualifies
- The mechanism being reasoned from is itself established at a higher rung, and that source is cited.
- The inference chain is written out explicitly.
- The step that has not been measured is named on the record.
Does not qualify
- Reasoning that originates from a seller's product page or marketing copy.
- Class inference presented as compound-specific — 'it is a GH secretagogue, therefore it does X' is not a claim about this molecule.
- A chain of inferences where any link is itself only theoretical.
What this rung does not license you to say
Nothing about magnitude, direction in a living body, or safety. A theoretical claim is a stated hypothesis with its author's reasoning attached, and it is rendered as one.
For example A compound binds a receptor whose activation is known to raise IGF-1; no study has measured IGF-1 after administering this compound.
computational
A prediction generated by a model, with no wet-lab or clinical measurement of the claimed outcome.
Qualifies
- Method and model named and versioned.
- Validation status stated — including 'not experimentally validated'.
- The prediction is reported as a prediction on the record.
Does not qualify
- An unvalidated prediction presented as a finding.
- Output from a model whose training data or method is undisclosed.
What this rung does not license you to say
A docking score is a hypothesis. It is the weakest thing on this ladder and it is on the ladder only so that predictions can be recorded honestly rather than quietly upgraded.
For example A molecular-docking study predicting binding affinity, with no binding assay performed.
Reported, not measured
Three tiers sit below the ladder, and they are not rungs. Nothing here was measured for this compound: it was summarised by someone else, reasoned across from adjacent data, or circulated among people using it. They exist so that a figure a reader has already seen elsewhere can be found here, next to what actually stands behind it — which is often nothing. A tier in this band never outranks a rung, and never enters a ranking at all.
review or editorial
A review article, narrative synthesis, editorial, letter or comment, indexed as such by its publisher. It may be authoritative and it may be wrong; either way the evidence it describes belongs to the studies it cites, which are graded on their own merits.
Qualifies
- Indexed by MEDLINE or the publisher as a review, editorial, letter or comment.
- Attributed to named authors in a named publication.
Does not qualify
- A systematic review or meta-analysis, which pools primary data and is graded at rung 2.
- A review presented as though it were the study it describes.
What this tier does not license you to say
A review may never be cited as the evidence for an effect. If a review is the strongest thing on file for a claim, the honest statement is that no primary study has been located — not that the claim is supported.
For example A narrative review of GLP-1 agonists in a specialty journal, summarising trials that are each graded separately here.
extrapolated
A figure carried across from a related compound, a different species, a different route, or a dose-scaling calculation. The reasoning is stated and the source of the original measurement is named.
Qualifies
- The measured source is cited, and what was actually measured is stated.
- The step taken is named — species, route, analogue or allometric scaling.
Does not qualify
- An extrapolation whose origin cannot be traced to a measurement.
- A number carried across silently, so the page reads as though it were measured here.
What this tier does not license you to say
An extrapolated figure may never be presented as a measurement, and may never be the basis of a dose shown to a reader without the step being visible beside it.
For example A human-equivalent dose calculated by body-surface-area scaling from a rat study, with the rat dose and the Km factors shown.
unverified community data
A dose, schedule, combination or claim that is stated widely in community guides, vendor protocol pages, forums or seller dosing calculators. It is recorded as a fact about what is being said, and it is set against what the corpus actually holds.
Qualifies
- The claim is stated on sources that can be linked, and it is quoted rather than paraphrased.
- How widely it circulates is stated.
- What the corpus holds on the same point is stated beside it — including 'nothing on file'.
Does not qualify
- A count, an average, a percentage or a ranking derived from community posts. There is no denominator and no verification; a number would manufacture authority this tier does not have.
- A claim recorded without what the record says, which turns a contrast into an endorsement.
- Anything sourced from this site's own journal or forum. A user's private record is never counted, averaged or rendered onto a compound page.
What this tier does not license you to say
Nothing. This tier licenses no claim about any effect, dose or safety, and it never enters a ranking, a comparison or a recommendation. Its only function is contrast: a reader who has seen a number elsewhere can find out here what stands behind it, which is frequently nothing.
For example A 250-500 mcg twice-daily BPC-157 protocol repeated across community guides, shown beside the record: one rat study at 10 mcg/kg, and no human trial.
