Fetal lung volume and quantification of liver herniation by magnetic resonance imaging in isolated congenital diaphragmatic hernia

Objective To determine associations between fetal lung and liver herniation volumes measured by magnetic resonance imaging (MRI) and mortality/need for extracorporeal membrane oxygenation (ECMO) in cases of isolated congenital diaphragmatic hernia (CDH). A secondary objective was to compare prenatal MRI parameters with two-dimensional ultrasound lung measurements. Methods A retrospective review of medical records of all fetuses with isolated CDH evaluated between January 2004 and July 2012 was performed. The following MRI parameters were measured at 20-32 weeks: observed/expected total fetal lung volume (o/e-TLV), predicted pulmonary volume (PPV), percentage of liver herniated into the fetal thorax (%LH) and the liver/thoracic volume ratio (LiTR). These were compared with the ultrasound-determined lung-to-head ratio (LHR) and the observed/expected LHR (o/e-LHR) in the same cohort. The predictive value of MRI and ultrasound parameters for mortality and the need for ECMO was evaluated by univariate, multivariate and factor analysis and by receiver-operating characteristics curves. Results Eighty fetuses with isolated CDH were evaluated. Overall mortality was 18/80 (22.5%). Two newborns died a few hours after birth. ECMO was performed in 29/78 (37.2%) newborns, with a survival rate of 48.3% (14/29). The side of the diaphragmatic defect was not associated with mortality (P = 0.99) or the need for ECMO (P = 0.48). Good correlation was observed among o/e-TLV, PPV, LHR and o/e-LHR as well as between %LH and LiTR (r = 0.89; P < 0.01); however, fetal lung measurements and measures of liver herniation were not correlated (all P > 0.05). All parameters were statistically associated with mortality or the need for ECMO. The best combination of measurements to predict mortality was o/e-TLV and %LH, with 83% accuracy. Conclusion Mortality and the need for ECMO in neonates with isolated CDH can be best predicted using a combination of MRI o/e-TLV and %LH.