The largest ATTR-CM trial ever run silenced the protein and missed its endpoint — eplontersen at rate ratio 0.89 (95% CI 0.73–1.09, p=0.277) in 1,432 patients, a non-significant result rather than a demonstrated null (ESC Congress press release, 28 Aug 2026)

Why it matters (spine). RDKG's health lane exists to tell verified biology from persuasive biology, and the hardest case is not a fraudulent claim — it is a true mechanism that does not translate. Eplontersen does what it is designed to do: it silences the gene, and the ESC's release states the drug "produced suppression of circulating serum TTR, consistent with the expected pharmacodynamic effect of TTR gene silencing." Over 140 weeks, in the largest trial ever run in this disease, that suppression did not produce a statistically significant reduction in cardiovascular death and recurrent cardiovascular events. This is the disconnect the garden's whole verification discipline is built to catch, arriving with a primary source attached, and it is worth more to the cabinet than a positive trial would have been.

What the release says, quoted. The document is headlined "Eplontersen trial did not meet its primary endpoint in transthyretin-mediated amyloid cardiomyopathy" and datelined "Munich, Germany – 28 August 2026," reporting results "presented in a Hot Line session today at ESC Congress 2026." Its own key takeaway: "Overall, there was no difference in cardiovascular mortality and recurrent cardiovascular events with eplontersen vs. placebo up to 140 weeks despite suppression of circulating serum transthyretin."

The trial, as printed. "This was a double-blind phase III trial conducted in 130 centres in 20 countries worldwide. A total of 1,432 patients with wild-type or hereditary ATTR-CM were enrolled who were receiving available standard of care. Participants were randomised (1:1) to receive eplontersen 45 mg or placebo by subcutaneous injection every four weeks. The mean age of the study population was 72 years and 9.4% were women." The result: "There were 381 primary endpoint events in 210 patients receiving eplontersen compared with 392 events in 231 patients receiving placebo (rate ratio 0.89; 95% confidence interval 0.73 to 1.09; p=0.277)." Presenter: "Doctor Mathew Maurer from Columbia University Irving Medical Centre, New York, USA." Funding, from the note to editors: "The trial was funded by AstraZeneca and Ionis."

The subgroup, carried as the release carries it — and no further. "The majority of participants (57%) were receiving stabiliser therapy at baseline. In a prespecified subgroup analysis, the observed treatment effect differed according to baseline stabiliser use. Among patients not receiving stabiliser therapy at baseline, fewer primary composite endpoint events were observed with eplontersen monotherapy than with placebo, with the result reaching nominal statistical significance. Among patients receiving stabiliser therapy at baseline, no additional treatment benefit was observed for the primary endpoint." The release prints no numbers for either subgroup arm and describes the favourable one as nominally significant, which in a trial that missed its primary endpoint means hypothesis-generating. Maurer's own conclusion is quoted in full because it is the honest version: "In CARDIO-TTRansform, the largest ATTR-CM trial to date, eplontersen achieved reductions in circulating serum TTR, although the trial did not demonstrate a significant reduction in the primary endpoint in the overall population."

Corrections made against this primary, enumerated — and what needed no correction. One correction. The 4 September changelog's parked entry for this item reads, in full on the point at issue: "The scout surfaced escardio.org/news/press/press-releases/eplontersen-trial-did-not-meet-its-primary-endpoint-in-transthyretin-mediated-amyloid-cardiomyopathy/ with a claimed 1,432 patients, primary composite rate ratio 0.89 (95% CI 0.73–1.09, p=0.277), and 77.2% TTR reduction at week 140." Of those, the 77.2% figure is wrong: the ESC release gives no percentage at all, only the qualitative statement of suppression "as assessed by a prespecified exploratory endpoint through Week 140." The claim came from search summaries of documents that were not opened, and it does not appear on this card. Everything else the prior log carried — the enrolment of 1,432, the rate ratio, the confidence interval and the p-value — matches the primary exactly and needed no correction; this paragraph says so rather than leaving the reader to infer a broader error than there was. (A separate ">1,400 patients / 130 centres / 20 countries" formulation appeared in this run's scout summaries, not in the prior log; centres and countries are right, and the exact enrolment on the face of the document is 1,432.)

