A new approach to risk stratification using fetal MRI to predict outcomes in congenital diaphragmatic hernia: the preliminary retrospective single institutional study

Congenital diaphragmatic hernia (CDH) is a condition with a wide range of severity. Prenatal diagnosis is essential to optimize postnatal management, especially for severe cases. The lung to head ratio (LHR) and liver herniation estimated by prenatal ultrasound has been used as prenatal predictors in CDH. However, reliability of these factors remains to be proven and prediction of outcome from prenatal imaging is still challenging. We propose our new stratification system using lung to liver signal intensity ratio (LLSIR) in fetal MRI, which has been shown to be related to pulmonary maturation.

Retrospective chart review was conducted on 25 infants with CDH treated from 2009 through 2016 in our hospital. We stratified patients according to fetal T2-weighted MRI as Grade I, detectable ipsilateral lung at the apex; Grade II, undetectable ipsilateral lung at the apex and contralateral LLSR ≥2.0; Grade III, undetectable ipsilateral lung at the apex and contralateral LLSR <2.0. To evaluate this stratification system, we analyzed survival, severity [inhaled nitric oxide (iNO) usage with or without extracorporeal membrane oxygenation (ECMO)], and requirement of patch closure.

All 15 patients survived in Grade I, while 2 out of 6 died in Grade II, and 3 out of 4 died in Grade III (P=0.003). Four were severe in Grade I, and all in Grade II and III who survived (P=0.007). One needed patch in Grade I, and all in Grade II and III (OR: 414,238,332; 95% CI, 0-∞). Liver herniation was noted in five patients, and significantly associated with survival (P=0.04), however, neither with severity (P=1.00) nor with the requirement of patch closure (P=0.52).

The risk stratification algorithm using contralateral LLSIR in fetal MRI could be useful and more reliable than liver herniation to predict survival, severity, and need of patch closure. Further investigation is warranted.