Successful validation of a survival prediction model in patients with metastases in the spinal column.

Abstract

The Dutch Bone Metastases Study Group developed a survival prediction model in patients with symptomatic spinal bone metastases to guide the treating physician. The objective of this study was to validate the Dutch model and compare with our previously developed survival model at the Rapid Response Radiotherapy Program (RRRP model). The following prognostic factors were extracted from a prospective database in an outpatient palliative radiotherapy clinic: Karnofsky Performance Scores (KPS), primary cancer site, and visceral involvement for the Dutch model; primary cancer site, site of metastases, KPS, fatigue, appetite, and shortness of breath scores in the Edmonton Symptom Assessment Scale for the RRRP model. Patients were assigned scores according to each model. The survival probabilities were generated and calibration was performed for each model. A total of 231 patients with spinal bone metastases from 1999 and 2002 were included in the analysis. The survival probabilities were similar to those in the original models. The calibration comparing actual survival with predicted survival from the Dutch and RRRP models gave R2 values of 0.90 and 0.86, respectively. The two models were successfully validated. The Dutch model using three clinical prognostic factors was easier to administer.