ObjectiveExtracorporeal membrane oxygenation (ECMO) has been used with increasing frequency in the treatment of acute respiratory failure in pediatric patients. Our objective in this study was to test the hypothesis that ECMO improves outcome in pediatric patients with acute respiratory failure.DesignMulticenter, retrospective cohort analysis.SettingForty one pediatric intensive care units participated in the study under the auspices of the Pediatric Critical Care Study Group.PatientsAll pediatric patients admitted to the participating institutions with acute respiratory failure during 1991 were included. Patients with congenital heart disease, contraindications to ECMO, or incomplete data were excluded, yielding a data set of 331 patients from 32 hospitals.InterventionsConventional mechanical ventilation, high-frequency ventilation, and extracorporeal membrane oxygenation.Measurements and Main ResultsMultivariate logistic regression analysis was used to identify factors associated with survival. In a second analysis, pairs of ECMO and non-ECMO patients, matched by severity of disease and respiratory diagnosis, were compared.The use of ECMO (p equals .0082), but not the use of high-frequency ventilation, was associated with a reduction in mortality. Other factors independently associated with mortality included oxygenation index (p less than .0001), Pediatric Risk of Mortality score (PRISM) (p less than .0001) and the Paco2(p equals .045). In 53 diagnosis- and risk-matched pairs, there was a significantly lower mortality rate (26.4% vs. 47.2%; p less than .01) in the ECMO-treated patients. When all patients were stratified into mortality risk quartiles on the basis of oxygenation index and PRISM score, the proportion of deaths among ECMO-treated patients in the 50% to 75% mortality risk quartile was less than half the proportion in the non-ECMO treated patients (28.6% vs. 71.4%; p less than .05). No effect was seen in the other quartiles.ConclusionsThe use of ECMO was associated with an improved survival in pediatric patients with respiratory failure. The lack of association of outcome with treatment in the ECMO-capable hospital or with another tertiary technology (i.e., highfrequency ventilation) suggests that ECMO itself was responsible for the improved outcome. Further studies of this procedure are warranted but require broad-based multi-institutional participation to provide sufficient statistical power and sensitivity to demonstrate efficacy.(Crit Care Med 1996; 24:323-329)