Criteria for weaning from venovenous extracorporeal membrane oxygenation (V-V ECMO) in acute respiratory distress syndrome (ARDS) remain poorly standardized. Some patients are decannulated despite persistent abnormalities in respiratory mechanics once oxygenation and carbon dioxide clearance are considered adequate. We aimed to compare outcomes associated with permissive versus conventional V-V ECMO weaning.

We conducted a retrospective single-center cohort study including adult patients with refractory ARDS treated with V-V ECMO between January 2015 and December 2022 who were successfully decannulated while receiving invasive mechanical ventilation. Permissive weaning was defined as ECMO decannulation despite persistence of at least one of the following criteria during the sweep gas-off trial: plateau pressure ≥30 cmH₂O, driving pressure ≥20 cmH₂O, positive end-expiratory pressure ≥15 cmH₂O, or tidal volume ≤5 mL/kg predicted body weight. The primary outcome was 90-day mortality after ECMO decannulation. Secondary outcomes included liberation from mechanical ventilation within 90 days, analyzed using competing-risk models with death as a competing event. Associations with mortality were assessed using landmark Cox proportional hazards models anchored at decannulation.

Among 469 patients treated with V-V ECMO, 280 successfully decannulated patients were included; 117 (42%) underwent permissive weaning and 163 (58%) conventional weaning. Baseline characteristics were broadly similar, although body mass index was higher and SARS-CoV-2 ARDS more frequent in the permissive group. Patients undergoing permissive weaning had longer ECMO support (19 [9–37] vs. 12 [6–27] days; p < 0.01). At the sweep gas-off trial, respiratory mechanics remained more impaired in the permissive group, whereas oxygenation parameters were comparable. In landmark analysis, permissive weaning was associated with higher 90-day mortality in both univariable (hazard ratio [HR] 3.53, 95% confidence interval [CI] 1.62–7.66; p < 0.01) and multivariable analyses (adjusted HR 4.45, 95% CI 1.98–9.97; p < 0.01). However, permissive weaning was not associated with delayed liberation from mechanical ventilation by day 90 (adjusted subdistribution HR 0.90, 95% CI 0.70–1.17; p = 0.4).

Among patients with severe ARDS successfully decannulated from V-V ECMO, permissive weaning was associated with increased 90-day mortality but not with delayed liberation from mechanical ventilation.