Corrections

How corrections work

When we find an error — or you report one — we fix it on the page, update the last-reviewed date, and add an entry below describing what changed and when.

A citation that does not support the claim beside it is a failure of our process. Tell us and we will correct it.

What counts as a correction

A correction is a change to something the page asserted that was not true: a claim the cited paper does not support, a number reported wrongly, a study described in the wrong species or the wrong population, a regulatory status stated inaccurately, or a statement that no evidence exists when evidence does. Those get an entry here, permanently.

An update is not a correction. Adding a newly published trial, rewriting a passage for clarity, or extending a page with a question it did not previously answer changes what the page covers without any earlier version having been wrong. Those change the last-reviewed date and are not logged below, because a log that records every edit records nothing.

What we will not do

We do not remove a correction once it is published, and we do not edit an error away without recording it. A page that quietly changed what it claimed would be indistinguishable from a page that had been right all along, which is the whole reason this page exists.

We also do not correct a disagreement. If a source says something we think is wrong, we report what it says and, where it matters, say what weighs against it. Our opinion of a paper is not a correction to our page.

How to report an error

Send the page, the sentence, and the paper. Contact us with all three and the fix is usually quick, because the claim is already tied to a specific quotation from a specific source and we can check it directly. A report that names the page without naming what is wrong with it takes considerably longer, and sometimes cannot be acted on at all.

The fastest corrections we have made came from someone pointing at a paper we had missed. That is the most useful thing anyone can send us.

Log

5 October 2026 — Natriuretic peptide and VIP: the medicines were missing

The natriuretic peptide and VIP pages were written from source packs that held almost nothing on the administered forms of either hormone, and several statements were wrong as a result. Natriuretic peptide said no one takes one and named carperitide as the only natriuretic peptide medicine; nesiritide, recombinant B-type natriuretic peptide, was approved in the US in 2001 for acutely decompensated heart failure, and its 7,141-patient outcome trial is now reported on the overview, safety, benefits and dosage pages. VIP said no VIP product was approved, that every human exposure was a single session, and that no study had looked at erectile dysfunction; the VIP-plus-phentolamine injection was approved in New Zealand and the UK in 2000, trials ran multi-day infusions and inhaled courses, and a 308-man series exists. Both packs were expanded with the trial literature on nesiritide and aviptadil, and all eight pages were rewritten from it. Found by our own review of which compounds lacked their drug forms.

5 October 2026 — GLOW comparison: registry entries miscounted

GLOW compared listed five registered trials across the blend’s three compounds — one for BPC-157, two for thymosin beta-4, one for the TB-500 fragment and one for GHK-Cu — and then summed them as “four entries”. It now says five. Found while reviewing the page for an evidence table.

5 October 2026 — Randomised trials labelled as smaller studies

Claims drawn from nine randomised trials, on 25 pages, carried a weaker evidence label than the trials deserved — “human pilot / early trial”, or in one case “primary research”. The largest group sits on the semaglutide and tirzepatide comparison pages, from the head-to-head trial of the two drugs in obesity; the others are on the Melanotan, Matrixyl, C-peptide, thymosin alpha-1, Cerebrolysin and larazotide pages, on the researched-effects overview and on peptides for skin. The labelling rule recognised a randomised design only when the title or the index entry named it, and these papers state it in the abstract. It now reads the abstract too, counting only a trial the paper itself ran — not a protocol, a drug overview, or a sentence saying such trials are needed. This understated the evidence rather than overstating it, and is logged because a label that misdescribes a study is an error either way.

5 October 2026 — Laboratory results labelled as human evidence

Four claims drawn from laboratory work were labelled as human studies. On Collagen and Collagen benefits, a finding from human skin explants — skin tissue kept alive in a dish — carried the label of the clinical trial it was published alongside. On Copper peptide benefits, a skin-delivery measurement made on excised skin was labelled a human pilot study. On Who should not take peptides, a result in cancer cells taken from two patients and grown in a dish was labelled a human pilot study, although no patient was given the peptide. Each is now labelled as laboratory work. The cause was in the rule that assigns evidence labels: it read the word “human” in phrases such as “human dermal fibroblasts” as a sign the study was done in people. The rule now recognises human cells and tissue as laboratory work, and nine other claims on the BAM15, Matrixyl, SS-31, argireline and KLOW pages were re-labelled with it. Found by our own writers while building new skin-care pages.

4 October 2026 — SS-31 dosage

SS-31 dosage stated that elamipretide “carries no approved label anywhere in the world” and “has never been approved by any regulator”. That was wrong. In 2025 elamipretide became the first disease-specific treatment approved for Barth syndrome, indicated for the improvement of muscle strength — a fact our own SS-31 overview page already reported correctly. The page now states the approval, and explains that it covers the approved product in that one disease and does not transfer to material sold as SS-31 or to any other use. Found during a full-site audit, when a newly written SS-31 safety page and the existing dosage page disagreed.

4 October 2026 — Retracted papers cited as evidence

Two papers that had been retracted were cited as evidence on three pages. VIP cited a 2022 study of 102 patients with constipation-predominant IBS; that study has been retracted, and the passage now rests on a different 2022 study. TB-500 safety and TB-500 comparisons cited a retracted 2023 mouse study for a link between thymosin beta-4 and insulin signalling; no other source behind those pages supports the point, so it has been removed rather than replaced. We now check every citation against its retraction status before a page can publish. Retracted work can still be cited to report that it was retracted — the PT-141 pages do this for a withdrawn 2008 trial — but never as evidence.

4 October 2026 — KPV and teduglutide

KPV described teduglutide as “an approved bowel medicine”. None of the sources behind the page says so, and the claim was unsupported. The sentence now says only what its source does: that teduglutide was loaded into the nanoparticles being studied.

4 October 2026 — Semax dosage

Semax dosage said its Russian clinical research had been repeated by “nobody outside Russia”. That is a claim about the whole world’s literature, and we can only speak for the research we read. It now says that no study among the sources behind the page repeats it outside Russia.

1 October 2026 — BPC-157 comparison

BPC-157 and TB-500: what the research compares stated that no published study had compared the two compounds against each other, in any species, for any outcome. That was wrong. A 2026 experiment in 32 rats ran four arms — control, BPC-157, TB-500 and the two combined — and measured tendon strength in each, which is a direct comparison. The sentence has been replaced with an accurate one and the page now points to our coverage of that experiment. A second instance of the same error survived that first fix: the page’s meta description — the sentence a search engine reprints — still carried the false claim, and was corrected separately later the same day. The first fix changed the page a reader sees; it did not change what the page says about itself. The error was found by cross-checking the page against our own Wolverine stack page, which cites the experiment: two pages on this site disagreed, and one of them was wrong.