An Externally Validated RPA-Based Pre-Treatment Decision-Making Tool Identifying Ideal Candidates for Spine SBRT

Abstract

<h3>Purpose/Objective(s)</h3> We sought to develop and validate a novel decision-making tool predictive of overall survival (OS) for patients receiving stereotactic body radiation therapy (SBRT) for spinal metastases. <h3>Materials/Methods</h3> Treatments were performed at two institutions and most commonly involved one or three fractions of spine SBRT. Recursive partitioning analysis (RPA) was conducted using decision-tree analysis. Potential correlations between variables were assessed with a correlation heatmap. OS was analyzed using the Kaplan-Meier method. Exclusion criteria included: treatments with proton therapy, benign primary histologies, missing variables of interest, and patients with the first course of spine SBRT outside of the participating institutions. <h3>Results</h3> The initial dataset consisted of 424 patients with 522 treated lesions. Of these, 361 patients were used for the RPA and initial model development because not all components of the SINS score were available for 63 patients. After consideration of 9 candidate variables, the final model consisted of the following variables and corresponding scores: SINS ≥ 6 (1); time from primary diagnosis < 21 months (1); ECOG performance status = 1 and ECOG > 1 (1 and 2, respectively); and > 1 organ systems involved (1). Each variable was an independent predictor of OS (p<0.001), and each 1-point increase in score was associated with a hazard ratio of 2.01 (95% confidence interval (CI): 1.79-2.25, p<0.0001) (Table). The concordance value was 0.75 (95% CI: 0.71-0.78). The scores were then discretized into 3 groups: favorable (score: 0-1), intermediate (score: 2), and poor (score: 3-5) survival, with 2-year OS of 84% (95% CI: 79-90%), 46% (95% CI: 36-59%), and 21% (95% CI: 14-32%), respectively (p<0.0001 for each). In the external validation set (182 patients), the score was predictive of OS (p<0.0001). The favorable, intermediate, and poor survival groups each demonstrated statistical significance (p<0.002), with 2-year OS of 77%, 50%, and 36%. Increasing SINS score was predictive of decreased OS as a continuous variable, with a hazard ratio of 1.17 (95% CI: 1.12-1.23, p<0.0001). Of the 6 components of the SINS score, 4 demonstrated associations with OS: pain, type of bone lesion, vertebral body collapse, and posterior spinal element involvement. <h3>Conclusion</h3> We propose this novel score as a decision-making tool to help optimize patient selection for spine SBRT. The SINS score may be an independent predictor of OS.