Randomized clinical trials are the most objective method for evaluating new therapies, but they are subject to the same biases as nonrandomized studies unless the principles of statistical design are observed at the planning stage. Estimation of sample size also requires early careful consideration, since studies of inadequate size will not have sufficient statistical power to detect meaningful treatment differences. For ethical reasons, interim data monitoring procedures should be used to detect early treatment responses that may lead to alteration or interruption of the planned study to give patients early benefit from a superior treatment or diminish their risk from ineffective or harmful treatment. Additional important aspects of clinical trial design that were not addressed in this report include: definition of study objectives and endpoints, description of data to be collected, details of the treatment regimens, informed consent and plans for data analysis. The science of clinical trial design is complex; only some of the key statistical issues have been addressed briefly in this report.