A Risk Stratification Model Based on a Population Analysis for
Predicting Cancer Specific Survival in Pediatric Brain Stem Glioma.
Abstract
Introduction: The aim of this study was to construct and validate a
nomogram and risk stratification model for predicting cancer-specific
survival (CSS) of pediatric brainstem glioma patients. Methods: Cases of
pediatric brainstem glioma patients (<12 years) from 1998 to
2016 were retrieved from the Surveillance, Epidemiology, and End Results
(SEER) database and demographic, clinicopathologic characteristics,
treatments, and survival outcomes were analyzed. The total cohort was
randomly divided into training and validation sets, followed by
univariate and multivariate Cox regression analyses. A nomogram was
constructed and risk stratification analysis incorporated using the
selected variables from the multivariate analysis. The accuracy of the
model was assessed using C-index and calibration curves. Results: A
total of 806 pediatric cases with histologically confirmed diagnosis of
brainstem glioma were selected and analyzed. Multivariate analysis
showed that age, race, tumor size, grade and radiotherapy
(P<0.05) were independent prognostic indicators of pediatric
gliomas. For prediction of CSS, the C-index of the nomogram was 0.75,
which shows a good predictive probability. Conclusion: The nomogram
developed in this study for predicting survival of pediatric patients
with histologically confirmed stem gliomas is the first to incorporate
risk stratification. Combining nomogram and risk stratification system
is a convenient tool to aid clinicians in the identification of
high-risk patients and to perform targeted adjuvant treatment.