Corrections
This site takes a hard line on other people's accuracy, which obliges it to be public about its own. Every substantive factual correction is logged here: what the page used to say, what it says now, and how the mistake got in. Typos, wording and design changes are not logged — only things that were factually wrong.
If you find an error, tell us. Corrections are made to the page and recorded here rather than edited away silently.
We stated that "no 6 mg dose was used in the published trials."
Now: That was wrong. The Phase 3 TRIUMPH trials use 6 mg as a four-week titration step on the route to both the 9 mg and 12 mg targets. What remains true, and is what we should have written, is that no trial has tested 6 mg as a maintenance dose. The pages now draw that distinction and show the full escalation schedule.
How it happened: We wrote the claim from the Phase 2 schedule, which escalated 2 → 4 → 8 → 12 mg, and did not re-check it against the Phase 3 protocol when Phase 3 results came out.
Our trial table described TRIUMPH-3 as studying "obesity + OSA."
Now: TRIUMPH-3 enrolled adults with severe obesity and established cardiovascular disease. Obstructive sleep apnoea was studied in nested sub-studies inside TRIUMPH-1 and TRIUMPH-2, not as its own trial. Both are now stated correctly, and the knee osteoarthritis trial is identified as TRIUMPH-4.
How it happened: The TRIUMPH programme covers several indications, and we attached a real endpoint to the wrong trial number.
Our comparison charts placed retatrutide’s 30.3% figure alongside tirzepatide, CagriSema and semaglutide, captioned as though all four came from the same analysis.
Now: They did not. Retatrutide’s 28.3% and 30.3% are the analysis that assumes participants stayed on the drug; the three comparators were the analysis that counts everyone randomised. Mixing them inflated retatrutide’s apparent lead by roughly two to three points. Every chart and ranking now uses the count-everyone analysis throughout, and both conventions are shown side by side so the gap is visible.
How it happened: This is the exact error this site was built to avoid, which makes it worth stating plainly rather than quietly fixing. The sponsors use six different names for two analyses, and we did not check the label behind each number before charting them together.
We quoted 30.3% at 104 weeks as a headline retatrutide result without qualification.
Now: That figure comes from an extension of 532 participants selected for a BMI of 35 or above, completion of the first 80 weeks, and tolerance of their assigned dose, then escalated to their maximum tolerated dose. It is a subgroup result. Every appearance of it now carries that caveat, and it is no longer used in cross-drug comparisons.
How it happened: The number was accurate; presenting it as a whole-population result was not.
We described survodutide as a Phase 2 compound and ranked it on a 46-week dose-finding result.
Now: Survodutide’s Phase 3 trial, SYNCHRONIZE-1, was published in the New England Journal of Medicine in June 2026: 725 adults, 76 weeks, 13.0% counting everyone randomised. Both pages now use that trial.
How it happened: The page had not been revisited since the Phase 3 publication.
Our CagriSema section covered only REDEFINE 1, where it performed well.
Now: REDEFINE 4 is now covered too. It tested CagriSema head to head against tirzepatide in 809 adults over 84 weeks, and CagriSema failed to meet its primary endpoint of non-inferiority. Omitting the trial it lost, while reporting the trial it won, was a real gap.
How it happened: The section predated the February 2026 result.
We wrote that approval was "expected around 2027."
Now: Eli Lilly has stated it plans to submit a Biologics License Application in the first quarter of 2027. That is a submission plan, not an approval date — review takes roughly six to ten months from acceptance, so a decision falls in late 2027 at the earliest and more plausibly 2028. The pages now say that.
How it happened: The original phrasing blurred submission and approval, which are more than a year apart.
The homepage advertised "0 claims without a citation."
Now: An absolute claim about our own accuracy is not something a reader can check, and this page is evidence it was not true. It has been replaced with counts that are verified automatically on every build.
How it happened: Claims about our own reliability should be verifiable or absent.
What we do to prevent this
- Trial figures are checked against the primary source — the publication, the registry record, or the sponsor's own release — rather than written from memory or from secondary coverage.
- Where a result exists under more than one statistical convention, we state which one we are using and never place two conventions in the same comparison.
- Results that come from press releases rather than peer-reviewed publication are labelled as such on the page, because they can change on the way to a paper.
- The counts on our homepage are recomputed from the built site on every deploy, so they cannot drift away from reality unnoticed.