Abstract
Background: Metastatic spinal cord compression (MSCC) carries a poor prognosis and management is based on the likelihood of maintaining mobility and predicted survival.
Patients and method: SCORAD is a randomised trial of 686 patients comparing a single dose of 8 Gy radiotherapy with 20 Gy in 5 fractions. Data was split into a training set (412, 60%) and a validation set (274, 40%). A multivariable Cox regression for overall survival (OS) and a logistic regression for ambulatory status at 8 weeks were performed in the training set using baseline factors and a backward selection regression to identify a parsimonious model with p ≤ 0.10. Receiver Operating Characteristic (ROC) analysis evaluated model prognostic performance in the validation set. Validation of the final survival model was performed in a separate registry dataset (n = 348).
Results: The survival Cox model identified male gender, lung, gastrointestinal, and other types of cancer, compression at C1-T12, presence of non-skeletal metastases and poor ambulatory status all significantly associated with worse OS (all p < 0.05). The ROC AUC for the selected model was 75% (95%CI: 69-81) in the SCORAD validation set and 68% (95%CI: 62-74) in the external validation registry data. The logistic model for ambulatory outcome identified primary tumour breast or prostate, ambulatory status grade 1 or 2, bladder function normal and prior chemotherapy all significantly associated with increased odds of ambulation at 8 weeks (all p < 0.05). The ROC AUC for the selected model was 72.3% (95% CI 62.6-82.0) in the validation set.
Conclusions: Primary breast or prostate cancer, and good ambulatory status at presentation, are favourable prognostic factors for both survival and ambulation after treatment.
Patients and method: SCORAD is a randomised trial of 686 patients comparing a single dose of 8 Gy radiotherapy with 20 Gy in 5 fractions. Data was split into a training set (412, 60%) and a validation set (274, 40%). A multivariable Cox regression for overall survival (OS) and a logistic regression for ambulatory status at 8 weeks were performed in the training set using baseline factors and a backward selection regression to identify a parsimonious model with p ≤ 0.10. Receiver Operating Characteristic (ROC) analysis evaluated model prognostic performance in the validation set. Validation of the final survival model was performed in a separate registry dataset (n = 348).
Results: The survival Cox model identified male gender, lung, gastrointestinal, and other types of cancer, compression at C1-T12, presence of non-skeletal metastases and poor ambulatory status all significantly associated with worse OS (all p < 0.05). The ROC AUC for the selected model was 75% (95%CI: 69-81) in the SCORAD validation set and 68% (95%CI: 62-74) in the external validation registry data. The logistic model for ambulatory outcome identified primary tumour breast or prostate, ambulatory status grade 1 or 2, bladder function normal and prior chemotherapy all significantly associated with increased odds of ambulation at 8 weeks (all p < 0.05). The ROC AUC for the selected model was 72.3% (95% CI 62.6-82.0) in the validation set.
Conclusions: Primary breast or prostate cancer, and good ambulatory status at presentation, are favourable prognostic factors for both survival and ambulation after treatment.
| Original language | English |
|---|---|
| Pages (from-to) | 77-83 |
| Number of pages | 7 |
| Journal | Radiotherapy & Oncology |
| Volume | 173 |
| DOIs | |
| Publication status | Published - 1 May 2022 |
