POET-II shortens endocarditis therapy; eplontersen misses in ATTR-CM

ESC Congress day one produced two readouts worth your attention, and Europe PMC finally resolved the last two leads that had been sitting on the outstanding list — both directly on scope.

1. POET-II — response-tailored antibiotic duration in left-sided endocarditis

Journal · Published: Presented in a Hot Line session at ESC Congress 2026 and published simultaneously in NEJM, 28 August 2026 — genuinely new, yesterday

Access note: numbers below are from the ESC and press releases; the NEJM paper itself was not reachable at the time of writing.

508 patients, mean age 70, 75% male, with left-sided infective endocarditis caused by Staphylococcus aureus, Enterococcus faecalis or streptococci. Randomised to a response-tailored strategy — antibiotics stopped once the patient reached a predefined minimum duration (2–4 weeks) and met the POET stabilisation criteria (clinical, biochemical and imaging) — or to standard therapy.

• Primary endpoint, days alive without antibiotic treatment within six months: 183 vs 169 days median, p<0.001
• Treatment duration: 26 vs 41 days median — 15 days shorter, p<0.001
• Safety (all-cause mortality, unplanned valve surgery, or symptomatic embolic events at six months): 8.2% vs 10.7%, meeting noninferiority
• Roughly half of patients with left-sided endocarditis could be eligible

Interpretation. POET (NEJM 2019) established that these patients can be switched from intravenous to oral antibiotics if they meet the POET criteria; the question it left open was how long, and duration has stayed anchored to custom rather than evidence. POET-II answers it with the same criteria repurposed as a stopping rule. Two weeks off a six-week course, without a safety signal, is a substantial change to a disease where prolonged intravenous therapy drives line complications, C. difficile, resistance and length of stay.

For cardiothoracic practice specifically: shorter courses change the arithmetic around timing of valve surgery and postoperative antibiotic planning. The caveat is the organism list — S. aureus, E. faecalis, streptococci — so this does not speak to the harder organisms, and the strategy is only as good as the stabilisation criteria that gate it. Worth reading the actual criteria before applying the result.

ESC press office · Coverage

2. CARDIO-TTRansform — eplontersen in ATTR cardiomyopathy

Published: ESC Congress 2026 Hot Line, 28 August 2026 — genuinely new, yesterday. Maurer et al.

Double-blind phase III trial, 130 centres in 20 countries. 1,432 patients with wild-type or hereditary ATTR-CM (mean age 72, 9.4% female), randomised 1:1 to eplontersen 45 mg or placebo subcutaneously every four weeks. Primary endpoint: a composite of cardiovascular mortality and recurrent cardiovascular events to 140 weeks.

• 381 events in 210 patients (eplontersen) vs 392 events in 231 patients (placebo) — rate ratio 0.89 (95% CI 0.73–1.09), p=0.277. Missed.
• Large, sustained reductions in circulating transthyretin, consistent with the silencer class — so the drug did what it was designed to do
• 57% were on a stabiliser at baseline, and in those patients there was no additional benefit
• In the monotherapy subgroup (no baseline stabiliser), fewer primary events with eplontersen, nominally significant

Maurer: “Baseline stabiliser use appeared to influence the primary results.”

Interpretation. A negative trial whose most interesting feature is why it might be negative. TTR silencing worked biochemically but did not translate — and the obvious candidate explanation is that most patients were already on a stabiliser, leaving little room to improve. That is a trial-design problem created by the field’s own success: standard of care moved while the trial ran. The monotherapy subgroup is hypothesis-generating only, and should be read as such rather than as a rescue of the primary result. Practically it leaves stabilisers as the anchor of ATTR-CM therapy and puts the burden on future trials to define whether silencing adds anything on top.

ESC press release

3. Induction agents for tracheal intubation of the critically ill

Journal · Published: Chest, 18 August 2026 — 11 days old McDougall, Forestell, Sadeghirad et al.

Systematic review and frequentist random-effects network meta-analysis, searched to 10 December 2025, GRADE-rated. 21 RCTs, 6,031 patients across ICU, emergency department, operating room and pre-hospital settings.

• Versus etomidate, ketamine probably causes more haemodynamic instability during intubation — RR 1.31 (95% CI 1.15–1.48), moderate certainty
• Ketamine-propofol (“ketofol”) may reduce haemodynamic instability versus etomidate (RR 0.44, 0.27–0.72) and versus ketamine (RR 0.34, 0.21–0.56) — both low certainty
• Versus etomidate, ketamine may decrease the need for post-intubation vasopressor infusions — RR 0.80 (0.50–1.28), low certainty
• Etomidate caused increased adrenal suppression compared with other agents

The authors’ own summary is admirably unresolved: etomidate probably causes less instability during intubation, but may increase the need for vasopressors after it, and increases adrenal suppression — “the clinical importance of these competing effects remains uncertain.”

Interpretation. This is the meta-analytic companion to the RSI trial (sent 22 August), and it points the same way: ketamine’s presumed haemodynamic advantage over etomidate does not survive contact with pooled randomised data. RSI found more cardiovascular collapse with ketamine (NNH 20); this finds RR 1.31 for instability. Two independent lines of evidence now contradict the received wisdom that grew out of the etomidate-adrenal-suppression literature.

The genuinely new signal is ketofol, with the largest effect sizes in the network — and the weakest evidence behind them. Worth knowing that the question is open rather than treating those numbers as a reason to change practice.

Note: this had been unresolvable across several runs because searches kept returning the adjacent Zampieri Critical Care network meta-analysis (DOI 10.1186/s13054-026-06067-w, May 2026), which reaches concordant conclusions — ketamine probably increases cardiovascular collapse versus etomidate, OR 1.44 (1.20–1.71).

Read the paper

4. Haemoadsorption during cardiopulmonary bypass

Journal · Published: British Journal of Anaesthesia, 21 July 2026 — ~5 weeks old Pittaway, Kelly, Price, Parekh, Howells

Systematic review and meta-analysis of RCTs in adults undergoing cardiac surgery with cardiopulmonary bypass, comparing haemoadsorption inserted into the CPB circuit against conventional CPB. 12 RCTs, 713 participants — only one at low risk of bias.

• Mortality: OR 1.08 (95% CI 0.59–2.01), p=0.75
• Hospital length of stay: mean difference 0.21 days (−1.89 to 2.30), p=0.82
• ICU length of stay: −0.36 days (−1.31 to 0.60), p=0.42
• No difference in adverse events, or in requirements for ventilation, cardiovascular support or renal replacement therapy
• Confidence intervals nonetheless include clinically important benefit or harm

Conclusion: no evidence of clinical benefit for routine haemoadsorption during CPB, though benefit or harm cannot be excluded; adequately powered trials targeting the highest cytokine-burden cases are needed.

Interpretation. 713 patients across 12 trials, one of them at low risk of bias — that is an evidence base thin enough that “no significant difference” mostly means “nobody has run the trial properly yet.” The honest reading is not that haemoadsorption fails but that it remains untested at scale, a decade after entering practice on mechanistic plausibility.

It joins CLEANSE (20 August), which found HA-330 in septic shock at HR 0.68 (0.44–1.07, P=0.09) and is now under formal critique. Same device class, same pattern: a signal that never quite reaches significance and never quite goes away. Note also the parallel narrative review (Ohri, Perfusion, May 2026) proposing tiered patient selection — endocarditis, complex aortic surgery under circulatory arrest, transplantation, emergency surgery on anticoagulants, vasoplegia — which is the field’s way of conceding that the unselected question is answered and the selected one is not.

Read the paper