• Baseline HRQOL

Patients were accrued to the study at their initial clinic consultation, before treatment was delivered. All demographic information and first HRQOL assessments were completed at this time. Kaplan–Meier OS curve was conducted in all patients, and in patients stratified by age or KPS, respectively. Log-rank test was performed to compare OS curves between age ≥65 and <65 years, or between KPS ≥70 and <70, as both age and KPS are significant prognostic factors in patients with brain metastases. A KPS of 70 was chosen as the cutoff because KPS <70 indicates special care is required for that individual. An age cutoff of 65 years was chosen based on the findings of the prognostic factor study conducted by Gaspar et al. [9]. Univariate Cox proportional hazard (PH) model of OS was conducted in all patients with demographic parameters, HRQOL and domain scores (0–100 continuous variables). The time (months) to death or the last follow-up was considered as the outcome variable. Patients alive or withdrew from the study had censored times based on the last follow-up date. To search for the parameters most predictive of time to death, univariate analysis was conducted. All variables with p < 0.10 obtained from the univariate analysis (including all potential variables with limited significance) were analyzed in multivariate analysis, using a backward stepwise selection procedure. A p-value <0.10 was used to include all potential variables with limited/borderline significant p-values from the univariate analysis. All variables with p < 0.05 in the multivariate analysis were retained and were the most predictive of prognosis.