Survival status of children with cerebellar gliomas

Abstract

✓ The goals of this study were: 1) to determine how well a multivariate analytic procedure shortly after surgery would have classified children with cerebellar glioma according to expected survival status (alive or dead) 5 years later, and 2) to identify histological, clinical, and other features that contributed most to the classification. Fifteen of the 76 (20%) children in this study died before the end of the fifth postoperative year. The linear discriminant function (LDF) constructed with the four histological features of microcysts, high cell density, oligodendroglia, and endothelial proliferation, misclassified outcome in only eight children and discriminated as well as the LDF constructed with 12 additional histological features. The LDF constructed with eight symptoms and 12 signs had a misclassification rate of 9.2%, and was no more successful than the LDF constructed with the single symptom of lethargy. The LDF constructed with all 42 variables misclassified only three children. The variables that contributed most to successful analysis were altered consciousness, microcysts, oligodendroglia, high cell density, endothelial proliferation, and perivascular desmoplasia. Multivariate analysis of histological features, symptoms, and signs appears to classify children with cerebellar glioma according to survival status with an acceptably low error.