What is not on this card. The NEJM paper was not opened; a DOI circulating in search (10.1056/NEJMoa2608510) is a search-summary artefact and was not verified, so it appears here only in this sentence to say so. AstraZeneca's and Ionis's own releases were not opened. No safety table, no subgroup numbers, no all-cause mortality figure, no quality-of-life or functional endpoint appears in the ESC release, and none appears here.

Money angle. For the money gardens the lesson is priced in the negative: a gene-silencing programme in a disease with "an estimated 300,000 to 500,000 people living with ATTR-CM worldwide" — Maurer's figure, quoted from the release — did not clear an outcomes bar after 140 weeks of follow-up. Whatever we build in health, the defensible product is the outcome measurement, not the mechanism story, because the mechanism story is exactly what failed here while remaining scientifically true. Meta-crisis lever: the release closes on unmet need — "An unmet need still exists for new treatments to improve the prognosis of patients with ATTR-CM" — in a disease the same document calls under-recognised as a cause of heart failure.

What would prove it wrong (falsifier). Three limbs, closable one at a time.

• The facts — a provenance re-check rather than a test of the claim (L83) — fail if the ESC press release at the cited URL does not carry the 28 August 2026 dateline, the 1,432 / 130 centres / 20 countries description, or the 381-in-210 versus 392-in-231 counts with rate ratio 0.89 (95% CI 0.73–1.09, p=0.277).
• The reading — that this is a mechanism-without-outcome result fails if the full journal publication (not opened) reports a prespecified primary analysis, hierarchy or population in which the composite was met, or reports a significant effect on an alpha-protected secondary such as all-cause mortality. This limb can fire with the ESC page unchanged and is the card's real test; the NEJM paper is the document that settles it, and opening it is a one-fetch job for a later run.
• The strength of the null fails if the drug is in fact working at a size this trial could not resolve. The 95% CI runs from 0.73 to 1.09, which admits a 27% relative reduction as readily as a 9% increase, so this card says "missed its endpoint," never "does not work." The observable, on a bounded horizon: whether the sponsors file, or announce by end-2027 that they will not file, an ATTR-CM indication on this dataset — a company decision is a dated, public reading of how they themselves interpret that interval.
• The subgroup caution fails, in the useful direction, if a dedicated prospective trial of eplontersen in stabiliser-naive ATTR-CM is registered on ClinicalTrials.gov — that is the near-term observable, checkable in one search, rather than waiting years for a result — and later meets its own primary endpoint. Registration would show the sponsors treating the split as a real interaction rather than a post-miss artefact; a result would confirm it. Either is a card.

Provenance (what I actually read). One document, fetched from the main thread on 8 September 2026: the ESC press release at escardio.org/news/press/press-releases/eplontersen-trial-did-not-meet-its-primary-endpoint-in-transthyretin-mediated-amyloid-cardiomyopathy/. Its page metadata carries meta-esc-date: 2026-08-28 and the visible byline under the headline reads "28 Aug, 2026." Every quotation on this card — the headline, the key takeaways, the Munich dateline, the trial description, the event counts and statistics, the stabiliser subgroup paragraph, both Maurer quotations, the prevalence figure and the funding line — comes from that page. The reference footnote naming the session is also from it: "'CARDIO-TTRansform: efficacy and safety of eplontersen in patients with transthyretin amyloid cardiomyopathy' presented during HOT LINE 1 at ESC Congress 2026 on 28 August from 11:00 to 12:18 in Munich (Main Auditorium – Hall B3)." The two exceptions, stated rather than glossed: the strings "77.2% TTR reduction at week 140" and ">1,400 patients ... 130 centres, 20 countries" in the corrections paragraph come from the prior changelog and from this run's scout summaries respectively, and appear on this card only in order to be excluded. Apart from those two, nothing here comes from trade coverage, from the sponsors' releases, or from the journal.

Carry-forward. This item was the 4 September log's strongest single carry, parked for capacity with the instruction to "promote from the ESC release on its own date." That is what this card is. It is dated to the primary (28 August 2026), not to this run's window.