A Comparison of Prognostic Models to Facilitate Surgical Decision-Making for Patients With Spinal Metastatic Disease

Abstract

Study Design. Retrospective cohort. Objective. Compare the performance of and provide cutoff values for commonly used prognostic models for spinal metastases, including Revised Tokuhashi, Tomita, Modified Bauer, New England Spinal Metastases Score (NESMS), and Skeletal Oncology Research Group model, at three-and six-month postoperative time points. Summary of Background Data. Surgery may be recommended for patients with spinal metastases causing fracture, instability, pain, and/or neurological compromise. However, patients with less than three to six months of projected survival are less likely to benefit from surgery. Prognostic models have been developed to help determine prognosis and surgical candidacy. Yet, there is a lack of data directly comparing the performance of these models at clinically relevant time points or providing clinically applicable cutoff values for the models. Materials and Methods. Sixty-four patients undergoing surgery from 2015 to 2022 for spinal metastatic disease were identified. Revised Tokuhashi, Tomita, Modified Bauer, NESMS, and Skeletal Oncology Research Group were calculated for each patient. Model calibration and discrimination for predicting survival at three months, six months, and final follow-up were evaluated using the Brier score and Uno's C, respectively. Hazard ratios for survival were calculated for the models. The Contral and O'Quigley method was utilized to identify cutoff values for the models discriminating between survival and nonsurvival at three months, six months, and final follow-up. Results. Each of the models demonstrated similar performance in predicting survival at three months, six months, and final follow-up. Cutoff scores that best differentiated patients likely to survive beyond three months included the Revised Tokuhashi score = 10, Tomita score = four, Modified Bauer score = three, and NESMS = one. Conclusion. We found comparable efficacy among the models in predicting survival at clinically relevant time points. Cutoff values provided herein may assist surgeons and patients when deciding whether to pursue surgery for spinal metastatic disease.