Page 95 - Predicting survival in patients with spinal bone metastasesL

In 1990 Tokuhashi et al. created the first predictive model for SBM19 and the 2005 revised version9 is still one of the most commonly reported models to date. It was based on a retrospective analysis of 246 patients and uses six items to assign patients into one of three categories. In this study, three of these items were not associated with survival. The primary tumor classification was based on average survival periods for each cancer, but due to the low number of patients this was not very accurate. Low scores for both the worst and best predictive groups make it difficult to justify the use of this model in a clinical setting.
The original Bauer model was based on a retrospective analysis of 88 patients with spinal metastases and 153 patients with bone metastases in the extremities. Due to the fact that the presence of a pathologic fracture negatively influenced survival only in the extremities group, Leithner et al. modified the score for use in SBM. Even though it is one of the easiest models to use, the simple primary tumor classification and the lack of a performance score resulted in a low accuracy.
The Tomita model was based on a retrospective analysis of 67 patients and essentially consists of four items, as the ‘bone metastases’ category includes both spinal and extraspinal metastases. The primary tumor classification in slow, moderate and fast growing tumors has been shown to be quite accurate. Unfortunately, this model also lacks the incorporation of a performance score, resulting in a c-statistic identical to the Tokuhashi and Bauer models.
The Van der Linden model was based on 342 patients treated for painful SBM by means of radiotherapy within a multicenter randomized phase III trial. The model contained all three risk factors that were shown to be associated with survival in this article and has a c-statistic of 0.66. It performs better than the previous models, especially when identifying patients with a longer expected survival.
The Rades model was based on a retrospective analysis of 2029 patients, all of whom were treated with radiotherapy. With seven items, it is the most elaborate score. The aim of this model was to identify patients with poor survival who may be candidates for best supportive care and it achieved this goal. In a clinical setting however, patients with a longer survival also need to be identified correctly and the model fails to do so. It is the only score that takes time to development of SBM
VI
SCORING SYSTEMS
93