The clinical utility of prognostic scoring systems in patients with brain metastases treated with radiosurgery

Abstract Purpose The RTOG recursive partitioning analysis (RPA) classification is the gold standard for assessing the prognosis of patients with brain metastases (BM). Newer prognostic scoring systems for BM patients have been proposed, but their superiority over RPA needs to be established for patients treated with radiosurgery. Methods 380 patients with 1–3 BM were treated at the VUmc with radiosurgery (RS) from 2002 to 2011. Using baseline characteristics, patient scores were calculated for RPA, the Rotterdam-system, the score index for radiosurgery (SIR), the basic score for BM (BSBM), the graded prognostic assessment (GPA), the diagnosis-specific GPA, the Rades score, and the Golden grading system (GGS) for comparison with survival time and survival classification (⩽3 months or ⩾12 months). Results Median survival after RS was 7.7 months, with 3-month and 1-year overall survival (OS) of 76% and 39%, respectively. Multivariate analysis confirmed the prognostic value of performance status, age, absence of extracranial metastases, primary tumor site, gender, and steroid response for OS. The percentage of patients included within the intermediate prognostic classes ranged from 48% to 77%, and was 64% for the RPA. All scoring systems highly correlated with OS ( p < 0.001). The specificity for predicting early death ranged from 85% to 98% (RPA 88%), with the unfavorable classes of Rades, GGS, BSBM and SIR performing best. The sensitivity for predicting long-term survival ranged from 10% to 69% (RPA 29%), and was highest for the favorable classes of Rades and GGS. Conclusions All prognostic scoring systems correlated very well with OS. All scores shared the limitation of unbalanced proportions of patients within the prognostic classes. As the clinical superiority of more recently developed prognostic scoring systems was only modest in predicting early death and long term survival, the well-known and easy to use RPA system currently remains the standard.