Page 94 - Predicting survival in patients with spinal bone metastasesL

CHAPTER VI
The most important limitation of this study is its retrospective design and the possibility of inaccurate interpretation of the source data. Also, the data were obtained from centers in two different European countries. Even though the quality of oncological care is considered to be excellent, differences in survival times between the centers were observed for some primary tumors, possibly resulting in lower predictive accuracy of the models. The multivariate analyses were adjusted for this center effect.
The Rades model was adjusted due to the fact that information concerning time to development of motor deficits could not reliably be obtained. Considering that neurologic deficit or ambulatory status were not associated with survival, it is unlikely that omitting this variable had a large effect on our results. Also, resectability of visceral metastases could not be assessed and all visceral metastases were considered unresectable. The effect of this on the accuracy of the Tokuhashi and Tomita models is unclear.
In order to predict survival, the evaluated models utilize a total of seven different variables. Each model has different guidelines on how to further categorize each variable, leading to different results for identical variables. As has been demonstrated, primary tumor, visceral metastases and performance status are associated with survival, whereas the number of SBM and the presence of extraspinal bone metastases are not.
On multivariate analysis, neurological functioning was also not associated with survival. Furthermore, including neurological functioning as a negative predictive factor in a model could prove to be disadvantageous. Restoration of ambulatory function in patients with acute deficit is the greatest gain possible in the treatment of SBM. If the estimated life expectancy would be too short based on a model incorporating neurological functioning, this patient would be denied the more aggressive, optimal treatment needed to improve his neurological functioning. Consequently, the use of a model would undermine, instead of improve the decision- making process. We therefore carefully suggest that neurological functioning and the duration of neurological deficit will be viewed as an indication for treatment, instead of a prognostic factor for estimating survival.
92