To determine the outcome of patients with metastatic malignant melanoma (MMM) treated with palliative whole brain radiotherapy (WBRT) and to identify factors that predict treatment outcome to assist future trial design, a retrospective study was performed on patients with MMM who received WBRT at the Royal Marsden Hospital between 1998 and 2003. Data regarding patient factors, tumour factors and survival were collected. A total of 112 patients were identified and full data were available for 102 patients. The median age was 53 years (range 25-81 years), 66.7% were male and 33.3% female. The median dose prescribed was 20 Gy in five fractions as a mid-plane dose. The median survival after WBRT for the whole group was 51 days (range 3-1386). In an attempt to define prognostic groups, we used the validated RTOG recursive partitioning analysis (RPA) classification for brain metastasis (class 1: Karnofsky Performance Score (KPS) >/=70%, age <65 years with no extracranial metastasis; class 3: KPS <70%; class 2: all others). The median survivals were 151, 71 and 21 days for RPA class 1, 2 and 3, respectively (P<0.001). Multivariate analysis showed that RPA class, leptomeningeal involvement, presence and number of extracranial metastatic sites and progressive disease in the brain on imaging before WBRT are important independent predictive factors. A prognostic index was derived from these factors that allowed identification of patients unlikely to benefit from WBRT. In conclusion, the RTOG RPA classification is valid when applied to patients with MMM. Patients in RPA class 1 and good prognosis class 2 are likely to benefit from palliative WBRT and should be considered for entry into trials that aim to improve duration of response. We identified that patients with RPA class 3, leptomeningeal involvement or RPA class 2 with poor prognostic index are unlikely to benefit from palliative WBRT.