Methodology

Everything on this site follows from two rules: grades are per claim, never per compound, and the species behind the evidence is never hidden. This page is the full contract — how grades are assigned, what can move them, and who is accountable.

The evidence grade — how much do we know?

Each claim (a compound → outcome pair, like "creatine → muscle mass") carries one letter grade. The grade measures the state of knowledge, not how good the compound is:

Multiple human RCTs, consistent direction, independently replicated, hard endpoints.

At least one well-conducted human RCT, or consistent human observational data with a plausible mechanism.

Human data only, but small, short, mixed, or limited to surrogate biomarkers.

Animal or in-vitro only. No human evidence for this claim.

Tested in humans and found not to work, or contradicted by the better studies.

A grade never collapses across claims. A compound with an A for one outcome and a D for another is shown exactly that way, because that is the state of knowledge.

Effect size — if it works, how much?

Shown separately as large / moderate / small / negligible / unclear, and never merged into the grade. A compound can be grade A with a negligible effect — rigorously proven to barely matter — and that combination is one of the most useful things this register can tell you. Where trials report standardised effect sizes we use them; where they don't, the call is the reviewer's and the studies to check it against are listed on the claim.

Flags

Small, factual annotations shown on both claims and individual studies:

industry funded: The study was paid for by a party selling the compound.
biomarker only: Measured a proxy, not an outcome anyone feels.
rodent only: The evidence for this claim comes from rodents.
dose mismatch: The studied dose is far from what supplements actually contain.
safety note: Documented upper limit, contraindication, or interaction class.
under review: New evidence has arrived and the grade hasn't been reassessed yet.

What we record per study

Species · design · sample size · duration · dose and form · outcome measured · funding source · trial registry ID · link to the primary source · the reviewer's note on what this study contributes to the claim. Every one of these is displayed, not summarised away. Mouse results and human results never share a visual treatment.

Grade movement

Every claim's grade is dated and carries its history. When new evidence moves a grade, that change is a first-class event: it appears on the compound's timeline, on the home page, and in the newsletter. The newsletter sends regrade events only — no digests, no news feed.

Who reviews, and corrections

Each compound page names its reviewer and the date of last review. Records marked "draft — unverified" have not yet been through full editorial review and should be read accordingly. If you find an error — a mis-cited study, a wrong number, a grade that doesn't survive contact with the literature — write to the address in the footer. Corrections are made in the open: material changes appear in the grade timeline, and the full edit history of every record is public in the site's source repository.

Funding and conflicts of interest

Evidence Stack sells nothing, reviews no brands, and carries no affiliate links or sponsorships. Product records exist only to compute cost per gram and are facts from labels, not endorsements. If that funding model ever changes, it changes on this page first.

The data is citable

Every record on this site is available as JSON: /api/compounds.json for the register index, and /api/compounds/<id>.json for a full compound record including claims, studies and grade history. No key, no rate limit beyond fair use. If you build on it, a link back is appreciated.

What this site is not

A summary of published research, however careful, is not medical advice. A grade measures evidence, not what you should do — that depends on your health, your medication, and your priorities, none of which this site knows.