Hyperoxia and neurologic complications in pediatric veno-arterial extracorporeal membrane oxygenation: a retrospective ELSO registry study

Hyperoxia can contribute to neurologic injury through oxidative injury and platelet activation leading to microvascular thrombosis in critical illness. We aimed to determine if hyperoxia exposure during ECMO is associated with increased neurologic complications. Secondarily, we analyzed subpopulations with cyanotic congenital heart disease (cCHD) and non-pulsatile blood pressure. We performed a retrospective cohort study utilizing the ELSO registry database, an international, multicenter data repository. We included pediatric patients from 2010 through 2023 who received veno-arterial ECMO. Hyperoxia exposure was defined as post-ECMO PaO 2 greater than 250 mmHg. The primary outcome was any neurologic complications. Our secondary outcomes were severe neurologic complications and mortality. We evaluated the primary and secondary outcomes by multivariable logistic regression. The final cohort included 12,753 encounters. Of these, 1,315 (10.3%) were exposed to hyperoxia. Any neurologic complications occurred in 2,333 patients (18.3%). Severe neurologic injury and mortality occurred in 1,899 patients (14.9%) and 4,608 patients (36.1%), respectively. Hyperoxia exposure was associated with increased neurologic complications (aOR 1.30; 95% CI = 1.13–1.51; p < 0.001), severe neurologic complications (aOR 1.28; 95% CI = 1.09–1.50; p = 0.003), and mortality (aOR 1.59; 95% CI = 1.40–1.80; p < 0.001). Hyperoxia was more strongly associated with neurologic complications in children with cyanotic congenital heart disease (aOR 1.61; 95% CI = 1.22–2.13; p < 0.001) and was not significant in the non-pulsatile subgroup (aOR 1.17; 95% CI = 0.92–1.50; p = 0.203). Hyperoxia exposure during V-A ECMO is associated with increase in neurologic complications and their severity, and mortality. These findings support targeting normoxia and avoiding hyperoxia during pediatric V-A ECMO. The 250 mmHg threshold derived from this cohort is hypothesis-generating and could serve as a pragmatic monitoring threshold, pending prospective confirmation. Not applicable.