Abstract
Background: Current prognostication models for survival estimation in patients with metastatic spine disease lack accuracy. Identifying new risk factors could improve existing models. We assessed factors associated with survival in patients surgically treated for spine metastases, created a classic scoring algorithm, nomogram, and boosting algorithm, and tested the predictive accuracy of the three created algorithms at estimating survival.
Methods: We included 649 patients from two tertiary care referral centers in this retrospective study (2002 to 2014). A multivariate Cox model was used to identify factors independently associated with survival. We created a classic scoring system, a nomogram, and a boosting (i.e., machine learning) algorithm and calculated their accuracy by receiver operating characteristic analysis.
Results: Older age (hazard ratio [HR], 1.01; p = 0.009), poor performance status (HR, 1.54; p = 0.001), primary cancer type (HR, 1.68; p < 0.001), > 1 spine metastasis (HR, 1.32; p = 0.009), lung and/or liver metastasis (HR, 1.35; p = 0.005), brain metastasis (HR, 1.90; p < 0.001), any systemic therapy for cancer prior to a surgical procedure (e.g., chemotherapy, immunotherapy, hormone therapy) (HR, 1.65; p < 0.001), higher white blood-cell count (HR, 1.03; p = 0.002), and lower hemoglobin levels (HR, 0.92; p = 0.009) were independently associated with decreased survival. The boosting algorithm was best at predicting survival on the training data sets (p < 0.001); the nomogram was more reliable at estimating survival on the test data sets, with an accuracy of 0.75 (30 days), 0.73 (90 days), and 0.75 (365 days).
Conclusions: We identified risk factors associated with survival that should be considered in prognostication. Performance of the boosting algorithm and nomogram were comparable on the testing data sets. However, the nomogram is easier to apply and therefore more useful to aid surgical decision-making.
