A Novel Prognostication System for Spinal Metastasis Patients Based on Network Science and Correlation Analysis

Area of Science:

• Oncology
• Medical Informatics
• Biostatistics

Background:

• Spinal metastases are a common complication of cancer, necessitating personalized treatment strategies.
• Existing risk calculators aid therapeutic decisions by estimating survival, but their predictive accuracy can be enhanced.
• Improved prognostic tools are crucial for optimizing personalized medicine in oncology.

Purpose of the Study:

• To develop a novel risk assessment system for predicting survival in patients with spinal metastases.
• To demonstrate a methodology enabling other institutions to create their own localized prognostic scores.
• To enhance the accuracy of survival prediction for personalized cancer therapy.

Main Methods:

• Retrospective analysis of 454 patients with spinal metastases.
• Prognostic factor selection using network science-based correlation analysis to maximize Uno's C-index.
• Validation through D-statistic, Integrated Discrimination Index, five-fold cross-validation, and integrated time-dependent Brier score.

Main Results:

• Identified five independent prognostic factors via multivariate Cox analysis for the risk calculator.
• The new system demonstrated superior predictive ability compared to six established systems, achieving an average C-index of 0.706 at 10 years.
• The network science-based approach showed encouraging training performance.

Conclusions:

• Accurate life expectancy estimation is vital for personalized cancer medicine.
• The developed risk assessment system shows improved predictive performance, benefiting personalized treatment selection.
• Future improvements may involve systematizing 'unknown' factors like radiological morphology to further enhance prediction accuracy.