Background: Adverse outcomes for infants born with left congenital diaphragmatic hernia (CDH) have been correlated with fetal imaging findings.
Objective: We sought to corroborate these correlations in a high-risk cohort and describe a predictive mortality algorithm combining multiple imaging biomarkers for use in prenatal counseling.
Materials and methods: We reviewed fetal MRI examinations at our institution from 2004 to 2016 demonstrating left-side CDH. MRI findings, hospital course and outcomes were recorded and analyzed using bivariate and multivariable analysis. We generated a receiver operating curve (ROC) to determine a cut-off relation for mortality. Finally, we created a predictive mortality calculator.
Results: Of 41 fetuses included in this high-risk cohort, 41% survived. Per bivariate analysis, observed-to-expected total fetal lung volume (P=0.007), intrathoracic position of the stomach (P=0.049), and extracorporeal membrane oxygenation (ECMO) requirement (P<0.001) were significantly associated with infant mortality. Youden J statistic optimized the ROC for mortality at 24% observed-to-expected total fetal lung volume (sensitivity 64%, specificity 82%, area under the curve 0.72). On multivariable analysis, observed-to-expected total fetal lung volume ± 24% was predictive of mortality (adjusted odds ratio, 95% confidence interval: 0.09 [0.02, 0.55]; P=0.008). We derived a novel mortality prediction calculator from this analysis.
Conclusion: In this high-risk cohort, decreased observed-to-expected total fetal lung volume and stomach herniation were significantly associated with mortality. The novel predictive mortality calculator utilizes information from fetal MR imaging and provides prognostic information for health care providers. Creation of similar predictive tools by other institutions, using their distinct populations, might prove useful in family counseling, especially where there are discordant imaging findings.
Keywords: Congenital diaphragmatic hernia; Fetus; Magnetic resonance imaging; Neonate; Perinatal outcome; Predictive mortality model.