PRAGUE-26 makes catheter thrombolysis look decisive, and ECMO's long-term reckoning

1. PRAGUE-26 — catheter-directed thrombolysis in intermediate-high-risk PE

Journal · Published: NEJM, 31 August 2026 — genuinely new, 1 day old

Access note: not yet in Europe PMC and the NEJM paper was not reachable. Design and numbers below come from the trial’s press coverage, which was detailed. Treat the figures as press-sourced until the paper can be read.

Multicentre, open-label randomised trial. 558 patients with intermediate-high-risk acute pulmonary embolism, randomised 1:1 to catheter-directed thrombolysis with alteplase plus anticoagulation, or anticoagulation alone. Median age 64, 41% women.

Primary endpoint — composite of all-cause death, PE recurrence, or cardiorespiratory decompensation or collapse within 7 days:

• 0.7% with CDT vs 6.8% with standard care
• Relative risk 0.10 (95% CI 0.02–0.44), p<0.001
• Deaths within 7 days: 4 (standard care) vs 1 (CDT, by 30 days)

Bleeding at 7 days: 4.6% vs 5.0% (p=0.846) — but two intracranial bleeds in the CDT arm versus none in the standard-care arm.

Interpretation. A relative risk of 0.10 is the kind of number that should make you look twice, and here the reason is visible in the denominator: the composite occurred in 19 patients in total. With events that scarce, the confidence interval spans 0.02 to 0.44 — the direction is convincing, the magnitude is not pinned down. Most of the composite is “cardiorespiratory decompensation,” a softer component than death, and in an open-label trial a decompensation endpoint is exactly where unblinded assessment can drift.

The bleeding result deserves more attention than its p-value suggests. Overall bleeding was identical, but two intracranial haemorrhages against zero in a 558-patient trial is not reassuring — it is simply too small to quantify. That has been the central problem with lytic therapy in submassive PE since PEITHO (NEJM 2014), where systemic tenecteplase reduced haemodynamic decompensation and caused a 2% intracranial haemorrhage rate. The whole promise of the catheter-directed approach is that lower doses avoid that trade-off, and the honest reading of PRAGUE-26 is that it is consistent with that promise without demonstrating it.

Still, this is a real advance on the pilot-scale evidence that preceded it, and the investigators’ framing — building a national network on the STEMI model — is the part that will shape practice, because access to CDT is the actual constraint.

Article · Coverage

2. Long-term mortality and functional outcomes after ECMO

Journal · Published: Critical Care Medicine, 31 August 2026 — genuinely new, 1 day old Stebbins, Yap, Segun-Beloved, Hodgson, Serpa Neto et al.

Systematic review and meta-analysis, PROSPERO CRD420251002639, searched to March

1. RCTs and observational studies reporting mortality or functional outcomes at or beyond 6 months in adults receiving ECMO. 163 studies, 78,053 adults, with 59,454 across 156 studies contributing to the mortality meta-analysis.

Pooled one-year mortality, by modality:

• Venovenous ECMO — 37.2% (95% CI 30.0–45.1)
• Venoarterial ECMO — 55.2% (50.2–60.1)
• Extracorporeal CPR — 74.4% (71.4–77.2)

Functional outcomes were reported in only 38 of 163 studies (23.3%), covering 7,876 survivors, using inconsistent instruments — Cerebral Performance Category in just over half, WHODAS 2.0 and modified Rankin in five studies each.

The authors conclude that standardised reporting and benchmarking are needed to inform prognosis, patient selection, and what might improve functional recovery.

Interpretation. The headline numbers are the most useful thing here: three modalities with one-year mortality ranging from roughly a third to roughly three-quarters, which is the sort of figure worth having to hand in a family conversation rather than reconstructed from memory. The ECPR figure of 74.4% is sobering but not surprising, and sits alongside the trial literature — ARREST, PRAGUE OHCA, INCEPTION — that has never resolved into a consistent benefit.

The more pointed finding is the 76.7% of studies that reported no functional outcome at all. A field that has invested this heavily in a therapy whose entire justification is survival with acceptable function has, three-quarters of the time, not measured the function. That is the paper’s real contribution, and it is a criticism of the literature rather than of the therapy.

Caveat worth stating: pooling observational studies across two decades of evolving practice, indication and selection produces a number with wide practical uncertainty. The between-modality ordering is robust; the precise percentages are not a prognostic calculator.

Read the paper · PMID 42671261

3. Timing of PCI in patients undergoing TAVI

Journal · Published: NEJM, 30 August 2026 — genuinely new, 2 days old Stähli, Ruschitzka, Westermann, Linke, Mangner, Van Mieghem et al.

About half of TAVI patients have concomitant coronary disease, and PCI is often done first, without established evidence for the sequence. International, open-label, noninferiority trial at 48 European centres. Patients with severe aortic stenosis and coronary artery disease randomised 1:1 to TAVI before PCI or PCI before TAVI.

Primary endpoint at 1 year: composite of death from any cause, non-fatal myocardial infarction, ischaemia-driven revascularisation, rehospitalisation related to the valve, procedure or heart failure, or life-threatening, disabling or major bleeding. Noninferiority margin 6.6 percentage points.

• 986 randomised — 498 TAVI-first, 488 PCI-first
• Primary endpoint: 22.2% (105) vs 24.2% (112)
• Risk difference −2.0 percentage points (95% CI −7.4 to 3.4), P<0.001 for noninferiority
• Serious adverse events: 264 vs 273

Interpretation. A clean answer to a genuinely unsettled sequencing question, and the direction of the point estimate slightly favours TAVI-first, though the trial was designed only to exclude meaningful inferiority.

The practical consequence is flexibility rather than a mandate. If TAVI-first is noninferior, the sequence can be chosen on the individual case — coronary anatomy, access, frailty, how urgent the valve is — instead of on a default that was never tested. For the anaesthetic and perioperative side, TAVI-first also means the coronary intervention happens in a patient whose outflow obstruction has already been relieved, which is the more forgiving haemodynamic setting of the two.

Note the 22–24% one-year composite event rate in both arms: this remains a high-risk population regardless of sequence, and the trial says nothing about whether the PCI was needed at all — a question COMPLETE-style trials in this population have yet to answer.

Read the paper · PMID 42670987