Surveying and optimizing the predictors for ependymoma specific survival using SEER data.

Area of Science:

• Neuro-oncology
• Epidemiology
• Biostatistics

Background:

• Ependymoma is a primary tumor of the central nervous system.
• Accurate prognostication and identification of outcome disparities are crucial for effective patient management.

Purpose of the Study:

• To develop predictive models for ependymoma outcomes using the Surveillance, Epidemiology and End Results (SEER) database.
• To identify potential disparities in ependymoma patient outcomes based on various factors.

Main Methods:

• Analysis of socio-economic, staging, and treatment factors from the SEER database for ependymoma.
• Generalized Linear Models (GLM) were used to predict cause-specific death.
• Receiver Operating Characteristic (ROC) curve analysis was employed to assess model performance.

Main Results:

• Age was the most significant predictor of outcome. Ependymoma grade significantly impacted survival, with higher grades showing increased risk of death.
• A 3-tiered grading model (ROC area 0.53) improved predictive accuracy compared to a 5-tiered model (ROC area 0.48).
• African-American patients exhibited a higher risk of death (21.5%) compared to other racial groups (16.6%). Radiation therapy (RT) was underutilized in cerebellar and spinal ependymomas.

Conclusions:

• Refined ependymoma grading systems can substantially enhance data modeling and predictive accuracy.
• Increased utilization of radiation therapy (RT) for spinal cord and cerebellar ependymomas may improve patient outcomes.