Tiers attach to claims, never to compounds
This is the rule the rest of the rubric depends on, and it is where almost every other peptide reference goes wrong. A compound does not have an evidence level. Each individual claim about it does.
Take a compound with a large randomised trial for one indication and nothing but a rodent study for another. A single page-level badge has to pick one, and either choice is a lie: it overstates the second claim or understates the first. Our nearest competitor prints one bare word for an entire 3,700-word page.
So a record here routinely shows a rung 1 claim and a rung 6 claim side by side, and the compound has no overall grade at all. What we show instead is the distribution — how many claims sit on each rung — because that is a real property and a single badge is not.
The rung is the strongest qualifying record, not an average
A claim sits on the highest rung for which at least one record qualifies. It is not an average, a score, or a vote count. Three animal studies do not add up to a human trial, and averaging would let volume substitute for quality.
The supporting records at lower rungs are still listed on the claim, in rung order, so you can see the whole basis rather than only its strongest part.
What is not evidence here, at any rung
These never enter the ladder. They may be recorded and labelled as what they are — see Reported, not measured, which is where a labelled record of one of them goes — but they never grade a claim, never outrank a rung, and never enter a ranking:
- Seller product pages, marketing copy and certificates of analysis. A COA describes one batch's purity. It is evidence about a vial, not about an effect.
- Scraped social media and community posts. Aggregating them into a count produces an authoritative-looking number with no denominator and no verification. Our nearest competitor does this and then disclaims it. The unverified community data tier records a circulated figure as one named, attributable thing so it can be shown beside the record; it never counts, averages or ranks anything, which is the difference.
- Press releases and pipeline announcements with no accompanying paper.
- Patents. A patent claims utility; it does not demonstrate it.
- AI-generated summaries, including our own. A summary inherits its source's rung and never improves on it.
“Not on file” is not a rung — and it is not the same as “no effect”
765 fields on this site read not on file. That means no record meeting any rung's criteria exists for that field. It is a statement about the literature, not a gap in our data entry, and it is browsable at the gap ledger.
It must not be read as evidence of absence. "No study has measured this" and "a study measured this and found nothing" are completely different statements, and the second is a graded claim like any other — a well-powered trial reporting no effect is a rung 1 claim, not a gap.
Where a field is empty, the record also names what evidence would fill it and at what rung, so absence is actionable rather than decorative.
When a record is downgraded, capped or removed
- Wrong route, formulation or salt. Evidence for an oral formulation does not grade a claim about a subcutaneous one. Either the claim is narrowed to match, or the record does not apply.
- Abstract only. Capped one rung below what the design would otherwise earn, until full methods are published.
- Retraction or expression of concern. The record is removed from the claim. Claims that depended on it are re-graded, and may fall to not on file.
- Undisclosed sponsor conflict. Recorded on the claim. It does not by itself change the rung, because study design determines the rung; it is shown so a reader can weigh it themselves.
When records disagree
The strongest rung stands, and the disagreement is displayed rather than resolved silently. If a human trial finds no effect where animal work found a large one, the claim reflects the trial and the animal record stays visible beneath it with its own rung.
Hiding the loser would make the record look tidier and be less true. A reader who only sees the winning record cannot tell a settled question from a contested one.
Challenging a grade
There is no button for this, and no ticket queue behind one. A grade is challenged by writing to [email protected] with the compound, the rung you think is wrong, and the citation you are reading it against. Naming the citation is what makes a challenge answerable: a rung is a statement about one study's design, so it is settled by reading that study, not by weighing opinions.
Rungs are derived from the study design recorded in the cited literature. Where a citation does not state its design clearly enough to place it, the claim is left ungraded rather than guessed at.
Changes to what a rung means are recorded in the revision history below, because every graded claim on this site inherits these definitions and a silent edit would retroactively change thousands of published grades. There is no separate feed of individual grade corrections: a re-graded record carries its new rung, and the citation that rung came from is on the record for you to check.
Revision history
Every change to a rung's meaning is recorded here. Grades already published inherit these definitions, so a silent edit would retroactively change what thousands of existing grades mean.
| Version | Date | Change |
|---|---|---|
| 1.0 | 2026-08-11 | First published. Seven rungs, per-claim grading, the exclusion list, and the separation of “not on file” from “no effect”. |