Source hierarchy
1. UK regulator (MHRA), NICE, NHS, ClinicalTrials.gov, or a peer-reviewed journal — the only tier we treat as settled.
2. Scientific-congress materials (conference abstracts, posters) — real data, not yet peer-reviewed.
3. Company investor updates or news releases — labelled "company topline" throughout, because the company has not yet published the underlying data for independent review.
4. Reputable journalism — used for context only, never as the primary source for a factual claim.
5. Never used as a source: vendor or seller pages, social media, forums, blogs, or AI-generated summaries.
Evidence tiers, in the Record's own tags
Regulator record — a primary filing on a UK or equivalent regulator's own register.
Peer-reviewed — published in, or accepted by, a peer-reviewed journal.
Congress presentation — presented at a scientific meeting, not yet in a journal.
Company topline — from the manufacturer's own release, ahead of independent publication.
Registry record — a trial registry entry, e.g. ClinicalTrials.gov.
Corrections
When a Record entry turns out to be wrong, outdated by a newer source, or missing important context, we mark it superseded or corrected rather than quietly editing it away. Every correction is logged, dated, and left visible on the corrections page.
Review cadence
Each entry carries a checked_on date and a next_review_on date. Entries overdue for review are tracked internally and re-verified against their primary source before that date passes — not silently left to go stale.