How we grade evidence, and what would change our mind
A grading system is only worth something if you can predict what it will say before you read it, and if the people using it have written down what would make them wrong. This page is that commitment.
- Seven evidence grades
- Nine regulatory standings
- Four review stages
Evidence grades
Seven grades, and the bar for each
Grades describe the strength of research. They are never a recommendation, and they say nothing about whether something is right for any individual.
Established
Supported by substantial human evidence for the stated context.
- What earns this grade
- Multiple adequately powered human trials, at least one of them randomised and controlled, with consistent direction of effect and independent replication.
- What would move it down
- A large well-conducted trial contradicting the effect, or a systematic review finding the earlier evidence was inflated by bias.
Tested, no benefit
Adequately tested in humans and did not show the claimed benefit. This is knowledge, not missing evidence — a different thing from Insufficient.
- What earns this grade
- An adequately powered human trial, or a consistent body of them, that tested the claimed benefit and did not find it. The programme has to have been capable of detecting an effect — a small underpowered study finding nothing is Insufficient, not this.
- What would move it down
- A better-designed trial in a population or at an exposure the earlier work did not cover, finding a real effect. Rare, but it is the only thing that should move this.
Promising
Positive human evidence, but limited in scale or replication.
- What earns this grade
- At least one adequately powered human trial with a positive result, but limited replication, limited duration, or a narrow population.
- What would move it down
- Failed replication, or a phase 3 result that does not reproduce the phase 2 effect.
Preliminary
Early or exploratory human evidence only.
- What earns this grade
- Small, exploratory or open-label human evidence. Enough to justify further study; not enough to justify a claim.
- What would move it down
- Nothing new for an extended period, or methodological problems found on close reading.
Preclinical
Primarily animal or laboratory evidence. No adequate human data.
- What earns this grade
- Animal or laboratory evidence only. No adequate human trial exists, regardless of how much is written about the compound online.
- What would move it down
- Nothing — this is a floor. It rises only when real human data appears.
Insufficient
Evidence is too weak, too small or too inconsistent to judge.
- What earns this grade
- Evidence exists but is too weak, too inconsistent, too old in method, or too dependent on a single unreplicated source to support any conclusion.
- What would move it down
- Not applicable — this grade already means we cannot judge.
Safety concern
Meaningful unresolved safety or regulatory concerns.
- What earns this grade
- A meaningful unresolved safety or regulatory concern, applied regardless of how strong the efficacy evidence is.
- What would move it down
- Resolution of the concern through adequate study or regulatory action.
Regulatory vocabulary
Nine standings, and why we refuse to collapse them
The most common way health content misleads people is by treating a regulatory process step as an endorsement. These words mean different things and we use them precisely.
- Approved
- A regulator has authorised a specific product, from a specific manufacturer, for a specific indication, in that country. Nothing broader.
- Approved — other indication
- Approved, but for something other than the use being discussed. Off-label use is a clinical decision, not an approval.
- Under review
- A submission is being assessed. The outcome is unknown.
- Under consideration
- A committee is discussing or has nominated the substance. This is the weakest possible signal and is routinely misreported as progress.
- Recommended
- An advisory body has recommended something. Regulators are not bound by advisory recommendations.
- Listed
- Appears on a list — a compounding category, a cosmetic ingredient inventory. A list is not a drug approval.
- Not approved
- A regulator has declined, or has acted against the product.
- Restricted
- Permitted only under specific conditions, licences or pathways.
- Unapproved / investigational
- No marketing authorisation. Legitimate only inside an authorised clinical trial.
Every regulatory statement on this site carries a jurisdiction and a date. Approval is always bound to a formulation, a manufacturer and an indication — we never write that a molecule is “approved” on its own, because that sentence has no meaning.
Where evidence comes from
We do not treat one agency's silence as a verdict
A large share of peptide research was done in Russia and the former Soviet Union, and some of it supports medicines registered there today. Judging that work by whether the FDA has looked at it would not be caution — it would be a category error.
Absence of an opinion is not a negative finding
The FDA has never evaluated most compounds in this library. That tells you nothing about whether they work. We report what each regulator has actually said, and say plainly when a regulator has said nothing at all.
Registration elsewhere is real registration
Semax and Selank are registered medicines in the Russian Federation. Thymosin Alpha-1 is approved in a number of countries. Those are genuine approvals by genuine regulators, and they appear in the regulatory table as such — not as 'unapproved' with a footnote.
The methodological bar does not move
Randomisation, blinding, power and independent replication are assessed the same way whatever the country of origin. Being even-handed means applying one standard to everyone, not a lower standard to work we find harder to check.
Hard to find is not the same as absent
Much Russian clinical literature is published in Russian, is not indexed in PubMed, and has never been through Cochrane-style synthesis. That is a discoverability and replication limitation, and we label it as exactly that — not as evidence of nothing. Where we could not read a primary source, the dossier says so.
The cost of getting this wrong runs both ways. Dismissing a decades-long research tradition because it is unfamiliar loses real information. Accepting a weak claim uncritically because rejecting it would feel parochial loses more. Where a body of work is genuinely thin, we say so — and we say the same about work from anywhere else.
Editorial process
Nothing publishes without four sets of eyes
- 01
Draft
A named writer produces the dossier or article against a fixed structure.
- 02
Scientific review
A reviewer with relevant expertise checks every claim against the cited source.
- 03
Medical review
A clinician checks safety framing, interactions and anything that could be read as advice.
- 04
Compliance review
A check for dosing guidance, treatment claims and jurisdictional accuracy.
- 05
Publish
With named author, named reviewers, review date and a change log that starts immediately.
Commitments
What we will not do
- Publish dosing, protocols or administration instructions for compounds that are not approved.
- Present a committee discussion, nomination or listing as though it were an approval.
- Use benefits language on a compound whose evidence base is animal data.
- Publish testimonials or before-and-after claims.
- Let a commercial relationship determine an evidence grade, or go undisclosed.
- Publish a health claim without a named reviewer and a source you can check.
- Quietly edit a page after a mistake — corrections are logged and dated.
Think we have got something wrong? Our corrections process · Who reviews our work