Prognostic factors in brain metastases: can we determine patients who do not benefit from whole-brain radiotherapy?

Abstract Aims Whole-brain radiotherapy (WBRT) is a standard treatment recommendation for patients with brain metastases. The goal of treatment is symptom control, which in the short run can be often achieved by steroids. Patients with a short life expectancy may not derive benefit from the addition of radiation. The ability to identify this group would aid the decision of whether to recommend WBRT. Materials and methods Data on all patients referred for WBRT to palliative radiotherapy teams at two comprehensive cancer centres were prospectively collected over a 2-year period. The most frequent radiation dose and fractionation was 2000cGy in 5 fractions over 1 week. Multivariate logistic regression analysis using forward stepwise selection process was used to develop a prediction model for early death (before 8 weeks). The variables considered were sex, performance status, primary disease, weight loss, age, time from primary diagnosis to brain metastases diagnosis, number of metastatic sites and Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis (RPA) status. Results Two hundred and seventy-five patients with brain metastases were assessed. The median follow-up was 19 weeks, and estimated median overall survival was 5.3 months. Multivariate regression revealed Eastern Cooperative Oncology Group (ECOG) performance status and number of metastatic sites to be significant predictors of early death. The odds ratios were 2.38 (95% confidence interval [CI] 1.77–3.19) and 1.39 (95% CI 1.07–1.81), respectively. Sixty-eight per cent of patients could be correctly classified; however, 55% would have been incorrectly predicted to die early. Conclusions Poor performance status and number of metastatic sites are useful predictors of early death. A regression model was highly predictive overall, but this was primarily due to a high negative predictive value of 86%. The ability to predict patients who would die early (positive predictive value) was only 45%. Thus, despite the understanding that we are over-treating a subset of patients, further research is required to identify patients who do not require radiotherapy.