Evaluation of novel therapies for brain tumours should logically consider quality and quantity of patient survival as primary endpoints. The urgency of the problem, however, frequently leads investigators to use surrogate endpoints and historical controls in order to more rapidly evaluate outcome. To examine the impact of the use of surrogate endpoints and historical controls on the evaluation of innovative brain tumour therapy, selective literature review of three content areas (intraarterial chemotherapy for malignant glioma, interstitial brachytherapy for malignant glioma and stereotactic radiosurgery for cerebral metastasis and malignant glioma) was carried out. The impact of surrogate outcome measures and use of historical controls was assessed by comparing the results of trials using these methods and randomised clinical trials. In the evaluation of both intraarterial chemotherapy and interstitial brachytherapy, promising results in early phase trials were not confirmed in randomised clinical trials. This result can be explained by selection bias and predicted by the use of controls carefully selected from large treatment data bases. In the evaluation of stereotactic radiosurgery, early phase trials are promising, but randomised clinical trials have not yet been done. Prior experience suggests that the early promising results with stereotactic radiosurgery should be subjected to randomised clinical trial validation before being considered proven. Careful selection of controls for early phase trials is necessary if erroneous conclusions are to be avoided.