Background: The treatment of patients with symptomatic spinal metastasis is challenging, and it warrants a multi-disciplinary approach. When surgery is considered, the expected survival time at three months might be an important argument in the discussion. With the advent of new treatment modalities, however, validation of an existing prediction model is warranted.
Methods: Validation study with inclusion of patients after informed consent from May 2021 through December 2023 in one academic hospital and two large non-academic medical centers. Information was collected on the following variables: sex, primary tumor type, treatment of the primary tumor with curative intention, cervical location of the metastasis, and the highest Karnofsky Performance Score in 24 hours before presentation. The validation set included 378 patients with symptomatic spinal metastasis. The derivation set consisted of 567 patients. The main outcome are C-index, calibration slope, D-statistic, R2D, Brier score, joint test for misspecification.
Reults: The model had a C-index value of 0.68 ± 0.02 (95% CI: 0.64-0.72). Calibration analysis yielded a calibration slope of 0.66 ± 0.09 (95% CI: 0.50-0.83). The D-statistic was 0.92, and the R2D value was 0.17. The model performed well, especially at three months, with a Brier score of 0.44. Although no mismatch was observed graphically, the joint test for misspecification yielded a statistically significant value. The model was therefore adjusted slightly, based on all 945 patients included in the original model and the current study.
Conclusion: The prediction model performed reasonably well in estimating survival at three months in patients with symptomatic spinal metastasis.
Keywords: prediction model; spinal metastasis; spinal surgery; survival.
© The Author(s) 2025. Published by Oxford University Press.