How we read the evidence
This library describes what has been studied about a substance or practice. It does not tell you what to take, and it does not rank things by how well they work — because the underlying data cannot support that claim.
Where the data comes from
Each profile is assembled from published research indexed on PubMed, pharmacology references, and reports collected from public community discussions. Every claim on a page carries the kind of source behind it, because those three are not interchangeable: a meta-analysis and a forum thread can point at the same outcome and mean entirely different things.
What a number on a card means
A count is a count of records, never a score. "88 options" means 88 supplements or practices in this library are linked to that outcome — not that 88 of them work. "28 human-evidenced uses" means 28 of a supplement's recorded uses cite evidence observed in people rather than in cells or animals. Nothing on a card is a recommendation.
Why rows are ordered the way they are
Within a panel, rows are grouped by the kind of evidence behind them before anything else is considered. A row observed in adults and not ruled out always sorts above one that is animal-only or a measured null result, however many citations the weaker row carries — because an animal study is genuinely stronger methodologically than a forum report, so any single continuous score would put it first. Only inside a group does a weighted score decide, and an absent field counts as no opinion rather than a zero.
What this library does not claim
Coverage is partial and always will be. An outcome with nothing on file means nothing has been recorded here, not that nothing exists — and for safety pages that distinction matters most: an empty interaction list is incomplete coverage, never a clearance. Dosages shown are what studies used, not instructions. Nothing here has been reviewed by a clinician for your situation.
Errors and corrections
Profiles are regenerated as the source pipeline improves, and a page can change between visits. If something here is wrong, tell us — the contact link is in the footer of every page.
The labels, in detail
The same key that appears on every profile page. Probability tiers, evidence types, study grades and confidence scores, in plain language.
How to read the evidence
Evidence tier — overall strength of the evidence for the whole compound
- StrongWell established — backed by strong, consistent human research.
- ModerateSome genuine support, but mostly from small or lower-quality studies. Promising, not proven.
- PreliminaryEarly-stage evidence — a promising hypothesis, not an established effect yet.
- LimitedVery limited evidence — treat this as a hypothesis, not proof.
- InsufficientNot enough good evidence to draw conclusions yet.
Certainty — how sure we can be that the effect is real
- Very highVery sure — strong, consistent evidence
- HighFairly sure — consistent evidence from good studies
- MediumModerately sure — new studies could change this
- LowNot sure — little or mixed evidence
- Very lowVery unsure — almost no usable evidence
Source — where the data comes from
- Meta-analysisPooled analysis of many trials — the strongest evidence
- RCTRandomized controlled trial — a strong human study design
- ClinicalA human clinical study
- Reference knowledgeFrom general health-science or pharmacology references, not a study of this exact claim
- Animal researchStudied in animals, not humans
- Community reportsAggregated user reports — anecdotal signal, not evidence
- TheoreticalMechanism-based reasoning, not yet tested
Study grade — quality of each underlying study (A best, D weakest)
- AHigh-quality — strong, consistent human trials
- BGood — solid evidence with some limitations
- CModerate — limited, small, or mixed-quality studies
- DWeak — preliminary, indirect, or animal-only
Confidence — how strongly the sources back a finding
A 0–100 score from study quality and agreement
Most supplement evidence is graded C–D. Read it as preliminary signal, not proof.