Joeky Senders

89 Glioblastoma survival calculator Introduction Glioblastoma is the most common primary malignant brain tumor with almost 12,000 new cases per year in the United States and a median survival of only a year after diagnosis. 1 Adequate survival prognostication is essential for informing clinical and personal decision-making. Although survival statistics are well-defined at the group- level, predicting individual patient survival remains challenging due to the heterogenous nature of the disease and significant variation in survival within strata. In recent years, numerous statistical and machine learning algorithms have emerged that can learn from examples to make patient-level predictions of survival. These algorithms can be particularly useful for tailoring clinical care to the needs of the individual glioblastoma patient. This study aims to compare the most commonly used statistical and machine learning algorithms in their ability to predict individual-patient survival in glioblastoma patients. In order to promote the reproducibility of the current study and facilitate external validation and implementation of the developed models, we deployed the best performing model as an online calculator that provides interactive, online, and graphical representations of personalized survival estimates. Methods Data and study population The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Statement was used for the reporting of this study. 2 Data was extracted from the Surveillance Epidemiology and End Results (SEER) database (2005- 2015). 3 The SEER registry compiles cancer incidence and survival data of 18 registries and covers 28% of the U.S. population from academic and nonacademic hospitals, and as such, is broadly representative of the U.S. population as a whole. 4 Patients who underwent surgery for a histopathologically confirmed diagnosis of a glioblastoma (International Classification of Diseases for Oncology-Third Edition [ICD-O-3] codes 9440, 9441, 9442) were included in the analysis. Patients were excluded from the analysis if they died in the direct postoperative period (≤30 days after surgery). Our institutional review board has exempted the SEER database from review and waived the need for informed consent due to the retrospective nature of this study.