Keep a claim ledger
Research notes record the source, supported claim, document date, access date and limits of interpretation. Sources are chosen for the question they can answer: a trial registry for design, a regulator for approval, a prescribing document for labeled use and a paper for reported outcomes.
Check more than the headline
- Approval: verify the country, regulator, product, formulation, indication and decision date.
- Clinical stage: distinguish a planned study from recruitment, completed follow-up and reported results.
- Numbers: check population, dose or intervention, duration, comparator, estimand and uncertainty. Avoid mixing different analyses.
- Safety: distinguish label warnings, observed trial events and hypotheses. Do not invent event rates.
- Comparisons: identify direct randomized comparisons; describe indirect comparisons as limited.
Uncertainty stays visible
Preliminary, investigational and unknown are useful distinctions. Absence of a verified record is reported as a limit of our check, not automatically as proof that an approval does not exist. A company statement is attributed to the company. We do not invent trial identifiers, quotes, experts or statistics.
Links and editorial checks
Before release we verify references, internal links, key takeaways, metadata, author identity, mobile presentation and clinical-status fields. A link returning successfully is not by itself proof that a claim is true; the claim must match the relevant document. Editorial evidence review uses research tools and does not represent an independent clinician review.
After release
New decisions, label changes and important trial results trigger a fresh check. Readers can identify a suspected error through Contact. We assess it under the corrections policy.