Reading key

How to read evidence levels

Every entry in this library carries a level, and the level is about one thing only: how good the human data are. It does not measure how promising a molecule sounds, how many people use it, or how confident a seller is. This page explains what each level requires, what it leaves out, and the questions that separate a real finding from a persuasive one.

The four levels

In descending order of certainty. The scale on each card shows where that level sits.

Strong evidence01

Several well-conducted human trials agree, the studies are large enough to be reliable, and the effect shows up on outcomes people can feel or measure rather than on a laboratory marker alone. For a licensed medicine, a regulator has independently reviewed that evidence.

Replicated, large human trials

What it requires: At least one large randomised controlled trial, ideally replicated by independent groups and summarised in a systematic review.
What it does not tell you: Only for the population and the outcome that were studied. It says nothing about a different use, a much higher dose, or a decade of exposure.

In this library: Semaglutide, tirzepatide, teriparatide and bremelanotide — each strong for its licensed indication, not for every claim made about it.

Moderate evidence02

Real human trials exist and generally point the same way, but something weakens them: small numbers, one centre, a short duration, a surrogate endpoint, or results that disagree between studies. Better than a promising animal study, not yet settled.

Several human trials, some inconsistencies

What it requires: At least one randomised or otherwise controlled human trial with a positive but not conclusive result, or consistent observational data.
What it does not tell you: A moderately supported effect is often smaller in reality than the trial headline. Replication by another group is what usually decides it.

In this library: Tesamorelin, thymosin alpha-1, salmon calcitonin, cagrilintide and collagen peptides — a mix of small trials, narrow populations and inconsistent outcomes.

Limited evidence03

Human data exist, but they are small, early, indirect or inconsistent — often single-dose studies showing a hormone or a marker moved, with no trial of the outcome people actually want. Animal work may be doing much of the persuading.

Small human studies or animal data only

What it requires: Small or uncontrolled human studies, or human biomarker data without outcome trials, usually supported by animal experiments.
What it does not tell you: That a mechanism is real and measurable. It does not show a benefit, an effective dose or a safety profile for long-term use.

In this library: BPC-157, GHK-Cu, ipamorelin, SS-31 and retatrutide — genuinely studied, but not yet on the outcome the marketing promises.

Anecdotal evidence04

No controlled human data. What exists is animal or laboratory work, a plausible mechanism, and user reports. That is not nothing — it is a reason to run a trial, not a reason to believe the claim today.

No controlled human data

What it requires: Nothing beyond cell or animal studies, theory, or self-reported experience in forums.
What it does not tell you: Almost everything that matters: whether it works in people, at what dose, and what it does over time. Products in this group are also the least likely to be what the label says.

In this library: MOTS-C, KPV, DSIP, Dihexa, humanin, and every multi-peptide blend in the library.

Six traps that make weak evidence look strong

Almost every exaggerated claim about a peptide leans on one of these. Checking for them is faster than reading the study.

Animal data is not human data

A rodent study can show that a mechanism exists. It cannot give you a human dose, and most surgical-injury models used for repair peptides have no equivalent in everyday human injury.

A marker is not an outcome

Growth hormone or IGF-1 rising is a measurement, not a benefit. Trials are what tell you whether the marker moved the thing you care about — strength, healing, pain, survival.

No control group, no conclusion

If everyone in a study received the peptide, any improvement could be natural recovery, regression to the average, or expectation. Small uncontrolled studies are a starting point, not a result.

Forums are not cohorts

Thousands of self-reports feel like data, but they are selected: people who did well tend to post, doses are remembered loosely, and nobody verifies what was in the vial.

Studied for is not shown to work

A published paper proves someone tested an idea. Plenty of peptides are in the literature because they were promising, and were quietly dropped when they were not.

The molecule and the vial are different questions

The literature describes a molecule made to pharmaceutical standards. An unregulated vial has unknown purity, dose and sterility — so even a well-studied peptide is an unstudied product in that form.

Method

How entries here are graded

Browse the library

01Human evidence sets the level. Animal and cell data can explain a mechanism, but they never raise a grade on their own.
02The strongest human data decides, weighed against everything else — not the single best-sounding result.
03A regulator's approval for a specific indication is the ceiling: it supports that indication strongly, and nothing beyond it.
04Popularity, price, packaging and a certificate of analysis are not evidence and never move a level.
05Small or mixed human data lands at limited; no human data lands at anecdotal. Nobody skips a step.

Evidence is not the same as approval

A peptide can have decent human data and no marketing authorisation, or a licence that covers only one narrow indication. Those are two separate labels, and reading them apart is what stops a strong study being used to sell an unrelated product.

Approved, investigational and unapproved