Machine Learning-Based Prognostic Scoring for Spinal Metastases: A JASA Multicenter Prospective Study Integrating...

Large multicenter prospective study.

We aimed to develop and validate a novel machine learning-based prognostic scoring system for spinal metastases.

Spinal metastases, common complications in patients with advanced cancer, significantly affect neurological function, pain, and quality of life. Although surgery plays a crucial role in selected cases, the accurate prediction of patient prognosis remains challenging. Traditional scoring systems, developed for older treatment paradigms, do not fully reflect the impact of modern oncologic therapies.

This multicenter prospective study, conducted by the Japan Association of Spine Surgeons with Ambition, included 401 patients who underwent surgery for spinal metastases at 35 medical centers between 2018 and 2021. Patient demographics, tumor burden, performance status, and treatment history data were collected. Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression was used to identify significant predictors of one-year survival, followed by stepwise variable selection. The model performance was assessed using the area under the receiver operating characteristic curve (AUROC) and calibration plots.

Among the 401 patients, 67.1% survived for one year, whereas 32.9% did not. Survivors had better performance status, lower tumor burden, and lower opioid use than non-survivors. LASSO regression identified five key predictors of one-year survival: age ≥75 years, poor performance status (≥3), presence of other bone metastases, preoperative opioid use, and lower preoperative Vitality Index. The final model demonstrated a strong predictive performance (AUROC=0.762). Based on the key prognostic factors, a simplified risk stratification system was developed to classify patients into low- (one-year survival 82.2%), intermediate- (67.2%), and high-risk (34.2%) groups.

We developed a clinically applicable prognostic scoring system for patients with spinal metastases using machine learning techniques to enhance predictive accuracy. This model provides a practical risk assessment tool to aid surgical decision-making and optimize postoperative management.

2.