BACKGROUND: Venovenous extracorporeal membrane oxygenation (VV-ECMO) is used when mechanical ventilation can no longer support oxygenation or ventilation, or if the risk of ventilator-induced lung injury is considered excessive. The optimum mechanical ventilation strategy once on ECMO is unknown. We sought to describe the practice of mechanical ventilation in children on VV-ECMO and to determine whether mechanical ventilation practices are associated with clinical outcomes. METHODS: We conducted a multicenter retrospective cohort study in 10 pediatric academic centers in the United States. Children age 14 d through 18 y on VV-ECMO from 2011 to 2016 were included. Exclusion criteria were preexisting chronic respiratory failure, primary diagnosis of asthma, cyanotic heart disease, or ECMO as a bridge to lung transplant. RESULTS: Conventional mechanical ventilation was used in about 75% of children on VV-ECMO; the remaining subjects were managed with a variety of approaches. With the exception of PEEP, there was large variation in ventilator settings. Ventilator mode and pressure settings were not associated with survival. Mean ventilator F IO 2 on days 1-3 was higher in nonsurvivors than in survivors (0.5 vs 0.4, P 5 .009). In univariate analysis, other risk factors for mortality were female gender, higher Pediatric Risk Estimate Score for Children Using Extracorporeal Respiratory Support (Ped-RESCUERS), diagnosis of cancer or stem cell transplant, and number of days intubated prior to initiation of ECMO (all P < .05). In multivariate analysis, ventilator F IO 2 was significantly associated with mortality (odds ratio 1.38 for each 0.1 increase in F IO 2 , 95% CI 1.09-1.75). Mortality was higher in subjects on high ventilator F IO 2 (6 0.5) compared to low ventilator F IO 2 (> 0.5) (46% vs 22%, P 5 .001). CONCLUSIONS: Ventilator mode and some settings vary in practice. The only ventilator setting associated with mortality was F IO 2 , even after adjustment for disease severity. Ventilator F IO 2 is a modifiable setting that may contribute to mortality in children on VV-ECMO.