Improved Survival in Left Liver-Up Congenital Diaphragmatic Hernia by Early Repair Before Extracorporeal Membrane Oxygenation: Optimization of Patient Selection by Multivariate Risk Modeling

Abstract

Delayed repair of congenital diaphragmatic hernia (CDH) for days or longer has become standard, allowing improved stabilization for many, but potentially complicating treatment in severely affected infants who require extracorporeal membrane oxygenation (ECMO) and arrive unrepaired. Survival in left liver-up CDH, the most severe anatomic subset, averages 45% in published studies, with deaths often occurring in patients who failed to improve on ECMO and are repaired late, or not at all. Reliable early prediction of ECMO risk in these patients could identify the best candidates for repair before ECMO. We sought to predict ECMO risk in left liver-up CDH, and to further evaluate survival stratified by surgical timing in these patients. We reviewed 298 single-center, consecutive CDH patients, focusing on 87 inborn left liver-up patients without associated lethal anomalies. Multivariate analysis using anatomic and physiologic markers of severity was performed to define associations with need for ECMO. Sixty of 87 ECMO-eligible inborn left liver-up CDH patients required ECMO (69%). Of these, 20 of 21 (95%) repaired in the first 60 hours and before ECMO survived; whereas 13 of 20 (65%) who had repair delayed and arrived to ECMO unrepaired survived (p= 0.018). Lung-to-head ratio, Apgar scores, Congenital Diaphragmatic Hernia Study Group-predicted survival, pH, PCO2, and PO2 at 1 hour of life all correlated strongly with risk for ECMO. Accurate multivariate models to predict ECMO (area under the receiver operating characteristic curve [AUC] 0.91 and 0.91) were successfully developed. Early repair of left liver-up CDH before ECMO results in improved survival. Multivariate models can accurately assess risk for ECMO at 1 hour of life, permitting stratification of CDH surgical timing to maximize survival potential while minimizing risk.