HG-106: A NOVEL TOOL TO PREDICT THE SURVIVAL OF DIFFUSE INTRINSIC PONTINE GLIOMA PATIENTS: EXTERNAL VALIDATION OF THE SURVIVAL PREDICTION MODEL USING THE INTERNATIONAL DIPG REGISTRY

PURPOSE: In this study, we aimed to determine the validity of the recently developed survival prediction model for DIPG through external validation in an independent patient cohort. Model performance was evaluated by analyzing the discrimination and calibration abilities. PATIENTS AND METHODS: The original survival prediction model was developed in a cohort of DIPG patients from the Netherlands, United Kingdom and Germany (n = 316). External validation was performed using patients from the International DIPG Registry, including patients from the United States, Canada, and Australia (n = 261). Basic comparison of the cohorts was performed using descriptive statistics and univariate or multivariable regression analyses. The generalizability of the original model was subsequently analyzed following a variety of analyses described previously by Royston et al. (2013). RESULTS: Baseline patient characteristics and results from the regression analyses in both cohorts were comparable. The slope on the prognostic index in the International Registry cohort was 0.72 (p < 0.01) and Harrell's c-index of concordance was 0.57. The chi² to test model misspecification was 9.77 (p = 0.002). Kaplan-Meier curves from both cohorts show well-separated lines of low, intermediate and high-risk groups, which were confirmed by similar values for the hazard ratios across these risk groups. CONCLUSION: The current study demonstrates successful validation of the survival prediction model for DIPG, which is able to discriminate between patients with very short, average, and increased survival based on three clinical and one radiological variable.