AIMS: Current prognostic models are not accurate enough to identify brain metastases patients with very short survival, i.e. <2 months, who are unlikely to derive major benefit from whole brain radiotherapy. Our aim was to develop a more reliable model. MATERIALS AND METHODS: This was a retrospective analysis of a German database, which was used to develop a score, and an additional database from Norway, which was used for validation purposes. RESULTS: The groups included 67 and 32 patients, respectively. An analysis of prognostic factors resulted in a risk score based on performance status, extra-cranial metastases, the interval from breast cancer to brain metastases and a need for corticosteroid treatment, which classified 63 of 67 test patients correctly. However, the validation failed and unfortunately the risk score that performed best in the Norwegian patients (31 of 32 correctly predicted) was not applicable to the German patients. CONCLUSIONS: The prediction of short survival is associated with several caveats and seems to result in an unacceptable risk of withholding radiotherapy in patients who actually survive for longer than 2 months.