To determine if brain imaging in fetuses that underwent prenatal repair of neural tube defect (NTD) can predict the need for postnatal hydrocephalus treatment (HT) in the first year postpartum. This was a retrospective study of fetuses diagnosed with open NTD that had in-utero myelomeningocele repair between April 2014 and April 2016. Independent variables were collected from four chronological sets of fetal images: presurgery ultrasound, presurgery magnetic resonance imaging (MRI), 6-week postsurgery MRI and predelivery ultrasound. The following independent variables were collected from all image sets unless otherwise noted: gestational age, head circumference, mean ventricular width, ventricular volume (MRI only), hindbrain herniation (HBH) score (MRI only), and level of lesion (LOL), defined as the upper bony spinal defect (presurgery ultrasound only). Based on these measurements, additional variables were defined and calculated including change in degree of HBH, ventricular width growth (mm/week) and ventricular volume growth (mL/week). The need for HT (by either ventriculoperitoneal shunt or endoscopic third ventriculostomy with choroid plexus cauterization) was determined by a pediatric neurosurgeon using clinical and radiographic criteria; a secondary analysis was performed using the MOMS trial criteria for hydrocephalus. The predictive value of each parameter was assessed by receiver-operating characteristics curve and logistic regression analyses. Fifty affected fetuses were included in the study, of which 32 underwent open hysterotomy and 18 fetoscopic repair. Two neonates from the open hysterotomy group died and were excluded from the analysis. The mean gestational ages for the presurgery ultrasound, presurgery MRI, postsurgery MRI and predelivery ultrasound were 21.8 ± 2.1, 22.0 ± 1.8, 30.4 ± 1.6 and 31.0 ± 4.9 weeks, respectively. A total of 16 subjects required HT. The area under the curve (AUC) of predictive accuracy for HT showed that HBH grading on postsurgery MRI had the strongest predictive value (0.86; P < 0.01), outperforming other predictors such as postsurgery MRI ventricular volume (0.73; P = 0.03), MRI ventricular volume growth (0.79; P = 0.01), change in HBH (0.82; P = 0.01), and mean ventricular width on predelivery ultrasound (0.73; P = 0.01). Other variables, such as LOL, mean ventricular width on presurgery ultrasound, mean ventricular width on presurgery and postsurgery MRI, and ventricular growth assessment by MRI or ultrasound, had AUCs < 0.7. Optimal cut-offs of the variables with the highest AUC were evaluated to improve prediction. A combination of ventricular volume growth ≥ 2.02 mL/week and/or HBH of 3 on postsurgery MRI were the optimal cut-offs for the best prediction (odds ratio (OR), 42 (95% CI, 4-431); accuracy, 84%). Logistic regression analyses showed that persistence of severe HBH 6 weeks after surgery by MRI is one of the best predictors for HT (OR, 39 (95% CI, 4-369); accuracy, 84%). There was no significant change in the results when the MOMS trial criteria for hydrocephalus were used as the dependent variable. Persistence of HBH on MRI 6 weeks after prenatal NTD repair independently predicted the need for postnatal HT better than any ultrasound- or other MRI-derived measurements of ventricular characteristics. These results should aid in prenatal counseling and add support to the hypothesis that HBH is a significant driver of hydrocephalus in myelomeningocele patients. Copyright © 2019 ISUOG. Published by John Wiley & Sons Ltd.
Read full abstract