Volumetric predictors for shunt-dependency in pediatric posterior fossa tumors

Abstract

Abstract Posterior fossa (PF) tumors are the most common neoplastic entity in pediatric neurosurgery. Children suffering from PF tumors regularly present with hydrocephalus and CSF diversion is a crucial point of treatment. There is an ongoing debate about external ventricular drainage (EVD) management before surgery and its influence on ongoing hydrocephalus treatment afterwards. Beyond onco-surgical aspects, the prevention of shunt-dependency is an important goal in posterior fossa surgery. Various predictors for shunt-dependency after posterior fossa surgery in children have been suggested. Because these predictors may only apply to small subsets of children, and their reliability has been questioned, we evaluated a straightforward, potentially automated, and unbiased method for shunt prediction. In this retrospective radiomic study we analyzed 60 pediatric patients with posterior fossa tumors. Exclusion criteria were age under two years, missing MRI data, tumor location non-exclusive to the PF, traumatic brain injury and less than 6 months follow-up. Ultimately, 36 children met the inclusion criteria. We performed a volumetric assessment of various skull and brain compartments before and after surgery focused on ventricle-brain ratio (VBR). We dichotomized for potential predictors and performed ROC analyses. We evaluated the prognostic parameters for shunt dependency, including supratentorial transependymal edema and VBR, as well as pre- and postoperative radiomic measurements as early prognostic tools. The cutoff in ventricle volume for CSF diversion was 60.9 ml (AUC 0.788, p = 0.001). The radiomic-based prediction of shunt dependency with VBR-scoring showed an AUC of 0.783. Postoperative reduction in ventricle size, depicted by the deltaVBR scoring, showed an AUC of 0.719 in predicting shunt-free survival. Perioperative CSF diversion did correlate with postoperative persistent HCP, whereas the odd’s ratio for shunting was decreased, but not significantly lower, when CSF diversion was undertaken perioperatively (AUC = 0.618, OR = 0.273, CI = 0.029–2.577). Ventricle-brain ratio may be a potential predictor for the necessity of CSF diversion. In our cohort, radiomic predictors performed better than hydrocephalus categorization, modified Canadian Preoperative Prediction Rule for Hydrocephalus (mCPPRH) or transependymal edema alone. VBR pre- and deltaVBR postoperatively may be potential tools to predict the need for shunting in pediatric posterior fossa tumor patients. The decision for pre- or intraoperative CSF diversion showed no correlation and no influence on persistent hydrocephalus.