OBJECTIVES To examine the Extracorporeal Life Support Organization (ELSO) registry database of infants and children with acute respiratory failure to compare outcome and complications of venovenous (VV) vs. venoarterial (VA) Extracorporeal Life Support (ECLS). DESIGN Retrospective cohort study. SETTING ELSO registry for pediatric pulmonary support. PATIENTS All nonneonatal pediatric pulmonary support ECLS cases treated at U.S. centers and reported to the ELSO registry as of July 1997. Patients were excluded if they had one or more of the following diagnoses: hematologic-oncologic, cardiac, abdominal surgical, burn, metabolic, airway, or immunodeficiency disorder. INTERVENTIONS Venoarterial or venovenous extracorporeal life support for severe pulmonary failure. MEASUREMENTS AND MAIN RESULTS From 1986 to June of 1997, 763 pediatric patients met the inclusion criteria. Overall, 595 were initially managed with VA bypass, and 168 with VV bypass. The VA group was younger (mean +/- SD, 26.1+/-42.2 months for VA vs. 63.5+/-68.7 months for VV) and smaller (11.8+/-15.1 kg vs. 22.9+/-23.8 kg) (p<.001). There were no differences between groups in number of days on mechanical ventilation before ECLS, number of hours on ECLS, or number of hours on mechanical ventilation post-ECLS in survivors. Mean pH and Paco2 values, positive end-expiratory pressure, and mean airway pressure just before placing the patient on ECLS were also similar. VA-treated patients had higher Fio2 requirements (p = .034), lower Pao2 (p = .047), and lower Pao2/Fio2 ratio (p = .014) just before cannulation. There was a trend of higher peak inspiratory pressure in VA-treated patients (p = .053). Overall, survival rate was not different for the two groups (55.8% for VA vs. 60.1% for VV; p = .33). Central nervous system complications were not different between the two groups. Examination of the same variables was then conducted after dividing the patients into four subgroups. There were no significant differences in survival or complications during bypass between VV and VA modes of ECLS in any subgroup. Stepwise logistic regression modeling was performed to control for variables associated with the outcome survival for VV and VA-treated groups, and variables measured before bypass were identified as being associated with improved survival. There was a trend of improved survival in the VV-treated patients (p = .12). CONCLUSIONS Overall survival of pediatric patients with acute respiratory failure supported by VA or VV ECLS was comparable. A randomized clinical trial may be useful in clarifying these observations.