EP-1252 Brain metastases postoperative stereotactic RT: WBRT free survival predicted by an external nomogram

Abstract

Purpose or Objective: Patients with brain metastases (BM) represent an extremely diverse group with substantial variability in risk of intra-cranial failure and survival. Patient selection for stereotactic radiotherapy (SRT) alone is complex and requires considering multiple predictive factors. Given that the ultimate goal of SRT is to prolong survival without whole brain radiotherapy (WBRT), a nomogram based on multi-institutional data was developed by another team to display 6 and 12 months WBRT-free survival (WBFS) probabilities (Gorovets, al. IJROBP 2018). The aim of this study was to externally validate this nomogram. Material and Methods: We retrospectively reviewed the data of 70 patients treated between 2008-2017 by SRT for resected BM. The primary endpoint was the WBFS. Two subgroups of 35 patients were constructed with respect to the patient score in the nomogram (superior and inferior to the median). In each group, the observed WBFS was plotted against the predicted WBFS. The ROC curve and AUC were calculated for both 6 and 12 months time points. Results: After a median follow up of 16.8 months, the 1-year local and distant brain failure rates were 14.3% and 35.0 %, respectively. The median time to salvage WBRT was 9.6 months. Median time to the first intracranial failure was 7.7 months. At the time of first recurrence, 90 % received local salvage therapy. We performed repeated SRT for 34% and salvage WBRT in 56%. After median time of 8.5 months, 10 patients experienced a second intracranial failure. Eight of 10 patients received further salvage therapy (5 WBRT, 3 SRT). The WBFS rates at 6 and 12 months were 87 % [IC 95% = 79-95 %] and 56 % [IC 95% = 45-69%], respectively. In terms of calibration, the 6 months rates were overestimated while they were accurate at 12 months (Fig. 1). It is reflected by the evolution of the cumulative proportion of WBRT or death in both subgroups with observed rates inferior to expectations at 6 months while they superimpose at 12 months (Fig. 2). A ROC curve was plotted for the 6 and 12 months nomogram predictions. AUC values were 0.47 and 0.62, respectively. [Figure Presented] [Figure Presented] Conclusion: In our population of patients with operated BM and then stereotactically irradiated, the nomogram predicted correctly the WBFS at 12 months but not at 6 months. The median time to salvage WBRT was 9.6 months in our series while it was 6.2 months in the series by Gorovets. The fact that one third of our patients were treated with SRT as first salvage treatment may have played a role. Patients were operated and this may also be a confounding factor. All in all, it reflects the inherent limitation of applying prediction models to highly heterogeneous populations. Including more patients and characteristics (e.g. molecular data) may improve the concordance index.