How we grade the evidence

• StrongConsistent trials6 claims

  Several good randomized trials or a high-quality meta-analysis agree, in people like the reader, at amounts people actually use.

• ModeratePromising, with caveats19 claims

  Randomized trials point the same way, but they are small, short, or mixed in quality, so the size of the effect is uncertain.

• LimitedEarly or mixed48 claims

  A few small trials, conflicting results, or evidence mostly from lab, animal or observational studies. Worth knowing about, not worth relying on.

• Not supportedTested and failed, or untested20 claims

  Good trials found no benefit, or the claim has never been tested in people. Popular is not the same as proven.

• Human studies first. Randomized controlled trials, and systematic reviews and meta-analyses that pool them (including Cochrane reviews), carry the most weight. Observational studies can suggest a link but cannot show cause, so on their own they earn no more than limited.
• Lab and animal research is background, not proof. A compound that calms inflamed cells in a dish may never reach those cells in a person at a dose anyone eats. We describe mechanisms where they help explain a result, and say plainly when that is all there is.
• Amount and form matter. Most herb trials test capsules or extracts. We say what was tested, and when the kitchen version has not been studied, we say that too.
• Authoritative summaries. NIH Office of Dietary Supplements fact sheets, the National Center for Complementary and Integrative Health, FDA, MedlinePlus and clinical guidelines are used for safety, interactions and context.

Food is not a drug, and this site does not say that any food or herb cures, treats or prevents a disease. When trials looked at people with a condition, we describe who was studied, what they took and what changed. If you have a condition, the right next step is your clinician; see the medical disclaimer.

Each page shows when it was last reviewed. If you know of a study we missed, or think a grade is wrong, send it through the contact form. We publish corrections on the page.