How we score the evidence
The score answers one question: how far up the evidence ladder has this remedy climbed for this condition?
| Score | What it means | How our headline starts |
|---|---|---|
| 1/5 Early | Only cell or test-tube studies. Many things that work in a dish never work in a body. | "Lab studies suggest …" |
| 2/5 Preliminary | Animal studies, and at most single case reports in people. No controlled human trial on this outcome. | "Animal studies and case reports suggest …" |
| 3/5 Limited | Some human evidence: small trials, trials on a related outcome, or population studies that disagree. | "Small or mixed human studies suggest …" |
| 4/5 Moderate | At least one well-run randomized trial in people on this outcome, with a positive result. | "Based on the evidence, …" |
| 5/5 Strong | Several consistent randomized trials, or a meta-analysis of them, on this outcome. | "Based on strong evidence, …" |
Where we look
Search engines rank pages, not evidence, so a lot of what is known about a remedy never reaches the first page of results. We search these sources directly:
- PubMed: the US index of published medical research; every human study read by us
- ClinicalTrials.gov: registered trials, including finished trials that never published results
- Drug codes and trade names: the same compound tested as a drug under another name
- FDA adverse-event reports (CAERS): what people and doctors reported to the FDA about the supplement
- Patent offices (via Europe PMC): experiments and claims filed as patents, often never published
- Preprint servers (via Europe PMC): studies posted before or instead of peer review
- NIH RePORTER: who is being paid to study it right now
- Public patient forums: first-person reports, every outcome counted
Being added next: WHO trial registry (China, India, Iran, Japan, EU registries); Non-English journals (J-STAGE, KoreaMed, SciELO, CyberLeninka, Chinese abstracts); Theses and dissertations; Conference abstracts; Traditional-use records (USDA ethnobotany databases, historic pharmacopoeias); Archived forums that no longer exist (Wayback Machine).
We do not use the dark web: what is there on medicine is mostly drug markets and fraud, and it cannot be checked.
How we find studies
We search PubMed, the US National Library of Medicine's index of medical research, for each remedy and condition, and link the exact search on every page. We remove papers that only share a word, such as the radioactive "phytate" tracer used in surgery. PubMed's labels sort the rest into lab, animal, review and human studies. Those labels are often wrong, so we read every human study ourselves and record what was measured, how many people took part, and the result. Only studies we have read move the score above 2.
How we count public reports
We read public first-person posts on forums and social media. We count a post only if the writer, or someone caring for a specific person, took the remedy and states an outcome. We count every outcome: helped, no effect, got worse, mixed or too early to tell, and every side effect mentioned. We never record names or usernames, and we link each report. Reports never change the score.
Wording rules
The headline's first words are fixed by the score, so a page can never claim more than its evidence. We never call anything a cure. A positive result in people needs a randomized trial before the word "effective" appears.
Money
If we ever earn from a product link, it will be labeled, it will only appear on remedies scoring 3 or higher, never on cancer pages, and it will never change a score.