Evidence methodology

Every compound profile grades what has been published about the compound into evidence levels. This page describes what each level means, how a level is assigned, and what has to happen before any of it reaches the public site. The editorial policy governs the language used around evidence; this page governs the grading itself.

The levels

Levels are ordered by how much they support an inference about human outcomes, from weakest to strongest. They are reported separately and never merged: a compound with strong animal data and one small human study is described as exactly that, not averaged into a middle grade.

• Evidence not yet assessed — The published literature for this compound has not yet been reviewed by an editor. No conclusion about evidence quality should be drawn from this label.
• In-vitro evidence — Findings come from cells, tissues or biochemical systems studied outside a living organism. Such findings describe what a compound can do in a controlled system, not what it does in a person.
• Animal evidence — Findings come from studies in living non-human animals. Species differences in physiology, metabolism and dosing mean animal results frequently do not carry over to humans.
• Early human evidence — Findings come from small, early-phase or observational human studies. These are typically short, sparsely replicated, and designed to examine tolerability or signal rather than to establish effect.
• Established clinical evidence — Findings come from adequately powered, replicated controlled trials in humans, generally supported by regulatory assessment or systematic review.

How a level is assigned

• Grading is done per body of findings, not per compound. Each evidence section of a profile — in vitro, animal, early human, clinical — carries its own citations, and a statement sits in the section its supporting studies belong to.
• The level reflects the study type of the cited work, not its conclusion. A promising result in cells is in-vitro evidence however striking the finding.
• A section with nothing behind it says so explicitly. “Evidence not yet assessed” means no editor has reviewed the literature for that record — it is a statement about this site’s progress, not about the compound.
• Citations follow the sourcing rules of the editorial policy: checkable identifiers, unverified citations visibly marked, access dates on sources that change.

When a level changes

• A level moves when a scientific re-review finds published work that changes the picture — new trials, a retraction, a failed replication. The profile records the date of its last scientific review.
• Levels move down as readily as up. A retracted or unreplicated study is removed and the grading re-assessed against what remains.
• An error in a grading is handled under the corrections policy: assessed against the cited source, corrected with a dated note where upheld.

Sign-off

No profile reaches the public site on grading alone. Approval requires a completed scientific review, a completed regulatory review with a cited source, a written safety interpretation and at least one reference — and the build enforces it: a record that has not been approved does not generate a page. The full requirements are set out in the editorial policy.