Between February 1990 and February 1993, 25 centres in the UK recruited 544 patients into a prospective randomized trial comparing two whole-brain radiotherapy regimens (30 Gy in ten fractions over 2 weeks versus 12 Gy in two fractions on consecutive days) for the treatment of patients with symptomatic cerebral metastases. Of these patients 533 were eligible for analysis: 270 assigned to the two-fraction arm and 263 to the ten-fraction arm. The two groups were well balanced with respect to patient characteristics. Median survival was 77 days with two fractions (95% CI 68-89) and 84 days for the longer schedule (95% CI 67-102). Analysis of the survival curves showed a marginal advantage for ten fractions (P = 0.04). Performance status (P = 0.0001), site of primary tumour (P = 0.006), dose of dexamethasone (P = 0.004), age (P = 0.04) and randomization treatment (P = 0.03) were independant factors associated with survival. The classification of patients into good or poor risk groups based on these factors, excluding treatment, showed highly significant differences in survival (P < 0.0001). Predictive models suggested that any benefit attributable to the longer radiotherapy schedule was confined to those in a good prognostic group (these patients formed 22% of the study population). Radiation related side effects, other than alopecia, were seen in 12% of patients receiving two fractions and 8% of those given ten fractions. The short survival of many patients hampered the assessment of response, but overall responses were seen in 39% of those given two fractions and 44% of patients receiving ten fractions. These results suggest that any increase in survival due to longer radiotherapy treatment is confined to good prognosis patients, but, for the majority, there is no advantage and the value of radiotherapy for these patients relates purely to the possibility of control or relief of distressing symptoms.