Prognostic models for patients with brain metastases after stereotactic radiosurgery with or without whole brain radiotherapy: a validation study Journal Article uri icon
  • PURPOSE/OBJECTIVE(S): To compare the performance of five prognostic models [RTOG recursive partitioning analysis (RPA), Score Index for Radiosurgery in Brain Metastases (SIR), Barnholtz-Sloan-Kattan nomogram (BSKN), diagnosis-specific Graded Prognostic Assessment (dsGPA), and Graded Prognostic Assessment for Lung Cancer Using Molecular Markers (Lung-molGPA)] against actual survival in patients with brain metastases treated with SRS +/- WBRT. MATERIALS/METHODS: 100 consecutive patients treated with SRS +/- WBRT between January 2006 and July 2012 were retrospectively analyzed. Patients were binned according to 33 percentiles of the predicted survival distribution for the BSKN and dsGPA models to compare with LungmolGPA, RPA and SIR. Pearson's correlation coefficients between predicted and observed survival were estimated to quantify the proportion of variance in observed survival. RESULTS: Median survival for the entire cohort was 13.5 months, with predicted vs actual MS by BSKN, SIR, dsGPA, RPA, adenocarcinoma Lung-molGPA, and nonadenocarcinoma Lung-molGPA was 3.8 vs 15.6 months, 7 vs 13.5 months, 9.4 vs 13.5 months, 10.3 vs 13.5 months, 13.7 vs 13.7 months, and 9.8 vs 9.7 months, respectively. The BSKN model and adenocarcinoma LungmolGPA created three groups with a statistically significantly different MS (p = 0.002 and p = 0.01, respectively). CONCLUSION: All models under-predicted MS and only the BSKN and Lung-molGPA model stratified patients into three risk groups with statistically significant actual MS. The prognostic groupings of the adenocarcinoma Lung-molGPA group was the best predictor of MS, and showed that we are making improvements in our prognostic ability by utilizing molecular information that is much more widely available in the current treatment era.

  • Link to Article
    publication date
  • 2018
  • published in
  • Brain Cancer
  • Models
  • Radiography
  • Retrospective Studies
  • Additional Document Info
  • 140
  • issue
  • 2