ECMO at one year: three very different prognoses under one name

Design

PRISMA 2020 systematic review and random-effects meta-analysis of randomised and observational studies, PROSPERO-registered, searched to March 2025

Population

163 studies, 78,053 critically ill adults treated with ECMO; 59,454 contributed to mortality pooling

Primary outcome

all-cause mortality at or beyond six months, and functional outcomes in survivors

Effect

one-year mortality 37.2% (95% CI 30.0-45.1) venovenous, 55.2% (50.2-60.1) venoarterial, 74.4% (71.4-77.2) ECPR

This review pooled 163 studies and 78,053 adults reporting mortality or function at six months or beyond after extracorporeal membrane oxygenation. The headline is the spread between modalities. One-year pooled mortality was 37.2% (95% CI 30.0 to 45.1) for venovenous support, 55.2% (95% CI 50.2 to 60.1) for venoarterial, and 74.4% (95% CI 71.4 to 77.2) after extracorporeal cardiopulmonary resuscitation.

The second finding is an absence. Only 38 of the 163 studies reported any functional outcome at all, in 7,876 survivors, and they used different instruments — Cerebral Performance Category in half, WHODAS 2.0 and the modified Rankin Scale in five studies each. So the literature can tell you how many patients are alive at a year and cannot tell you, in any poolable way, how they are.

For the conversation at the bedside this is directly usable, with a caveat. These are pooled observational cohorts, not a trial, and case mix drives much of the difference between modalities: ECPR is offered to arrested patients, venovenous support largely to respiratory failure with a survivable trajectory. The numbers are not causal comparisons between techniques. What they legitimately give a family is a realistic frame for the indication their relative was cannulated for, rather than a single institutional figure for 'ECMO'.

• Quote the modality-specific figure when counselling families, not a pooled ECMO survival rate
• State explicitly that one-year survival is known far better than one-year function
• Record a functional score at discharge and at follow-up using one instrument your unit standardises on
• Substantial heterogeneity in the venovenous estimate (CI 30.0 to 45.1) means local audit still matters
• Do not read the modality gradient as evidence that one technique outperforms another — indications differ

The statistics, in plain English

These are pooled proportions from observational cohorts, so the confidence intervals describe uncertainty about the average across studies, not the range of risk for an individual patient. The venovenous interval is wide (30.0 to 45.1%) because the contributing studies disagreed — different populations, eras and selection criteria — so a single number understates how much your own unit's result may differ. Only 39% of venovenous studies contributed to the one-year estimate, which introduces the possibility that studies with better follow-up were also studies with better outcomes.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.