Page 122 - Predicting survival in patients with spinal bone metastasesL

CHAPTER IX
four categories; luminal A, luminal B, HER2 and triple negative. The luminal A subtype is ER positive and/or PR positive without HER2 overexpression. The luminal B subtype is ER positive and/or PR positive with HER2 overexpression. The HER2 subtype is ER and PR negative, with HER2 overexpression. Finally, the triple negative subtype is negative for all three receptors. Median survival times in the luminal A, luminal B and HER2 categories were not significantly different, whereas the median survival time in the triple negative category was much shorter. Median survival in the triple negative category was 5.5 months (95%CI 2.0-9.0), whereas survival in the other three combined receptor positive phenotypes was 23.4 months (95%CI 19.0-27.8) (log-rank test p<0.001). Based on this information we changed our primary tumor classification, by moving triple negative breast cancer patients into the ‘moderate’ clinical profile. All receptor positive phenotypes remained in the ‘favorable’ clinical profile. The adjustment resulted in an increase of 5% of the c-statistic calculated for this specific cohort.
Analogous to the study described above, Willeumier et al. studied the effect of epidermal growth factor receptor (EGFR) mutations in patients presenting with bone metastases from non-small cell lung cancer.15 It was shown that survival in patients with EGFR-positive tumors was much longer, due to the advent of new systemic treatment options.
As stated above, the aim was to develop a more accurate model than the ones available at the time. In order to ascertain whether we had achieved this, in Chapter VI we compared our own model to the ones created by Tokuhashi, Bauer, Tomita, Van der Linden and Rades.7-11 A database containing 1379 patients was created by combining our own data with data from Graz, Austria. All models were fitted with the available patient information and survival times and c-statistics were calculated for each model. Results showed that all models performed within a similar range (c-statistic 0.64-0.69), with our own model slightly outperforming the older models. A recently published study by Nater et al. evaluated a total of eight predictive models based on prospectively collected data in 142 surgically treated patients.16 C- statistics found in this study ranged from 0.61 to 0.68, similar to our own study.
In 2015, the Dutch national guideline on spinal metastases was updated and one of the topics addressed was the estimation of survival. In order to investigate which
120