Corrections policy

Anything that would leave a reader with a false impression of the evidence. In practice that means:

• a misreported number — sample size, effect size, dose, duration, endpoint, or trial phase;
• a citation that does not support the sentence attached to it, or points at the wrong paper;
• an approval or availability claim that is out of date or wrong for a jurisdiction;
• an absence claim that is no longer true, because the trial reported or the paper was published;
• a mischaracterisation — animal data described as though it were human, or an early signal written up as established;
• an undisclosed commercial relationship, or a link that misrepresents one.

Disagreement about interpretation is welcome too, and we will engage with it, but it is handled as editorial argument rather than as a correction.

Write to research@peptideworth.com. Include the page URL, quote or describe the specific claim, and say what you believe the correct position is. If you can point at the source — a PubMed record, a registry entry, a label, a regulator decision — include it, and the fix usually happens in the same sitting.

You do not need to be an expert, and you do not need to be certain. “This number looks wrong to me” is a perfectly good report; checking it is our job, not yours.

We acknowledge reports within a few business days. We then re-open the primary source ourselves rather than deciding from the report, because a correction made from a second-hand account of a study is how a second error gets introduced.

• If the claim is wrong, we correct the page and, where the error was material, mark it visibly — what the page previously said and what it says now. We do not silently overwrite a substantive claim.
• If the claim is right, we say so, and explain which source supports it. If it was right but ambiguously written, that is still a fix, and we rewrite it.
• If it cannot be resolved from the sources, the claim comes out or is rewritten to state the uncertainty explicitly. An unsupportable sentence does not get to stay on the page while we think about it.

Corrections that materially change what a page concludes are made promptly, ahead of other work. A correction is never traded away for a commercial relationship — see the disclosure.

This is the failure mode particular to how this site is built, and it does not announce itself. Three cases, three different responses:

• Retracted or subject to an expression of concern. The citation is removed as support. Any claim that rested on it is removed with it unless another source independently carries it — a retracted paper does not get quietly swapped for a similar one. Where the retraction is itself part of the story, we say that the paper was retracted and when.
• Superseded. A larger trial reports, or an early result is contradicted. The page is rewritten around the better evidence, and the older result stays visible in context rather than disappearing — how a finding changed is often the most useful thing on the page. Anything that turned on the old figure, including the evidence tier, is re-derived.
• Moved or dead. A DOI resolves somewhere new, a registry record is renumbered, a label is replaced, a URL rots. We re-resolve the reference against the database that owns it and record what it now points at. We never find-and-replace a link to make it resolve, because a link that loads is not the same thing as a source that still says what we claimed.

An update adds new information to a page that was accurate: a trial reported, an approval changed, a section got expanded. A correction means something we published was wrong. Both move the date at the top of a page, and only one of them is an admission — so we do not use “updated” as a softer word for “we were wrong”.

A date only moves when a page has actually been re-read against its sources. We do not restamp pages to look fresh; a freshness date that means nothing is its own small lie, and the editorial policy explains why the re-verification behind it is real work.