Prognostic factors for survival with metastatic spinal cord compression in the SCORAD randomized trial

Abstract

Purpose/Objective(s): SCORAD randomized 686 patients with metastatic spinal cord compression (MSCC) to receive either a single dose of 8Gy or 20Gy in 5 fractions; no significant difference in survival, recovery of ambulatory status or pain control was seen (Hoskin et al, 2019). Survival with MSCC after radiotherapy is poor and variable. The aim of this study was to establish a prognostic model for predicting overall survival in patients with MSCC. Materials/Methods: The median overall survival in SCORAD is 13 weeks. This time point has been used for the analysis. SCORAD data was randomly divided into a training set (412, 60%) and a validation set (274, 40%) matched for key characteristics: treatment, ambulatory status, primary tumor and extent of metastases. Multivariable Cox analysis used twelve factors on the training set to select prognostic factors for overall survival: treatment, sex, extent of metastases, number of MSCC sites, baseline bowel and bladder function, prior chemotherapy, radiotherapy or hormonal status, baseline ambulatory status, primary tumor and location of MSCC. Using backward elimination with a significance level of 0.20 and keeping treatment in the model, five factors were retained as prognostic factors: sex, extent of metastases, primary tumor, baseline ambulatory status and location of MSCC. Receiver operating characteristic (ROC) analysis was performed to evaluate the accuracy of the reduced model in predicting 13-week overall survival. Specific cut-off values associated with sensitivity rates for fixed false positive rate values were identified. Results: The model indicates that, adjusted for treatment, females had decrease risk of death compared with males (HR:0.66: 95% CI 0.47 to 0.92, p Z 0.02) as well as patients with SCC located in L1-S2 (HR:0.69: 95% CI 0.52 to 0.92, p Z 0.01) and T6-L5 (HR:0.71:95% CI 0.44 to 1.16, p Z 0.17) compared with C1-T12. Poor survival was associated with presence of nonskeletal mets [HR: 1.31: 95%CI 1.04 to 1.65 , p Z 0.02], having lung (HR:3.98: 95% CI 2.85 to 5.54, p<0.001), gastrointestinal cancer (HR: 2.97, 95% CI 2.02 to 4.37, p<0.001) or a cancer other than prostate or breast cancer (HR:2.4: 95% CI 1.67 to 3.44, p<0.001). Patients classified with impaired ambulatory status at baseline were also at increased risk of death (grade 2, HR: 1.72: 95% CI 1.27 to 2.34, p Z p<0.001; grade 3, HR: 2.5, 95% CI 1.79 to 3.48, p<0.001; grade 4, HR: 2.04: 95%CI 1.3 to 3.2, p Z 0.002). The prognostic performance of the model was associated with a 13-weeks OS ROC area under the curve (AUC) of 0.75, 95%CI: 0.69 to 0.81. The risk score 0.996 for the 13- weeks OS had a false positive rate of 30% and corresponding sensitivity of 65.63%. Conclusion: The accuracy of this prognostic model developed in SCORAD is limited and is below the recommended level of AUC >0.85 for a strong prognostic performance. However, this model will guide patient management and provide the basis for stratification factors in further studies.