Racial and socioeconomic disparities in long term survival after surgery and radiation for spinal cord hemangioblastoma.
Journal:
Scientific reports
Published Date:
Aug 21, 2025
Abstract
Spinal cord hemangioblastomas are rare, benign, intradural tumors that, despite their nonmalignant histopathology, can lead to substantial neurological morbidity. While disparities in outcomes based on race and socioeconomic status have been well-documented in other spinal tumor populations, their role in spinal cord hemangioblastoma remains poorly understood. In this study, we utilize the National Cancer Database (NCDB) to evaluate the influence of race, socioeconomic factors, and healthcare access on survival outcomes in patients with spinal cord hemangioblastoma. Additionally, we explore the utility of machine learning-based survival models to improve individualized risk prediction and to identify key clinical and sociodemographic determinants of long-term survival. Patients diagnosed with spinal cord hemangioblastoma were identified from the National Cancer Database (NCDB) using ICD-O-3 histology and topography codes. Demographic, socioeconomic, and clinical variables were compared across racial groups (White, Black and Asian). Long-term overall survival (OS) was defined as survival beyond 10 years. Kaplan-Meier and multivariable Cox regression analyses were used to evaluate survival outcomes and identify independent predictors of mortality. Tumor size was stratified using the cohort-wide mean (62.2 mm) for interpretability. Temporal trends in racial distribution and surgical technique (open vs. MIS) were assessed using Mann-Kendall trend testing. Gradient Boosting Survival, Cox proportional hazards, and Random Survival Forest models were developed and validated for mortality prediction. The best-performing model was interpreted using SHAP analysis. A total of 716 adult patients with spinal cord hemangioblastoma were analyzed, with the majority being White (83.7%), followed by Black (12.3%) and Asian (4%). Significant differences were observed across racial groups in age, insurance status, income quartiles, and comorbidity scores, though sex distribution and facility type utilization were comparable. Most patients were treated at academic centers, and surgery alone was the predominant treatment modality, with no racial disparities in extent of resection or use of radiation. Kaplan-Meier analysis showed significantly higher 10-year and long-term mortality in White patients; however, race was not an independent predictor in multivariable Cox regression, where increased age, higher CDCC scores, urban residence, and treatment at comprehensive community cancer centers were associated with worse survival. Surgery, with or without radiation, was protective compared to radiation alone. Temporal analysis showed stable racial distribution and minimal uptake of minimally invasive surgery from 2010 to 2017. The Gradient Boosting Survival model achieved the highest predictive performance (AUC = 0.8214; C-index = 0.7817), with age, facility type, and comorbidity burden identified as the strongest predictors of mortality in SHAP analysis. A publicly available web-based calculator was developed based on this model to provide individualized survival estimates. Racial and socioeconomic disparities were associated with differences in clinical outcomes on univariate analysis. However, race and insurance status were not independent predictors of mortality in multivariable-adjusted models. This suggests that the observed survival differences may be explained by confounding factors, such as comorbidity burden, treatment modality, or access to specialized care. Notably, poorer survival was independently associated with treatment at Comprehensive Community Cancer Programs and with higher comorbidity scores, underscoring the importance of ensuring equitable access to high-volume, specialized centers. Lastly, the Gradient Boosting Survival model enhanced mortality risk prediction by incorporating both clinical and socioeconomic variables, supporting its potential utility in guiding targeted interventions to improve long-term outcomes.
