Survival in Patients With Spinal Metastatic Disease Treated Nonoperatively With Radiotherapy: Are the SORG-ML Algorithms Relevant?

Summary of background data: The SORG-ML algorithms for survival in spinal metastatic disease were developed in patients who underwent surgery and were externally validated for patients managed operatively.

Objective: To externally validate the SORG-ML algorithms for survival in spinal metastatic disease in patients managed nonoperatively with radiation.

Study design: Retrospective cohort.

Methods: The performance of the SORG-ML algorithms was assessed by discrimination [receiver operating curves and area under the receiver operating curve (AUC)], calibration (calibration plots), decision curve analysis, and overall performance (Brier score). The primary outcomes were 90-day and 1-year mortality.

Results: Overall, 2074 adult patients underwent radiation for spinal metastatic disease and 29% (n=521) and 59% (n=917) had 90-day and 1-year mortality, respectively. On complete case analysis (n=415), the AUC was 0.76 (95% CI: 0.71-0.80) and 0.78 (95% CI: 0.73-0.83) for 90-day and 1-year mortality with fair calibration and positive net benefit confirmed by the decision curve analysis. With multiple imputation (n=2074), the AUC was 0.85 (95% CI: 0.83-0.87) and 0.87 (95% CI: 0.85-0.89) for 90-day and 1-year mortality with fair calibration and positive net benefit confirmed by the decision curve analysis.

Conclusion: The SORG-ML algorithms for survival in spinal metastatic disease generalize well to patients managed nonoperatively with radiation.